iraq casualties and the 2006 senate elections - francis x. …€¦ · · 2015-06-06iraq...

507Iraq Casualties and Senate Elections

LEGISLATIVE STUDIES QUARTERLY XXXII 4 November 2007 507

DOUGLAS L KRINERBoston University

FRANCIS X SHENHarvard University

Iraq Casualties and the2006 Senate Elections

Prior scholarship on the effects of war casualties on US elections has focusedon large-scale conflicts For this article we examined whether or not the much-smallercasualty totals incurred in Iraq had a similar influence on the 2006 Senate contestsWe found that the change in vote share from 2000 to 2006 for Republican Senatecandidates at both the state and county level was significantly and negatively relatedto local casualty tallies and rates These results provide compelling evidence for theexistence of a democratic brake on military adventurism even in small-scale warsbut one that is strongest in communities that have disproportionately shouldered awarrsquos costs

In the immediate aftermath of the Democratsrsquo sweeping victoryin the 2006 midterm elections many political punditsmdashlike modern-day augurs divining auspices from exit-polling datamdashwere quick topronounce the elections a resounding referendum on the Bushadministrationrsquos conduct of military operations in Iraq While the Iraqwarrsquos electoral consequences appear obvious to mainstream newsoutlets a fervent debate continues among political scientists aboutwhether or not foreign affairsmdasheven major warsmdashhave significanteffects on federal elections (Aldrich Sullivan and Borgida 1989 GelpiReifler and Feaver 2007 Hess and Nelson 1985 Nincic and Hinckley1991)

This debate is particularly important to revisit within the contextof the current Iraq war because this conflict has involved considerablysmaller casualty totals than other major American wars Scholarshipto date on the relationship between war casualties and congressionalelectoral fates rests on data from the Civil War and Vietnam (Carsonet al 2001 Gartner Segura and Barratt 2004) Yet Vietnam involved17 times more casualties than Iraq and the Civil War an astounding170 times the Iraq tally1 Both the Civil War and Vietnam also involved

508 Douglas L Kriner and Francis X Shen

conscription and significant draft resistance escalating to violence(Foley 2003 Schecter 2005) The relatively low number of casualtiessustained in Iraq coupled with the absence of large-scale resistanceon par with the draft riots of earlier eras raises questions about scaleDoes a threshold exist below which casualties will not affect senatorsrsquoelectoral fates Has the casualty count in Iraq reached that thresholdWe confront both questions in this article

In addition to Iraqrsquos critical importance in the off-year electionsthe 2006 midterms were also marked by an unusually high percentageof Americans responding that national not local concerns were themotive forces behind their congressional votes The growing national-ization of congressional elections since Tip OrsquoNeil coined his apho-rism ldquoall politics are localrdquo is well documented (Brady Cogan andFiorina 2000 Jacobson 2004) For this study however we investi-gated the possibility that even the most national of issuesmdashthe war inIraqmdashmay have a strong local component Previous studies havedemonstrated the influence of local casualties on public opinion(Gartner and Segura 1998 Gartner Segura and Wilkening 1997) butthere is scant evidence that local casualties at lower levels of aggrega-tion than the state or congressional district influence electoral outcomesWe correct this deficit by exploring how the Iraq warrsquos influence onvoting returns was critically mediated by local casualty rates at boththe state and county levels

We begin our discussion by examining the influence of a statersquosshare of Iraq casualties on the change in that statersquos Republican sena-torial vote share from the 2000 to the 2006 elections2 We then shift tothe county level where there is considerably greater variance in wartimeexperiences to examine the relationship between local casualties andchanges in Republican electoral fortunes at a lower level of geographicaggregation Finally we narrow our focus to the effect of local casualtieson the 14 Republican incumbent senators seeking reelection in 2006all but two of whom voted to authorize military action against Iraq in20023 We show that Republican senatorial candidates lost ground fromtheir 2000 performance in states and counties hit hard by the war inIraq but generally fared no worse in states and counties that had thusfar emerged from the war relatively unscathed Our findings suggest aremarkable degree of casualty sensitivity among the American elec-torate The historically modest number of casualties suffered in theIraq war has not spurred riots in the streets but it has produced signifi-cant negative reactions in the voting booth


Theory

Aside from 911 the most frequently repeated number inAmerican politics today is the number of American soldiers slain inIraq There are many metrics on which the public might evaluate thewar and its conduct from dollars spent to strategies employed to Iraqicivilian lives lost The number of American casualties however is themost concrete and publicly visible measure of the warrsquos costs andconsequences Indeed war casualties lie at the heart of political sciencetheories in multiple subfields In international relations research theadverse reaction of the public to combat casualties is central to manyinstitutionally based theories of the Democratic peace (Maoz andRussett 1992 Ray 1995 Reiter and Stam 2002 Russett and OrsquoNeal2001 Siverson 1995)

Similarly a lengthy literature in American politics examines theimpact of foreign policy in general and wartime casualties in particu-lar on presidential approval (Eichenberg Stoll and Lebo 2006 Hurwitzand Peffley 1987 Mueller 1973 Nickelsburg and Norpoth 2000)Scholars of public opinion have also examined the effect of casualtiesand casualty framing on support for the military campaign (Berinskyand Druckman 2007 Boettcher and Cobb 2006 Feaver and Gelpi 2005Gartner and Segura 1998 Gelpi Feaver and Reifler 2005 Larson1996) Other research has demonstrated casualtiesrsquo influence on moretangible political outcomes specifically on presidential election results(Cotton 1986 Gelpi Reifler and Feaver 2007 Karol and Miguel 2007)

Whether in international relations or American politics thecasualties hypothesismdashthat casualties may have significant bearingon political outcomesmdashinvolves a question of scale Starting withMuellerrsquos (1973) consideration of cumulative casualty counts scholarshave discussed the possibility of a casualty threshold casualties mayaffect opinion and elections more once these deaths rise past a certainlevel Indeed the scholarly debate in International Security betweenKlarevas Gelpi and Reifler (2006) focused in large part on Feaverand Gelpirsquos (1999) claim that the American publicrsquos casualty thresh-old for Iraq was ldquonot just hundreds but thousandsrdquo (B3) The casualtythreshold debate has for the most part surfaced only in the context ofbattle deathsrsquo influence on public opinion and not on electoral out-comes The comparatively few scholars who have examined theinfluence of casualties on congressional elections have yet to adequatelyaddress the scale question Carson et al (2001) found a strong rela-tionship between district-level casualties in the Civil War and votingpatterns in the House midterm elections of 1862ndash1863 Gartner Segura


and Barratt (2004) found that state-level casualties in Vietnamnegatively affected incumbent senatorsrsquo vote shares in the 1966ndash1972elections

Our study with its focus on the Iraq war extends the literatureby testing theories of casualtiesrsquo political import in a conflict withconsiderably smaller casualty totals With the number of casualties inIraq almost 20 times lower in 2006 than the number of US casualtiessustained in Vietnam it is not clear a priori that casualties shouldnecessarily have a sizeable effect on incumbent Republicansrsquo voteshares in the 2006 elections Our study asks if the same dynamicsbetween local casualty rates and electoral behavior that were presentin prior high-casualty conflicts continue to operate in the context ofthe war in Iraq

In addition to addressing this question of scale we also explorewhether or not the wartime experiences of local communitiesmdashnotonly statesmdashaffect voting behavior Prior research from the Vietnamera suggests that at both the individual and aggregate levels publicopinion on the war was highly responsive to variations in county-levelcasualties (Gartner and Segura 1998 2000 Gartner Segura andWilkening 1997) To explain how local casualties might influencepolitical attitudes and behavior scholars have posited at least threeplausible mechanisms The first stresses direct personal contact withthe costs of war Voters from high-casualty communities have a greaterprobability of direct personal contact with the human costs of warthrough their social networks (Moody 2006) The second positedmechanism is casualtiesrsquo indirect influence on the public through theirinfluence on political elites (Berinsky Nd Brody 1991 Larson 1996Zaller 1994) Local elites may respond to casualties within their con-stituencies and in turn influence mass opinion and behavior A thirdmechanism emphasizes the role of local media coverage from whichthe majority of Americans obtain their news (Gilliam and Iyengar 2000)If local news outlets adjust the scope and tone of their war coverage tofit the wartime experience of the local community then individualsfrom high-casualty communities may be exposed to a greater volumeof negative coverage of the war and its human costs than individualsfrom low-casualty communities (Gartner 2004)

All three of these mechanisms offer reasons why voters mayrespond to casualties sustained at the state level but they all also suggestthat casualties suffered at the community or county level may affectelectoral outcomes Prior studies have only demonstrated the influ-ence of variance in casualty figures on voting behavior at higher levelsof aggregation (Carson et al 2001 Gartner Segura and Barratt 2004)


The one study that has examined the influence of county-level casualtieson election outcomes Karol and Miguelrsquos (2007) analysis of the 2004election found a strong negative relationship between state-level Iraqcasualties and the change in George W Bushrsquos vote share but no relation-ship between county casualties and electoral support for the president

Nevertheless there are strong reasons to believe that the conse-quences of casualties were significantly different in 2006 than twoyears earlier at the time of Karol and Miguelrsquos study By 2006American casualties in Iraq had mounted and conditions on the groundworsened Through the end of 2004 the United States had suffered1334 casualties in Iraq in 2005 and 2006 the United States sufferedan additional 1670 fatalities By the midterm elections there werealso increasing calls from both inside and outside of Washington forsignificant changes in military strategy Moreover because levels ofpublic information are considerably lower on average in midterm thanin presidential elections the simple retrospective frame of whetherthe situation in Iraq had improved or worsened may have been evenmore powerful in 2006 than it was in 2004

We conducted a new test of whether or not both state- and county-level casualties can affect congressional electoral outcomes specificallyRepublican senatorial fortunes in the 2006 midterms We expectedthat voters in those localities that had suffered the largest numbers andhighest rates of casualties in Iraq would punish Republican senatorialcandidates the most particularly incumbents who voted for the warand continued to support the president

2006 Midterm Elections

Iraq was the centerpiece of virtually all US Senate elections in2006 Casualties which had been 1334 at the end of 2004 but hadmore than doubled to nearly 3000 by the midterms were necessarilya part of the debates NBC host Tim Russert stated the issue clearly inhis first question during a nationally televised debate betweenMarylandrsquos Senate candidates ldquoVoters in Maryland all across thecountry say the big issue for them this year is Iraqrdquo4 In anothernationally televised debate this time between Missourirsquos Senatecandidates Russert brought the casualties issue center stage when hesaid to Senator Jim Talent ldquoHerersquos the headline in todayrsquos paper lsquoUSCasualties in Iraq Rise Sharplyrsquo The number of people American troopsbeing killed and attacked every 15 minutes and yoursquore saying itrsquosgoing wellrdquo5 Similar questions were asked of candidates across thecountry


Consistent with Jacobson and Kernellrsquos (1981) theory of strategicchallengers Democrats challenging Republicans tried to link the sit-ting politicians with the growing body count In the Missouri Senatedebate the Democratic challenger Claire McCaskill described the Iraqwar as ldquoa failed policy where wersquore mired in a civil war where we arelosing lives every day and innocent Iraqi livesrdquo6 In Ohio Democraticchallenger Sherrod Brown made similar strategic moves to connectthe policy of his opponent sitting Republican senator Mike DeWineto Ohio casualties As part of a press release to back up the facts of atelevision advertisement attacking DeWine on Iraq Brownrsquos campaignwrote ldquoFACT Mike DeWine Still Supports lsquoStay the Coursersquo in Iraqrdquoand followed that with the number of fatalities casualties and Ohiofatalities7 Brownrsquos mention of the 123 Ohio fatalities that had occurredup to that time (September 29 2006) is evidence that the Browncampaign believed that local casualty counts would play to voter sym-pathies more than aggregate national figures The ad itself whichfeatured a local woman whose son was in Iraq without proper bodyarmor reinforced the theme the choices senators make about interna-tional conflicts have local consequences

Democrats also tried to paint Republicans as being politicallymotivated in ignoring the casualty count In New Jersey incumbentDemocratic senator Robert Menendez argued that his Republicanopponent and President George W Bush were ldquoliving in an alternativereality where intelligence findings donrsquot matter mounting casualtiesdonrsquot count and rhetoric about the war on terror is more importantthan resultsrdquo8 The Senate race in neighboring Pennsylvania saw asimilar theme emerge Criticizing Republican senator Rick Santorumrsquosrecord on Iraq the Democratic challenger Bob Casey Jr attackedSantorumrsquos silence amidst growing Pennsylvania casualties ldquoHerepresents the state that has the biggest National Guard contingentover there the state that ranks fourth in the number of casualties Yethe hasnrsquot been able to muster one word of criticism Maybe he doesnrsquothave the independence to ask the tough questionsrdquo9

In all of these races the Democratic candidates attempted to bringthe casualty questionmdashwith a particular focus on the losses sufferedby their respective statesmdashcloser to the foreground The fact thatcasualties were an issue so central to these campaigns lends credenceto our theory that local casualties are likely to be a significant factor inexplaining Republican losses in 2006

With both strategies the message these candidates were sendingto voters was clear if you vote for my Republican opponent wersquoregoing to experience more casualties than if you vote for me Each of


these four Democratic Senate candidates eventually won their racesWhether or not local casualties were part of the reason for thesevictories is the focus of our empirical analysis

Data and Methods

While the Iraq war has certainly affected public opinion andpolitical conditions nationwide the most direct cost of the warmdashitshuman tollmdashhas been borne unequally across society As of November2006 Wyoming had suffered the fewest casualties seven and Cali-fornia had suffered the most 298 In terms of casualty rates throughNovember 2006 the average state had suffered just under 11 casualtiesper one million people but there was also considerable variance aroundthat mean As of the 2006 midterms Vermont had paid the highestprice per capita with a casualty rate of almost 30 deaths per millionresidents Conversely New Jersey had the lowest casualty rate at justover 5 deaths per million At the county level the disparities wereeven more dramatic More than half of all counties had not sufferedany casualties in Iraq while Los Angeles County had suffered 74 Evenafter one controls for population differences across counties thedisparities remain extreme More than 70 of counties had experienceddeath rates in Iraq of less than 1 per 100000 residents But 13 ofcounties had suffered casualty rates of more than 3 per 100000 andmore than 70 counties had suffered casualty rates of greater than 10per 100000

To examine the effects of this uneven geographic distribution ofthe Iraq warrsquos costs on the 2006 midterm elections we constructedmodels of the change in vote share of Republican senatorial candi-dates from 2000 to 2006 at both the state and county level Iraqcasualties might have affected the calculus of American voters at leasttwo ways First the total number of combat fatalities suffered in Iraqmight have encouraged voters to abandon the Republicans who despitesome internal divisions within both parties remained the most stead-fast supporters of the presidentrsquos course in the Middle East If thiswere the only mechanism by which the war affected the election out-comes then Iraqrsquos adverse effect on Republican vote shares shouldhave been felt nationwide with little or no geographic variance Insuch a world we would find no evidence that Republican candidatesdid any better or worse on average in high-casualty statescountiesthan in low-casualty statescounties

Alternately although sensitivity to American casualties as a wholeundoubtedly influenced voting decisions the publicrsquos perspectives on


the war might also have been moderated by the experience of theirlocal communities10 If so then residents of states and counties thatsuffered disproportionately high casualty totals or rates might also havefelt the warrsquos costs more acutely and punished the ruling Republicansdisproportionately

Because both mechanisms may have been operative any evidenceuncovered for state and local casualtiesrsquo influence would be a conser-vative estimate of the warrsquos total effect on the election since themounting costs of the conflict may have had an additional uniformeffect on voters as a whole Still evidence for the continued influenceof state and local casualties above and beyond any national reaction tothe casualty total would greatly strengthen the theoretical contentionthat Americansrsquo attitudes toward war are critically mediated throughthe lens of their local communities Even an issue as national as thewar in Iraq may have a strong local component

Our empirical analysis proceeds in three stages the first discussesthe casualtiesrsquo effects on all senatorial election results at the state levelthe second reveals the influence of casualties on every Senate contestat the county level and the third focuses narrowly on casualtiesrsquo effecton the county-level returns for the 14 Republican incumbents seekingreelection in 2006 In the first two stages we included all states withsenatorial contests except for Connecticut and Vermont Because thesecontests were complicated by strongmdashindeed favored third-partycandidatesmdashthey were excluded from the analysis11

At both the state and county levels we modeled the change inRepublican senatorial vote share as a function of state-level casualtiesand a number of political economic and demographic control variablesdrawn from prior research An extensive literature has identified op-ponent quality (Green and Krasno 1988 Jacobson 2004 Squire 1992)and campaign spending (Abramowitz 1989 Gerber 1998 Jacobson1978 1990) as two of the most important predictors of a candidatersquoselectoral fortunes12 To account for changes in opponent quality wecoded each Republicanrsquos opponent according to Green and Krasnorsquos(1988) eight-point ordinal scale and we calculated the change in thismeasure across the two electoral cycles To control for the influenceof campaign expenditures we included the change in the percentageof total campaign expenditures spent by the Republican candidate from2000 to 200613

In addition to factors specific to the Senate race at hand scholarshave long documented the connections between presidential perfor-mance and the success of his copartisans in presidential elections evenin midterm contests (Abramowitz and Segal 1986 Campbell 1991


Campbell and Sumners 1990 Carsey and Wright 1998) To accountfor this relationship in the current context we included a measure ofPresident George W Bushrsquos share of the two-party vote in each stateor county in the 2004 election

Additionally a number of previous studies have explored anddebated the relative importance of economic conditions for congres-sional election outcomes (see Squire 1995 for a review) To controlfor economic factors we included measures of the change in the stateand county unemployment rates (obtained from the Bureau of LaborStatistics) over the year preceding the 2006 midterm elections Votersin areas with increasing unemployment rates may be more likely topunish Republican candidates in this era of unified Republican controlof Congress and the presidency

Finally our models also controlled for two important demographicconstituency characteristics that might be correlated with consider-able change in Republican electoral fortunes from the peacetimeelection of 2000 to the wartime 2006 contest the percentage of resi-dents aged 18 to 64 serving in the military and the percentage of allresidents who were veterans of the armed forces We constructed thedemographic controls from the US Census Bureaursquos summary files(sf3) for the 2000 Census Conventional wisdom suggests that militarycommunities have largely rallied around the president and thepresidentrsquos policies if so then Republican candidates may haveperformed better relative to their 2000 baseline in these areas than inotherwise comparable communities Additionally an extensiveliterature regarding political elites has examined the different perspec-tive that veterans bring to questions of military policy (see for exampleFeaver and Gelpi 2005) Yet expectations for electoral behavior in statesor counties with large veteran contingents at the mass level are lessclear Communities with large contingents of veterans like those withhigh percentages of active-duty personnel and their families may haverallied around the president and the Republicans in the 2006 midtermsor they may have viewed the war and the administrationrsquos militarypolicies through a distinctly different and more critical lens and adjustedtheir voting behavior accordingly We tested these competing hypotheses

As for the explanatory variable of interest Iraq casualty data weobtained information on each soldierrsquos home state and county of recordfrom the Statistical Information Analysis Division of the Departmentof Defense14 Because geographic data is frequently unavailable forsoldiers wounded in Iraq we limited our definition of casualties tothose killed in action15 For both the state- and county-level analyseswe employed two operationalizations of a localersquos war losses the raw


casualty count and the casualty rate per one million and per 10000residents for states and counties respectively16

We estimated all models with ordinary least squares (OLS)regressions and Hubert-White heteroskedasticity-consistent standarderrors (White 1980) according to the following specification

(GOP Senate Vote 2006)i ndash (GOP Senate Vote 2000)i= α + β1 (Iraq Casualties)i + β2 (ΔOpponent Quality)i

+ β3 (Δ GOP Campaign Expenditures)i+ β4 (Bush Vote 2004)i + β5 (ΔUnemployment Rate)i+ β6 ( 18ndash64 in Armed Forces)i + β7 ( Veterans)i + εi

Results and Discussion

State Level

At first blush there is considerable evidence that local casualtieshad a significant negative effect on Republican electoral fortunes inthe 2006 Senate races The scatterplot in Figure 1 suggests a strongnegative relationship between a statersquos casualty rate and the Republicansenatorial candidatersquos electoral fortunes17 This simple bivariateanalysis indicates that an increase in a statersquos casualty rate of fivecasualties per million residents (approximately one standard deviation)cost the Republican candidate about five percentage points at theballot box

The negative relationship also appears robust at the county levelConsider the following numbers By November 2006 10 of countieshad suffered two or more casualties in Iraq since the war began inMarch 2003 Republican senatorial candidates captured 55 of thevote in these counties in 2000 A year and a half into the war in 2004President Bush secured 54 of the two-party vote in these localesBut a mere two years later Republicans won only 48 of the vote inthe Senate contests Contrast this precipitous decline with the perfor-mance of Republican candidates in the counties that experienced nocasualties in Iraq prior to the election In these counties the Republicancandidate won 57 of the vote in 2000 President Bush won handilyin these areas in 2004 garnering 62 of the vote And in 2006Republican candidates continued to do well earning 55 of the two-party vote share Limiting the analysis to the 993 counties in which 14incumbent senators ran for reelection in 2006 reveals a seven percentagepoint decrease from their 2000 totals in the two-or-more casualtycounties In counties that experienced no casualties in Iraq the


Republican candidates gained 65 of the vote on average in both the2000 and the 2006 elections Certainly something seems afoot

To explore casualtiesrsquo effects on the midterm elections moresystematically we examined a series of models for both the state andcounty levels Results for the change in GOP vote share at the statelevel appear in the first two data columns in Table 1

Even after controlling for the political economic and demo-graphic factors already discussed we found the coefficients for boththe statersquos casualty tally and rate to be negative as expected althoughonly the coefficient for the casualty rate per one million residents isstatistically significant More importantly the empirical model indicatesthat the substantive size of a state casualty ratersquos effect on the changein GOP vote share is considerable a finding consistent with the bivariaterelationship illustrated in Figure 1 A one standard deviation increaseof 46 casualties per million residents cost the Republican candidateon average over seven and one-half percentage points at the pollsThe size and robustness of this result strongly suggest that as they didin the Vietnam years (Gartner Segura and Barratt 2004) state-levelcasualties strongly influenced Senate electoral dynamics in 2006

FIGURE 1Scatterplot of State-Level Casualty Rates

and Change in GOP Senate Vote Share

AZ

CA

CT

DE

FL

HI

INMA

MDME

MIMN

MO MS MT

ND

NE

NJ

NM

NV

NY

OH

PARI

TN

TXUT VA

VT

WAWI

WV

WY

Cha

nge

in G

OP

Vote

Sha

re

Casualties per Million Residents

20

10

0

ndash10

ndash20

ndash30

5 10 15 20 25 30


TABLE 1The Effect of State and County Casualties

on the Change in GOP Senate Vote Share 2000ndash2006(standard errors in parentheses)

State State County County GOP Inc GOP Inc

Iraq State Count ndash001(002)

Iraq State Rate ndash146(59)

Iraq County Count ndash023 ndash041(013) (017)

Iraq County Rate ndash006 ndash099(059) (047)

Change in Opponent Quality ndash069 ndash030 ndash182 ndash182 019 020(054) (058) (011) (011) (011) (011)

Change in GOP Spending 25 33 03 03 73 75(18) (14) (03) (03) (06) (06)

Bush 2004 044 078 011 012 021 023(042) (037) (002) (003) (003) (003)

Change in Unemployment 971 1001 ndash014 ndash016 ndash016 ndash025(594) (507) (048) (048) (057) (057)

in Military 306 443 ndash013 ndash018 021 016(329) (224) (014) (015) (013) (011)

Veterans ndash109 28 ndash031 ndash028 ndash081 ndash081(105) (123) (010) (010) (012) (012)

Constant 170 ndash1269 ndash638 ndash775 253 114(1667) (1370) (221) (231) (257) (257)

Observations 31 31 1856 1856 993 993R2 29 41 15 15 21 20

p lt 10 p lt 05 p lt 01 (all significance tests are two-tailed)

From these strong results at the state level we believe that votersdemonstrated a remarkable degree of casualty sensitivity The resultssuggest that the United States need not suffer 50000 casualties or morebefore the public rises up and turns against those in power Rathereven a war with comparatively modest levels of casualties can have asubstantial effect on congressional elections with ruling-party candi-dates from states that have suffered the heaviest losses bearing thebrunt of the popular backlash


Turning to the political control variables we find that most ofthe relationships are in the expected direction and many are statisti-cally significant In both state-level models strong support for PresidentBush in 2004 is positively correlated with increases in Republicansenatorial vote share and in the second specification the coefficientis statistically significant Similarly in both models the coefficientfor the change in the opponent quality variable is negative as expectedRepublican candidates tended to lose ground when they faced a tougheropponent in 2006 than in 2000 however there is considerableuncertainty around the estimates of both coefficients Also consistentwith theoretical expectations and prior studies emphasizing theimportance of campaign spending both specifications detect a stronglink between relative campaign expenditures and the change inRepublican vote share The second state-level model suggests that aone standard deviation increase in the percentage of total campaignexpenditures spent by the Republican candidate produced a fourpercentage point increase on average in GOP vote share from 2000to 2006

Economics also appear to have had some influence on Republicanelectoral fortunes yet far from being punished electorally in areas ofincreasing unemployment as the party in power the models suggestthat Republicans actually performed better in these areas on averagethan they did in the 2000 contests To explore this relationship furtherwe reestimated the two state-level models disaggregating the changein unemployment measure by the partisanship of the incumbent Thisadditional step revealed that rising state-level unemployment onlyincreased Republican vote share when the Republican faced anincumbent Democrat the coefficients for the effect of changingunemployment on incumbent Republicansrsquo electoral fortunes arenegative but statistically insignificant All other results remainedunchanged

Finally turning to the military-related demographic characteristicsof the states themselves we found some evidence of states with largeactive-duty military populations rallying around the Republican PartyIn both models the coefficient is positive and in the second specifi-cation it is statistically significant This model suggests that a onepercentage point increase in the statersquos active-duty military popula-tion results in a 4 increase in GOP vote share from the peacetime2000 contest to the 2006 election Yet neither model finds an effect forthe size of a statersquos veteran population on the change in GOP voteshare


County Level

The next set of models in Table 1 sharpens the scope of ouranalysis by demonstrating the influence of the geographic distributionof Iraq war casualties on Republican vote shares at the county levelThe dependent variable here is the change in county-level vote sharefrom 2000 to 2006 in all 1856 counties from the 31 states with sena-torial contests (excluding Vermont and Connecticut where BernieSandersrsquos and Joe Liebermanrsquos Independent candidacies complicatecross-election comparisons) The results at this lower level ofgeographic aggregation also strongly suggest that local casualtiesinfluenced Republicansrsquo electoral fates

The first county-level model shows a strong negative relation-ship between the number of Iraq battle deaths for that county and thechange in Republican vote share Substantively the size of the effectis modest yet still of political import a two standard deviation increasein a countyrsquos casualty tally cost the Republican candidate on averagemore than one percentage point at the polls

Unlike the models at the state level the second county modelprovides little evidence of a strong relationship between a countyrsquoscasualty rate and GOP electoral fortunes The coefficient is negativeas expected but the correlation is not statistically significant At thecounty level the casualty rate may not be nearly as important as thesimple fact of a casualty from the votersrsquo local community After all amajority of counties as of November 2006 had not suffered a singlebattle death in Iraq As a result whether a community had suffered adisproportionate share of the burden in Iraq in terms of its casualtyrate may have been considerably less important to many of its votersrsquoelectoral choices than whether voters had experienced the costs of warthrough the lens of their local community at all Alternatively aspreviously discussed the considerable variance in county-level casualtyrates particularly the presence of low-population outlier communitiesthat had suffered one or two casualties may be skewing the resultswhen we assume a linear relationship To account for this possibilitywe reestimated the model using the logged casualty rate In thisspecification the relevant coefficient is negative as expected andstatistically significant p lt 10 on a one-tailed test Although far fromconclusive evidence the logged casualty rate specification is at leastsuggestive of a relationship between county casualty rates and changein Republican vote share across all Senate contests Nevertheless thenumber of casualties incurred by a county appears to be the strongestcorrelate of changing GOP electoral fortunes at the county level


In both county-level models the political control variables closelyfollow theoretical expectations The coefficient for increasing opponentquality is negative as expected and highly statistically significant Aone point increase in the caliber of the Republican opponent on theGreen and Krasno scale decreased the Republicanrsquos vote share byalmost two percentage points Similarly the coefficient for the shareof campaign expenditures disbursed by the Republican is positivealthough it fails to reach conventional levels of statistical significancein either specification And finally both models suggest that Republicansenatorial candidates reaped modest gains over their 2000 showingsin counties that strongly supported George W Bush in the 2004 electioncontest

In the economic realm the coefficients for change in a countyrsquosunemployment rate are negative but statistically insignificant Againfurther analysis suggests that the relationship is contingent on thepartisanship of the incumbent senator Disaggregating the unemploy-ment measure by partisanship shows that rising unemployment bolstersthe Republican candidatersquos fortunes when he or she challenges a sittingDemocrat but depresses the GOP vote share when the Republican isthe incumbent

Finally turning to the two military demographic variables wefind no evidence at the county level of communities with largeconcentrations of active-duty military personnel rallying behind theRepublican Party In both specifications however the coefficients forthe percentage of veterans in a county are negative and statisticallysignificant The models suggest that the Republican candidate faredalmost two percentage points worse in counties with veteran popula-tions that were two standard deviations above the mean in 2006 thanthey fared in 2000 Considered in conjunction with the state-levelanalyses these results imply that communities with large veteran popu-lations approached the 2006 midterms differently than did those withlarge active-duty military populations18

Republican Incumbent Races at the County Level

The models of election results from all states and countiesinvolved in the 2006 elections offer considerable evidence that theexperience of votersrsquo state and local communities in Iraq influencedtheir electoral calculations in the 2006 midterm elections Because thefirst four models in Table 1 do not differentiate among electoralcontests however it is possible that they underestimate local casualtiesrsquoeffects on the Senate races For example in the Tennessee Senate race


it is not clear that Bob Corker the former Chattanooga mayor andRepublican nominee should have performed worse than the 2000Republican candidate in counties that experienced higher casualtiesin Iraq If anything Harold Ford who voted to authorize the war whilein the House might stand to bear the brunt of any voter dissatisfactionregarding Iraq Corker acknowledged that mistakes had been made inIraq and emphasized the need for a change in strategy to get the jobdone and bring the troops home Because Corker was unsaddled bythe baggage of voting for the authorization to use force against Iraq orthe need to support the presidentrsquos policies on the Senate floor there islittle reason to expect the effects of Iraq on his candidacy to have beenas acute

Taking this distinction into account the third set of models inTable 1 focuses exclusively on the county-level election results for the14 incumbent Republican senatorsmdashall but two of whom voted toauthorize the war in Iraqmdashrunning for reelection in 2006 For thissubset of elections the dependent variable measuring the change inRepublican vote share from the previous election is cleanest More-over it is for these senators that the expectations of a strong effect forIraq casualties on electoral success are most robust19

In this critical test of the electoral import of local casualties themodels uncover a strong relationship between both the county casu-alty tally and rate and the change in vote share for the Republicanincumbent A two standard deviation increase in a countyrsquos casualtycount cost the Republican incumbent more than two percentage pointsat the polls Similarly a two standard deviation increase in the countyrsquoscasualty rate decreased the Republican incumbentrsquos expected vote shareby almost one percentage point from his or her 2000 performance Bysome accounts these effects are rather modest still a two- to four-point swing could have meant the difference in a number of contestsin 2006 particularly in the hotly contested races in Montana MissouriVirginia and Tennessee

Moreover the effect of county-level casualty tallies and rates isrobust even after one controls for state-level casualty figures Reesti-mating the models with both state- and county-level casualty talliesand rates reveals a strong relationship between county-level casualtymeasures and the change in GOP vote share

The control variables with one exception again largely accordwith theoretical expectations For this subset of counties the coeffi-cient for change in opponent quality is now actually positive althoughthis anomaly is most likely due to idiosyncratic factors in the smallernumber of Senate contests in the restricted sample For example the


largest change in Republican opponent quality was in Virginia whereGeorge Allen ran against incumbent senator Charles Robb in 2000and then against James Webb who had never held elected office in2006 On the Green and Krasno scale which fails to capture Webbrsquosformidability as a candidate in the 2006 election cycle Webb scoresconsiderably lower than many candidates running for Senate Yet theother controls follow expectations closely The greater the change inthe share of total campaign expenditures spent by the Republican thebetter the Republican candidate performed Incumbent senators weremost likely to gain ground from their previous elections in countieswhere George W Bush performed well in the 2004 presidential raceFinally rising unemployment is negatively correlated with the changein Republican vote share although the relationship is not statisticallysignificant

We find more evidence of differential voting behavior in areaswith high concentrations of active-duty military personnel and veteransAs in the state models the coefficient for active-duty military popula-tion is positive and in the first specification it is statistically signifi-cant Yet as in the model of all county returns the coefficient for thepopulationrsquos veteran percentage is negative and significant in bothspecifications With all appropriate caveats about the dangers ofecological inference we note that the evidence is at least suggestivethat areas with large concentrations of active-duty soldiers and veteransviewed the Iraq war very differently Counties with large shares ofactive-duty service members rallied slightly behind the GOP whereascounties with strong veteran presences abandoned the Republicans

In sum at both the state and county levels the models providecompelling evidence across a wide range of specifications that bothstate- and county-level Iraq casualties depressed voting for Republicansenatorial candidates The war was indeed a national issue of thegreatest import but its electoral consequences appear to have been atleast in part a function of the distribution of the warrsquos costs across thecountry

Conclusion

This article has demonstrated that in the 2006 midterm electionscounty- and state-level casualties from the Iraq warmdashdespite their smallnumbers compared to previous major conflictsmdashhad a significant andnegative effect on the electoral fate of Republican candidates for USSenate When we isolate the incumbent Republican senators themagnitude of the effects of local casualties becomes even larger In


these races with a Republican incumbent a two standard deviationincrease in a countyrsquos casualty tally cost the incumbent more than 2on average at the polls A similar increase in the county casualty rateresulted in a one percentage point swing in Republican vote share

These findings which are consistent with the campaign strategyof 2006 Democratic Senate candidates are an important contributionto the emerging literature on American wartime casualties and elec-toral outcomes beyond those for the commander in chief GartnerSegura and Barrattrsquos (2004) study of the negative effect of state-levelVietnam casualties on senatorsrsquo vote shares from 1966 to 1972 leftopen the question of thresholds At what threshold will voters respondto casualties The current Iraq conflict which so far has less than one-fifteenth of Vietnamrsquos casualty total provides an important test caseOur analysis suggests that voters are sensitive to casualties in theircounty and state even when average state casualty rates are 11 battledeaths per million residents

Furthermore consistent with theories of the importance of localcasualties to public-opinion formation our analysis also finds thatcounty-level casualty tallies and rates influenced voting behavior inthe 2006 midterms In contrast to Karol and Miguel (2007) whosecounty-level analysis did not find a significant relationship betweencounty-level casualties and President Bushrsquos vote share in 2004 wefound strong negative relationships between a countyrsquos casualty tallyand rate and the change in Republican vote share from the 2000 to the2006 Senate races What explains these divergent results

One possible explanation is the change from 2004 to 2006 indissatisfaction with the war in Iraq In 2004 the country was roughlysplit in their opinion of President Bushrsquos handling of Iraq By 2006less than 30 of the populace approved and over 60 disapproved20

Gelpi Feaver and Reifler (2005) have argued that public confidencein the success of a mission is directly related to casualty toleranceWhen confidence is high as it was for Bush in many segments of thecountry in 2004 they contend that casualties will have little effect onpolitical outcomes Our empirical analysis strongly suggests that thereverse is also true when confidence in a military venture and its leadersis low as it was for most Americans considering Iraq in 2006 casualtieswill have a significant negative effect on the electoral fates of thosepublic officials tied most directly to the war and its conduct

In addition to its contribution to the existing literature on casualtysensitivity among the American electorate and the influence of localcasualties on congressional elections our research also has importantimplications for recent scholarship emphasizing congressional


importance in military affairs A critical component of many theoriesproclaiming presidential dominance in foreign policy is the assump-tion that Congressmdashcomposed of 535 single-minded seekers ofreelection (Mayhew 1974)mdashwillingly and logically defers to the presi-dent in military matters (Gowa 1998 Meernik 1995 Peterson 1994Wildavsky 1966) Yet a growing number of scholars have challengedthis president-centered conception of foreign policy (Clark 2000Howell and Kriner 2007 Howell and Pevehouse 2005 2007 Johnson2006) Implicit in their arguments is the understanding that undercertain conditions members of Congress stand to reap political gainsor insulate themselves from political fallout by challenging presiden-tial discretion in military affairs Our results offer considerable supportfor this perspective by documenting that senators do incur politicalcosts from deferring to the president even tacitly in an unpopularwar even when casualty totals are orders of magnitude smaller thanthose sustained in Vietnam

Finally our study paves the way for a number of additionalexplorations Two lines of future analysis seem most promising Firstqualitative work can be carried out to study further the mechanisms bywhich casualties affect electoral outcomes News of casualties is filteredthrough the media experienced through social networks and framed(in contrasting ways) by partisan campaigns It is important to knowhow these three streams interact to produce the casualty effect we haveobserved in our data Recent work by Voeten and Brewer (2006)suggests that at the presidential level the connections betweencasualties and approval are not as direct as previous scholarship hasconcluded At the congressional levels too it may be that there iscomplexity in the pathways through which casualties influence elec-toral outcomes Second as the Iraq conflict seems destined to carry onthrough the next election cycle political scientists can monitor whetheror not rising casualties lead to effects of larger magnitudes in 2008 Itis not clear with a Democratic House and Senate how the public willallocate political blame for further casualties

Douglas L Kriner ltdkrinerbuedugt is Assistant Professor ofPolitical Science Boston University 232 Bay State Road Boston MA02215 Francis X Shen ltfxshenfasharvardedugt is a doctoralfellow in the Harvard Multidisciplinary Program in Inequality andSocial Policy 1737 Cambridge St CGIS N-151 Cambridge MA02138


NOTES

1 This ratio is based on the May 2007 Iraq casualty count of 3422 the esti-mate of Vietnam casualties of 58219 from Department of Defense statistics and theestimate of Civil War casualties of 620000 (Beer 1983)

2 The dependent variable for all models is the change in Republican voteshare from 2000 to 2006 with one exception In 2002 James Talent defeated incumbentDemocratic senator Jean Carnahan who was appointed to the seat following herdeceased husbandrsquos narrow victory over John Ashcroft in 2000 For Missouri we examinedthe change in Republican vote share from 2002 to 2006 and used the appropriate controlsAll of the model results remain the same if the 2000 to 2006 data is used

3 Senator Lincoln Chaffee voted against the authorization and Senator James Talentof Missouri did not hold his seat at the time of the authorization vote Replicating these modelswithout Missouri and Rhode Island yields even stronger results for both casualty measures

4 ldquoMeet the Pressrdquo Transcript October 29 2006 httpwwwmsnbcmsncomid15473528 (September 25 2007)

5 ldquoMeet the Pressrdquo Transcript October 8 2006 httpwwwmsnbcmsncomid15116699page2 (September 25 2007)


7 Sherrod Brown ldquoSherrod Brown lsquoWent to Batrsquo for Our Troopsrdquo press release29 September 2006 httpsherrodbrowncompressreleases675 (September 25 2007)

8 Jeff Whelan ldquoMenendez Renews His Iraq Attack on Keanrdquo New JerseyStar-Ledger 27 September 2006 httpoperationhousecallorgarticlephpid=749(September 25 2007)

9 Bob Casey Jr Interview with Philadelphia Jewish Voice 2005 httpwwwpjvoicecomv44800wordshtml (September 25 2007)

10 Recent experimental research by Adam Berinsky (Nd) also raises questionsabout the influence that casualty totals have on public opinion Berinsky demonstratesthat in 2004 most Americans held wildly varying estimates of how many casualties theUnited States had suffered in Iraq with Republicans dramatically underestimating thetrue number and Democrats systematically overestimating the figure

11 In Connecticut political newcomer Ned Lamont ran against incumbent JoeLieberman to protest Senator Liebermanrsquos support for the Iraq war Although Lamontwon the primary Lieberman successfully ran as an Independent and held his Senateseat by garnering 50 of the vote to Lamontrsquos 40 Vermont presents a more-difficultcase Independent candidate Bernie Sanders won the Democratic primary but declinedthe nomination Sanders defeated his Republican rival Richard Tarrant for the seatvacated by Independent senator James Jeffords by securing 65 of the vote To checkthe robustness of our results we conducted additional analyses including these stateswhich yielded virtually identical results across specifications In a similar vein Indianawas an outlier being the only race not contested in 2006 by both major parties ExcludingIndiana from the analysis also yields virtually identical results across specifications

12 An additional political factor that may have influenced the change in GOPvote share is any change in the incumbency status of the Republican candidate fromthe 2000 to the 2006 campaigns All models were reestimated with two dummy variablesindicating if the GOP candidate went from being a challenger (either facing an incumbent


or vying for an open seat) to being an incumbent from 2000 to 2006 or vice versa Allof our results remained virtually identical in this expanded specification

These augmented models show the expected negative relationship between a shift fromincumbent to challenger status and GOP vote share at both the state and county levels A comple-mentary shift from challenger to incumbent status however had no effect at the state level andcontra expectations a negative correlation with the change in GOP vote share at the countylevel The relationship is almost certainly spurious Only three states involved a Republicanchallenger from 2000 (2002 for James Talent) running in 2006 as an incumbent VirginiaNevada and Missouri In the Virginia race George Allen lost to James Webb in Nevada JohnEnsign handily beat Jack Carter but not by the same margins as he trounced his Democraticopponent who lacked a presidential name in 2000 and the Missouri races were decided byrazor-thin margins in 2000 2002 and 2006 A confluence of national trends and idiosyncraticfactorsmdashnot any change in incumbency statusmdashdetermined these three electionsrsquo end results

13 Because Krasno and Greenrsquos scale was designed to measure challenger qualityit required one minor modification If the Republican candidate faced an incumbent senatorwe coded the opponent-quality score at its maximum value of 8 Prior studies have adoptedvaried operationalizations of relative campaign spending To control for several outliersin Republican-opponent spending we took the log of both major candidatesrsquo FederalElection Commission-reported expenditures and calculated the percentage of this totalspent by the Republican All of our results are robust across other operationalizationssuch as the change in the percentage of unlogged total expenditures spent by theRepublican candidate and the change in the ratio of Republican to Democratic spendingFollowing Jacobson Green and Krasno and others we recoded the handful of missingexpenditure data points as $1000 All of these data points represent minor dark-horsecandidates who had little in the way of a formal campaign apparatus

14 We downloaded all casualty data in November 2006 from httpsiadappdmdcosdmilpersonnelCASUALTYcastophtm

15 This method is consistent with many other studies of casualtiesrsquo (ie battledeathsrsquo) effects on electoral outcomes and public opinion (inter alia Eichenberg 2005Feaver and Gelpi 2005 Gartner Segura and Barratt 2004)

16 Casualty figures particularly at the county level exhibited considerablymore variance For example at the county level the standard deviation for casualtyrates per 10000 residents was 3 times the mean value and a small number of outlyingcounties mostly in very sparsely populated areas had casualty rates more than 50times the mean value To mitigate these extreme outliers we replicated all of the modelsat both the state and county levels using logged tallies and logged casualty rates Inalmost every specification the observed relationships between casualties and changein Republican vote share were even stronger when we used the logged measures

17 The bivariate relationship is statistically significant p lt 05 on a two-tailed test18 Veteran populations and large active-duty military populations are positively

correlated but the correlation is not high (r = 16)19 As mentioned in note 3 Lincoln Chaffee and James Talent may not fit this

mold Replicating this final set of models at the county level without Rhode Island andMissouri yields even stronger results for both casualties measures

20 Adam Nagourney and Megan Thee ldquoBushrsquos Public Approval at New LowPointrdquo New York Times 9 May 2006 httpwwwnytimescom20060509washington09cnd-pollhtml (September 25 2007)


REFERENCES

Abramowitz Alan 1989 ldquoCampaign Spending in US Senate Electionsrdquo LegislativeStudies Quarterly 14 487ndash507

Abramowitz Alan and Jeffrey Segal 1986 ldquoDeterminants of the Outcomes of USSenate Electionsrdquo Journal of Politics 48 433ndash39

Aldrich John John Sullivan and Eugene Borgida 1989 ldquoForeign Affairs and IssueVoting Do Presidential Candidates lsquoWaltzrsquo before a Blind Audiencerdquo AmericanPolitical Science Review 83 123ndash41

Beer Francis A 1983 ldquoTrends in American Major War and Peacerdquo Journal of ConflictResolution 27 661ndash86

Berinsky Adam Nd ldquoAssuming the Costs of War Events Elites and American PublicSupport for Military Conflictrdquo Journal of Politics Forthcoming

Berinksy Adam J and James N Druckman 2007 ldquoPublic Opinion Research andSupport for the Iraq Warrdquo Public Opinion Quarterly 71 126ndash41

Boettcher William A III and Michael D Cobb 2006 ldquoEchoes of Vietnam CasualtyFraming and Public Perceptions of Success and Failure in Iraqrdquo Journal ofConflict Resolution 50 831ndash54

Brady David John Cogan and Morris Fiorina 2000 Continuity and Change in HouseElections Stanford CA Stanford University Press

Brody Richard 1991 Assessing the President The Media Elite Opinion and PublicSupport Stanford CA Stanford University Press

Campbell James 1991 ldquoThe Presidential Surge and its Midterm Decline 1868ndash1988rdquoJournal of Politics 53 477ndash87

Campbell James and Joe Sumners 1990 ldquoPresidential Coattails in Senate ElectionsrdquoAmerican Political Science Review 84 513ndash24

Carsey Thomas and Gerald Wright 1998 ldquoState and National Factors in Gubernatorialand Senatorial Electionsrdquo American Journal of Political Science 42 994ndash1002

Carson Jamie Jeffrey Jenkins David Rohde and Mark Souva 2001 ldquoThe Impact of NationalTides on District-level Effects on Electoral Outcomes The US CongressionalElections of 1862ndash1863rdquo American Journal of Political Science 45 887ndash98

Clark David 2000 ldquoAgreeing to Disagree Domestic Institutional Congruence andUS Dispute Behaviorrdquo Political Research Quarterly 53 375ndash401

Cotton Timothy 1986 ldquoWar and American Democracy Electoral Costs of the LastFive Warsrdquo Journal of Conflict Resolution 30 616ndash35

Eichenberg Richard 2005 ldquoVictory Has Many Friends US Public Opinion and theUse of Military Forcerdquo International Security 30 140ndash77

Eichenberg Richard Richard Stoll and Matthew Lebo 2006 ldquoWar President TheApproval Ratings of George W Bushrdquo Journal of Conflict Resolution 50 783ndash808

Feaver Peter and Christopher Gelpi 1999 ldquoHow Many Deaths are Acceptable ASurprising Answerrdquo Washington Post 7 November B3

Feaver Peter and Christopher Gelpi 2005 Choosing Your Battles American Civil-MilitaryRelations and the Use of Force Princeton NJ Princeton University Press

Foley Michael S 2003 Confronting the War Machine Draft Resistance during theVietnam War Chapel Hill NC University of North Carolina Press

Gartner Scott 2004 ldquoMaking the International Local The Terrorist Attack on the USS ColeLocal Casualties and Media Coveragerdquo Political Communication 21 139ndash59


Gartner Scott and Gary Segura 1998 ldquoWar Casualties and Public Opinionrdquo Journalof Conflict Resolution 42 278ndash320

Gartner Scott and Gary Segura 2000 ldquoRace Casualties and Opinion in the VietnamWarrdquo Journal of Politics 62 115ndash46

Gartner Scott Gary Segura and Bethany Barratt 2004 ldquoWar Casualties Policy Posi-tions and the Fate of Legislatorsrdquo Political Research Quarterly 53 467ndash77

Gartner Scott Sigmund Gary M Segura and Michael Wilkening 1997 ldquoAll PoliticsAre Local Local Losses and Individual Attitudes toward the Vietnam WarrdquoJournal of Conflict Resolution 41 669ndash94

Gelpi Christopher Peter Feaver and Jason Reifler 2005 ldquoSuccess Matters CasualtySensitivity and the War in Iraqrdquo International Security 30 7ndash46

Gelpi Christopher Jason Reifler and Peter Feaver 2007 ldquoIraq the Vote Retrospec-tive and Prospective Foreign Policy Judgments on Candidate Choice and CasualtyTolerancerdquo Political Behavior 29 151ndash74

Gerber Alan 1998 ldquoEstimating the Effects of Campaign Spending on Senate ElectionOutcomes Using Instrumental Variablesrdquo American Political Science Review92 401ndash11

Gilliam Franklin and Shanto Iyengar 2000 ldquoPrime Suspects The Influence of Local Televi-sion News on the Viewing Publicrdquo American Journal of Political Science 44 560ndash73

Green Don and Jonathan Krasno 1988 ldquoSalvation for the Spendthrift IncumbentReestimating the Effects of Campaign Spending in House Electionsrdquo AmericanJournal of Political Science 32 884ndash907

Gowa Joanne 1998 ldquoPolitics at the Waterrsquos Edge Parties Voters and the Use ofForce Abroadrdquo International Organization 52 307ndash24

Hess Stephen and Michael Nelson 1985 ldquoForeign Policy Dominance and Decisive-ness in Presidential Electionsrdquo In The Elections of 1984 ed Michael NelsonWashington DC CQ Press

Howell William and Douglas Kriner 2007 ldquoBending so as Not to Break What theBush Presidency Reveals about Unilateral Actionrdquo In The Polarized Presidencyof George W Bush ed George Edwards and Desmond King Oxford UK OxfordUniversity Press

Howell William and Jon Pevehouse 2005 ldquoPresidents Congress and the Use ofForcerdquo International Organization 59 209ndash32

Howell William and Jon Pevehouse 2007 While Dangers Gather Princeton NJPrinceton University Press

Hurwitz John and Mark Peffley 1987 ldquoThe Means and Ends of Foreign Policy as Determi-nants of Presidential Supportrdquo American Journal of Political Science 2 236ndash58

Jacobson Gary 1978 ldquoThe Effects of Campaign Spending in Congressional Elec-tionsrdquo American Political Science Review 72 769ndash83

Jacobson Gary 1990 ldquoThe Effects of Campaign Spending in House Elections NewEvidence for Old Argumentsrdquo American Journal of Political Science 34 334ndash62

Jacobson Gary 2004 Politics of Congressional Elections New York PearsonLongman

Jacobson Gary and Samuel Kernell 1981 Strategy and Choice in CongressionalElections New Haven CT Yale University Press

Johnson Robert David 2006 Congress and the Cold War Cambridge UK CambridgeUniversity Press


Karol David and Edward Miguel 2007 ldquoThe Electoral Cost of War Iraq Casualtiesand the 2004 US Presidential Electionrdquo Journal of Politics 69 633ndash48

Klarevas Louis Christopher Gelpi and Jason Reifler 2006 ldquoCorrespondenceCasualties Polls and the Iraq Warrdquo International Security 31 186ndash98

Larson EV 1996 Casualties and Consensus The Historical Role of Casualties inDomestic Support for US Military Operations Santa Monica CA RAND

Maoz Zeev and Bruce Russett 1992 ldquoNormative and Structural Causes of the Demo-cratic Peacerdquo American Political Science Review 87 624ndash38

Mayhew David 1974 Congress The Electoral Connection New Haven CT YaleUniversity Press

Meernik James 1995 ldquoCongress the President and the Commitment of the USMilitaryrdquo Legislative Studies Quarterly 20 377ndash92

Moody James 2006 ldquoFighting a Hydra A Note on the Network Embeddedness of theWar on Terrorrdquo Structure and Dynamics eJournal of Anthropological andRelated Sciences Vol 1 No 2 Article 9 httprepositoriescdliborgimbssocdynsdeasvol1iss2art9 (September 25 2007)

Mueller John 1973 War Presidents and Public Opinion New York WileyNickelsburg Michael and Helmut Norpoth 2000 ldquoCommander-in-Chief or Chief

Economist The President in the Eye of the Publicrdquo Electoral Studies 19 313ndash32Nincic Miroslav and Barbara Hinckley 1991 ldquoForeign Policy and the Evaluation of

Presidential Candidatesrdquo Journal of Conflict Resolution 35 333ndash55Peterson Paul ed 1994 The President Congress and the Making of US Foreign

Policy Norman OK University of Oklahoma PressRay James Lee 1995 Democracy and International Conflict An Evaluation of the

Democratic Peace Proposition Columbia University of South Carolina PressReiter Dan and Alan Stam 2002 Democracies at War Princeton NJ Princeton

University PressRussett Bruce and John R OrsquoNeal 2001 Triangulating Peace New York NortonSchecter Barnet 2005 The Devilrsquos Own Work The Civil War Draft Riots and the

Fight to Reconstruct America New York Walker amp CoSiverson Randolph 1995 ldquoDemocracies and War Participation In Defense of the

Institutional Constraints Argumentrdquo European Journal of International Relations4 481ndash89

Squire Peverill 1992 ldquoChallenger Quality and Voting Behavior in Senate ElectionsrdquoLegislative Studies Quarterly 17 247ndash63

Squire Peverill 1995 ldquoCandidates Money and Voters Assessing the State ofCongressional Elections Researchrdquo Political Research Quarterly 48 891ndash917

Voeten Erik and Paul R Brewer 2006 ldquoPublic Opinion the War in Iraq and Presi-dential Accountabilityrdquo Journal of Conflict Resolution 50 809ndash30

White Halbert 1980 ldquoA Heteroskedasticity-consistent Covariance Matrix Estimatorand a Direct Test for Heteroskedasticityrdquo Econometrica 48 817ndash38

Wildavsky Aaron 1966 ldquoThe Two Presidenciesrdquo Trans-Action 4 7ndash14Zaller John 1994 ldquoElite Leadership of Mass Opinion New Evidence from the Gulf

Warrdquo In Taken by Storm Media Public Opinion and US Foreign Policy inthe Gulf War ed W Lance Bennett and David L Paletz Chicago IL Universityof Chicago Press


conscription and significant draft resistance escalating to violence(Foley 2003 Schecter 2005) The relatively low number of casualtiessustained in Iraq coupled with the absence of large-scale resistanceon par with the draft riots of earlier eras raises questions about scaleDoes a threshold exist below which casualties will not affect senatorsrsquoelectoral fates Has the casualty count in Iraq reached that thresholdWe confront both questions in this article

In addition to Iraqrsquos critical importance in the off-year electionsthe 2006 midterms were also marked by an unusually high percentageof Americans responding that national not local concerns were themotive forces behind their congressional votes The growing national-ization of congressional elections since Tip OrsquoNeil coined his apho-rism ldquoall politics are localrdquo is well documented (Brady Cogan andFiorina 2000 Jacobson 2004) For this study however we investi-gated the possibility that even the most national of issuesmdashthe war inIraqmdashmay have a strong local component Previous studies havedemonstrated the influence of local casualties on public opinion(Gartner and Segura 1998 Gartner Segura and Wilkening 1997) butthere is scant evidence that local casualties at lower levels of aggrega-tion than the state or congressional district influence electoral outcomesWe correct this deficit by exploring how the Iraq warrsquos influence onvoting returns was critically mediated by local casualty rates at boththe state and county levels

We begin our discussion by examining the influence of a statersquosshare of Iraq casualties on the change in that statersquos Republican sena-torial vote share from the 2000 to the 2006 elections2 We then shift tothe county level where there is considerably greater variance in wartimeexperiences to examine the relationship between local casualties andchanges in Republican electoral fortunes at a lower level of geographicaggregation Finally we narrow our focus to the effect of local casualtieson the 14 Republican incumbent senators seeking reelection in 2006all but two of whom voted to authorize military action against Iraq in20023 We show that Republican senatorial candidates lost ground fromtheir 2000 performance in states and counties hit hard by the war inIraq but generally fared no worse in states and counties that had thusfar emerged from the war relatively unscathed Our findings suggest aremarkable degree of casualty sensitivity among the American elec-torate The historically modest number of casualties suffered in theIraq war has not spurred riots in the streets but it has produced signifi-cant negative reactions in the voting booth


Theory






















Data and Methods





















State Level









AZ

CA

CT

DE

FL

HI

INMA

MDME

MIMN

MO MS MT

ND

NE

NJ

NM

NV

NY

OH

PARI

TN

TXUT VA

VT

WAWI

WV

WY

Cha

nge

in G

OP

Vote

Sha

re


20

10

0

ndash10

ndash20

ndash30

5 10 15 20 25 30











Bush 2004 044 078 011 012 021 023(042) (037) (002) (003) (003) (003)





Observations 31 31 1856 1856 993 993R2 29 41 15 15 21 20








County Level




















Conclusion














NOTES

























REFERENCES


































































Theory






















Data and Methods





















State Level









AZ

CA

CT

DE

FL

HI

INMA

MDME

MIMN

MO MS MT

ND

NE

NJ

NM

NV

NY

OH

PARI

TN

TXUT VA

VT

WAWI

WV

WY

Cha

nge

in G

OP

Vote

Sha

re


20

10

0

ndash10

ndash20

ndash30

5 10 15 20 25 30











Bush 2004 044 078 011 012 021 023(042) (037) (002) (003) (003) (003)





Observations 31 31 1856 1856 993 993R2 29 41 15 15 21 20








County Level




















Conclusion














NOTES

























REFERENCES



















































































Data and Methods





















State Level









AZ

CA

CT

DE

FL

HI

INMA

MDME

MIMN

MO MS MT

ND

NE

NJ

NM

NV

NY

OH

PARI

TN

TXUT VA

VT

WAWI

WV

WY

Cha

nge

in G

OP

Vote

Sha

re


20

10

0

ndash10

ndash20

ndash30

5 10 15 20 25 30











Bush 2004 044 078 011 012 021 023(042) (037) (002) (003) (003) (003)





Observations 31 31 1856 1856 993 993R2 29 41 15 15 21 20








County Level




















Conclusion














NOTES

























REFERENCES


















































































State Level









AZ

CA

CT

DE

FL

HI

INMA

MDME

MIMN

MO MS MT

ND

NE

NJ

NM

NV

NY

OH

PARI

TN

TXUT VA

VT

WAWI

WV

WY

Cha

nge

in G

OP

Vote

Sha

re


20

10

0

ndash10

ndash20

ndash30

5 10 15 20 25 30











Bush 2004 044 078 011 012 021 023(042) (037) (002) (003) (003) (003)





Observations 31 31 1856 1856 993 993R2 29 41 15 15 21 20








County Level




















Conclusion














NOTES

























REFERENCES







































































AZ

CA

CT

DE

FL

HI

INMA

MDME

MIMN

MO MS MT

ND

NE

NJ

NM

NV

NY

OH

PARI

TN

TXUT VA

VT

WAWI

WV

WY

Cha

nge

in G

OP

Vote

Sha

re


20

10

0

ndash10

ndash20

ndash30

5 10 15 20 25 30











Bush 2004 044 078 011 012 021 023(042) (037) (002) (003) (003) (003)





Observations 31 31 1856 1856 993 993R2 29 41 15 15 21 20








County Level




















Conclusion














NOTES

























REFERENCES











































































Bush 2004 044 078 011 012 021 023(042) (037) (002) (003) (003) (003)





Observations 31 31 1856 1856 993 993R2 29 41 15 15 21 20








County Level




















Conclusion














NOTES

























REFERENCES






































































County Level




















Conclusion














NOTES

























REFERENCES

















































































Conclusion














NOTES

























REFERENCES













































































NOTES

























REFERENCES













































































REFERENCES

































































iraq casualties and the 2006 senate elections - francis x. …€¦ · · 2015-06-06iraq...

Documents