The Effect of Belief of Victory on Third-Party Vote Share: Duverger's Law & Why Evan McMullin Lost Utah in 2016

Transcription

1 Brigham Young Univrsity BYU ScholarsArchiv Undrgraduat Honors Thss Th Effct of Blif of Victory on Third-Party Vot Shar: Duvrgr's Law & Why Evan McMullin Lost Utah in 2016 John Gilman Brigham Young Univrsity Follow this and additional works at: Part of th Amrican Politics Commons, Modls and Mthods Commons, and th Political Thory Commons BYU ScholarsArchiv Citation Gilman, John, "Th Effct of Blif of Victory on Third-Party Vot Shar: Duvrgr's Law & Why Evan McMullin Lost Utah in 2016" (2018). Undrgraduat Honors Thss This Honors Thsis is brought to you for fr and opn accss by BYU ScholarsArchiv. It has bn accptd for inclusion in Undrgraduat Honors Thss by an authorizd administrator of BYU ScholarsArchiv. For mor information, plas contact

2 Honors Thsis THE EFFECT OF BELIEF OF VICTORY ON THIRD-PARTY VOTE SHARE: DUVERGER S LAW & WHY EVAN MCMULLIN LOST UTAH IN 2016 by John Gilman Submittd to Brigham Young Univrsity in partial fulfillmnt of graduation rquirmnts for Univrsity Honors Political Scinc Dpartmnt Brigham Young Univrsity April 2018 Advisor: David B. Maglby Radr: Joshua Gublr i

3 ABSTRACT THE EFFECT OF BELIEF OF VICTORY ON THIRD-PARTY VOTE SHARE: DUVERGER S LAW & WHY EVAN MCMULLIN LOST UTAH IN 2016 John Gilman Political Scinc Dpartmnt Bachlor of Arts A ky rason Duvrgr s Law is valid is votr s blif that a third-party dos not hav a chanc at winning an lction in a first past th post lctoral systm. Duvrgr s Law has traditionally bn xplaind through two rasons a mchanical factor and a psychological factor. Th mchanical factor focuss on aspcts of lctoral systms that work against third partis, whil th psychological factor focuss on what votrs think and fl about third partis. In th 2016 prsidntial lction in th Unitd Stats, votrs in th stat of Utah dmonstratd that thir prcption of th lctability of a third-party candidat has a substantial ffct on th third-party vot shar. Th Utah Collgs Exit Poll survyd Utah votrs during th 2016 lction, asking if thy would vot for a candidat othr than Hillary Clinton or Donald Trump if th votr thought a third-party candidat could win Utah. My rsarch indicats that if popl had blivd that a third-party candidat could win th stat of Utah and votd for thir prfrrd candidat, Evan McMullin ii

4 potntially could hav won th stat of Utah and gaind 6 vots in th Elctoral Collg. If that had happnd, Trump still would hav had th 270 lctoral vots ndd to scur th prsidncy, but McMullin would hav bn th first third-party candidat to win vots in th Elctoral Collg sinc Gorg Wallac in This finding dmonstrats th importanc of th psychological factor in Duvrgr s Law. Duvrgr s Law was powrful, vn with two vry dislikd candidats from th two major partis. iii

5 ACKNOWLEDGEMENTS I would lik to thank my mntor, Profssor David Maglby, for how much h has taught m ovr th past thr yars. H taught m th finr dtails of political scinc and how to b a good Christian man. His frindship and mntoring has shapd th rst of my lif. I would also lik to thank Profssor Joshua Gublr for th hours h spnt taching m how to rason and how to writ. H also taught m to kp lif in prspctiv and to rmmbr th trnal things. I would also lik to thank my fathr, David, for supporting m, inspiring m, and srving as a copy ditor on this thsis. My mothr, Elizabth, has bn just as important. Sh has lovd m, hlpd m through strssful momnts, and introducd m to th world of books. I would also lik to thank my fiancé, Sarah, for ncouraging and bliving in m. I would doubl th lngth of my thsis if I listd off all of th family, frinds, coworkrs, and fllow studnts who hlpd m on this projct. This projct would not b possibl without thir lov and support. Thank you. iv

6 TABLE OF CONTENTS Titl... i Abstract... ii Acknowldgmnts... iv Tabl of Contnts...v List of Tabls and Figurs... vi Introduction...1 Litratur Rviw and Explanation of Duvrgr s Law...2 Mchanical Factor... 5 Psychological Factor...7 Contributions of this Thsis... 9 Contxt of th 2016 Elction in Utah Mthodology Findings and Discussion Th Conflictd Votrs Th Flings Thrmomtr...21 How to Rapportion Th Conflictd Votrs? In What Scnario Could McMullin Hav Actually Won Utah? Implications Conclusion Appndix A Front Sid of Blu Form Appndix B Confidnc Intrvals of th Dmographics of th Subst and th Rst of th UCEP to Prov Rprsntativnss of th Subst Nots v

7 LIST OF TABLES AND FIGURES Tabl 1- Votrs from Survy Tabl 2- Support Gaind from Conflictd Votrs Tabl 3- Flings of Votrs Towards Trump, Clinton, and McMullin Figur 1- Th Diffrnc of Man Flings That Conflictd and Non-Conflictd Votrs Hav Towards Thir Chosn Candidat and to Evan McMullin...25 Tabl 4- Votrs from Survy...27 Tabl 5- Votrs from Survy, Moving Conflictd Trump Votrs to Third-Party Candidats Tabl 6- Votrs from Survy, Moving Conflictd Trump and Clinton Votrs to Third- Party Candidats Tabl 7- Votrs from Survy, Moving Trump and Clinton Votrs According to th Flings Thrmomtr vi

8 Introduction Evry yar thr ar lctions hld in th Unitd Stats. Thousands of offics ar filld according to th will of th popl. With so many diffrnt offics to fill and th wid varity of idas, on could rasonably xpct to find many political partis vying for ach sat. Howvr, thr ar ffctivly only two th Rpublican Party and th Dmocrat Party. Thr ar othr third partis in th Unitd Stats, such as th Grn Party, th Constitution Party, and th Librtarian Party. Thr also ar individuals who run as indpndnts, lik Ross Prot. Ths partis do not far wll and rarly win sats, spcially on th national lvl. Why dos this happn? Why do only two partis consistntly win lctions? Th answr can b found by xamining Duvrgr s Law. Duvrgr s Law stats that in a singl-majority singl-ballot systm, lik that in th Unitd Stats, thr will only b two comptitiv political partis at th sam tim. 1 Th only xcptions to this law ar during transition tims, whn on party is fracturing and anothr is coming into powr (such as th Rpublicans and Whigs in th 1800s), and whn thr ar vry strong, thnically-basd political groups, such as in Canada and India. 2 Th durability of this law was sn in th Unitd Stats in th 2016 lction. Whil Donald Trump and Hillary Clinton wr historically unpopular candidats, no nationally viabl third-party option aros. Howvr, thr wr som third-party candidats who found limitd succss. Evan McMullin was on of th third-party candidats who ran against Donald Trump and Hillary Clinton. Whil h finishd with a lowr prcntag of th popular vot than Gary Johnson or Jill Stin, h cam th closst to winning a stat Utah, with 21.3% prcnt of th vot. 3 1

9 Using uniqu data gathrd from th Utah Collgs Exit Poll (UCEP), this thsis dmonstrats how Evan McMullin could hav won th stat of Utah if popl had votd for thir prfrrd candidat. This finding quantitativly xplains th psychological causal mchanism of Duvrgr s Law, somthing that has not bn don to this xtnt bfor. This thsis bgins with a litratur rviw to provid contxt on Duvrgr s Law, dscribs th situation of Utah in th 2016 lction, tracs th mthodology usd to discovr why Evan McMullin could hav won th stat of Utah, and thn closs with a gnral discussion on th findings and implications. Litratur Rviw and Explanation of Duvrgr s Law In th 1950s, Mauric Duvrgr prsntd what bcam known as Duvrgr s Law. H wrot th simpl-majority singl-ballot systm favours th two-party systm. Of all th hypothss that hav bn dfind in this book, this approachs th most narly prhaps to a tru sociological law. 4 H qualifid his claim by saying Th lctoral systm [simpl-majority singl-ballot] works in th dirction of bipartism; it dos not ncssarily and absolutly lad to it in spit of all obstacls. Th basic tndncy combins with many othrs which attnuat it, chck it, or arrst it. With ths rsrvs w can nvrthlss considr that dualism of partis is th brazn law of th simpl-majority singl-ballot lctoral systm. 5 A simpl-majority singl-ballot systm was his trm for lctoral systms in which vryon has on vot and do not hold runoff lctions if a candidat did not rciv an absolut majority. Ths systms do not rsult in coalition govrnmnts. Th ballot systm h dscribs is th lctoral systm usd to form th Unitd Stats govrnmnt. 2

10 Othr political scintists had writtn about this sam dual-party trait in a simplmajority singl-ballot systm bfor Duvrgr. Thy includ Frdinand Hrmns in 1941, Hrman Finr in 1949, Carl Fridrich in 1950, and V.O. Ky Jr. in William Rikr xplaind why th law is namd aftr Duvrgr instad of ths arlir political scintists, writing that It is customary to call th law by Duvrgr s nam, not bcaus h had much to do with dvloping it but rathr bcaus h was th first to dar to claim it was a law. Th mmorial honors, thrfor, a trait of charactr as much as a scintific brakthrough. 7 Consquntly, Duvrgr rtains th fam for having com up with a prominnt law in political scinc. As with all significant claims, it was immdiatly challngd by othr political scintists. Thir challngs wr undrstandabl; Duvrgr had mad a bold claim. Political scinc is oftn rfrrd to as a soft scinc bcaus of th lack of concrt laws. Human natur, although it has prdictabl tndncis, can still b rratic and is gnrally unquantifiabl. Most social scintists acknowldg this problm and rfr to thos unquantifiabl charactristics as unobsrvabls. Th standard approach is to attmpt various statistical mthods to account for that problm. 8 For Duvrgr to claim that this tndncy was a brazn law was sn as particularly xtraordinary, and not in a good way. Ths political scintists thn hld th law up to th cold light of xamination. As tim wnt on, th most succssful attack was don by political scintists who contrastd th xampls of Canada and India. According to Knnth Bnoit, [both countris hav] singl-mmbr district plurality lctoral systms but both supporting mor than two partis Conducting th most systmatic rviw (up to his tim) of vidnc for and against Duvrgr s law, Ra (1971) found that Canada 3

11 offrd rgular xcptions Ra suggstd a rvision to Duvrgr s law, assrting that plurality formula ar always associatd with two party comptition xcpt whr strong local minority partis xist (Ra, 1971, 95). Subsqunt attmpts to amnd th laws hav tndd to do so by offring similar qualifications of conditions, or by wakning th catgorical languag suggsting univrsal applicability. 9 Whil Duvrgr latr claimd in a 1986 ssay that h had not intndd for his law to hav th dtrministic significanc latr attributd to it, h clarly had mad a stir in th political scinc community and nshrind a law that still stands with only a fw tackd-on qualifications. 10 For thos who liv in a simpl-majority singl-ballot systm it is important to undrstand why this law works. Elctd officials and citizns both nd to know th causal mchanism bhind th law to fully undrstand th ffcts that Duvrgr s Law has on lctions. Duvrgr wrot, [that this law] is itslf th rsult of two factors working togthr: a mchanical factor and a psychological factor. Th mchanical factor consists in th undrrprsntation of th third, i.. th wakst party, its prcntags of sats bing infrior to its prcntag of th poll Th psychological factor is ambiguous in th sam way. In cass whr thr ar thr partis oprating undr th simplmajority singl-ballot systm th lctors soon raliz that thir vots ar wastd if thy continu to giv thm to th third-party: whnc thir natural tndncy to transfr thir vot to th lss vil of its two advrsaris in ordr to prvnt th succss of th gratr vil. 11 4

12 Ths two factors th mchanical and th psychological ar th rasons that Duvrgr s Law stands. Both aspcts ar important and crat Duvrgr s Law as w know it. Mchanical Factor Duvrgr points out that thr is a mchanical factor that plays a rol in limiting th powr of third partis. Th mchanical aspct is th ruls and conditions that st up th lctoral structur in a mannr unfavorabl to third partis. Thr ar lots of obstacls that limit th lctability of third-party candidats. Thr ar laws rquiring a crtain numbr of signaturs to b placd on th ballot. 12 Signatur gathring rquirs funding and tim, two things that third partis oftn lack. Thr oftn ar filing fs for candidats, somthing that consultants warn potntial third-party candidats about. 13 Stats hav diffrnt filing dadlins, crating a logistical nightmar of kping track of th fforts in diffrnt stats, as wll as limiting th accss of candidats who don t ntr th rac soon nough. Ths xtrnalitis from lction laws mak it hard for third partis to appar on th ballot to run against th major two partis. Ths obstacls all hurt Evan McMullin during his candidacy. H missd th dadlin for filing in Pnnsylvania (August 1) and Wisconsin (August 2) and consquntly did not appar on th ballots for thos two stats. H also did not mak it onto th Michigan ballot. Ths thr stats ar all swing stats. His prsnc on th ballot could hav swung th rsults in thos stats. In som stats h got on th ballots aftr gtting 5,000 signaturs, such as in Kntucky and Virginia. All in all, 5

13 McMullin only got on th ballots of 11 stats, making it hard to b a viabl nationwid candidat. 14 Thr also ar obstacls that arn t codifid into law, but ar crtainly part of th lction structur. Ths includ things such as nding to hav a crtain prcnt of th population supporting you to qualify for nationally tlvisd dbats and finding nough donors to fund both TV ads and a ground gam. Anothr obstacl is rciving insufficint mdia attntion th mdia in a hors rac fashion focuss thir covrag on th top two candidats, limiting th amount of unarnd mdia that third-party candidats rciv. Th simpl-majority singl-ballot itslf maks it difficult for third-party candidats to succd. In a first past th post systm, votrs naturally coalsc around two partis. Whil som viw this bhavior as psychological, it is clar that ths ar th mchanical ffcts of th systm itslf. 15 Ths mchanical ffcts includ th ffcts that th sats availabl in a district and th magnitud of sats in th national assmbly hav on ncouraging or dissuading third party candidats, somthing that gos byond th obstacls mntiond abov. Ths mchanical ffcts hav bn studid by many political scintists, including Douglas Ra, Michal Gallaghr, Arnd Lijphart, and Gary Cox. 16 Thy all agr with Duvrgr that mchanical ffcts play an important rol in th duality of lctoral systms with a winnr tak all systm. Othr political scintists, such as Lsli Lipson, disagrd with th findings of mchanical ffcts and claimd that th numbr of partis had stabilizd du to social quilibrium. 17 This thory has largly falln out of favor, as dmonstratd by th xtnsiv, and ongoing, 6

14 rsarch of Rikr, Octavio Nto and Gary Cox, and Bnoit. 18 Thos political scintists catalogu th history of rsarch into Duvrgr s Law in thir litratur rviws, going all th way up to 2006 in th cas of Bnoit. Political scintists hav trid to crat prdictiv modls and quations on th numbr of comptitiv partis that will b prsnt in any givn country. Rin Taagpra and Matthw Shubart argu that numbr of sats in a district and th ovrall assmbly siz hav th most impact on dtrmining th amount of ffctiv partis in a country. 19 Psychological Factor Th psychological factor focuss on why votrs mak th dcisions that thy do, diffring from th mchanical factor which focuss on stting th structural conditions for third partis to fail. Th ida of a psychological factor was strongly attackd by som political scintists whn Duvrgr introducd his law. John Grumm in 1958 wrot that th xamination of th voting statistics of ths countris supplis almost no vidnc of th xistnc of th hypothtical psychological factor. 20 Othr political scintists hav not dismissd it so radily and hav trid to study th psychological factor bhind Duvrgr s Law. Thy hav mt with numrous problms as thy hav attmptd to study th actual chancs third-party candidats hav in an lction. On problm consistntly ncountrd was th lack of quality data on th subjct. This was th problm Grumm xprincd in his analysis in

15 Political scintists hav com up with crativ ways to gt around this problm. Som rly on qualitativ obsrvations and hypothtical thorms, such as Rogr Myrson and Robrt Wbr. 21 Othrs hav usd cultural clavags as thir psychological factor to dtrmin th numbr of political partis that a country can hold. In 1997, Nto and Cox found that th numbr of cultural clavags, combind with th lctoral structur, ar indicativ of th numbr of partis. 22 This supportd arlir findings by Ptr Ordshook and Olga Shvtsova. 23 Still othrs hav spculatd that th gratr th distanc btwn th votr and th narst major party candidat, th mor likly it is that th votr will look for a third-party altrnativ. 24 Political scintists who follow this school of thought look at votr satisfaction with candidats as thy crat thir modls of votr bhavior. 25 Othrs, such as Paul Abramson, us fling thrmomtrs of favorability instad of satisfaction to dtrmin which candidat should hav fard th bst. Abramson, using his fling thrmomtr mthod, concludd that at last som votrs rspondd to th wastd vot argumnt [but that] th mchanical ffcts of th plurality-vot-win systm had mor of an impact than its psychological ffcts. 26 His findings, as with othrs using similar mthods, ar full of cavats bcaus thy us data that only indirctly masurs th lctability of third-party candidats. 8

16 Contributions of this Thsis With th cry for a third-party that is so oftn rpatd, particularly in th currnt politically polarizd climat in th Unitd Stats, why has no group succssfully formd a comptitiv third-party? 27 It cannot b a compltly structural thing, as th litratur rviw has dmonstratd. Thr is no law against a third-party bing formd and many hav bn formd throughout history. Looking for othr rasons, thr must b a psychological factor that affirms Duvrgr s Law. Th clar majority of Amricans do not know what Duvrgr s Law is but hav obsrvd it for dcads with thir voting bhavior. This indicats that popl hav a similar mntality about voting for third partis, a mntality that causs thm to shy away from voting from third-party candidats. My thsis focuss on this psychological factor. I askd votrs Would you vot for a candidat othr than Hillary Clinton or Donald Trump if you thought thy had a chanc to gt th most vots in Utah? A similar qustion about third-party prfrncs, in a campaign with a srious third-party contndr, has only bn askd onc bfor. During a Votr Rsarch and Survys (VRS) xit poll in th 1992 lction, Gordon S. Black, thn CEO of th Gordon S. Black Corporation (which was latr mrgd with Harris Insights and Analytics) askd 3,900 popl if thy would hav votd for Ross Prot if thy thought h had a chanc to win. H found that Prot would hav won th lction with 40% of th vot. 28 Black did not us this to xplain Duvrgr s Law. H concludd that th impropr polling had ld th public to bliv Prot did not hav a chanc, which is why popl did not vot for him. Gordon Black usd this conclusion to chastis th pr-lction 9

17 pollstrs across th country and nvr did anything mor with th data. 29 His finding hints at th psychological aspct, but th connction was nvr xplicitly mad. His data and survy mthods ar not publicly availabl. Bcaus his data and survy mthods ar not publicly availabl, it is hard to hold up his findings as conclusiv proof of th psychological ffct of Duvrgr s Law. For th rasons statd in th sctions abov, it is hard to quantitativly tst this aspct of Duvrgr s Law. Bcaus of my qustion in th UCEP, th strngth of th mthods usd to survy votrs, and th simplicity of my analysis, I hav a uniqu data st that quantitativly provs th impact of th psychological factor on Duvrgr s Law. Contxt of th 2016 Elction in Utah Th lction of 2016 was on of th most unusual lctions that th Unitd Stats has vr sn bcaus Hillary Clinton and Donald Trump wr viwd mor unfavorably than any othr st of prsidntial candidats sinc Gallup bgan tracking candidat favorability ratings in On th wk lading up to Elction Day, Gallup rportd that Clinton s 52% unfavorabl scor was th scond wors favorability scor vr by a candidat, batn only by Trump s 61% unfavorabl scor. 30 Many popl wr rplld by th prsonality and bhavior of Donald Trump, th Rpublican candidat. H rgularly mad offnsiv commnts about Mxicans, Muslims, and womn. H also ran on a populist, protctionist platform that diffrd from traditional Rpublican stancs, causing many Rpublicans to shy away from voting for him. On th othr hand, Hillary Clinton was sn by many as a dply flawd candidat. Hr dcads in th public y had ld to 10

18 an accumulation of a lot of baggag, including hr commnts about young black mals as supr-prdators, hr mishandling of classifid mails, th Bnghazi incidnt, and hr marriag to formr prsidnt Bill Clinton. Th unpopularity of ths candidats pushd votrs into an uncomfortabl zon whr many votrs in th middl flt thy had to pick a candidat whos baggag thy flt most comfortabl with. In Utah th cross prssur was vn mor xtrm. Th majority of popl in th stat ar mmbrs of th Church of Jsus Christ of Lattr-day Saints (also rfrrd to as th Mormon or LDS Church). Traditionally, most mmbrs of th LDS Church lan to th right of th political spctrum. This has ld to Utah voting Rpublican in prsidntial lctions. In fact, th last tim a Dmocrat won Utah s lctoral vots was in In addition to a tndncy to vot Rpublican, Mormons ar known for bing a moral popl. 31 On of thir Articls of Faith stats W bliv in bing honst, tru, chast, bnvolnt, virtuous, and in doing good to all mn 32 Donald Trump s bhavior, both bfor and aftr h bcam th Rpublican nomin, did not rflct thos virtus. Howvr, most popl in th stat of Utah did not want to vot for Hillary Clinton. Th Clinton imag was nvr good in th stat of Utah, du to a mix of thir political viws, th immorality of Bill Clinton whil h was in offic, th xcutiv dcision by thn Prsidnt Clinton to crat th Grand Staircas-Escalant National Monumnt, and th scandals that marrd Hillary Clinton s tim as Scrtary of Stat. 33 This frustration was flt not just by avrag Utah votrs, but by Utah s lctd officials, such as Govrnor Gary Hrbrt, Snator Mik L, and Rprsntativs Jason Chafftz, Mia Lov, and Chris Stwart. This frustration boild ovr aftr th Accss Hollywood taps with Trump s xtrmly lwd and graphic commnts about womn 11

19 wr rportd in th nws in Octobr Upon haring Trump s rcordd commnts, prominnt Utahns ld th way in withdrawing thir ndorsmnts and asking him to stp asid. Govrnor Gary Hrbrt was th first lctd official to pull his ndorsmnt from Donald Trump. 35 Rprsntativ Jason Chafftz was th first mmbr of th US Hous of Rprsntativs to withdraw his ndorsmnt of Donald Trump, stating I m out. I can no longr in good conscinc ndors this prson for prsidnt I can't tll th good popl of Utah that I ndors a prson who acts lik this. 36 Formr Utah govrnor and 2012 prsidntial candidat Jon Huntsman also withdrw his support, as did Utah s favorit son Mitt Romny. 37 Th backlash against Trump in Utah was so intns that th Dsrt Nws, a mdia outlt ownd by th Church of Jsus Christ of Lattr Day Saints, brok with an 80-yar tradition of rfraining from prsidntial ndorsmnts to publish an ditorial calling on Mr. Trump to stp asid. 38 Th majority of votrs could not choos which candidat thy flt lss uncomfortabl with! Thy didn t want to go against gnrations of Rpublican voting but didn t fl comfortabl with Trump s morality. Thn, if thy wr wavring on voting for Trump, thy wr facd with th prospct of voting for Clinton, a daunting prospct that was a tough pill to swallow for many votrs. Donald Trump s running mat, Michal Pnc, rmindd Utahns of what was at stak, promising that thy would fight against abortion, mak it asir for rligions to spak in th public forum, and appoint a consrvativ, pro-lif justic to th Suprm Court. 39 Thy also lackd clar cus from thir lctd officials about which candidats thy should support. Whil Snator Orrin Hatch and Rprsntativ Rob Bishop continud to support Donald Trump s candidacy, thy wr fairly mutd in thir support 12

20 as thy both strongly disapprovd of Trump s statmnts towards womn. 40 Utahns also rcivd mixd cus from othr lctd officials. Govrnor Hrbrt dclard that h would not vot for Donald Trump or Hillary Clinton, but did not say who h would vot for. Rprsntativ Jason Chafftz said h would not ndors Trump but would vot for him bcaus HRC is that bad. 41 Caught in this cross prssur, many popl flt unsur what to do, and nighbors bgan having discussions, both in prson and via social mdia, about choosing th lssr of two vils. 42 With no party agrmnt or statwid consnsus about which candidat flt lik th lssr of two vils, many Utahns flt trappd, frustratd with a systm that had ld thm to having to choos btwn two candidats that thy did not lik. In August 2016, a potntial altrnativ aros. Evan McMullin, a mmbr of th LDS faith and a graduat of Utah-basd Brigham Young Univrsity, announcd his candidacy. H ran as an Indpndnt, mphasizing that h did not hav th baggag that Trump and Clinton wr prcivd as having. His talking point that rsonatd bst with popl was I m not Trump or Clinton! H prsntd himslf as a modrat who was frustratd with th candidats that th primary systm had cratd. In practic, h was a typical Rpublican. H had bn th chif policy dirctor to th Hous Rpublican Confrnc and workd with GOP havywights lik Paul Ryan and John Bohnr. H also had bn a mmbr of th CIA, which boostd his appal in th patriotic stat of Utah. 43 Evan McMullin bcam a much discussd third option in th stat of Utah and popl wondrd if thy should vot for him so that thy would fl comfortabl with thir vot. McMullin s candidacy gaind ground and multipl polls suggstd that h was ithr tid with Trump or within striking distanc of winning Utah. 44 McMullin s 13

21 popularity surg in Utah amazd national obsrvrs who wr stunnd that a third-party candidat was putting togthr a comptitiv campaign. This is th contxt of Utah lading up to Elction Day in Utah had always bn a stat that Rpublicans could count on for lctoral vots in th prsidntial rac. Howvr, du to th cross prssurs Utahns flt as thy dcidd btwn Donald Trump and Hillary Clinton, thr was no crtainty on how Utah would vot, a stark diffrnc to dcads of prvious prsidntial lctions. Mthodology I participatd in th 2016 Utah Collgs Exit Poll (UCEP). Th Utah Collgs Exit Poll has bn run sinc 1982 and has bn th most accurat xit poll in th stat of Utah. It has bn rfrrd to both as th gold standard of polling and as th Cadillac of xit polling by Fritz Schurn, formr had of th Amrican Statistics Association. 45 Th UCEP is run by Profssor David Maglby of Brigham Young Univrsity. H tachs th classs of studnts that plan th xit poll, including writing th poll itslf, crating th statistical modl of how th stat of Utah will b sampld, rcruiting and training voluntr studnt pollstrs, assigning th pollstrs to Elction Day locations, inputting th data, and solving criss that invariably aris on Elction Day itslf. All of this is don in conjunction with othr collgs and univrsitis in th stat of Utah. In 2016, this includd Brigham Young Univrsity, th Univrsity of Utah, Utah Vally Univrsity, Utah Stat Univrsity, Wbr Stat Univrsity, Southrn Utah Univrsity, and Dixi Stat Univrsity. Th UCEP also workd in conjunction with th Liutnant Govrnor s Offic of th Stat of Utah. All told, thr wr approximatly 2400 voluntrs that participatd 14

22 as pollstrs on Elction Day, which givs an ida of th magnitud of th Utah Collgs Exit Poll. 46 Dr. Maglby xplaind how th sampling was don in his papr to th Amrican Political Scinc Association in H wrot, Th sampl is a multi-mod stratifid random sampl of Utah Votrs dsignd to rprsnt th population of votrs in ach of Utah s four congrssional districts. Th sampl is not rprsntativ of stat lgislativ districts and not all countis ar in th sampl Th mods of th sampl wr a sris of random sampls of arly and by-mail votrs drawn from th list of votrs whos ballots had bn rturnd to our sampl countis. Thy wr contactd by postcard asking thm to go onlin and complt th poll, and to th xtnt w could accss phon numbrs for ths sampls multipl attmpts wr mad to contact ths votrs by tlphon. Our Elction Day sampl was stratifid at th congrssional district and county lvls, and for in-prson voting countis th voting placs wr randomly slctd (using a probability proportional to siz (PPS) mthod) as was th sampl of votrs (using an intrval mthod). For vot-by-mail countis w rrd on th sid of including all vot cntrs in th Salt Lak, Davis, Wbr, and Cach countis. 47 ing wights ar applid to account for th PPS sampling dsign. 48 Ths sampling wights wr don by Profssor Dan Williams and som of his statistic studnts that wr involvd with th Exit Poll. My study uss this wightd data to nsur accuracy of th findings across th stat of Utah. Th wording of th third-party qustion was Would you vot for a candidat othr than Hillary Clinton or Donald Trump if you thought thy had a chanc to gt th 15

23 most vots in Utah? Th answrs wr Ys, No, or Don t Know. To maximiz th qustions that could b askd in th Exit Poll whil minimizing survy fatigu by thos rsponding, w dcidd to us four diffrnt forms whit, grn, yllow, and blu. Each rspondnt only rcivd on of th four forms. Whil many of th qustions wr th sam, such as dmographic qustions and thir vots for thir spcific racs, othr acadmic qustions, lik min, only appard on on form. Min appard on th blu form. Our pollstrs wr givn spcific instructions. If thy askd a prson to participat in our voluntary poll and th individual dclind, our pollstrs wr to fill out a survy form that includd th tim of contact and a fw facts about th individual who had just turnd down th survy. Ths nonrspons forms wr includd in th total n of th survy to nsur that thr was no systmatic bias in non-rsponss. Including thos nonrsponss to th survy, 51,269 popl wr survyd in th 2016 Utah Collgs Exit Poll. I droppd all th nonrsponss which lft m with 31,778 rsponss. From thos 31,778 rsponss, I droppd all th nonrsponss to thos who did not answr my qustion on th blu form. This liminatd individuals who wr givn th whit, grn, and yllow forms, as wll as individuals who did answr my qustion on th blu form. I also droppd individuals who did not rcord who thy votd for in th prsidntial rac. This lft m with 7,747 popl in my subsampl. Whil this appars to b a lot of drops of data points, thr ar rasons for ths drops. I cannot run tsts on individual rsponss for popl s viws of third-party candidats if thy did not spcify thir viws on third-party candidats. As such, all th drops wr don to narrow in on th spcific subst of popl who had rspondd to my 16

24 qustion. This narrowing in procss liminatd thr-quartrs of thos who had participatd in th Utah Collgs Exit Poll and ovr four-fifths of th total popl survyd. Thortically, ths drops mak sns. Whn xit polling is occurring, a ncssary assumption is mad by th pollstrs that thr is no systmatic non-rspons bias. To b positiv that thr was no systmatic non-rspons bias, I usd a datast that was wightd in th wks aftr th lction to accuratly rflct th final vot totals of th stat of Utah. This wighting was don so that political scintists would b abl to us th rsults of th Exit Poll aftr th lction for projcts lik this. Th nxt qustion is whthr or not it is appropriat to drop all th othr forms as I cratd my subst sampl. Th blu form was givn out to vry fourth prson survyd, and ach prson survyd was chosn basd on a prst intrval. Bcaus th blu form was givn to individuals who wr randomly chosn, th blu form thortically rprsnts an accurat sub-sampl of th ntir sampl. Two of th ky assumptions that my thsis rsts on ar that a) th survy was proprly don and that b) my subsampl is rprsntativ of th sampl as a whol. Th statistics profssor rsponsibl for wighting this datast, Dan Williams of BYU, wrot that th data was wightd to mak it rprsntativ of Utah s tn most populous countis. H also wrot that What w hav found is that th diffrnc in xcluding th last populatd countis affcts th vot totals only vry minimally - not nough to caus concrn that our data stimats particularly on th vot prcntag stimats would not b accurat. 49 This wighting mthod, in addition to th propr survy mthodology, nsurs that our sampl rprsnts an accurat viw of Utah votrs. 17

25 I also statistically validatd th rprsntativnss of my subsampl (th sampl I usd aftr all th drops) by comparing that subsampl to th rst of th sampl. Using STATA, I gnratd mans and confidnc intrvals for a varity of dmographic charactristics for my subsampl and for th othr rspondnts to th survy. 50 I xamind th 95% confidnc intrvals to nsur that thr was ovrlap. Th 95% confidnc intrvals on Othr for Party ID do not lin up. This is not much of a problm bcaus on vry othr dmographic variabl th confidnc intrvals ovrlap, maning that p valus would b gratr than.05 on vry othr variabl whn comparing th subst to th rst of th sampl. This strngthns my cas that my subsampl is rprsntativ of th sampl as a whol. To s ths 95% confidnc intrvals, plas rfr to Appndix B. In summary, my subsampl dos not includ rspondnts who did not fill out th blu form (which w know was random) and popl who did not answr that qustion but did fill out th rst of th blu form (which w hav to assum was unsystmatic, spcially sinc thr is no rason to think that it was systmatic). I also droppd thos who did not say who thy votd for in th prsidntial rac. Ths drops wr statistically appropriat and my subsampl still rprsnts th stat of Utah. Aftr all thos drops, I hav 7,747 rsponss to that qustion, which is a larg numbr of rspondnts. This subsampl is th group that I usd to xamin th strngth of Duvrgr s Law. 18

26 Findings and Discussion Th Conflictd Votrs Tabl 1 shows th distribution of prsidntial vots from th sampl of popl who filld out th blu form and answrd my qustion. Tabl 1- Votrs from Survy Candidat Survy Vot Prcntags Donald Trump 40% WINNER Hillary Clinton 31% Gary Johnson 4% Evan McMullin 21% Somon ls 3% Don t rmmbr.4% Did not vot.6% 100% I cratd a nw variabl to pull out th votrs who votd for Trump or Clinton but prfrrd a third-party candidat. I codd th variabl so that an individual would b markd as a 1 if thy had votd for Trump but would hav votd for a third party if thy 19

27 thought th third-party candidats had a chanc of winning. This cratd a catgory of Trump votrs who would hav changd thir vot. For Clinton votrs, I codd th variabl so that an individual would b markd as a 2 if thy had votd for Clinton but prfrrd a third-party candidat. This cratd a catgory of Clinton votrs who would hav changd thir vot if thy had blivd thir prfrrd choic could win. Whn ths individuals markd as a 1 or 2 ar summd up, thy ar th total of votrs who votd for Clinton and Trump but would hav votd for a third candidat, assuming that third-party candidat was lctabl. I could hav mad thm both th sam numbr but I wantd thm codd diffrntly bcaus thy had actd diffrntly in which mainstram candidat thy had votd for. Aftr isolating th Trump votrs who prfrrd McMullin but didn t think McMullin could win (who I will rfr to as conflictd Trump votrs), I found thr wr 1,210 votrs who would hav changd thir vot. Aftr isolating th Clinton votrs who didn t think a third-party candidat could win. I found thr wr 860 Clinton votrs who prfrrd a third-party candidat but would hav changd thir vot (who I will rfr to as conflictd Clinton votrs). Ths wr surprisingly larg numbrs. Clarly a larg numbr of votrs wantd to dfct from thir candidats but had not don so. Tabl 2 shows what happns to th total vot shars of Trump and Clinton in th prsidntial rac if, hypothtically, thir conflictd votrs dcidd to not vot for thm. 20

28 Tabl 2- Support Gaind from Conflictd Votrs Trump Total Vot Shar Clinton Total Vot Shar With 40% 31% Conflictd Votrs Without 24% 20% Conflictd Votrs Clinton and Trump both pickd up a larg numbr of votrs who would hav votd for somon ls if thy thought th othr candidat could win. Th fact that thy had so many conflictd votrs voting for thm is a tstamnt to th strngth of Duvrgr s Law. Evn whn votrs flt a strong dislik for both candidats, thy still chos on or th othr instad of voting for a third-party candidat. Thy did not bliv a third-party candidat could win and consquntly did not vot for a third-party candidat. This validats th psychological factor Duvrgr pointd to as bing an intgral part of his law. Th Flings Thrmomtr On of th assumptions that ths findings of a svrly diminishd vot shar for Trump and Clinton rst on is that votrs who indicatd that thy would vot for an lctabl third-party would actually do so. If thy would hav don what thy said, thn it 21

29 is compltly appropriat to rmov thir vots from thir chosn candidat and rallocat thir vot to a third-party candidat s vot shar for this analysis. This sction xamins th liklihood that conflictd votrs would actually hav votd for a third-party candidat. Thir flings towards McMullin ar th basis for analysis in this sction for two rasons. First, McMullin was th most popular third-party candidat in th stat of Utah and it is natural to assum mor of th Rpublican vots would hav gon to him than to othr third-party candidats. His popularity is shown by th fact that h had bn tid with Trump and Clinton in many of th polls right bfor th lction. 51 H lost by a significantly largr margin than polls had shown in th wks lading up to th lction, which this data indicats is bcaus votrs dcidd that h did not hav a chanc to win in Utah. Th scond rason is that data on flings towards Johnson and Stin wr not gathrd by th UCEP, which is a limitation of this study. It is thrfor unknown which third-party candidat conflictd Clinton votrs would hav prfrrd. This ncssary assumption that popl who indicatd that thy would hav votd for an lctabl third-party candidat, such as McMullin, is strngthnd by looking at th flings of votrs towards Trump, Clinton, and McMullin. On of th othr qustions on th UCEP askd how votrs viwd Donald Trump, Hillary Clinton, and Evan McMullin. Using th catgoris of conflictd and non-conflictd votrs, I compard th flings btwn th groups that thy hav towards thir prfrrd candidat and towards Evan McMullin. If th conflictd votrs who rspondd that thy would vot for a thirdparty candidat viw McMullin mor favorably than thir chosn candidat (Trump or Clinton), thy ar mor likly to hav actually votd for McMullin if thy blivd h could win. Thir answrs, aggrgatd by groups, ar blow. 22

30 Tabl 3- Flings of Votrs Towards Trump, Clinton, and McMullin 52 All Trump Votrs All Clinton Votrs All McMullin Votrs Non- Conflictd Trump Votrs Conflictd Trump Votrs Non- Conflictd Clinton Votrs Conflictd Clinton Votrs n 3,108 2,403 1,587 1,898 1,210 1, Man Flings towards Trump Mdian Flings towards Trump Standard Dviation, Flings towards Trump Man Flings towards Clinton Mdian Flings towards Clinton Standard Dviation, Flings towards Clinton Man Flings towards McMullin Mdian Flings Towards McMullin Standard Dviation, Flings towards McMullin

31 Thr ar stark diffrncs btwn conflictd Trump votrs and non-conflictd Trump votrs. Non-conflictd Trump votrs had a man fling of towards Trump, whil conflictd Trump votrs had a man fling of This is a larg diffrnc, a diffrnc which is statistically significant with a p valu of lss than.001. Thr is also a significant diffrnc btwn non-conflictd Trump votrs and conflictd Trump votrs flings towards McMullin. Non-conflictd Trump votrs had a man fling of towards McMullin, whil conflictd Trump votrs had a man fling of This larg diffrnc also has a p valu of lss than.001. This indicats that conflictd Trump votrs viwd McMullin as an accptabl altrnativ to Trump, strngthning my claim that thy would hav actually votd for McMullin if thy thought h had a chanc to win Utah. Ths findings ar also consistnt whn comparing th man flings of nonconflictd Clinton votrs and conflictd Clinton votrs. Non-conflictd Clinton votrs had a man fling of towards Clinton, whil conflictd Clinton votrs had a man fling of Th p valu of that diffrnc in mans is lss than.001. Whn comparing th two groups flings towards McMullin, non-conflictd Clinton votrs had a man fling of 2.564, whil conflictd Clinton votrs had a man fling of This is also statistically significant, with a p valu of lss than.05. Whn comparing th man flings of conflictd Clinton votrs btwn McMullin and Clinton, conflictd Clinton votrs did fl mor favorably towards Clinton than thy did towards McMullin. Howvr, that diffrnc in th mans is lss, and th mdians ar th sam. This diffrnc can b xplaind by McMullin s d facto status as a Rpublican and would hav a fw diffrnt policy prfrncs than som conflictd Clinton votrs. Bcaus th 24

32 diffrnc is so small and is xplainabl, it still is ralistic that at last som conflictd Clinton votrs would hav votd for a third-party candidat. This was dmonstratd in th flings thrmomtr apportionmnt mthod. Ths diffrncs btwn conflictd and non-conflictd votrs ar shown in th following figur. Figur 1- Th Diffrnc of Man Flings That Conflictd and Non-Conflictd Votrs Hav Towards Thir Chosn Candidat and to Evan McMullin Groups of Votrs Clinton Votrs Flings to Clinton Trump Votrs Flings to Trump Clinton Votrs Flings to McMullin Trump Votrs Flings to McMullin Diffrnc in Flings This figur illustrats th diffrncs that conflictd votrs fl compard to nonconflictd votrs. On th top lin, w s th diffrnc of conflictd Clinton votrs and non-conflictd Clinton votrs to McMullin. Th othr labls xplain which groups th diffrnc of mans and confidnc intrvals ar for. A point along th zro lin would indicat that conflictd and non-conflictd votrs flt th sam way about a candidat. Whn th point and its confidnc intrval ar on th ngativ sid of th chart, it 25

33 indicats that conflictd votrs fl lss favorabl towards th candidat than th nonconflictd votrs. Convrsly, whn th point and its confidnc intrval ar on th positiv sid of th chart, th conflictd votrs had a mor favorabl viw of th candidat than th non-conflictd votrs. Figur 1 dmonstrats that conflictd Trump votrs had significantly lowr viws of Trump than non-conflictd Trump votrs did (Group 1). Th scond point shows that conflictd Trump votrs had highr viws of McMullin than non-conflictd Trump votrs did (Group 2). Groups 3 and 4 dmonstrat similar findings for Clinton votrs viws on thir chosn candidat and McMullin. Bcaus thr is such a strong diffrnc in flings btwn conflictd Trump votrs and non-conflictd Trump votrs both of thir viws on thir chosn candidat and th most popular altrnativ (McMullin), it is likly that thos conflictd votrs would actually hav votd for a third-party. This finding is also tru for Clinton votrs. Thr is such a strong diffrnc in flings btwn conflictd Clinton votrs and non-conflictd Clinton votrs towards Clinton and towards McMullin that thy likly would hav actually votd for a third-party candidat. Th assumption that thos conflictd votrs would hav actually votd for Evan McMullin if thy blivd h was lctabl was also supportd by his favorability rating in Utah. In th middl of Octobr, a month bfor th lction, Rasmussn Rports found that 51% of likly Utah votrs had a favorabl or somwhat favorabl viw of McMullin. Only 26% had a vry unfavorabl or somwhat unfavorabl viw of McMullin. 53 With such a high favorability rating, and th larg diffrnc in flings btwn conflictd votrs and non-conflictd votrs, it is vry likly that thos conflictd votrs would hav actually votd for a third-party candidat. 26

34 How to Rapportion Th Conflictd Votrs? If, hypothtically, th conflictd votrs had followd thir prfrnc and votd for a third-party candidat, which on would thy hav votd for? This is important bcaus thir dcision on which third-party candidat thy votd for would dtrmin who wins th lctoral vots of Utah. I usd two mthods to rapportion conflictd votrs to third-party candidats. Both mthods hav bnfits and drawbacks. Th first way of rapportioning th conflictd votrs is by rapportioning thm proportionally. Tabl 4 shows th initial distribution of prsidntial vots from th sampl of popl who filld out th blu form and answrd my qustion. Tabl 4- Votrs from Survy Candidat Survy Vot Prcntags Donald Trump 40% WINNER Hillary Clinton 31% Gary Johnson 4% Evan McMullin 21% Somon ls 3% Don t rmmbr.4% Did not vot.6% 27

35 100% Bcaus w don t know which third-party candidat th conflictd votrs would hav gon for, w will rapportion th conflictd Trump votrs and conflictd Clinton votrs btwn McMullin, Johnson, and othrs in a proportional mannr. In th initial vot count in our subsampl, McMullin had 1,587 vots, Johnson had 303, and somon ls had 270. W will not apportion any to thos who said, Don t rmmbr and thos who did not vot for U.S. prsidnt. Proportionally, of th third-party vot, McMullin had 73%, Johnson had 14%, and somon ls had 13% (all roundd). Of conflictd Trump votrs, McMullin pickd up 883 vots (1,210*.73), Johnson pickd up 170 vots (1,210*.14), and somon ls pickd up 157 (1210*.13). (I gav Johnson th xtra vot bcaus h was closst to rounding up. Othrwis it would hav bn 169 for Johnson, which is 1,209 ovrall). Tabl 5 holds th rsults of what happns whn w rapportion thos who votd for Trump bliving that McMullin had littl chanc of winning. Tabl 5- Votrs from Survy, Moving Conflictd Trump Votrs to Third-Party Candidats Candidat Survy Vot Prcntags Donald Trump 24% 28

36 Hillary Clinton 31% Gary Johnson 6% Evan McMullin 32% Somon ls 6% Don t rmmbr.4% Did not vot.6% 100% As you can s, McMullin would hav com clos to winning th stat of Utah with just conflictd Trump votrs. Th nxt stp is to factor in conflictd Clinton votrs. Of conflictd Clinton votrs, McMullin would hav pickd up 628 vots (860*.73), Johnson would hav pickd up 120 vots (860*.14), and somon ls would hav pickd up 112 vots (860*.13). Tabl 6 adds th conflictd Clinton votrs into th rsults in Tabl 4, which alrady includs Conflictd Trump votrs aftr thy wr rapportiond. 29

37 Tabl 6- Votrs from Survy, Moving Conflictd Trump and Clinton Votrs to Third- Party Candidats Candidat Survy Vot Prcntags Donald Trump 24% Hillary Clinton 20% Gary Johnson 8% Evan McMullin 40% WINNER Somon ls 7% Don t.4% rmmbr Did not vot.6% 100% This finding has som bnfits and drawbacks. This mthod of proportional rapportionmnt is don basd on th proportion of th third-party vot shar that McMullin, Johnson, and othrs rcivd in th 2016 lction. Th rasoning bhind this proportional rapportionmnt is that it is logical to assum that th third-party vot shar in th lction is rprsntativ of how Utah votrs flt about ach candidat. This mthod is also simpl. 30

38 Howvr, thr ar no guarants that this is xactly th prcntags of popl that would hav votd for McMullin, Johnson, and Stin, although it is plausibl. Th biggst drawback with this mthod is rapportioning Clinton votrs. It is vry likly that conflictd Trump votrs would hav transfrrd thir vot to McMullin if thy blivd McMullin could hav won. It is much lss likly that Clinton votrs would hav transfrrd thir vot to McMullin. Many conflictd Clinton votrs would hav had substantiv policy diffrncs with McMullin, who ssntially was a traditional Rpublican. As conflictd Clinton votrs thy would hav had lss rason than conflictd Trump votrs to throw thir support bhind McMullin as thir third-party candidat, it is likly that th abov tabl xaggrats th numbr of vots McMullin would hav rcivd from conflictd Clinton votrs. Bcaus this mthod probably ovrstats th support McMullin would hav gottn from conflictd Clinton votrs, I rxamind vot shars for third party candidats by using a diffrnt mthod to rapportion conflictd votrs. Instad of using th qustion Would you vot for a candidat othr than Hillary Clinton or Donald Trump if you thought thy had a chanc to gt th most vots in Utah? I usd th fling thrmomtr. This is similar to th mthodology usd by Abramson in his study. 54 I xamind th flings that all Clinton votrs had for McMullin and that all Trump votrs had for McMullin. If a votr had highr flings for McMullin than thir prfrrd candidat, I movd thir vot from Clinton or Trump into McMullin s total vot shar. Aftr doing so, I found that 1,242 Trump votrs prfrrd McMullin ovr Trump. This is highr than what w found using th proportional rapportionmnt mthod in Tabls 4, 5, and 6. Only 200 Clinton votrs prfrrd McMullin ovr Clinton, which is 428 lss than in th 31

39 proportional rapportionmnt mthod. This may b closr to th tru amount of Clinton votrs that would hav votd for McMullin if thy thought h had a chanc to win. Tabl 7 has th rsults aftr rapportioning th votrs according to th flings thrmomtr, pulling out th 1,242 Trump votrs and th 200 Clinton votrs to add to McMullin s vot shar. Votrs wr not askd about thir flings towards Johnson and Stin, which is a limitation of this analysis, as w cannot prdict how many conflictd votrs would hav votd for Johnson or Stin. Tabl 7- Votrs from Survy, Moving Trump and Clinton Votrs According to th Flings Thrmomtr Candidat Survy Vot Prcntags Donald Trump 24% Hillary Clinton 29% Gary Johnson 4% Evan McMullin 39% WINNER Somon ls 3% Don t.4% rmmbr Did not vot.6% 32

40 100% In th flings rapportionmnt mthod, McMullin pickd up much mor of Trump votrs than h did whn proportionally splitting th vots with othr third-party candidats. McMullin rcivd substantially lss support from Clinton votrs, which maks mor sns considring that Clinton votrs hav littl rason (othr than th fact McMullin is not Clinton) to support a third-party candidat who acts lik a traditional Rpublican. Thr ar drawbacks to this mthod though. First, all rapportionmnt dcisions ar mad using th flings thrmomtr. Th UCEP data that asks votrs whthr or not thy would hav votd for an lctabl third-party candidat is not usd in this mthod. Bcaus this mthod of rapportionmnt dos not us that data, th analysis of Duvrgr s Law is mor roundabout than if w usd th qustion about third-party lctability. Anothr drawback is that th votrs wr not askd about thir flings towards Gary Johnson and Jill Stin. Bcaus thy wr not askd, w cannot s what proportion of vots should b rapportiond to thos two candidats. In What Scnario Could McMullin Hav Actually Won Utah? Th two mthods of rapportionmnt, th proportional mthod and th fling thrmomtr mthod, both hav bnfits and drawbacks. Th drawbacks ar significant nough that it is unwis to us ithr mthod as dfinitiv proof that McMullin would hav won th stat of Utah if popl had votd for thir prfrnc. Howvr, it is within th ralm of rality that Evan McMullin potntially could hav won th stat of Utah if h had procurd nough vots from th conflictd votrs. In Tabl 1, I found that aftr 33

41 rmoving conflictd votrs from Trump and Clinton s vot shar but not rapportioning thm to spcific third-party candidats, Trump had 24% of th total vot and Clinton had 20%. McMullin alrady had 21% of th total vot. If vn a simpl majority of conflictd Trump votrs mov ovr to vot for McMullin, h would hav won th stat of Utah, rgardlss of which third-party candidat conflictd Clinton votrs support. Implications Ths rsults ar informativ. A significant numbr of thos who votd for Trump and Clinton did not want to do so. Thy would hav votd for a third-party candidat if thy had blivd a third-party candidat could hav won th stat of Utah. Rgardlss of th mthod usd, whn ths conflictd votrs ar rapportiond to third-party candidats Evan McMullin would hav won significantly mor vots if popl had blivd h was lctabl. H clarly was mor prfrrd by votrs than th actual vot count rvals. Th fact h faild to win rflcts th strngth of th psychological factor of Duvrgr s Law popl will not vot for third-party candidats if thy do not think thos candidats can actually win. Th xtrnal validity of this study is strong. Th candidats and vnts of th 2016 prsidntial lction, combind with th dmographic makup of Utah, wr a prfct storm of vnts that gav th UCEP this opportunity to viw how votrs thought about third partis in a situation whr a third-party could hav bn sn as a viabl altrnativ. This situation is not likly to b oftn rpatd. In lctions whr th candidats of th two main partis ar mor popular and th dmographic is diffrnt, third-party candidats may not b lookd on as favorably as Evan McMullin was in

42 This dos not wakn th xtrnal validity bcaus th lssons larnd about third-party candidats and Duvrgr s Law ar still applicabl vn though most votrs will nvr b this cross prssurd. Ths findings ar gnralizabl to othr situations. On of th bnfits to th fdralist systm in th Unitd Stats is that ach stat acts as a laboratory of dmocracy. 55 This is th situation in our study. In this cas, Utah placd Duvrgr s Law undr xtrm strss. Utahns traditionally vot Rpublican, strongly dislikd th Rpublican candidat, strongly dislikd th Dmocrat candidat, and had a fairly appaling candidat in McMullin. Evn in this situation, a lot of popl did not vot for thir tru prfrnc bcaus thy did not think a third-party candidat could actually win. This finding that Duvrgr s Law holds up undr xtrm strss is gnralizabl to othr stats, th Unitd Stats on a national lvl, and to othr countris. On lsson from ths findings ar changs that third-party candidats nd to mak in thir mssaging whil thy campaign. Th rsults mak it clar that popl knw who Evan McMullin was. H had a high rcognition lvl in th stat and was considrd favorably by a majority of th votrs. This nam rcognition and favorability was not nough to win though. Bcaus votrs did not bliv that McMullin had a chanc of winning, th powr of his nam rcognition and favorability was mutd. Whil McMullin did mak som attmpts to tll votrs that thir vot mattrd, that Utah had th powr to dny Trump and Clinton th ncssary 270 vots to scur th prsidncy, and that a Utah poll as lat as Novmbr 4 showd that h had a chanc to hang with Trump in th Utah lction, most of his branding was focusd on th fact that h was not Donald J. Trump. 56 That is ultimatly what stuck with votrs whn thy wnt to th polls. 35

43 Futur third-party candidats nd to mak sur that a substantial portion of thir mssaging and branding is focusd on th viability of thir candidacy. A third-party candidat will bnfit mor from convincing votrs that thir campaign can win than from chasing 100% nam rcognition with votrs. Th countris that hav notabl xcptions to Duvrgr s Law, Canada and India, do so bcaus thy hav strong local minority partis 57 If a third-party is to succssfully stablish itslf in th Unitd Stats, thy would hav to start on a local lvl and convinc th popl in on particular rgion that thy ar a viabl altrnativ and can win, rathr than focusing on th ntir country at onc. A scond lsson from this study is that Duvrgr s Law sms to work in part bcaus popl do not bliv that a third-party can actually win an lction. If thy did, thy would vot for a third-party candidat. This finding dos not com from a fancy rgrssion or a complicatd indx of votrs. Rathr, it coms from a simpl qustion. Would you vot for a candidat othr than Hillary Clinton or Donald Trump if you thought thy had a chanc to gt th most vots in Utah? This finding is validatd by th flings thrmomtr, showing that conflictd votrs viwd McMullin favorably nough to hav actually votd for him. Th sampl siz of th popl that answrd was larg nough that by th Law of Larg Numbrs w can hav a significant amount in th accuracy of th rsults. This study statistically validats Duvrgr s thory that his law workd bcaus of psychological aspct of th lctoral systm, not mrly bcaus of mchanical factors in th systm itslf. 36

44 Conclusion A ky rason that Duvrgr s Law stands is bcaus popl will vot for partis that hav a chanc at winning th lction. This is why in a first past th post systm th lctorat coalscs into two distinct partis, with som fring partis that do not ralistically compt. In th 2016 Utah lction, th votrs wr xtrmly cross prssurd bcaus thy wr trappd btwn two candidats with srious flaws Donald Trump and Hillary Clinton. In th Utah Collg Exit Poll, w found that mor popl would hav votd for McMullin if thy thought h had a chanc at winning th lction. Whil thr ar too many obstacls to dfinitivly say if h could hav actually won or not, th data is clar that a significantly largr numbr of popl would hav votd for him than actually did. This dmonstrats th rol that th psychological ffct plays in Duvrgr s Law stands bcaus in this xampl, vn though Evan McMullin potntially had th vots to win Utah, h lost bcaus th votrs wr convincd that a third-party candidat did not stand a chanc at winning. 37

45 Appndix A- Front Sid of Blu Form 38