ncumbency,Redistricting,and the Decline ofmpetition in U.S.House Elect - PDF document

 ncumbency,Redistricting,and the Decline ofmpetition in U.S.House ElectionsEmory UniversityEmory UniversityEmory Universityompetition in U.S.House elections has been declining for more than 50 years and,based on both incumbent reelec-ion rates and the percentage ofclose races,the 2002 and 2004 ! .    ,    ,    $198,000 dollars.Even in potentially competitive dis-icts,challengers have been falling farther behindincumbents in campaign spending.order to estimate the effects ofdistrict parti-sanship,incumbency,and campaign spending on thempetitiveness ofHouse elections since the 1970s,nducted regression analyses ofthe Democraticshare ofthe major party vote in contested races in eachelection from 1972 through 2002.The results oftheseession analyses are summarized in Table 4.Wehave grouped the results into four time periods witheach time period consisting offour elections:1972Ð78,1980Ð86,1988Ð94,and 1996Ð2002.For each independent variable during each timeperiod,we present the average unstandardized regres-sion coefÞcient,the average score on the variable forall winning candidates,and the average impact ofthevariable on the vote share for winning candidates.Thelatter statistic is computed by multiplying the regres-sion coefÞcient for each variable by the average advan-tage ofwinning candidates on that variable.In the caseofincumbency and district partisanship,the raw scoreis a measure ofadvantage.Incumbency is coded as for incumbents,0 for open seat candidates and for challengers,so the average advantage score forinning candidates is the proportion ofwinningincumbents minus the proportion ofwinning chal-lengers.District partisanship is measured by the normalized presidential vote so the average advantagescore for winning candidates is the average normal-ized presidential vote for the winning candidateÕsparty in each district.Campaign spending is measuredeach candidateÕs percentage ofcombined Democ-tic and Republican campaign spending so theerage advantage score for winning candidates is theerage percentage ofcampaign spending by theinning candidate in each district minus 50.The results in Table 4 indicate that over the past30 years,the direct effect ofincumbency on the out-mes ofHouse elections has actually been shrinking.r controlling for district partisanship and cam-paign spending,the estimated effect ofincumbencyon the vote in contested House elections fell from anerage of8.5 percentage points to an average ofonly4.5 percentage points.Therefore,despite an increasein the mean score ofwinning candidates on theincumbency variable (meaning that a larger propor-tion ofwinners were incumbents and a smaller pro-inancial Competitiveness ofHouse Challengers by Political Experience,1972Ð2002 oportion Financially Competitive InexperExper Note:Experienced challengers are those who have held any elected ofÞce.Source:Data compiled by authors.r additional documentation ofthis trend,see Figure 3 in theb appendix to this article on the urnal ofPoliticsweb site:http://www.journalofpolitics.org.  ,   ,          ..   ofincumbents in these contests were defeated.In con-ast,only 7% ofcontests in which the challenger spentless than halfa million dollars were highly competitiveand not one incumbent in these contests was defeated.The reason so few incumbents in these high risk dis-icts faced stiffcompetition is that only 17% oftheirhallengers spent more 68% spent less than halfa million dollars.Over time,the cost ofrunning a competitive cam-paign for the House ofRepresentatives has increaseddramatically.Over the last three decades,the medianspending ofwinners in House elections has grown atmore than double the rate ofinßation,going from$41,885 in 1974 to $657,359 in 2002.While incum-bents have been able to increase their fundraising toep up with rising campaign costs,challengers havebeen much less successful.Figures 4 and 5 display theend in the percentage ofchallengers who were Þnan-cially competitiveÑaccounting for at least 40% ofotal spendingÑin House elections between 1972 and2002.In Figure 4 the trend is broken down by the competitiveness ofthe district.In Figure 5 theend is broken down by the political experience ofthehallenger.Challengers running in high-risk districts andhallengers who have held elected ofÞce should bemore capable ofwaging serious campaigns againstincumbents than challengers in low- to moderate-riskdistricts or challengers who have not held electedofÞce.However,the results displayed in Figures 4 and5 show that even among these potentially formidablehallengers there has been a fairly drastic decline inÞnancial competitiveness since the 1970s.Takengether,these results demonstrate that even thosehallengers with the greatest potential for threateningincumbents are having more and more difÞcultyaising the funds necessary to wage competitive campaigns.ther evidence ofthe declining Þnancial com-petitiveness ofchallengers can be seen by examiningthe trend in median expenditures ofincumbents andhallengers in high-risk districts.These are districtswhose partisan composition is most favorable to thehallenger.However,between the 1992Ð1994 electioncycle and the 2000Ð2002 election cycle,median spend-ing by incumbents in high-risk districts rose from$596,000 to $910,000 while median spending by chal-lengers in high-risk districts fell from $229,000 toinancial Competitiveness ofHouse Challengers by Type ofDistrict,1972Ð2002 Year2002200019981996199419921990198819861984198219801978197619741972Percentage Financially Competitive5550353025201550 Note:Challengers classiÞed as Þnancially competitive ifthey accounted for at least 40 percent ofcombined challenger and incumbentspending.High-risk districts are those in which vote percentage ofvote for presidential candidate ofincumbentÕs party is less tpercentage.Source:Data compiled by authors.  .    ,    ,    the twenty-Þrst century,however,only about 15% ofraces with incumbents were at least moderatelycompetitive.y do so few incumbents face stiffcompetitionen in districts whose partisan composition couldmake them vulnerable? Part ofthe answer appears tobe that the large majority ofchallengers in these raceslack the Þnancial resources to wage competitive campaigns.The data displayed in Table 3 shows that in elections between 1998 and 2002 in high-risk districtsÑthose with the most unfavorable partisanmposition for incumbentsÑthere was a very stronglationship between the challengerÕs campaign spend-ing and the competitiveness ofthe race.Seventy-sixpercent ofcontests in which the challenger spent atleast a million dollars were highly competitive and 27%ompetition in Incumbent and Open Seat House Races by Decade Decade2002–041992–001982–901972–801962–701952–601946–50Percentage Won with LT 60% Source:Data compiled by authors.ompetition in Marginal House Districts by Type ofContest,2002Ð2004 Type ofContinnerÕs % ofRunning ouse VoteOpen SeatIncumbentUncontestedor 703%20%60Ð70295555Ð60231950Ð55455otal100%100%(n)(31)(197) Note:arginal districts are those in which Democratic presiden-tial candidateÕs percentage ofmajor party vote is within 5 pointsofnational percentage.Source:Data compiled by authors.ompetition in High-Risk IncumbentDistricts by Challenger Spending, 1998Ð2002 Challenger SpendingChallengerÕs$500,000Ð % ofVote$0Ð499,999999,999$1,000,000LT 3016%0%0%30Ð404825340Ð4529292174676inners0%14%27%(n)(132)(28)(33) Note:Risky districts are those in which percentage ofmajor partyfor presidential candidate ofincumbentÕs party is less thannational percentage.Source:Data compiled by authors.  ,   ,          ..   mpetition in marginal districts has beendeclining for decades.Figure 2 displays the trend sincethe 1950s in the percentage ofhighly competitivees,those decided by less than 10 percentage points,in marginal and safe House districts.Throughout thistime period the percentage ofhighly competitive raceswas substantially greater in marginal districts than insafe districts.However,notwithstanding a briefsurgein competition during the 1990s,the percentage ofhighly competitive races in marginal districts fell dra-matically between the 1950s and the Þrst decade ofthetwenty-Þrst century.By 2002Ð2004,the percentage ofhighly competitive races in marginal districts was onlyslightly greater than the percentage ofhighly compet-itive races in safe districts during the 1950s.One possible explanation for the lack ofcompeti-tion in many marginal districts is the advantage ofincumbency.Because oftheir ex-ofÞcio advantages(Mayhew 1974),their ability to raise huge campaignwar-chests,and the inability ofmost oftheir chal-lengers to raise the funds required to mount seriouscampaigns,the large majority ofHouse incumbents,en in marginal districts,win reelection by widemargins.When the advantage ofincumbency ismoved,however,there is a much greater chance ofa competitive contest.The data displayed in Table 2show that in these marginal districts there was a hugedifference in competitiveness between contests withunning incumbents and contests for open seats.Indistricts with running incumbents,only 5% ofthees were highly competitive and only 19% weremoderately competitive.In contrast,in districtsithout incumbents,45% ofthe races were highlympetitive and another 23% were moderately competitive.Figure 3 displays the trend in the percentage ofmpetitive open seat and incumbent races since theend ofWorld War II.In the Þrst three elections ofthepostwar era there was very little difference in the com-petitiveness ofopen seat and incumbent races.Sincethe 1950s,however,open-seat House races have gen-erally been much more competitive than those withincumbents.Moreover,while the extent ofcompeti-tion for open seats has changed very little,there hasbeen a fairly steady decline in competition for seats inhich incumbents were running.During the 1950s,about 40% ofraces with incumbents were at leastmoderately competitive.In the Þrst two elections ofompetition in Marginal and Safe Districts,1952Ð2004 Decade2002–041992–001982–90 1972–801962–701952–60Percentage Highly Competitive4030200 Safe Note:arginal districts are those in which Democratic presidential candidateÕs percentage ofmajor party vote is within 5 points ofnational percentage;safe districts are those in which Democratic presidential candidateÕs percentage ofmajor party vote is morepoints above or below national percentage.Source:Data compiled by authors.  .    ,    ,    ealignment in the South:the trend was evident inngressional districts outside ofthe South as well asthose in the South.This growing consistency appearsßect an increase in partisan voting in both presi-dential and congressional elections since the 1980s(Bartels 2000). a result ofboth increased partisan polarizationand increased partisan consistency in voting behavior,far fewer Representatives now occupy high-risk dis-icts,districts that are less supportive oftheir partythan the national average,and far more Representa-tives now occupy low-risk districts,districts that thatare at least 10 percentage points more supportive oftheir party than the national average.Table 1 displaysthe average percentage ofDemocratic and Republicanmembers occupying high-risk and low-risk districts ineach decade since the 1970s.During this time period,the percentage ofRepublicans in high-risk districtshas fallen from 22% to 11% and the percentage ofDemocrats in high-risk districts has fallen from 41%16%.At the same time,the percentage ofRepubli-cans in low-risk districts has risen from 23% to 41%and the percentage ofDemocrats in low-risk districtshas risen from 24% to 51%.Between 1972 and 2004,the total number ofmembers representing high-riskdistricts fell from 157 to 97 while the total number ofmembers representing low risk districts rose from 50to The changing partisan composition ofHouse dis-icts has important implications for competition inouse elections.Compared with 30 years ago,a muchsmaller proportion ofmembers now represent dis-icts which,based on presidential voting patterns,favor the opposing party:fewer Republicans representDemocratic-leaning districts and far fewer Democratspresent Republican-leaning districts.These high-isk districts account for a disproportionate share ofincumbent defeats and party turnover in House elec-tions.For example,in the 1994 midterm election inwhich Republicans regained control ofthe House ofpresentatives,32% ofDemocratic incumbents inhigh-risk districts were defeated compared with only7% ofDemocratic incumbents in all other districts.Only 34% ofDemocratic seats in 1994 were in high-isk districts,but 70% ofDemocratic seat lossesoccurred in these high-risk districts.mpared with the 1970s and 1980s,a muchlarger proportion ofmembers now represent districtshich,based on presidential voting patterns,stronglyfavor their own party:more Republicans representsolidly Republican districts and more Democrats present solidly Democratic districts.These low-riskdistricts account for a disproportionate share ofuncontested and one-sided races.Even under the mostunfavorable circumstances,incumbents in such dis-icts are rarely defeated.For example,in the 1994election not one ofthe 69 Democratic incumbentspresenting a low-risk district was defeated.ing partisan polarization means that bothparties have far fewer seats at risk in House electionsthan they did 30 years ago or even 10 years ago.the increased number ofsafe districts doesnot completely explain the lack ofcompetition innt House elections.The partisan composition ofaouse district is not the only factor determiningwhether that district is going to have a competitiven in marginal districts,those with the mosten balance ofparty support,only a small minorityofHouse contests in 2002 and 2004 were actuallympetitive.Only 11% ofcontests in these marginaldistricts were decided by less than 10 percentagepoints and only 20% were decided by between 10 and20 percentage points.More than two-thirds ofHousees in marginal districts were blowouts in which theinning candidate was unopposed or received at least60% ofthe major party vote.ercentages ofDemocratic and Republican Incumbents in Low-Risk and High-Risk Districts by Decade 1972Ð19801982Ð19901992Ð20002002Ð2004w-Risk23.231.535.040.8h-Risk21.516.116.411.3Democratsw-Risk24.524.539.551.1h-Risk41.435.425.516.4 Note:Low-risk districts are those in which share ofmajor party vote for presidential candidate ofincumbentÕs party was at least 10 pntage points greater than national vote share.High-risk districts are those in which share ofmajor party vote for presidentiadate ofincumbentÕs party was less than national vote share.Source:Data compiled by authors.  ,   ,          ..   
The evidence displayed in Figure 1 shows that atthe national level redistricting had a negligible impacton the competitiveness ofHouse districts in each ofthe last three redistricting cycles.This does not meanthat redistricting was never used to reduce the com-petitiveness ofHouse districts in a state.In California,for example,the number ofmarginal districts wentfrom 14 in 2000 to only four in 2002.But this was not part ofa national trend.Outside ofCalifornia,the number ofmarginal districts increased slightlybetween 2000 and 2002,going from 109 to 112.Our evidence shows that control ofredistrictinghad no effect on change in the proportions ofsafe andmarginal districts between 2000 and 2002.Regardlessofwhether there was one-party control ofredistrict-ing,divided-party control,or nonpartisan/judicialntrol,there was very little change in the proportionsofsafe and marginal districts.In states in which redis-icting was done by nonpartisan commissions orourts,the proportion ofmarginal districts decreasedfrom 25% in 2000 to 24% in 2002 while the propor-tion ofsafe districts increased from 44% in 2000 to51% in 2002;in states in which redistricting was donepartisan state legislatures,the proportion ofmar-inal districts decreased from 29% in 2000 to 28% in2002 while the proportion ofsafe districts decreasedfrom 46% in 2000 to 45% in 2002.There is no evi-dence that redistricting by nonpartisan redistrictingmmissions or courts resulted in more competitivedistricts than redistricting by partisan state legislatures.The most signiÞcant changes in the competitive-ness ofHouse districts occurred between redistrictingcycles.This pattern is consistent with the partisanpolarization hypothesis.As a result ofpopulationmovement,immigration,and ideological realignmentithin the electorate,Republicans are increasinglysurrounded by other Republicans and Democrats byother Democrats (Oppenheimer 2005).This trend hasbeen evident since the 1970s,but it appears to haveaccelerated in recent years.Between 1992 and 2004,the number ofmarginal districts fell from 157 to 112while the number ofsafe districts rose from 156 toThe effect ofthis increase in partisan polarizationhas been magniÞed by the growing consistency ofting behavior between presidential and House elections.The correlation between the Democraticpercentage ofthe House vote and the Democratic percentage ofthe presidential vote in House districtsincreased from .58 during the 1970s to .82 in2002Ð2004.This was not simply a result ofpartisanumbers ofSafe and Competitive Districts Before and After Redistricting,1980Ð2002 Year200220001992199019821980210200190180170160140130120100 Competitive Source:Data compiled by authors.r additional evidence ofthe failure ofnonpartisan redistrict-ing commissions to increase competition in congressional andstate legislative elections,see Hill (2005).
 .    ,    ,    ouse election between 1946 and 2004,the party afÞliation ofthe incumbent Representative in eachdistrict,whether the incumbent was running foreelection,the previous political experience ofthehallenger,the percentage ofthe major party vote wonthe Democratic and Republican presidential can-didates in every House district in the current or mostnt presidential election in every election between1956 and 2004 except for the 1962 midterm election,and total campaign spending by the Democratic and Republican candidates in every House contestbetween 1972 and 2002.use the normalized presidential vote in eachouse district in the current or most recent presiden-tial election as a measure ofthe partisan compositionofthe district.The normalized presidential vote ismputed by subtracting the Democratic presidentialcandidateÕs percentage ofthe major party vote in theentire nation from his percentage ofthe vote in thedistrict.For the 2004 election,we used the results ofthe 2000 presidential election to construct thismeasure because 2004 presidential election resultsnot yet available for individual House districts.The normalized presidential vote provides us witha measure ofdistrict partisanship that is comparableacross districts and elections and independent ofthe results ofcongressional elections themselves.Anxamination ofthe relationship between the nor-malized presidential vote and the results ofHouseelections during the 1980s,1990s,and the Þrst twoelections ofthe twenty-Þrst century strongly supportsthe validity ofthe normalized presidential vote as ameasure ofdistrict partisanship.The Democraticshare ofthe normalized presidential vote was stronglylated to the Democratic share ofthe vote in con-ested House races during each decade and this rela-tionship became stronger over time.The correlation(PearsonÕs r) between the normalized Democraticpresidential vote and the Democratic share oftheouse vote increased from .65 during the 1980s to .78during the 1990s and .82 in the 2002Ð2004 elections.a result,the percentage ofsafe districts won by thecandidate ofthe majority party increased from 89%during the 1980s to 94% during the 1990s and 97% in2002Ð2004,the percentage ofleaning districts won bythe candidate ofthe majority party increased from72% during the 1980s to 80% during the 1990s and85% in 2002Ð2004,and the percentage ofmarginaldistricts won by the candidate ofthe majority partyincreased from 59% during the 1980s to 61% duringthe 1990s and 69% in 2002Ð2004.Findingsince the 1994 election,Republicans have maintaineda narrow majority in the House ofRepresentatives.The number ofRepublican seats has ranged between223 and 232 while the number ofDemocratic seats hasanged between 203 and 212.Because ofthe smallsize ofthe Republican majority,control ofthe Houseappears to be at stake in every election.Despite theappearance ofnational competitiveness,however,thember ofcompetitive House contests has fallen since1994,reaching record low levels in 2002 and 2004.Onepossible explanation for this decline in competition inindividual House races is that there are fewer marginalouse districts than in the past.Figure 1 displays the numbers ofmarginal andsafe House districts before and after each recent redis-icting cycle.Marginal districts are those in which thetwo-party division ofthe presidential vote was withinplus or minus 5 percentage points ofthe two-partydivision ofthe national presidential vote;safe districtsare those in which the presidential vote was at least 10percentage points more Democratic or Republicanthan the national presidential vote.By using the normalized presidential vote to measure the partisanmposition ofHouse districts before and after redis-icting,we can evaluate the impact ofredistricting onthe numbers ofmarginal and safe districts.Ifredis-icting were responsible for the decline in the com-petitiveness ofHouse districts,we would expect to seea substantial decrease in the number ofmarginal dis-icts and a substantial increase in the number ofsafedistricts immediately after redistricting.ould like to thank Gary Jacobson for providing us with thedata for the 1946 through 2000 House elections.Data on the 2002and 2004 House elections were collected by the authors fromvarious sources including the 2004 edition ofThe Almanac oferican Politicsand the c-span.org web site.Between 1946 and1968 a number ofstates with more than one House district madeuse ofat-large House districts.For an explanation ofthe treatmentofat-large House elections between 1946 and 1968,see the webappendix to this article which is available on the urnal ofPoliticsb site:http://www.journalofpolitics.org.The validity ofthe normalized presidential vote as a measure ofdistrict partisanship does not appear to be affected by the pres-ence ofa strong third party or independent presidential candidateon the ballot.For a discussion ofthis issue,see the web appendixthis article on the urnal ofolitics web site:http://www.journalofpolitics.org.ount Rep.Bernie Sanders ofVermont,who is technically anindependent but caucuses with the Democrats,as a Democrat.Sanders has had either no Democratic opposition or nominalDemocratic opposition in recent elections.Excluding his seat fromthe analysis would have no effect on our results.  ,   ,          ..   

the number ofmarginal districts in the electionsimmediately after redistricting: 1982,1992,and 2002.The partisan polarization hypothesis offers asomewhat different explanation ofthe decline inmpetition in House elections.According to thispothesis,House districts have become less compet-itive over time,but this trend is not a result ofparti-san gerrymandering.Instead,Democratic districtshave become more Democratic,Republican dis-icts have become more Republican,and marginaldistricts have been disappearing as a result ofpower-ful forces at work in American society,including internal migration,immigration,and ideologicalealignment within the electorate.Like the redistricting theory,the partisan polar-ization theory has received considerable support in thepopular press.In a cover story in The Atlantic,Bro(2001) argued that profound cultural differences basedon education,income,religion,and numerous otherfactors were leading to a growing partisan divide in thenited States.Other commentators have accepted thisisdom,sometimes coloring it with references thatfurther exaggerate the degree ofdivision.One colum-The Washington Postput it this way:Òthe redstates get redder,the blue states get bluer,and the polit-ical map ofthe United States takes on the colorationofthe Civil WarÓ(Dionne 2003,A31).Some in academia have repeated this characteri-zation ofa Òdeep cultural divide between the red statesand the blue statesÓ(Mannan 2004,1).One widely dis-cussed academic work goes so far as to argue that cul-tural divisions are so deep they have put the UnitedStates on a path toward political violence (Hunter1995).However,other observers have argued that theidea ofan increasingly polarized America is a Òmyth,Óand that Americans are remarkably moderate andincreasingly tolerant in their political views (Fiorina,ans,and Pope 2004).Staking out something ofamiddle ground in a detailed study ofvoting trends intwelve states,Gimpel and Shuknecht (2003) make astrong case that a range ofcomplex geographically-linked factors such as immigration,migration,educa-tion,income,and religion are contributing to growinggeographic divergence in party loyalties.Stonecash,ewer,and Mariani (2003) and Oppenheimer (2005)have proposed similar explanations for growing par-tisan divergence among congressional districts.Evi-dence ofthis trend can also be seen at the county levelere the number ofcounties dominated by one partyand the proportion ofvoters living in such countieshave increased dramatically over the past severaldecades (Bishop 2004).There is also evidence that growing ideologicalpolarization at the elite level has made it easier forrs to choose a party identiÞcation on the basis oftheir ideological preferences (Abramowitz and Saunders 1998;Jacobson 2000).Therefore what manyobservers describe as ÒpolarizationÓmight more accurately be described as ÒsortingÓas voters bringtheir policy and partisan preferences into alignment (Levendusky 2004).As a result ofthis sorting,south-ern and border states that once regularly elected con-servative Democrats have been trending Republican(Black and Black 2004) while urban and suburbanareas in the North that once regularly elected moder-ate and liberal Republicans have been trending Demo-cratic (Paulson 2004).pothesis 2:Ifthe partisan polarization hypothesis iscorrect,we should Þnd that the number ofsafe districtshas been steadily increasing and the number ofmar-nal districts has been steadily decreasing,with mostthis change occurring between redistricting cycles.ntrast to both the redistricting hypothesisand the partisan polarization hypothesis,the incum-bency hypothesis argues that declining competition inouse elections is due less to change in the partisanmposition ofHouse districts than to the growingadvantages ofincumbency.There are two types ofadvantages to which incumbent politicians might owetheir extraordinarily high reelection rates:advantagesthat derive from holding ofÞce,and campaign-relatedadvantages.Although the debate is ongoing,recentesearch points strongly toward the second set ofadvantages,especially those involving challengeresources,as the cause ofthe recent spike in incum-bent electoral performance (Abramowitz 1991;Campbell 2002,2003;Cox and Katz 1996;Kazee 1983;itt and Wolfram 1997).pothesis 3:Ifthe incumbency hypothesis is correct,we should Þnd that even in marginal districts,compe-ion is now relatively rare because ofthe inability ofchallengers to compete Þnancially.We should also Þndthat the decline in competition in House elections hasbeen most evident in races involving incumbents,withcompetition declining less drastically in open seat races.Data and Measureshave collected data on competition in U.S.Houseelections since the end ofWorld War II.These datainclude the percentage ofthe major party vote won bythe Democratic and Republican candidates in every
 .    ,    ,    tition since the end ofWorld War II:the proportionofHouse races decided by less than 10 percentagepoints fell from 22% between 1946 and 1950 to 21%between 1952 and 1960,17% between 1962 and 1970,16% between 1972 and 1980,12% between 1982 and1990,15% between 1992 and 2000,and 7% in theEven moderately competitive races have becomeincreasingly rare.Between 1946 and 1960 about 40%ofHouse races were decided by a margin ofless than20 percentage points.In the Þrst two elections ofthetwenty-Þrst century,however,only 17% ofHousees were decided by a margin ofless than 20 per-ntage points.Despite the relatively small size ofthepublican majority in the House ofRepresentatives,the close division ofthe electorate between Demo-cratic and Republican identiÞers,and the intensempetition that has characterized the last two presi-dential elections,the vast majority ofindividualouse races in 2002 and 2004 were all but decidedbefore the fall campaigns even began.Explaining Declining Competitionthe remainder ofthis article we test three hypothe-ses that attempt to explain declining competition inouse elections:the redistricting hypothesis,the partisan polarization hypothesis,and the incum-bency hypothesis.The redistricting hypothesis,whichappears to enjoy the status ofconventional wisdomamong media commentators and editorial writers,argues that declining competition is due mainly to the effects ofpartisan or bipartisan gerrymandering.ing to this hypothesis,state legislatures usingsophisticated new computer-based technology havebeen skillfully drawing congressional district lines toeither maximize partisan gains,creating as many dis-icts as possible that favor the majority party bypacking minority party voters into as few districts aspossible,or to protect incumbents ofboth parties.Ineither case,the results ofthis process,according to thedistricting hypothesis,have been an increase in thember ofdistricts that are safe for one party and adecrease in the number ofmarginal districts.As theNew York Timeseditorial page recently opined,Òbothparties have succeeded in drawing district lines inways that cement their current power by eliminatingcontested elections (w York Times 2004).ÓSimilarly,David Broder,the dean ofWashington politicallumnists,has argued that district lines in ÒmoststatesÓwere drawn to protect incumbents from Òtheinconvenience ofcompetitionÓ(2004,A37).hile the refrain in popular media outlets isnearly unanimous regarding the negative conse-quences ofredistricting,the scholarly literature ismore divided.Following MayhewÕs (1974) originalobservations on the Òvanishing marginals,Óone polit-ical scientist found that redistricting was the cause ofmost ofthe decline in marginal districts during the1960s (Tufte 1973).Similarly,a statistical analysisshowed that incumbents beneÞted from reducedmpetition in 1992 as a result ofredistricting,butthat this occurred mainly in states where the redis-icting was bipartisan (Lyons and Galderisi 1995).One expert on mapmaking has argued that the adventofgeographic information systems (GIS) technologyhas made it easier than ever for politicians to draw district lines to achieve their political goals,whetherthose goals are to maximize majority party seats or toprotect incumbents (Monmonier 2001).On the other hand,some studies have concludedthat redistricting has a neutral or positive effect onmpetition (Glazer,Grofman,and Robbins 1987;Gopoian and West 1984).This is not surprisingbecause partisans drawing district lines face a funda-mental tension between incumbent protection andmaximizing their partyÕs electoral potential.Moreoften than not,the only way to shift marginal districtsward the party is to cut the safety margins ofincum-bents by moving reliable partisans out oftheir dis-icts.For this reason,it is often the case that partisandistricting has the effect ofreducing the safety ofincumbents,thereby making elections more competi-tive (Gelman and King 1994).Our focus in this study is the impact ofredis-icting on the partisan composition ofHouse dis-icts,not its impact on competition in the next roundofHouse elections.We classify a district as safe ifitspartisan composition strongly favors one party;weclassify a district as marginal ifits partisan composi-tion is relatively evenly balanced between the twomajor parties.ether the impact ofredistricting on competi-tion is immediate or delayed may depend on factorsother than the partisan composition ofthe new districts such as whether incumbents affected bydistricting choose to run for reelection or retire.hanges in the partisan composition ofdistricts should be evident immediately and have the potential to affect competition for an entirepothesis 1:Ifthe redistricting hypothesis is correct,we should observe a substantial increase in thenumber ofsafe districts and a substantial decrease in  .    ,    ,    bson,Gary C.2000.ÒParty Polarization in National Politics:The Electoral Connection.ÓIn larized Politics: Congress and the President in a Partisan Erads.Jon R.Bond and Richard Fleisher.Washington:Congressional Quarterly Press,pp.9Ð30.e,Thomas A.1983.ÒThe Deterrent Effect ofIncumbency oncruiting Challengers in U.S.House Elections.Óislativeudies Quarterly8 (3):469Ð80.ndusky,Matthew S.2004.ÒSorting,Not Polarization:TheChanging Nature ofParty ID and Ideology in the U.S.Elec-esented at the Annual Meeting ofthe American olitical Science Association.itt,Steven D.,and Catherine D.Wolfram.1997.ÒDecomposingthe Sources ofIncumbency Advantage in the U.S.House.Óislative Studies Quarterly22 (1):45Ð60.ons,Michael,and Peter F.Galderisi.1995.ÒIncumbency,Reap-portionment and U.S.House Redistricting.Ólitical ResearchQuarterly48 (4):857Ð71.annan,Rossana.2004.ÒUCI Professors Examine Political Divideong States.ÓNew University,November 8:News.yhew,David R.1974.ongress: The Electoral Connection.Newn:Yale University Press.nmonier,Mark.2001.Bushmanders and Bullwinkles: Howliticians Manipulate Electronic Maps and Census Data to WinElections.Chicago:University ofChicago Press.New York Times2004.ÒElections with No Meaning.ÓNew YorkTimesebruary 14,p.A14.Oppenheimer,Bruce I.2005.ÒDeep Red and Blue CongressionalDistricts.ÓIn ongress Reconsidered8th ed.,eds.Lawrence C.Dodd and Bruce I.Oppenheimer.Washington:CongressionalQuarterly Press,pp.135Ð57.aulson,Amanda.2004.ÒSuburb Shift Turns State Blue.ÓChrist-ian Science Monitor,July 16,p A1.Stonecash,Jeffrey M.,Mark D.Brewer,and Mack D.Mariani.2003.verging Parties: Social Change,Realignment,and Party Polar-izationoulder:Westview Press.ufte,Edward R.1973.ÒThe Relationship Between Seats and Votesin Two-Party Systems.Óican Political Science Review(2):540Ð55.Alan I.Abramowitz is professor ofpoliticalscience,Emory University,Atlanta,GA 30322.BradAlexander is a Ph.D.Student in political science,Emory University,Atlanta,GA 30322.Matthewunning is a Ph.D.Student in political science,Emoryersity,Atlanta,GA 30322.  ,   ,          ..   
tisan voting:voting in House elections is now muchmore consistent with voting in presidential elections.a result ofboth ofthese trends,districts held byDemocrats are now more strongly Democratic than inthe past and districts held by Republicans are nowmore strongly Republican than in the past.There arefar fewer Democrats and Republicans representinghigh-risk districts,districts whose partisan composi-tion favors the opposing party,and far more Democ-ats and Republicans representing low-risk districts,districts whose partisan composition strongly favorstheir own party.The effects ofincreasing partisan polarizationhave been reinforced by the second trend uncoveredour studyÑthe decreasing Þnancial competitive-ness ofHouse challengers.Even in the remaininghigh-risk districts most challengers lack the Þnancialesources needed to wage competitive campaigns.ether this trend reßects growing ability ofincum-bents to deter strong challengers,growing reluctanceofcontributors to offend powerful incumbents,orsome other set offactors is not clear.The reasons forthe declining Þnancial competitiveness ofHouse hallengers certainly should be explored in futureesearch.The increasing proportion ofsafe districts and thedeclining Þnancial competitiveness ofchallengershave important implications for congressional poli-tics.The decreasing proportion ofDemocrats andpublicans representing marginal or high-risk dis-icts means that there are fewer members who havean incentive to cross party lines on issues in order toappeal to supporters ofthe opposing party in theirdistricts.The result is increased party-line voting andduced bipartisan cooperation on major issues.ncreasing polarization and declining competitionalso mean that both parties have fewer seats at risk inelections.The result is diminished seat turnover andsmaller seat swings.This helps to explain why,despitethe relatively small size oftheir majority,Republicanshave been able to maintain control ofthe House sincetheir 1994 takeover.knowledgmentsould like to thank the three anonymous review-urnal ofPoliticsand Linda Fowler for theirhelpful comments and suggestions.Elizabeth Frickerprovided valuable editorial assistance in the prepara-tion ofthe manuscript.An earlier version ofthisarticle was presented at the 2005 Annual Meeting ofthe Southern Political Science Association,Interconti-nental Hotel,New Orleans,Louisiana.nuscript submitted 30 March 2004nuscript accepted for publication 10 June 2005Referencesamowitz,Alan I.1991.ÒIncumbency,Campaign Spending,andthe Decline ofCompetition in U.S.House Elections.ÓJourofPolitics53 (1):34Ð56.amowitz,Alan I.,and Kyle Saunders.1998.ÒIdeologicalealignment in the U.S.Electorate.Óurnal ofPoliticsBartels,Larry M.2000.ÒPartisanship and Voting Behavior,1952Ð1996.Óican Journal ofPolitical Science44 (1):35Ð50.ishop,Bill.2004.ÒThe Schism in U.S.Politics Begins at Home.Óustin American-Statesman,April 4,p.A1.Black,Earl,and Merle Black.2004.The Rise ofSouthern Republi-ambridge:Harvard University Press.oder,David.2004.ÒNo Vote Necessary.Óashington Postmber 11,p.A37.ooks,David.2001.ÒOne Nation,Slightly Divisible.Ótlanticecember:53Ð65.Campbell,James E.2002.ÒIs the House Incumbency Advantageostly a Campaign Finance Advantage?ÓPaper presented atthe Annual Meeting ofthe New England Political Science Asso-ciation,Portland,Maine.Campbell,James E.2003.ÒCampaign Financing and the Stagna-tion ofCongressional Elections.ÓIn Life After Reform: Foresee-ing the Bipartisan Campaign Reform Act in Practiceael Malbin and Robin Kolodny.Lanham,MD:Rowmanand LittleÞeld,pp.141Ð58.x,Gary W.,and Jonathan N.Katz.1996.ÒWhy Did the Incum-bency Advantage in U.S.House Elections Grow?Óurnal ofPolitical Science40 (2):478Ð97.Dionne,E.J.2003.ÒOne Nation Deeply Divided.Óashington Postmber 7,p.A31.Fiorina,Morris P.,Samuel J.Abrams,and Jeremy C.Pope.2004.ulture War? The Myth ofa Polarized America.New York:earson Longman.Gelman,Andrew,and Gary King.1994.ÒEnhancing Democracythrough Legislative Redistricting.Óican Political ScienceReview88 (3):541Ð59.Gimpel,James G.,and Jason E.Schuknecht.2003.PatchworkNatSectionalism and Political Change in American Politicsnn Arbor:University ofMichigan Press.Glazer,Amihai,Bernard Grofman,and Marc Robbins.1987.ÒPar-tisan and Incumbency Effects of1970 Redistricting.Óurnal ofPolitical Science31 (3):680Ð707.Gopoian,J.David,and Darrell M.West.1984.ÒTrading Securityfor Seats:Strategic Considerations in the Redistrictingocess.Óurnal ofPolitics46 (4):1080Ð96.ll,Steven.2005.ÒSchwarzenegger vs.Gerrymander.ÓNew Yorkebruary 19,p.A15.unter,James Davison.1995.fore the Shooting Begins: Search-ing for Democracy in AmericaÕs Culture Wars.New York:Free  .    ,    ,    Democratic candidateÕs percentage ofcombinedDemocratic and Republican campaign spending inthe race.The results ofthese path analyses are sum-marized in Table 5.These results indicate that district partisanshipand incumbency have strong indirect as well as directeffects on House elections.In addition,the results inble 5 reveal several important trends in the effectsofour three independent variables on election results.During the 1970s,incumbency had a much strongerdirect effect on House election results than districtpartisanship,a slightly stronger direct effect than cam-paign spending,and by far the strongest total effect ofthe three independent variables.During the mostnt period,however,incumbency had a mucheaker direct effect on House election results thaneither district partisanship or campaign spending anddistrict partisanship had by far the strongest totaleffect ofthe three independent variables.The declinein the direct effect ofincumbency may itselfreßect theing importance ofdistrict partisanship sinceincumbents who represent safe districts have littleincentive to cultivate their constituencies in order toxpand their electoral coalitions beyond their partyThe growing correlation among district partisan-ship,incumbency,and campaign spending means thatthe effects ofthese three variables tend to reinforceeach other to a greater extent than in the past.Betweenthe 1972Ð78 time period and the 1996Ð2002 timeperiod the correlation between district partisanshipand incumbency increased from .38 to .59,the corre-lation between district partisanship and campaignspending increased from .45 to .69,and the correla-tion between incumbency and campaign spendingincreased from .66 to .86.Not only do a much largerproportion ofHouse districts now strongly favor oneparty,but the majority party is more likely to be rep-esented by an incumbent and to enjoy an over-elming advantage in campaign spending.Thesult is a pattern ofreinforcing advantages that leadsto extraordinarily uncompetitive elections.The evidence presented in this article indicates thatdeclining competition in U.S.House elections isxplained by two major factors:a shift in the partisanmposition ofHouse districts and a decline in theability ofchallengers to compete Þnancially withincumbents.Since the 1970s,and especially since1992,there has been a substantial increase in partisanpolarization among House districts.The number ofmarginal districts has been declining while thember ofdistricts that are safe for one party has beenincreasing.Redistricting appears to have little ornothing to do with this trend:almost all ofthe changein district partisanship has occurred between redis-icting cycles.Along with the increasing polarization ofHousedistricts,there has been a substantial increase in par- Direct and Indirect Effects on Vote in Contested House Elections during Four Time Periods Effects ime PeriodVariableDirectIndirectTotal1972Ð1978District Partisanship.202.336.538ncumbency.451.234.685Campaign Spending.421Ñ.4211980Ð1986District Partisanship.238.433.671ncumbency.314.356.670Campaign Spending.520Ñ.5201988Ð1994District Partisanship.283.409.692ncumbency.249.397.646Campaign Spending.508Ñ.5081996Ð2002District Partisanship.337.490.827ncumbency.224.345.569Campaign Spending.498Ñ.498 Source:Data compiled by authors.Note:efÞcients shown are path regression coefÞcients.CoefÞcients for election year dummy variables are not shown.  ,   ,          ..   portion were challengers),the estimated net impact ofincumbency on winning candidatesÕmargins fell fromabout 6.4 percentage points in the 1970s to just over3.7 percentage points in the most recent period.These results are consistent with Þndings reported Oppenheimer (2005) based on an examination ofends in the size ofthe sophomore surge and retire-ment slump during the same time period.The declining contribution ofincumbency toinnersÕmargins has been more than offset,however,the increasing contributions ofdistrict partisan-ship and campaign spending.Both the average regres-sion coefÞcient for the district partisanship variableand the average advantage ofwinning candidates onthis variable increased between the 1970s and the late1990s.As a result,the net impact ofdistrict partisan-ship on winning candidatesÕmargins increased fromabout 1.4 percentage point in the 1970s to about 3.9percentage points in the most recent period.The average regression coefÞcient for the cam-paign spending variable changed very little betweenthe 1970s and the late 1990s.However,the averageadvantage ofwinning candidates on this variableincreased dramatically over this time period.Duringthe 1970s,winning candidates accounted for anerage ofabout 69% oftotal campaign spending.Inthe 1998Ð2002 elections,however,winning candidatesaccounted for an average ofalmost 82% oftotal cam-paign spending.As a result,the net impact ofcam-paign spending on the winning candidatesÕmarginsincreased from less than 5 percentage points in the1970s to almost 8 percentage points in the most recentperiod.The results in Table 4 indicate that since the 1970scampaign spending has replaced incumbency as themost important factor contributing to the margins ofinning House candidates.However,the results inble 4 reßect only the direct effects ofthese variableson the outcomes ofHouse elections,not their indirecteffects.Incumbency may affect the outcomes ofelections not only directly but indirectly through itsinßuence on campaign spending:an important and,according to the results presented earlier in this article,ing advantage ofincumbency is the ability ofincumbents to dominate their challengers Þnancially.nd district partisanship may affect the outcomes ofelections not only directly but indirectly through itsinßuence on both incumbency and campaign spend-ing:the larger the majority that a party enjoys amongrs in a district,the more likely it is that the districtll be represented by an incumbent from the major-ity party and the larger the expected Þnancial advan-tage ofthe majority partyÕs candidate.order to examine the direct and indirect effectsofdistrict partisanship,incumbency,and campaignspending on the outcomes ofHouse elections,we nducted path analyses ofHouse election resultsduring each ofthe time periods examined in ble 4.The dependent variable in the path analysesis the Democratic share ofthe major party vote in all contested House races.Independent variables aredistrict partisanship,measured by the normalizedpresidential vote,incumbency (coded as 1 for con-ests with Democratic incumbents,0 for open seates,and 1 for contests with Republican incum-bents) and campaign spending,measured by thege Effects ofIncumbency,District Partisanship,and Campaign Spending on Winning CandidatesÕVote during Four Time Periods iable1972Ð19781980Ð19861988Ð19941996Ð2002ncumbencyCoefÞcient8.506.354.644.45innersÕMean.75.78.78.84t Impact6.384.953.623.74DistrictCoefÞcient.34.38.41.46isanshipWinnersÕMean4.225.705.838.36t Impact1.432.142.403.86Campaign$CoefÞcient.26.29.25.25entageWinnersÕMean69.0876.6879.4181.88t Impact4.877.747.267.97 Source:Data compiled by authors.Note:Based on races with two major party candidates.Dependent variable is winning candidateÕs percentage ofmajor party vote.Coef-Þcients are unstandardized regression coefÞcients.District partisanship measured by normalized presidential vote.Net impact is based on coefÞcient mean value for winning candidates.Mean value used to compute net impact ofcampaign spending is winnerÕs