University of Southern California Law School€¦ · Show Me the Money: “Dark Money” and the...

University of Southern California LawSchool

Legal Studies Working Paper Series

Year Paper

Show Me the Money: “Dark Money” and theInformational Benefit of Campaign Finance

Disclosure

Abby K. Wood∗

∗University of Southern California, [email protected] working paper is hosted by The Berkeley Electronic Press (bepress) and may not be commer-cially reproduced without the permission of the copyright holder.

http://law.bepress.com/usclwps-lss/254

Copyright c©2017 by the author.

Show Me the Money: “Dark Money” and theInformational Benefit of Campaign Finance

Disclosure

Abby K. Wood

Abstract

Campaign finance disclosure is under threat. While the Court continues to up-hold mandatory disclosure, it has also eviscerated much of the legal justificationfor it. Simultaneously, gaps in the legal framework mean that some campaign ac-tivity is subject only to voluntary disclosure – consider “dark money” groups andunregulated Internet campaign advertising. In upholding the parts of the campaignfinance regime that mandate disclosure, the Court has assumed that disclosure pro-vides valuable policy information to voters, but it has not considered non-policyinformation that voters learn about candidates from the choice to disclose morethan is legally required. This article provides new survey and experimental evi-dence that voters value disclosure of campaign finance information and will re-ward voluntary disclosure while punishing candidates supported by dark moneygroups. Voluntary disclosure signals transparency and thus trustworthiness. Theimportance of the second kind of information has not been previously recognizedand suggests a role for voluntary as well as mandatory disclosure.

Electronic copy available at: https://ssrn.com/abstract=3029095

1

SHOWMETHEMONEY:“DARKMONEY”ANDTHEINFORMATIONALBENEFITOFCAMPAIGNFINANCEDISCLOSURE*AbbyK.Wood†ABSTRACT

Campaign financedisclosure isunder threat.While theCourt continues toupholdmandatorydisclosure, it has also evisceratedmuchof the legal justification for it.Simultaneously, gaps in the legal frameworkmean that some campaign activity issubject only to voluntary disclosure – consider “dark money” groups andunregulatedInternetcampaignadvertising.Inupholdingthepartsofthecampaignfinance regime that mandate disclosure, the Court has assumed that disclosureprovidesvaluablepolicyinformationtovoters,butithasnotconsiderednon-policyinformation that voters learn about candidates from the choice to disclose morethanislegallyrequired.Thisarticleprovidesnewsurveyandexperimentalevidencethat voters value disclosure of campaign finance information and will rewardvoluntarydisclosurewhilepunishingcandidatessupportedbydarkmoneygroups.Voluntary disclosure signals transparency and thus trustworthiness. Theimportanceof the secondkindof informationhasnotbeenpreviously recognizedandsuggestsaroleforvoluntaryaswellasmandatorydisclosure.

* This work benefitted from helpful conversations with and feedback from Sam Erman, Mike Gilbert, Benjamin Graham, Christian Grose, Erin Hartman, Daniel Klerman, Morris Levy, Dan Simon, Doug Spencer, seminar participants at Pepperdine and USC, my fellow members of the Federal Bipartisan Campaign Finance Task Force, and members of the Google policy team. Dara O’Neill, Oliver Wu, Danielle Stierna, Mimi Sanicola, and Rachel Abanonu provided excellent research assistance. I gratefully acknowledge support from the Hewlett Foundation and the Democracy Fund. † Associate Professor of Law, Political Science, and Public Policy, University of Southern California.

Hosted by The Berkeley Electronic Press

Electronic copy available at: https://ssrn.com/abstract=3029095

2

1.INTRODUCTION Campaign financehas changed enormously in thepast decade.One of the biggestchanges has been the amount of political spending that happenswithout anydisclosure.Manypoliticalactorsandmanytimesofpoliticalmessagingremainanonymous,withtheirsourcesoffundingonlyrarelydisclosedvoluntarily.Mostoftheso-called“darkmoney”isspentbygroupswhoorganizeundersection501(c)of the taxcode(usually501(c)4and501(c)6). Since Citizens United v. FEC (558 U.S. 310 (2010)), these “social welfare”organizationshavebeenabletomakeunlimitedexpendituresinelections.Theycanreceiveunlimitedcontributionsfromindividuals,corporations,andunions,andtheydonothavetodisclose thesourcesof theirmoney.Theirpoliticalactivity issupposedtobe limited,buttheIRSdoeslittletoenforcelimitsonpoliticalactivitybysocialwelfareorganizations.Thegroupsbuytelevisionads,placeadsonline,andassistwithget-out-the-voteefforts. Inthe2016U.S.electioncycle,$183millionwasspentviadarkmoneyavenuesfortelevision and radio advertisements, up from $6 million in 2006, but down from $300million in2012(Center forResponsivePolitics2017).Presidentialracesattract themostdarkmoney, followedby Senate races, thenHouse races. In2016, non-disclosing groupsranapproximatelyone thirdofallads inHouseandSenateraces thataired twoormoremonthsbeforethegeneralelection.3 Online political advertising has exploded since 2008, but mandatory disclosurerequirementscoveronlyasmallfractionofonlinepoliticalads.Thedisclosuresforpoliticaladvertising include disclaimers, the “stand by your ad” requirements at the end of anadvertisement. They include the name and address of the group running the ad, or thecandidate saying “I’m so-and-so, and I approve thismessage.”We lack ameasure of theamountofonlineadvertisingwithoutdisclaimers,butitisbothvastandgrowing.Therisein online advertisements provides additional opportunities for voters to receiveanonymouscampaignmessages. Nocampaignfinancelawprohibitscandidatesorgroupsfromvoluntarilydisclosingthe sources of their support. Indeed, two well-known 501(c)4 groups do voluntarilydisclose. Do voters care about voluntary campaign finance disclosure? Transparency iswidely believed to be popular with voters. But what will voters do when faced with apolitically attractive, but non-transparent, candidate or campaign? Will they punish acandidatewhose supportinggroupsdonotdisclose the sourcesof their funds?Will theyrewardacandidatewhodisclosesmorethanthelawrequires? Researchers have largely ignored these questions. We have few studies aboutwhethervotersreacttolearningaboutdarkmoney,andresultsaremixed(Brown&Martin2015,Theodoridis&Spencer2017,Rhodesetal2017).Iarguethatvoluntarydisclosureofcampaign finance reveals valuable information to voters about candidates’ non-policycandidate traits.Even inavoluntarydisclosure regime,voters can learnabout candidatetransparencyfromthecandidate’schoicetodisclose.

I present results from a survey and experiment to address this argument. Theresultsdemonstrate thatcandidateandgroupchoicesaboutcampaign financedisclosureaffecthowpeoplevote.Respondentsobserve twocandidatesof thesameparty,onewho3DarkmoneyactivityismostcommoninthetimebeforetheFEC’smandatorydisclosurewindowsat30daysbeforeaprimaryand60-daysbeforeageneralelection.


3

voluntarily discloses more than is required by law and one who is supported by darkmoney groups. Respondents reward the more transparent candidate and punish thecandidateconnectedtodarkmoneygroups.Thesizeoftheeffectislargeenoughtoswinganelection–around15percentagepoints. Whydoesthismatter?Evenasithasstruckdownstatutesandregulationsaimedatlimiting campaign financing, the Supreme Court has long upheld mandatory disclosureregulations, partly on an assumption that the disclosed information improves votercompetencebyhelpingvoterstoknowaboutthekindsofpoliciesacandidatewillpursueonce in office. The results presented here enrich our understanding of the informationvoters glean around campaign finance disclosure: in addition to whatever policyinformation voters learn from disclosures, candidate and group decisions to voluntarilydisclose(ornot)informthemaboutcandidatetrustworthiness.

The jurisprudential landscape could soon change. The Court recently lost adisclosure advocate in Justice Scalia. His successor, Justice Gorsuch, ismore skeptical ofcampaign finance disclosure, as revealed in his confirmation hearing. The politicallandscapehasalsochanged.Thesamelawyerswhousedtourgethecourtstostrikedowncampaign finance limits and bans (e.g. bans on corporate independent expenditures inCitizens United) while upholding disclosure are now turning their sights on disclosure.Morecourtchallengeswillundoubtedlyfollow.

Candidatesandgroupscancapitalizeon the findingspresentedherebydisclosingvoluntarily.Candidateswhodisclosemorethanrequiredbylawcanalsocampaignagainstcandidateswhoaresupportedbydarkmoneygroupsbyhighlightingtheiropponent’slackoftransparencyandthuslackoftrustworthiness.Theresultscanalsohelppolicymakers,astheyattempttooptimizethedesignofcampaignfinancedisclosureregimes.Policymakersmusttradeofftwoeffects:providinginformationaboutpoliciesthecandidatemightpursueonce inoffice inaneffort topleasesupporters,andprovidingsignalsabout transparencyand trustworthiness of the candidate who chooses to voluntarily disclose. A completelyvoluntary regime (with no mandatory disclosures) would maximize information abouttransparency/trustworthiness,but itwould reduce informationaboutpolicy in theeventthat candidates and groups continue to take advantage of loopholes in the system thatallow fornondisclosure. Conversely, a regimewith comprehensivemandatorydisclosurewouldmaximize information about candidates’ future policies butminimize informationabout transparency and trustworthiness. An optimal campaign disclosure regimewouldstrikeabalancebetweentheseextremes. 2.BACKGROUND Disclosure is the last robust regulationofcampaign finance.Our federalcampaignfinance transparency regime has plenty of mandatory disclosure, but there are severalareasinwhichdisclosureisvoluntary,suchasdisclosureby501(c)groups.Priorresearchsuggests that voters can benefit from additional information about candidates or ballotinitiatives, but whether campaign finance disclosures themselves improve votercompetence has rarely been examined. Campaign finance disclosures can improve votercompetencebyenhancingthecredibilityofthecandidateoroutsidegroup’spromises.Therestofthissectionexplainsinmoredetail.


4

2.1DISCLOSURE:THELASTREGULATIONSTANDING Littlebylittle,thecourtshaveemptiedthecampaignfinanceregulatorytoolkit,byruling that campaign finance regulations unconstitutionally infringe on the FirstAmendment rights to free speechor free association.Most famously, the SupremeCourthas recently lifted aggregate contribution limits (McCutcheon v. FEC, 134 S. Ct. 1434(2014))andremovedbansoncorporateindependentexpenditures(CitizensUnitedv.FEC558U.S.310(2010)).Asaresult,regulationofmoneyinelectionscentersmoreandmoreheavily on disclosure (Levinson 2016; Shaw2016).While the existing disclosure regimehassignificantloopholes,theCourtseemsgenerallybullishondisclosure. Disclosure comes in several forms. Campaigns for federal office must gatherinformationaboutdonors(name,address,amountcontributed,andemployer).Ifadonor’saggregate contributions reach the $200 mandatory disclosure threshold, the donor’sinformationisreportedtotheFECinthecampaign’speriodicfilings.Moststatesfollowasimilar system, thoughmandatorydisclosure thresholds vary considerably.Expendituresare also subject to disclosure requirements. Disclaimers attached to “electioneeringcommunications” are another common type of disclosure. The “Stand by Your Ad”requirements,withthecandidatesaying,“I’mSo-and-So,andIapprovethismessage”arethemostfamiliarexample.Disclaimersarerequiredonmanykindsofpoliticalmessaging,likemailersandsome(butfarfromall)onlineads.Disclosureanddisclaimerrequirementsapplytopartiesandsomeoutsidegroups,likeSuperPACs,aswell. The Court analyzes campaign finance regulations under a First Amendmentframework. It prefers disclosure to monetary limits and bans for regulating campaignfinance, as disclosure regulations are a less restrictive means to meet importantgovernmentalinterests.Disclosurelaws“imposenoceilingoncampaignrelatedactivities”(Buckleyat64)and“donotpreventanyonefromspeaking”(McConnellat201).Indeed,theCourt followed the urging of deregulation proponents in adopting this “deregulate anddisclose”approach.However,aftertheirsuccessesinthecourts,deregulationproponentsnowarguethatdisclosingtheidentityofdonorsimpermissiblyburdensFirstAmendmentspeech(Lowensteinetal2012;see,e.g.Primo2011;ACLU2010).2.2JURISPRUDENTIALLANDSCAPE The Supreme Court has upheld disclosure laws three times in the past six years,thoughithasdonesowith“reasoningmorebroadthandeep.”(Shaw2016).Thecourtalsorecentlydeniedcertiorarionadisclosurechallenge.4Thecourt’srecent,briefrejectionsofdisclosure challenges are not necessarily a sign that the court will continue to embracedisclosure,especiallyinthisderegulatorypoliticalclimate.

4Del.StrongFamiliesv.Denn, 136 S. Ct. 2376 (2016). TheDelaware Strong Families case is a familiar factpattern,inwhichaperfectlysympatheticplaintiffwillbeburdenedbydisclosurerequirementsestablishedbythestateandsochallengestherequirements.Inthisparticularcase,anonprofitorganizationputoutavoterguidein2012andplannedtodosoagainin2014.Intheinterveningtime,inthewakeofthesurgeindarkmoneyenabledbyholes in the federaldisclosureregime,Delawareamended itsdisclosure laws torequiredisclosureofdonorsorpersonssupplying “electioneeringcommunications”,which include thevoterguide.DelawareStrongFamilieschallengedthelaw,becauseitdidnotwanttodisclosedonorinformation.


5

TheCourthaslongassumed,withoutconclusiveempiricalevidence,thatdisclosure“chills” political speech protected by the First Amendment.5Because of the (assumed)intrusion on a constitutionally protected right, the court uses an amorphous “exactingscrutiny” standard to analyze the constitutionality of disclosure regulations. Exactingscrutiny requires that governmentsdefending theirdisclosure lawsmust identify a stateinterest sufficient to overcome the burden. Depending on the case, the interestmust be“overriding” (McIntyre v. Ohio Elections Comm’n, 514 U.S. 334 (1995)), “sufficientlyimportant”(CitizensUnited),oreven“compelling”(McCutcheon).Finally,thecourtanalyzeswhether the regulation is “substantially related” (Buckley v. Valeo, 424 U.S. 1 (1976),CitizensUnited),oreven“narrowlytailored”(McIntyre)tothegovernmentinterest. The Buckley court listed three governmental interests that courts since that timehave adopted without alteration. They are: enforcement of campaign finance laws,deterringcorruptionortheappearanceofcorruption,andprovidinginformationtovoters. Enforcement. There are few substantive campaign finance regulations left. Theprimarybansthatremainarethebanoncorporateandunioncontributionsandthebanonforeign donors. Themost important limit that remains is the contribution limit to eachcandidateineachcycle.Practicallyeverythingelse–aggregatecontributionlimits,limitsonindependentexpenditures,bansoncorporateandunion independentexpenditures–hasbeen struck down under the First Amendment. The Buckley court explained thatinformation from disclosure can help aid enforcement of the limits that remain, butenforcementisnotanimportantdriverofdisclosurejurisprudence.Indeed,nodisclosurechallengehasturnedonthewaydisclosureaidsthegovernment inenforcingsubstantivelimits and bans. Andwhile FEC conducts enforcement actions based on the informationrevealedincampaignfinancedisclosures,FECenforcementisbothslowandrare(McGehee2015).

Corruption.Deterringcorruptionandtheappearanceofcorruption is important tovoters’ faith in democracy. Nevertheless, under Chief Justice Roberts, the Court hasnarrowed the definition of “corruption” such that the government’s interest in deterringcorruptionisonlyinpreventingquidproquobribery(CitizensUnitedat136,McCutcheonv.FEC,134S.Ct.1434at1435-36,1454-56).Quidproquobriberyrequiresatleasttwopeople–onetoofferthebribe,andonetoacceptit.Thenarrowingofthedefinitionof“corruption”thereforebringsintoquestionwhetherthestate’sinterestincombattingcorruptionoritsappearancecanbemetwhenthereisnocandidateinvolved.

Forexample, thecourthasruledthatthemoneygiveninsupportoroppositionofballotinitiativescannotcorrupt.Aballotinitiativeisnotacandidate,soaballot initiativecannot receive abribe. Similarly, outside groups are formally independentof campaigns,duetotheprohibitiononcoordinationbetweencampaignsandsupportingoutsidegroups.RelyingonthebanoncoordinationandsimilarlanguagefromBuckley(repeatedinCitizens5Whetherdisclosure,onnet,chillsorthawsspeechisanopen,empiricalquestion,andtheCourt’sassumptionhassofarbeentestedonlyrarely.See,e.g.,Wood&Spencer2016,LaRaja2014.InthelanguageofMichaelGilbert,disclosureprobablythawssomespeech.Somethawingresultsfromdisclosurerevealingcrosspressuresonthecandidatesandcandidates’credibilityinagivenpolicyarea(Gilbert2013).Otherthawingresultsfrompeoplewhoareeagerto“goonrecord”assupportingacandidate(Wood&Gilbert2016).


6

United), the D.C. Circuit has ruled that independent expenditures cannot corrupt(SpeechNow.orgv.FEC,389U.S.App.D.C.424,431(2010)).Disclosurehasnotbeenstruckdown ballot initiatives or independent expenditures, but if the court’s definition ofcorruption is now “quid pro quo” in the campaign finance context, a state attempting todefend its disclosure regime would do well to not rely solely on the state’s interest inreducingcorruptionoritsappearance. Information. With the enforcement and anticorruption interests unavailable,disclosureregulations“hangonthissinglethread”oftheinformationbenefit(Ortiz2012).Aninformedelectorateiscrucialtothehealthofarepresentativedemocracy.Whilevotercompetence– theabilityof voters to choose thepeopleandpolicies that represent theirvalues – is notmentioned explicitly in the constitution, scholars have argued that it is aconstitutional value (Garrett 1999), and that an “effective accountability” canon or“democracy” canon should be recognized in statutory interpretation (Elmendorf 2010;Hasen2009). The informational interest is, essentially, the government’s interest in ensuringvotersarecompetenttovoteinlinewiththeirinterests,thewaytheywouldiftheyhadfullinformation(Garrett1999;GerberandLupia1999).AccordingtotheBuckleymajority,theinformationdisclosedpursuanttocampaignfinancedisclosureregulations“allowsvotersto place each candidate in the political spectrum more precisely than is often possiblesolely on the basis of party labels and campaign speeches. The sources of a candidate'sfinancialsupportalsoalertthevotertotheintereststowhichacandidateismostlikelytobe responsive, and thus facilitatepredictionsof futureperformance inoffice” (Buckleyat67).Thecourt’sunderstandingisone-dimensional,left-to-rightpolicydimensiononwhichvotersplacecandidates. The disclosure holdings inMcIntyre and Citizens United engaged the informationrationale,thoughneitherdidsoveryextensively.InMcIntyre,theliberalwingofthecourtvoted to strike down an Ohio law requiring disclosure on pamphlets. Justice Stevens,writingforthemajorityandjoinedbyJusticesO’Connor,Souter,andBreyer,withJusticesGinsburg and Thomas concurring, ruled that disclaimers on pamphlets are not thatinformative to voters, and even if they were, the informational interest is “plainlyinsufficienttosupportthestatute’sdisclosurerequirement,sincethespeaker’sidentityisno different from other components of a documents’ contents that the author is free toincludeorexclude,andtheauthor’snameandaddressaddlittletothereader’sabilitytoevaluatethedocumentinthecaseofahandbillwrittenbyaprivatecitizenunknowntothereader” (McIntyre at 348-49). Justice Thomas concurred, saying that he would haveanalyzed the question of disclosure for a pamphleteer using only the originalunderstandingof“freedomofspeech,orofthepress”,whichhebelievedledtotheresultinthemajority opinionoverturning thedisclosuremandate. JusticeGinsburg’s concurrenceemphasized “in fora calf isnotalways in fora cow”and that inothercircumstances thestate may require a speaker to “disclose its interest by disclosing its identity.” Notably,conservative justicesScalia andRehnquist recognized the importanceofdisclosure toaninformed electorate. They challenged the idea that there has ever existed a “hithertounknown right-to-be-unknown”when engaging in political speech in a “free, democraticelection”.


7

The Citizens United majority emphasized that the information provided bydisclosure isevenmorepowerful in theageof the Internet, “becausemoderntechnologymakes disclosures rapid and informative.” According to themajority, “this transparencyenables the electorate to make informed decisions and give proper weight to differentspeakersandmessages”(CitizensUnitedat364).

All three Justices from theMcIntyreerawho remain on the court today voted tooverturn the statute inMcIntyre, but only Justice Thomas has continued to insist thatdisclosurelawsareanunconstitutionalinfringementontheFirstAmendmentrighttofreespeech (CitizensUnitedat480,McConnell at264).RecentCourt turnovermatters. JusticeScaliawasparticularlyfondofdisclosure,saying“requiringpeopletostandupinpublicfortheir political acts… fosters civic courage,withoutwhich democracy is doomed. Formypart, I do not look forward to a societywhich, thanks to the Supreme Court, campaignsanonymously…. [A country that allows anonymous campaigning] does not resemble theHomeof theBrave” (Doev.Reed, 561U.S. 186 (2010)) JusticeGorsuch, his replacement,recentlyjoinedJusticeThomasonadissentfromasummaryaffirmanceinRepublicanPartyof Louisiana v. FEC, a case challenging the ban on corporate contributions to parties,signaling his de-regulatory inclinations in the realm of campaign finance. During hisconfirmationhearing,JusticeGorsuchsignaledhiswillingnesstolookcloselyatdisclosurelaws,sayingthattheFirstAmendmentcontainscompetinginterests.HesaidthatBuckleystandsforthenotionthatCongresscanregulatemoneyinpolitics,andthatthere“maybelimitswhen it chills expression, as it did in theNAACP case, andwehave to beworriedaboutthat….”.6

6Theentireexchangewasfascinatingforastudentofcampaignfinancelaw.Itwasoneofthemorerevealingexchangesinthethree-dayhearing.Itcontinuedasfollows:Sen.Whitehouse:“Ifwehavetobeworriedaboutthechillingofexpression,whichisavaluepropositionthatyouhavejustenunciated,shouldwenot–amInotalsoentitledtoaskthequestionaboutwhetherweshouldbeworriedabouttheinfluenceofdarkmoneyessentiallycorruptingourpolitics?It’stakingalongtimetogettowhatIwouldthinkwouldbeasimpleanswer.”J.Gorsuch“I’msorry,Senator,butIdon’tthinkthisissimplestuffatall.Ithinkyou’vegotFirstAmendmentconcernsandprecedentsintheareathatwouldhavetobeconsidered.We’dhavetoseewhatlawcongressenacted.I’dthenwanttogothroughthefulljudicialprocess….”Whitehouse: “But you just asserted right here that the value of ‘not chilling speech’ is something thatweshouldconsider,right?”Gorsuch:“IsaidthattheSupremeCourtoftheUnitedStatesinNAACPrecognizedthattheFirstAmendmentprotectionsweallaspeople in thiscountryenjoy–canbechilledsometimes. It’saFirstAmendmentrightwe’retalkingabout,Senator.”Whitehouse:“Andwheredoesanonymity–lets’sayabilliondollarsinanonymousfundingintoourelections.Wheredoesthatfitin,intothevaluesthatyoubringtothis?”Gorsuch: “In the first instance senator, it’s for this body to legislate. Then itwould come to court, and therecordwouldbemade--”Whitehouse:“OfcourseCitizensUniteddidoverrulealawthatwehadmade,sothat’shardlyabeallandendall.”


8

2.3INCOMPLETETRANSPARENCY:GAPSINCAMPAIGNFINANCEDISCLOSUREREGIME Thecurrentdisclosure frameworkproduces incomplete transparencyofcampaignfinances.Indeed,ifCongress’sgoalwasfulldisclosure,theBuckleycourtdidawaywiththepossibilityof fulldisclosurefourdecadesagowhenit limitedtheapplicationofcampaignfinanceregulations– includingdisclosureregulations–topoliticalcommunicationsusingthe eight “magic words”.7Since that time, clever donors have found other ways to skirtdisclosurerules,tobecome“veiledpoliticalactors”(GarrettandSmith2005). Thegapsinthemandatorydisclosurelawsthatexistarenotpartofagranddesign.Few gapswerewritten into FECA or BCRA, for example. Instead, the FEC created someexemptions,technologicalchangecreatedopportunitiesforotheranonymouscampaigningthat congress and regulators have not kept upwith, and the Supreme Court lifted banswhich Congress and the FEChave not reacted to by creating disclosure requirements toapplytonewactors. Intentional “gaps” in thedisclosure framework includenot requiringdisclosureofdonordetailswhenacontributionisbelow$200.Criticsarguethatdisclosureofsuchsmallcontributionsdoesnotfurtherstateinterests(thoughtheyfocusoncorruption).8 Intentional gaps with unintended consequences include requiring campaigns oroutside groups only to include disclaimers on online ads if they are “publiccommunications”. “Public communications” on the Internet are only required to have adisclaimer attached if they are “placed for a fee on another person’s website” (11 CFR100.26).Uploadingads toYouTube,creatingmemes thatgoviral,nativeadvertising,andmany other activities online do not fall under the FEC’s definition of “publiccommunications”notbecauseoftheircontent–manyareexpressadvocacy–butbecauseoftheirtheywerenot“placedforafee”(Dykhne2017;Persily2015).Asmoreandmorecampaign advertisingmoves online,more andmore online ads are not required to havedisclaimers. Finally,theCourt,inconjunctionwithcongressandtheFEC,createdthebiggestgapin themandatorydisclosureregime: the lackofdisclosurerequirements for independentexpenditures by corporations, including nonprofit organizations. Independentexpendituresby thesegroupsare relativelynew,havingbeenallowedbyCitizensUnited.Before CitizensUnited, corporations and unions were banned frommaking independentexpenditures from their treasuries. Now they can, but they have no disclosurerequirements. It is unlikely that the current Congress would amend disclosure laws to

(JusticeGorsuchisreferringtoNAACPv.Alabama,357U.S.449(1958))7Buckleyv.Valeo,424U.S.1,fn.52(“Thisconstructionwouldrestricttheapplicationof§608(e)(1)tocommunicationscontainingexpresswordsofadvocacyofelectionordefeat,suchas"votefor,""elect,""support,""castyourballotfor,""SmithforCongress,""voteagainst,""defeat,""reject.’“)8WeknowfromexperimentalresearchbyKallaandBroockman(2014)thatlegislatorsdograntmoreaccesstodonors.Ifgrantingaccessrelatestotheappearanceofcorruption–andtheonlystudytotestthissofardidnotshowaneffect(SpencerandTheodoridis2017)–thenaccessforsmalldonorscanaffecttheappearanceofcorruption.Apubliclistofdonorshelpsusmonitorwhetherdonorsdogetmoreaccess,butwhetherthisisenoughtojustifylowdisclosurethresholdsisanopenquestion.


9

increasemandatorydisclosure,especiallyfortax-exemptorganizations.Indeed,thecurrentappropriations bill again prohibits the SEC from regulating “the disclosure of politicalcontributions, contributions to tax exempt organizations, or dues paid to tradeassociations”(P.L.115-31sec.635(May5,2017)).TheFECissimilarlyunlikelytoact.Itisnotoriouslydeadlockedonthedarkmoneyissue(Wilson2015). Anonymous corporate political advertising happens one of two ways. Donorsseekinganonymityhavefoundit inthepast fewelectionsbydonatingto“socialwelfare”organizations created under sections 501(c)4 and 501(c)6 of the tax code. The 501(c)organizationsusethedonationstomakeindependentexpendituresonbehalfoforagainstcandidatesorballotinitiatives–andthe501(c)organizationsarenotrequiredtodisclosetheir donors for anything but “electioneering communications”.9These social welfareorganizations also donatemoney to SuperPACs. The SuperPACs are required to disclosetheircontributorstotheFEC,butwhenthecontributorisa501(c)organization,thepublicisunabletofollowthemoneytoitsoriginalsource.Reformerssee501(c)disclosureasanobviousnextstepinregulatingcampaignfinance. Not all 501(c)4 groups take advantage of their right to anonymity.MoveOn.org, a504(C)4 group, discloses all contributions above $5,000 used for political messaging.Similarly, Bernie Sanders’ new socialwelfare group,OurRevolution, announced in 2016that itwoulddisclosedonorsaswell (Gaudiano2016),a thresholdcurrentlysetat$250.OurRevolutionalsolimitsannualaggregatecontributionsto$5000,“unlessapprovedbyamajorityvoteoftheboard.”10 A second way to avoid disclosure is to establish an anonymous LLC. Now thatcorporationscangiveunlimitedamounts toSuperPACs,giving toaSuperPACthroughananonymousLLCisanotherwaytodonatewithoutexposingthedonor’sidentity(Briffault2012). Anonymous LLCs could also hide illegal foreign involvement in our elections(Weintraub2017).Importantly,LLCsdonothavetobefullyanonymousforcontributionsfrom them to be uninformative to the average voter. Naming an LLC somethinguninformativewillachievemostoftheidentityprotectionsoughtbythoseseekingtokeeptheirpoliticalcontributionsprivate.ASuperPACdisclosingitsdonorscansimplydisclosethe uninformative name (which could be anything from someone’s intials to somethingvaguelypatriotic, like“AmericansforAmerica,LLC”).Thatkindofdisclosureprovidesnorealinformationtovotersabouttheissuesorcandidatesintheadvertisement,anditforcesthepublic(oroppositionresearchers)totrackdowntherealpeoplebehindtheLLC. As commonly happens with campaign finance reform, enterprising stategovernments have led the way, under the assumption that voters will use informationaboutdisclosure to formopinionsabout candidatesand issues.Forexample, the stateofMontana recently adopted a proposal by Heather Gerken and coauthors to requiredisclaimersaboutnon-disclosureoffundingsources.Montanaadoptedastatuterequiringgroups who do not disclose their donors to “clearly and conspicuously include in allcommunicationsadvocatingthesuccessordefeatofacandidate…thefollowingdisclaimer‘This communication is funded by anonymous sources. The voter should determine the

9Avoidingelectioneeringcommunicationsiseasy.Thirtydaysbeforeaprimary,or60beforeageneralelection,501(c)organizationsstopadvertisingactivitiesandinsteadswitchtoothermeansofsupport,likeget-out-the-votecampaigns(WesleyanMediaProjectandCenterforResponsivePolitics2016).10https://ourrevolution.com/donors/


10

veracityofitscontent.’”11(Gerkenetal2014;Mont.CodeAnn.§13-35-237(2016)).Somestates now regulate online adsmore comprehensively than theFEC. For example,Mainerequiresadisclaimerfora“communicationexpresslyadvocatingtheelectionordefeatofaclearlyidentifiedcandidatethrough…publiclyaccessiblesitesontheInternet“(Me.Rev.Stat.tit.21-A,§1014),andMinnesotarequiresthatonlinebanneradsandothersmalladsdo not require a disclaimer on their face as long as they link directly to a sitewith thedisclaimer(Minn.Stat.Ann.§211B.04).2.4PRIORRESEARCHONCAMPAIGNFINANCEDISCLOSUREANDVOTERCOMPETENCE Political scientists dispute how competent voters are (Zaller 1992; Lupia 1994).Votershavelongbeenconsidered“rationallyignorant”(Downs1957).Theyoperatebasedon heuristics, or informational shortcuts, like a candidate’s party identification (Lupia1994).

Inageneralelectionbetweencandidatesfromdifferentparties,partyidentificationshould be amore powerful heuristic than campaign finance disclosure. Ifmy own partyidentification is strongenough,nomatterwho funds the candidate frommyparty, Iwillprefermyparty’scandidatetotheotherparty’scandidate.Butnotallelectionsarebetweencandidates fromopposing parties. Consider party primary elections, nonpartisan judicialelections,nonpartisancityelections,ballotinitiatives,andelectionsinstatesdominatedbyonepartythathavetop-twoprimaries(likeCalifornia’sSenateelectioninNovember2016,between Kamala Harris and Loretta Sanchez, both Democrats). Voters cannot use partyidentificationasaheuristicinanyofthesekindsofelections.Withoutpartyidentificationasanavailableshortcut,voters looktootherheuristics, likeendorsementsandcampaignfinanceinformation.

Information is valuable only where it improves knowledge (Lupia & McCubbins1998).Campaignfinanceinformationwillonlybeusedbyvotersforwhomit isvaluable.Most voters arenot undecided – they know forwhomorwhat initiatives theywill vote.Only those who remain undecided stand to benefit from campaign finance information.That means that the effect sizes we observe from offering campaign finance disclosureinformationtoallvoters(andnotjustundecidedvoters)willnecessarilybelimited.Thisisespecially true because, among the subset of voters who seek out campaign financeinformation,manywillbesophisticatedvoterswhohavemadetheirchoiceandaremerelycurious about the campaigns (Zaller1992). Small effect sizesdonotminimize thepolicyimportanceofcampaignfinancedisclosures.

Few studies have attempted to quantify the size of the informational benefit.Knowingwhichofseveralballot initiatives issupportedbyindustrywasenoughtoboostvotercompetenceinLupia’sstudyofCaliforniainsurancereform(Lupia1994).AndwhileexperimentalworkhasexaminedwhethertheCourthasfallenoutofsyncwiththepublicby narrowing the legal definition of “corruption” (Spencer and Theodoridis 2017;11 The statute states: “If a political committee claims to be exempt from disclosing the name of a person making a contribution to the political committee, the committee shall clearly and conspicuously include in all communications advocating the success or defeat of a candidate, political party, or ballot issue through any broadcasting station, newspaper, magazine, outdoor advertising facility, direct mailing, poster, handbill, bumper sticker, internet website, or other form of general political advertising or issue advocacy the following disclaimer: ‘This communication is funded by anonymous sources. The voter should determine the veracity of its content.’” MONT. CODE ANN. § 13-35-237 (2016).


11

Robertsonetal2016;BrownandMartin2015),Iamonlyawareofoneexperimentontheinformation benefit. Primo’s 2013 experiment showed no additional boost in votercompetence from campaign finance information. Nevertheless, his study involved anextraordinarily information-rich environment in whichmost subjects did not access thedisclosure informationatall. It isnotsurprising thathe foundno impactofdisclosure insuchanenvironment.Afterthatproject,wewereunabletosaywhethertherereallyisnoinformational benefit of campaign finance disclosure orwhether the null effect resultedfrom the information-rich environment in the design.Here, I use the opposite approach,providingonlyoneortwopiecesofinformationtorespondents.12IfurthernotethatPrimowasfocusedonmeasuringpolicy-relevant learning,andI focus insteadonlearningaboutnon-policytraits.

2.5 PRIOR RESEARCH ON NON-POLICY CAMPAIGN ISSUES AND A THEORETICALEXTENSION Voters care about more than policy. They also want elected officials to becompetent, persuasive, and honest. The political science literature lumps non-policycandidate and campaign traits into an umbrella category called “valence” characteristics(Stone&Simas2010;Stokes1963;Bianco1994;McCurley&Mondak1995;Mondak1995).Valence traitsarenon-policy traits thatvotersseek tomaximize, regardlessof thepolicypreferences of who is in office (Ansolabehere & Jones 2011; Wittman 2005; Enelow &Hinch 1984). Stone and Simas describe “campaign valence” as the ability to raise fundsfromothers,abilitytofundone’scampaign,currentnamerecognitionindistrict,abilitytoattractattention,abilitytobepersuasiveinpublic,abilitytorunaprofessionalcampaign,and overall strength as a campaigner. “Character valence” includes personal integrity,abilitytoworkwellwithother leaders,abilitytofindsolutionstoproblems,competence,grasp of the issues, qualifications to hold public office, and overall strength as a publicservant(Stone&Simas2010). Scholars have not analyzed voter preference around candidate legal compliancegenerally, compliance with campaign transparency rules specifically, or voluntarytransparencyofanykind.Thisprojectisafirststepinaddressingthegapintheliterature. There are fewopportunities in campaigns to signal transparency and compliance.Campaignfinancedisclosureprovidesanopportunitytosignaltovotershowtransparentlythecandidatewillgovern,asignalthatisparticularlyimportantforachallengertobeableto send. If a campaign files disclosures on time, and if the disclosures do not reveal anyprohibited contributions or gaps in required information, voters receive a message ofcompliance,transparency,cleangovernance,andcompetence.Thislogiccanbeextendedtooutside groups as well. If outside groups supporting the candidate disclose theircontributors, even where the law permits donor anonymity, voters can infer that thecandidate’ssupportershavenothingtohide.Conversely,ifthecandidate’sdisclosuresareamess,eitherbecausetheyare filed late,containprohibitedcontributions,orhavegaps inrequiredinformation,voterscaninferthatthecandidateisincompetent,doesnotprioritizetransparency, or both. If supporting groups hide their donors – even where they do so

12Alldesignshavelimitations.IdiscussthelimitationsofmydesigninSection4.4.


12

lawfully – voters can infer that the supporters, and perhaps the candidate aswell, havesomethingtohide.

Votersshouldprefermoregovernmenttransparencytoless,soforanygivensetofproposedpolicies,theyshouldpreferacandidatewhosignalsthatshewillgovernopenlybybeingmore transparentduringher campaign.13Moreover, foragivensetofproposedpolicies, voters will prefer a candidate who they believe will comply with the law. Inaddition to rule of law and other higher-order considerations, voter concern withcompliance is pragmatic: legal compliance avoids scandal and enables governance toproceedmoresmoothly.

Itwouldnot requirea strongpreference for transparency itself or strongdistasteforscandaltoobservethisresult.Disclosureinformationlendscredibility(orundermines)candidate claims about policies, because it reveals whether the donor’s interests andcandidate’sstatedpolicypreferencesmatch(Gilbert2013).Adislikeforuncertaintyalongthe policy dimension alone could lead voters to abandon candidates who receive darkmoney support. If darkmoney groups support a candidate, voters cannot see all of theinterest groups towhom the candidatemaybe responsive onpolicymatters as a publicofficial, thus increasing uncertainty and undermining credibility. Therefore, voters couldabandon candidates who are exactly aligned with their political policies (according tostump speeches and disclosed contributors) but potentially responsive to undisclosedgroupswithwhom the voterdisagrees, in order to vote for a candidatewho is lesswellaligned with the voter’s policy preferences but a “known quantity” because all of hersupportersaredisclosed.Putsimply,votersmightpreferthelesspoliticallyaligned“surething”tothemore(seemingly)alignedwildcard. Credibilityofmandatorydisclosuresresultsfromfinancialpenaltiesforviolationsofthedisclosure requirements.Credibilityof voluntarydisclosures canbeestablishedviaathird-partyaudit.Audits,especiallywhenconductedbyatrustedorneutralparty,enhancedisclosurecredibility(HealyandPaleu2001).

Some observers, particularly those who have expertise in transparency, alreadyfollow disclosures closely in key races. For example, the Center for Responsive Politicsmakesdataformostfederalracesavailable,andtheyhavetoolslikean“anomalytracker”tohelpvoters, journalists,andevenoppositionresearchersfollowthemoney.Sometimesjournalists take deepdives that examinedonors to particular candidates or spending bygroups.Thesekindsofinvestigationsarehelpful,andtheycanuncoveraspectsofcampaignfinance that I examine here, such as support by “dark money” groups or voluntarydisclosurebeyondwhat the lawrequires.However,onlymoreextensiveaudits, inwhichauditors examine the campaign’s books, bank statements, and similar documentation,would be able to confirm that what the candidate voluntarily discloses constitutes thecompletesetofpotentialexpendituresandcontributions.

13 Obviously, transparency can impede effective governance in some settings, like in the case of national security. I do not argue here that voters prefer transparency in all policy settings, but that they do prefer transparency in most policy settings. If governing transparently takes time away from policy tasks, they constituents might not want their public officials maximizing transparency, but instead optimizing it subject to resource constraints.


13

2.6SUMMARYANDEXPECTATIONS Insum,campaignfinancedisclosureisoneofthefewcampaignfinanceregulationstosurviverepeatedFirstAmendmentchallenges,butfuturechallengestodisclosureareonthehorizon.Courtshaveuphelddisclosure lawsbasedon the informational interest andthenow-narrowedanticorruptioninterest.Theinformationalinterest,asidentifiedbytheBuckleycourt, concerns the credibilityof candidate claimsaroundpolicies.But given thenumberofexemptionsandloopholesinourcurrentmandatoryframework,itisimportanttounderstandwhethervoterslearnfromvoluntarydisclosureaswell. Voluntary disclosures should cause voters to update their valence evaluations ofcandidates and campaigns. Existing studies on valence effects do not includeconsiderations of transparency and legal compliance, but our theories about voterpreferences and valence predict that voters should reward candidates who are moretransparentandwhomodellegalcompliance.Thesizeofanypunishmentsorrewardswillprobablybesmall,becausesofewvotersactuallyneedcampaignfinanceinformation.

Allofthisleadstothreeexpectations:1. Voterswillrewardover-disclosurewhereathird-partygroupidentifiesit.2. Voterswillpunishalackoftransparencywhereathird-partygroupidentifiesit.3. Voterperceptionsofcandidatetrustworthiness(credibility)areonemechanism

ofrewardsandpunishments.

To my knowledge, no one has asked voters whether they use campaign financedisclosuresandhowimportanttheythinkcampaignfinancetransparencyis.Beforetestingthis theoryand inorder to temperourexpectationsabouteffect sizes, I surveyedvotersabouttheiruseofcampaignfinanceinformation.3.VOTERDEMANDFORCAMPAIGNFINANCEINFORMATION:SURVEYRESULTS Someobserverssuspectthatcampaignfinanceinformationisnotofinteresttothepublic.Itistoocomplicated,tooarcane,ornotinformativeenoughtobeworththepublic’stime (Primo 2011; Eber 2006; Overman 2002; Smith 1998; cf. Briffault 2010). Politicaljournalists work hard to distill the information for public consumption, but is anyonepayingattention?

Asafirstcutatthesequestions,Iconductedanonlinesurveyconcerningcampaignfinancedisclosure.Thesurveyresultssuggestthatasignificantportionofthepublicbothdemands and uses campaign finance information. They provide a foundation for theexperimentalresultsthatfollow.3.1SURVEYDESIGNANDPARTICIPANTS Every fall for the past decade, political scientists have combined resources toconduct the Collaborative Congressional Research Survey. In November 2015, theBipartisanCampaignFinanceTaskForce(ofwhichIamamember)surveyed2000onlinerespondents, seeking their views about campaign finance.14The survey firm, YouGov,

14 The task force’s participation in the survey was funded by the Hewlett Foundation and Democracy Fund.


14

drawsstratifiedsamplesthatapproximatethecharacteristicsofrandomsamplesoftheU.S.population. The sampling frame is designed to match the population in the AmericanCommunitySurveyconductedbytheU.S.Censusandhasbeenaugmentedwithvoterandconsumerdatabases. The survey was conducted at a moment when campaign finance disclosureinformation was relatively low-salience in our political discourse. It was an off-year forfederalcongressionalelections.ThePresidentialprimarieswereunderway,butmostofthepublicdoesnotfollowthePresidentialraceuntilaftertheconventions,whichwerestill8-9monthsaway.Therefore,responsestothesurveyprobablyunderestimatethedemandforcampaignfinanceinformation.

3.2SURVEYQUESTIONSANDRESULTS Respondents answered questions about how often they seek campaign financeinformationfromcampaignsandoutsidegroups,howimportanttheinformationistothemcompared to other non-policy traits, andwhether they are satisfiedwith the amount ofinformationavailabletothem.TheresultsarepresentedinFigures1-4. Figures1and2presenttheresultsfromrelatedquestions:“Howoftendoyouseekinformationaboutcandidate funding”and“Howoftendoyouseekout informationaboutindependentgroupsthatfundpoliticaladvertising?”Respondentswereprovidedaslidingscale where the left endwas “Never” and the right endwas “Always”. Each respondentclicked a spot on the scale, and a number between 0 (never) and 100 (always) wasrecordedasherresponse. TheFigurespresenthistogramsshowingthefrequencyofeachresponseamongthe2000respondents.Thedegreetowhichthepublicreliesonthedisclosedinformationwasunknownuntilnow.AsFigures1and2show,around75%ofrespondentsreportseekingcampaign finance disclosure information at least some of the time. Over 25% ofrespondentsneverseekitout.

Figure 1. Frequency of responses to the question “How often do you seek informationaboutcandidatefunding”,0=Never,100=Always.Dashedlineismedianresponse.Solidlineismeanresponse.Source:CooperativeCongressionalElectionStudy(“CCES”)2015.

Responses

Freq

uenc

y

0 20 40 60 80 100

010

030

050

0

Mean (27.7)Median (24)


15

Figure2.Frequencyofresponsestothequestion“Howoftendoyouseekoutinformationabout independent groups that fund political advertising?” 0 = Never; 100 = Always.Dashedlineismedianresponse.Solidlineismeanresponse.Source:CCES2015. Ofcourse,respondentsmightseekoutcampaignfinancedisclosure informationtolearn about candidate policy responsiveness to contributors on the single dimensionenvisioned by the court in Buckley. But how important is the non-policy informationconveyedbycampaignfinancedisclosures?Iconstructedaquestiontoincludesomeofthecharacteristics consideredbyother scholars, and I added campaign financedisclosure tothelist. Thequestionaskedrespondents“Whenyouthinkaboutthestrengthofacandidateforelectedoffice,howimportanttoyouarethefollowingconsiderations?”Thesurveythenpresented the considerations below one at a time, in random order. The new item is inboldface.

• Theamountofmoneyraisedbythecampaign• The amount of money raised by independent groups in support of the

candidate• The amount of information the candidate provides about his or her

sourcesoffunding• Theabilityofthecandidatetobepersuasiveinpublic• The candidate’s professional background (government experience, private

sectorexperience,etc.)• Thecandidate’sgraspoftheissues

Respondentsansweredthateachconsiderationwas“Unimportant”,“OfLittleImportance”,“ModeratelyImportant”,“Important”,or“VeryImportant”.Respondentscouldalsoanswer“Idon'tknow”.Figure3shows,foreachquestion,whatpercentchose“VeryImportant”or“Important” (dark gray), “Moderately Important” (gray), or “Of Little Importance /Unimportant”(lightgray)foreachitem.Eachcolumnaddsupto100%ofrespondents. As is clear in Figure 3, voters place a good deal of importance on the amount ofinformationprovidedabouteachcandidate’s sourcesof funding.TheFigurepresents theconsiderationstherespondentsconsideredleastimportanttomost.Fortythreepercentof

Responses

Freq

uenc

y

0 20 40 60 80 100

010

030

050

0Mean (28.9)Median (24)


16

respondents chose “important” or “very important” for the amount of money raised bycampaigns, and 44 percent answered the same for money raised by outside groups.Strikingly, respondents considered candidate transparency about campaign finance asmoreimportantthantheamountofmoneyraised,with66percentofrespondentssayingtheamountof informationdisclosed is importantorvery important,andonly14percentsayingthatcampaignfinanceinformationisunimportantoroflittleimportance. Some of the usual non-policy factors were considered more important thancampaign finance transparency. The candidate’s ability to be persuasive in public wasconsidered important or very important by 72%, a candidate’s professional backgroundwas considered important or very important by 79%, and the candidate’s grasp of theissueswasconsideredimportantorveryimportantby91%.

Figure 3. Percent of respondents who responded “Important” or “Very Important”;“ModeratelyImportant”;or“Unimportant”or“OfLittleImportance”tothequestion“Whenyouthinkaboutthestrengthofacandidateforelectedoffice,howimportanttoyouarethefollowingconsiderations?”SourceCCES2015. Presentingconsiderationsoneatatime,asIdidwiththequestion,risksinflatingtheimportance of each item. But the bias should inflate all items equally, preserving therelativeorderingof items.Ofthe2000respondents,710(35.5%)ratedcampaignfinancedisclosureasmoreimportantthanallotheritemsranked,orofequalimportancewithallother items ranked.15The takeaway is that voters care about the amount of informationdisclosedbycandidatesconsiderablymorethantheycareabouttheamountofmoneythecampaigns or outside groups raise. Indeed, they value campaign finance transparencyalmostasmuchasthecandidate’sabilitytobepersuasiveinpublic.

15 Raw numbers are interesting here: 566 respondents rated disclosure “very important”; 74 rated disclosure as “important” and nothing else higher; 58 rated disclosure as “moderately important” and nothing else higher; 5 rated disclosure as “of little importance” and nothing else higher; 7 rated everything as “unimportant”.

0.00

0.25

0.50

0.75

1.00

camp.

money

group

mon

ey

disclo

sure

persu

asiven

ess

prof b

ackgro

und

grasp

of iss

ues

Perc

ent

responseImpt / Very Impt

Moderately Impt

Unimpt / Little Imptce


17

Notonlydoescampaignfinancetransparencymatter,butthepublicwantsaccesstomorecampaignfinanceinformation. Iaskedrespondents“Howsatisfiedareyouwiththeamount of information that you can find about who funds political advertising?” Thequestion concerns advertising because the vast amount of spending is on politicalmessaging, usually television ads. Figure 4 displays the frequency of responses to thequestion.Respondentsagainclickedonascale,whichcanbeinterpretedastheirpercentsatisfaction.“Notatallsatisfied”iscodedas0,and“completelysatisfied”is100.

Figure4.Frequencyofresponsestothequestion,“Howsatisfiedareyouwiththeamountof information that you can find about who funds political advertising?” 0 = Not at allsatisfied,100=completelysatisfied.Source:2015CCES.As Figure 4makes clear, some respondents are completely dissatisfied, around25%areneitherdissatisfiednorsatisfied,andafewarecompletelysatisfied. Thesurveyresultsimplythatcampaignfinancedisclosuresareimportanttovoters.Most voters report seeking campaign finance information at least sometimes, and theyreport seeking it from campaigns and outside groups. They consider the amount ofinformation disclosed to be more important than the amount raised, and almost asimportantas theabilityof thecandidate tobepersuasive inpublic. Inotherwords, theygleannon-policyinformationfromdisclosures.Finally,votersarenotentirelysatisfiedwiththeamountofinformationavailableaboutcampaignfinance.ScholarsandtheCourthaveoverlooked theweight voters give to disclosures independent of any policy informationthatdisclosuresmightreveal.4.EXPERIMENT Aseparategroupof1000respondentsfromthesamesurveypopulationcompletedasurveyexperimenttoestimatehowmuchpeoplerewardandpunishtransparencyorlackthereofincampaignfinancedisclosures.

Responses

Freq

uenc

y

0 20 40 60 80 100

010

020

030

040

0

Mean (41.1)Median (48)


18

4.1STIMULUSANDINSTRUMENT Respondentsreadavignetteaboutanelection,andatranscriptofanattackad.Thecontext is a primary election for an open seat for the state senate.16The candidates aresimilar in their favored policies. This context amplifies the valence effects of campaignfinance transparency. Policy differences between candidates are minimized where thecandidatesarefromthesameparty,sotransparencyevaluationsshouldbemostsalientinprimary elections. Incumbents have higher valence evaluations than challengers(Groseclose2001).Toneutralize the incumbentvalenceadvantage, thevignette is foranopenseat. Hereisthefirstpartofthevignette:

The words and phrases in brackets are populated automatically, based on respondents’answerstopriorquestionsinthesurveyabouttheirpartyaffiliationandideology.Botharein the respondent’sparty (true independentswere randomlyassignedaparty).Whereastherespondentranksherselfona1-10scale,thecandidatesarepresentedona1-100scale.The vignette presents Johnson as exactly the respondent’s level of liberalism orconservatism(times10,toscaleappropriately).

ThegroupsobservedConleyaseither10or20pointsmoreextreme inherviewsthanJohnson,unlesstherespondentisextreme(a9or10),inwhichcaseConleyis10or20pointsmoremoderate.Asanexample,considerarespondentwhoidentifiesasamoderateRepublican,givinghimselfa“5”outof10onhowconservativeheis,whoisassignedtoagroup in which Conley is 20 points more conservative than Johnson. Our hypotheticalrespondentwouldseethetextthisway:

Therespondentswhowererandomlyassignedtothetransparencytreatmentthenreadonthesamescreenabouttransparencyevaluationsofthecandidates.16 Nebraska has a unicameral legislature, and so respondents in Nebraska saw “Legislature” where respondents in other states saw “Senate”.

Julia Johnson and Pearl Conley are competing in the [PARTY] primary for an open seat in the state Senate. A local political scientist working for the [PARTY] party has evaluated the candidates on their policies, with 1 being not at all [liberal if PARTY = Democrat/conservative if PARTY = Republican] and 100 being very [liberal/conservative, as above]. The political scientist scored Johnson a [SCORE] and Conley a [SCORE].

Julia Johnson and Pearl Conley are competing in the Republican primary for an open seat in the state Senate. A local political scientist working for the Republican Party has evaluated the candidates on their policies, with 1 being not at all conservative and 100 being very conservative. The political scientist scored Johnson a 50 and Conley a 70.


19

After reading the vignette, respondents answered a series of questions onfavorability,votechoice,andtrustworthinessthatarecommonintheliterature(Gerberetal2013;Settleetal2016;Wrightetal2016).First, theyanswer“[H]owfavorablywouldyouratethecandidates?Pleasedraganddropthemontothescalebelow.”Thescaleisfrom“Highly unfavorable” to “Highly favorable”, and the order of the candidate names ispresentedinrandomorder. Respondents thenanswer “[h]ow likelywouldyoube tovote for”eachcandidate,again,presentedinrandomorder.Thescaleisfrom“Highlyunlikely”to“Highlylikely”.Thescreenalsofeaturesatableremindingtherespondentoftherelevantinformationfromtheprior screen. The control group’s table has one row, for policy ratings. The treatmentgroup’s table has both thepolicy rating rowand a transparency grade row.A treatmentgrouptableforourhypotheticalmoderateRepublicanisshowninTable1.

Johnson ConleyPolicy 50 70Transparency C A

Table1:TreatmentgrouptableforhypotheticalmoderateRepublican. Finally, respondents are asked how trustworthy the two candidates are.Respondents are asked, in random order “How trustworthy is [Candidate]?” The scaleprovidesrangesfrom“ExtremelyUntrustworthy”to“ExtremelyTrustworthy”.

A. TreatmentConditions Each of the 1000 respondents was randomly assigned to one of four treatmentconditions: a control group with Conley 10 points from Johnson, and Johnson at therespondent’sidealpoint(“C10”);acontrolgrouplikeC10,butwithConley20pointsfromJohnson (“C20”); a treatment group that saw the policy scores in C10 as well astransparencyscores (“T10”);anda treatmentgroup that saw thepolicyscores inC20aswellasthetransparencyscores(“T20”).Thetransparencyscoresdonotchangebetweenconditions. Johnson always receives a “C”, and Conley always receives an “A”. TheexperimentisdesignedtoallowexaminationofwhethertransparencyallowsConleytobemore extreme in her policy positions. Answers from respondents in C10/T10 werepractically identical to answers from respondents in C20/T20, so to ease exposition, Ipresent results from C20 and T20 in the main text. Results from C10/T10 are in the

A group of nonpartisan transparency advocates rates candidates based on how much of their campaign finance information is disclosed to the public. They call their ratings “transparency grades”. A grade of “A” means high transparency, and a grade of “F” means no transparency. Johnson received a transparency grade of “C” and Conley received a transparency grade of “A”. Johnson received a low grade because her campaign has received support from independent groups that allow their donors to remain anonymous. Conley received a high grade because both she and the independent groups supporting her candidacy disclose more campaign finance information than the law requires.


20

appendix. Randomization on characteristics that might predict reactions to policy ortransparencydifferencessucceeded.Randomizationresultsareavailableintheappendix. Table2summarizestheexperimentaldesignandexpectations.Therightmostcolumformalizes my expectations. “J” is “Johnson”, “C” is Conley, “T” is treatment, and “C” iscontrol.Theoutcomesexaminedineachcellofthecolumnareaveragefavorability[0-100],averageprobablevoteforeachcandidate[0-100],andaveragetrustworthinessevaluations[0-100].

Condition JohnsonPolicy

JohnsonTransp.

ConleyPolicy

ConleyTransp.

N.Obs

Expectedeffect

C10 Same asRespondent

NA J+10 NA 248 J-C>0

T10 Same asRespondent

“C” J+10 “A” 248 JT-JC<0CT-CC>0

C20 Same asRespondent

NA J+20 NA 258 J-C>0

T20 Same asRespondent

“C” J+20 “A” 246 JT-JC<0CT-CC>0

Table 2. Experimental design. N.Obs is number of observations, or group size. Eachparticipanthada25%chanceofbeingassignedintoeachcondition,andgroupsizeswereallowedtovary.Therightmostcolumnpresentsthemainexpectationsforfavorability,votechoice,andtrustworthiness. Theexperimentaldesignallowsustolearnwhethervoterscareabouttransparencyandtowhatdegreetheyarewillingtotradepolicyforit.Inotherwords,asdesigned,theexperiment should give us insight into the shape of respondent indifference curvesbetween policy and transparency. Political scientists assume that voters weigh policyconsiderations much more strongly than valence considerations (Page & Jones 1979;Krosnick1988).Asithappens,netfavorabilitybetweenJohnsonandConleyinthecontrolgroup was basically constant regardless of whether Conley was 10 or 20 points fromJohnson(and,therefore,therespondent).Thetypicalrespondentpreferredthecandidatesequallyorslightlypreferredthecandidatewhowasmoreextremethanthey,themselves,professedtobe.17 Because thepolicydistancemanipulationdidnothave its intendedeffect and thegroupsarethereforebasicallyexchangeable, IcouldarguablypoolC10andC20,T10andT20, data and analyze them together. Doing so would have the effect of making theestimatesmore precise. Even though circumstances prevented timely pre-registration oftheresearchdesign,18Ididnotplantheexperimentwithpooleddata,andintheinterestofmethodologicaltransparencyandrigor,Idonotpoolthedataformyanalysis.

17Respondents’failuretorecognizethedifferencebetweenConleybeing10pointsmoreextremeandherbeing20pointsmoreextremepresentsapuzzlethatislargelybeyondthescopeofthispaper,exceptthatitprecludesusfromlearningabouttheshapeofrespondentindifferencecurvesbetweenpolicyandtransparency.Theresultsinformusaboutthedegreetowhichvotersdemandcampaignfinancetransparency,andfutureworkwillinvestigatethepolicy/transparencytradeoff.18Myfamilyexpandedamonthbeforeexpected!


21

4.2RESULTS Themainexperimentalresultsarepresented inFigure5.Thetoptworowsof theFigureshowrespondents’evaluationsofJohnson.Themiddletworowsshowrespondents’evaluationsofConley.Thedifferenceinmeansbetweenthetreatmentandcontrolgroupsis represented by an open circle. The 95% confidence interval on the estimate isrepresentedbyablackline.Thet-testp-valueisreportedaboveeachestimate.Allresultsarestatisticallysignificant.

A. VotersPunishLackofTransparency We expect that Johnson will suffer favorability and vote choice losses in thetreatmentgroup, compared to the control group. Inotherwords, voterspunisha lackoftransparency,evenwhenthelackoftransparencydoesnotviolatethelaw,andevenwhenitisnotthecandidate’sfault.TheexperimentalresultsareinthedirectionoftheexpectedeffectsinTable2. The top two rows in Figure 5 indicate that respondents punished Johnson.Compared to respondents who were only provided Johnson’s policy information,respondentstoldthatJohnson’ssupportive“independentgroups”chosedonoranonymitypunishedJohnson.Theyratedher9.4percentagepointslowerintermsoffavorabilityandexpressedlesswillingnesstovoteforherby9.2percentagepoints. Figure5presentsdifferences,buttheunderlyinglevelsareinterestingaswell.ThetreatmentgroupgaveJohnsonafavorabilityratingof45.9,andthecontrolgroupratedher55.3, which results in the statistically significant difference of 9.4 percentage pointsreportedintheFigure.Theresultsforvotechoiceweresimilar.Respondentswhosawthetransparencyscoresreported45.6percent likelihoodofvoting forher,comparedto54.8percentlikelihoodofvotingforheramongcontrolgrouprespondents.


22

Figure 5. Differences in the mean favorability and probable vote choice among thetreatmentgroup,minusthemeanfavorabilityandprobablevotechoiceamongthecontrolgroup, for Johnson (top two rows) and Conley (middle two rows). The net effect ispresentedinthebottomtworows.Theverticallineisat0,wheretheestimateswouldbeifrespondentswerenotaffectedbyseeinganevaluationofthecandidates’campaignfinanceinformation. The horizontal axismeasures percentage points. The open circles show thedifference inmeans,and thep-values reportedare fora two-sided t-test.Theblack linesshowa95%confidenceinterval.Allfindingsarestatisticallysignificant. Without real-world referents for transparency scores, I tread carefully ininterpretationoftheresults.Therespondentsinthesurveyareaverycloseapproximationtothevoting-ageAmericanpublic.Theirreactiontolearningmoreaboutcampaignfinancedisclosurepracticesofthecampaignsandsupportinggroupssuggestthattheycareaboutandreacttoinformationaboutcampaignfinancetransparency. The law prohibits coordination between independent groups and the candidatestheysupportandassumesthatcorruptionbetweenanoutsidegroupandcandidatecannothappen. However, respondents in this study reacted similarly to respondents in relatedstudies with regard to coordination (Brown & Martin 2015; Spencer 2017). Therespondentsheremayhaveassumedcoordinationbetween thegroupand thecandidate.Another possibility is that respondents assumed that “clean” candidates attract “clean”money, and “dirty” candidates attract “dirty”money, coordination aside. Future researchcanhelpclarifytheassumptionsheldbyrespondentsinthesesituations.

−20

−10 0 10 20

●

●

●

●

●

●Johnson − Favorability (T−C)

Johnson − Probable Vote (T−C)

Conley − Favorability (T−C)

Conley − Probable Vote (T−C)

Net Favorability (J−C)

Net Vote (J−C)

0.0000

0.0000

0.02

0.03

0.0000

0.0000

P ValueOutcome


23

B. VotersRewardTransparency Mytheoryalsopredictsthatvotersrewardtransparency.AsthesecondtworowsinFigure 5 show, voters rewarded Conley’s transparency. They rated Conley 55.3 forfavorabilitywhenonlypresentedpolicyinformation,and59.9whenpresentedpolicyandtransparency information, a difference of 4.6. For probable vote choice, voters said theywere58.4%likelytovoteforConleywhenpresentedonlypolicy information,and62.3%likelywhenpresentedpolicyandtransparencyinformation,adifferenceof3.9,asshowninthefourthrowofthefigure.Thedifferencesarestatisticallyandsubstantivelysignificant;votersrewardtransparency. ThebottomtworowsofFigure5showtheneteffectsandcombinethefavorabilityandvotechoice losses to Johnsonwith the favorabilityandvotechoicegains toConley. Itake the net favorability (Johnson – Conley) in the treatment group and subtract netfavorability(Johnson-Conley)inthecontrolgroup.Netfavorability inthecontrolgroup,whereonlycandidatepoliciesarecompared,is-0.43.19Forthetreatmentgroup,whichalsoobserved differences in disclosure practices, net favorability is -14.4, the result of apunishment of Johnson and a “reward” for Conley. The difference in the net favorabilitymeasuresis-14,asshowninthefifthrowofFigure6.RelativetoConley,Johnsonsuffersanetfavorabilitylossof14percentagepoints.Doingasimilaranalysisforthevotelikelihoodyieldsnetvotechangesof-13.3,asshowninthesixthrow. In the vignette, Conley and her supporting groups are not only complying withdisclosure laws, theyareovercomplyingwithdisclosure laws.20Inthe interestofkeepingtreatment and control vignettes approximately the same length, the type of overcompliancebyConleyandsupportinggroupswasleftvagueinthetreatmentvignette.Asdiscussed above, it is important to think about useful over-compliance versus over-compliancethatwouldallowacandidatetotakecreditforover-complyingbutnottransmitinformationthatwouldimprovevotercompetence.Increaseddisclosureoftheentitiesandpeople contributing to outside groups would be socially useful. Over-compliance thatprovidesactualnamesoftheactualpeoplebehindthemoneyspentforpoliticalmessagingcould improve voter competence. Finally, over-disclosure is a relative concept.Whetheroverdisclosureisusefuldepends,ofcourse,onwhatthelawrequires.Noneoftheseissueswith over-disclosure are captured in the vignette here, and all are avenues for futureresearch.

C. Interpretingthemaineffects How should we think about the scale of these effects? A comparison to negativefavorability effects in the experimental context is instructive. The largest campaignexpenditure,atleastforstateornationaloffice,isadvertising.Studiesofnegativeadshaveshown favorability differences, which vary based on the source of the ad (Dowling &Wichowsky 2015; Ridout et al 2014). Candidates who run negative ads generally

19 The median difference between candidates in the control group was 0, as expected. 20 This raises the possibility of observational equivalence – voters might not be rewarding transparency per se; they might be rewarding over compliance with the law. I asked respondents to fill in a text box about their vote choice. Many of them mentioned transparency, and I did not see over compliance mentioned, so I do not think that voters are rewarding over compliance more generally. Nevertheless, future work can further tease apart these two mechanisms.


24

experience backlash for doing so, with recent scholarship estimating the backlash of acandidate-run negative ad as causing, on average, = a 20 percentage point drop incandidatefavorability(Dowling&Wichowsky2015).21Thatstudymodifiedanactualvideoad from a prior state election to alter the voiceover and disclaimers, but it’s the closestanalogy in the literature to thisproject. For Johnson, being supportedbyoutside groupswhodonotdisclosetheirdonorsresultedinafavorabilitydropofabouthalfthemagnitudeofrunninganegativead. How should we think of the positive favorability boost for Conley? Dowling andWichowsky’sstudyagainprovidesausefulcomparison.Conley’s4percentagepointboostisaboutthreetimessmallerthanthebenefittoacandidateoflettingoutsidegroupsdothe“dirtywork”ofrunningnegativeadsfoundinthatstudy(13percentagepoints). VotersrewardConleyfortransparency lessthantheypunishJohnsonfora lackoftransparency,thoughthatdifferenceisnotstatisticallysignificant.Futureresearcherscanexplorethisbehavioralreactionmoredeeply.Generallyspeaking,respondents’eagernessto punish more than reward is not particularly surprising. Behavioral economists andpsychologists have previously shown that there is an asymmetry between reward andpunishment (Eisenberger et al 2004; Fehr & Gachter 2000, c.f. Wang & Leung 2010).Nevertheless, thesizeof thepunishmentof Johnson is striking– respondentspunishherquiteseverelyfortheperfectlylegalandcommonplacedisclosurechoicesofasupportiveindependentgroup.

D. TransparencyCreatesTrust;LackofTransparencyCreatesDistrust Howcanweexplaintherespondents’willingnesstoswitchtheirvotesfromJohnsonto Conley once transparency information was revealed? One mechanism is trust.Transparencyengendersvotertrust;lackoftransparencyengendersvotermistrust. In response to the question about each candidate’s trustworthiness, treatmentgroup respondents rated Johnson as3.8percentagepoints less trustworthy than controlgroup respondents, a result that is shown on the first row of Figure 6. The rawtrustworthiness ratings underlying that -3.8 measure are 49 (treatment) and 52.8(control). Conley, on the other hand,was rated 8.9 percentage pointsmore trustworthyamongrespondentsinthetreatmentgroup,comparedtorespondentsinthecontrolgroup,as shown in the second row of Figure 6. The raw trustworthiness ratings were 63.1(treatment) and 54.1 (control). Both differences are statistically significant. Finally, thetrustworthinessgapbetweenJohnsonandConleywasincreasedwhenrespondentsviewedthetransparencyscores.TheJohnson–Conleygapamongthoseviewingonlypolicyscoreswas-1.3.22Amongrespondents inthetreatmentgroup, thegapwas-14.1, forabetween-groupdifference(treatmentgap–controlgap)of-12.8.Again,thedifferenceisstatisticallysignificant.

21 The effect is much higher among co-partisans of the attacked candidate, of course. 22 Based on policy alone, I expected the Johnson-Conley gap to be greater than or equal to zero. Johnson looked exactly like the respondents, in terms of policy, and Conley was 20 percentage points more extreme. The median difference was 0.


25

Figure 6. This figure presents the differences in the mean trustworthiness evaluationsamong the treatment group, minus the mean trustworthiness evaluations among thecontrolgroup,forJohnson(toprow)andConley(middlerow).Theneteffectispresentedin the bottom two row. That quantity is the effect for Conley (Treatment grouptrustworthiness–Controlgroup trustworthiness)subtracted fromtheeffect for Johnson.Theverticallineisat0,wheretheestimateswouldbeifrespondentswerenotaffectedbyseeinganevaluationofthecandidates’campaignfinanceinformation.Thehorizontalaxismeasures percentage points. As before, the open circles show the estimate, which ismeasuredasthedifferenceinmeansbetweenthecontrolandtreatmentgroups.P-valuesreportedareforatwo-sidedt-test.Theblacklinesshowa95%confidenceintervalaroundtheestimate.Allfindingsarestatisticallysignificant. Clearly, trust is implicated in the favorability shifts we observe when includinginformation about campaign finance transparency. But whereas in the favorability andvotingcontext,thenegativeeffectsoutweighedthepositiveeffects,whenevaluatingtrust,the positive rewards to transparency outweigh the lack of trust generated by lack oftransparency.Whilemoreresearchwouldbeneeded toknowwhy the trusteffectswerelargerforConleythanJohnson,itcouldbearesultofrespondentsophistication.Presentedwith a supporting group using anonymous donors, respondents were willing to rateJohnsonlessfavorablyandreportthattheywouldbelesswillingtovoteforher.ButtheirimpressionofJohnson’s(personal)trustworthinessisdifferent.VotersmayhavebeenabletodisentangleJohnson’spersonaltrustworthinessfromthegeneraldeclineinfavorabilitytowardherasacandidatewhoreceivessupportfrom“darkmoney”groups. TherewardstoConleytelltheoppositestory.Whereassheonlyreceiveda3.9-4.6percentage point boost in favorability and vote choice from her transparency score,respondents in the treatment group found her to be 8.9 percentage points moretrustworthyiftheylearnedthatsheover-disclosed.Thetrusteffectwasmorethandouble

−20

−10 0 10 20

●

●

●Johnson − Trustworthiness (T−C)

Conley − Trustworthiness (T−C)

Net Trustworthiness (J−C)

0.03

0.0000

0.0000

P ValueOutcome


26

thefavorabilityandvotingeffect.Thetrustboostfromover-disclosureisinteresting,anditechoes results found in themedical context. Disclosures of potential bias in themedicalcontext leadpatients to trustdoctorswho reveal theirbiasmore than thosewhodonotrevealtheirbias(Sah,etal.2016). In sum, the results suggest that voluntary campaign finance disclosure enhancesvoterperceptionofcandidatetrustworthiness,andfailuretovoluntarilydisclosesreducestheperceptionofcandidatetrustworthiness.Theasymmetryingainingalargeamountoftrustfortransparencybutlosingonlysometrustforalackoftransparencycouldrelatetoabroader behavioral impulse. Or it could be explained by respondent sophistication inunderstandingthatthelackoftransparencyisactuallyonthepartofthesupportivegroup,not Johnson herself. Respondents might not have thought that a group’s lack oftransparency implicated Johnson’s trustworthiness, but that it raised enough questionsabout her policy preferences that they found her less favorable andwere lesswilling tovoteforherforotherreasons.4.3LIBERALSVS.CONSERVATIVES,OREXTREMISTSVS.MODERATES? Can we identify particular characteristics of respondents that might drive theresults? Scholars have argued that liberal voters aremost likely to support anddemandtransparency(Piotrowski&VanRyzin2007;Marguardt2013),andmembersofthemediamakethesameassumption(see,e.g.,Faturechi2016).Here,Ipresentexploratorysubgroupanalysisbasedonrespondentideology.23 Figure7 shows the samemeasuresused in Figures5 and6, bypolitical ideology.TheFigureispresentedintwocolumns,Johnson(left)andConley(right),andthreerows,oneforeachoftheoutcomesmeasured.Theaverageofresponsesinthecontrolconditionis shownwithopen symbols, a triangle for Johnson, andapoint forConley.Theaverageresponse among treatment condition respondents is shown with a solid version of thesymbol.Ideologyispresentedonthehorizontalaxis,anditismeasuredwith1=extremelyliberal, 4 =middle of the road, and 7 = extremely conservative. The average evaluationfromeachgroupisontheverticalaxis.ThearrowsinFigure7aredrawnfromthemeanresponseinthecontrolgrouptothemeanresponseinthetreatmentgroup.TheU-shapeofresponses for Johnson in the Figure is striking. Ideologically extreme respondents gavehigherevaluationstoJohnsonthantheirmoremoderatecounterparts.“Middleoftheroad”respondents gave the lowest evaluations among control group respondents and oftenamongtreatmentgrouprespondentsaswell. Whenweexaminepunishmentof Johnson– as the sizeof thedifferencebetweencontrol and treatment favorability, vote choice, and trustworthiness evaluations – theliberal vs. conservative difference predicted by existing scholarship is not apparent. Butinterestingly, what is apparent is that the extreme respondents punished Johnsonmorethan theirmoremoderate counterparts did – the arrow length is longer for “extremelyliberal”or“extremelyconservative”respondents.

When it comes to rewarding Conley’s transparency, the patterns are murkier. Ingeneral,left-leaningrespondentsweremorerewardingoftransparencythanright-leaningrespondents.Thearrowsarelongeramongrespondentsinthe2-3ideologicalrangethanin23 I note again that I was unable to file a pre-analysis plan for this experiment. This is the only subgroup exploration I have conducted, and it’s the only subgroup exploration I intended to do from the outset.


27

the5-6range.Yetpoliticallyextremerespondentsstillbehavedifferentlythanmoderates,andagain,extremeconservativesandextremeliberalsappearmoresimilartoeachotherthan to their co-partisans. They did not reward Conley for her transparency – theypunishedheronallthreemeasures. A full explanation of this behaviorwill require follow up studies. For now, a fewthings couldbe true.First, peopleon theextremeendsof the ideological spectrumreactstronglywhen an ideologically aligned candidate receives darkmoney support. But theyrate ideologically aligned candidatesmore highly in both control and treatment groups.Their ratings of Johnson across conditions, and across thematter under evaluation, arehigher than those of their more moderate fellow respondents. Perhaps as evidence ofpolicy’s primacy among extreme respondents, they refused to credit Conley for hertransparency, evenusing it against her. Second, it seems that both left and right leaningvotersarewillingtopunishtheuseof“darkmoney”.Finally,rightleaningvotersseemlessimpressed with transparency than left-leaning voters. Conley’s transparency does notimproveherfavorabilityestimatesasmuchamongconservativesasamongliberals.Theyviewhertransparencyfavorably,buttheeffectismoremutedthantheeffectamongtheirleft-leaningcounterparts.All inall, theredoesseemtobeanideologicaleffect,but itcutstwoways,andpolicyextremityseemstobeanimportantpartofthestory.


28

Figure7:Averagetreatmentandcontrolgroupevaluationsonfavorability(toprow),votechoice (center row) and trustworthiness (bottom row), for Johnson (left column) andConley(rightcolumn).Groupresponsesareshownbyideologyfrom1=veryliberalto7=very conservative, with 4 being “middle of the road”. Control group evaluations arepresented as the tail of the arrows. Treatment group evaluations are presented as theheadsofthearrows.

Ideology

Favo

rabi

lity −

John

son

1 2 3 4 5 6 7

40

50

60

70

80

Ideology

Favo

rabi

lity −

Con

ley

1 2 3 4 5 6 7

40

50

60

70

80

Ideology

Vot

e C

hoic

e −

John

son

1 2 3 4 5 6 7

40

50

60

70

80

Ideology

Vot

e C

hoic

e −

Con

ley

1 2 3 4 5 6 7

40

50

60

70

80

Ideology

Tru

stw

orth

ines

s −

John

son

1 2 3 4 5 6 7

40

50

60

70

80

Ideology

Tru

stw

orth

ines

s −

Con

ley

1 2 3 4 5 6 7

40

50

60

70

80


29

4.4STRENGTHSANDLIMITATIONS As with any experiment, this one contains design tradeoffs. Future researchersmightmakedifferentdesignchoices.First,thetreatmenttextissparsetokeepthelengthofthe vignettes as similar aspossiblewhile still including the treatmentmanipulation.Thevignette contains carefully selected details. First, in the vignette, no one has violateddisclosure rules. The vignette is set up to offer voters a chance to punish a candidate(Johnson)forachoicethatisneithermadebythecandidatenorillegal.Candidatescannotcontrol anything about the outside organizations that choose to support them or attacktheiropponents.TheFederalElectionCampaignActandrelatedFECregulationsprohibitcandidatesfromcoordinatingwithsupportiveoutsidegroupsafterannouncingcandidacies(2U.S.C.§431(1972),amendedbyPub.L.No.93-443,88Stat.1289(1974),implementedin11CFR109.21–109.23).Theactionsofanoutsideorganizationshouldhave,atmost,atenuous connection to the candidate in the eyes of the voter – indeed, the lack ofcoordinationbetweenthecandidateandoutsideorganizationunderliethemainholdingsofSpeechNowandCitizensUnited.AnypunishmentofJohnsonisaresultofvotersignoring(orbeingignorantof)the“wall”prohibitingcoordinationtopunishthecandidateforreceivingsupport from a group that is choosing, in a voluntary disclosure regime, not to disclose.There is no evidence presented that Johnson attempted to illegally coordinate with theoutsidegroup. Second, Conley has over complied with disclosure requirements by voluntarilydisclosing.Inreality,somekindsofvoluntarydisclosurearemorehelpfulthanothers.Forexample, if Conley discloses the names of all donors, not just those above the reportingthreshold,somescholarsmightarguethatsheisnotincreasingvotercompetencebecauseinformation about small-time donors is of low informational value (Cain 2010; Hasen2010).Ontheotherhand,ifConleyisinvesting“costlyeffort”intodisclosure,maintainingrecordsofalldonationsanddisclosingtheidentityofeachdonormaytellthevoteraboutthecandidate’scommitmenttotransparency. Eventhosewhodisagreethatdisclosureofsmalldonorsisusefulmightbeabletoimaginewaysthatcandidatesmightovercomply,inthesensethattheyreleaseinformationtheyarenotrequiredtorelease,inawaythatincreasesvotercompetence.Forexample,acandidatemightincludeageo-coded“heatmap”ofzipcodesinherdistrictorstate,colorcodedby thenumberofcontributionsshehasreceived fromeachzipcode,regardlessofamount. Voters would understand whether people in their area were contributing andwhetherthecandidateattractedbothurbanandruralsupporters,whichcouldhelpinformthevoter’schoice.Ifsupportiswideenough,thecandidatemightstandtogainfromusingcontribution information in this way, despite not being required to provide it herself,beyond reporting addresses to the election regulator for contributors whose aggregatecontributions have exceeded a certain amount. Finally, votersmight reward this level oftransparency when C is competing with a less transparent but ideologically similarcandidate. Voterpreferencesovercampaignfinancetransparencymaybestableovertime,ortheymaychange.Wehavesolittledatathatitishardtoknow.Thecurrentadministrationhasmadeseveralmovestoreducetransparency,suchasrefusingtoreleaseWhiteHousevisitor logs (then reluctantly agreeing to release partial logs), increasing the number ofdailybriefingsthatareoff-camera,andnotreleasingPresidential taxreturns.Allof theseissueshavemade it into themedia. Furthermore, campaign finance is in thenews lately


30

withtheallegationsthattheTrumpcampaignandRussiamayhavecoordinatedtoopposethecandidacyofHillaryRodhamClinton.Becausetransparencyandcampaignfinancearemuchmore in the news now, compared to the time the surveywas fielded in 2015, theresultsIfindherecouldunder-estimatetheeffectwemightobserveifthesamesurveyandexperimentwereruntoday.5.IMPLICATIONS5.1FORCOURTS AlmostnoneoftheCourt’sassumptionsaboutthecostsandbenefitsofdisclosureinBuckleyhaveundergonethoroughempiricalanalysis.Thejurisprudencehasnotchangedin40years,butsocialscienceprogressonthequestionhasbeenalsobeenslow.Thesurveyresultsindicatethevotersvaluetheamountofinformationcampaignsprovideoverotherindicia of candidate strength, such as the size of the war chest. Moreover, voters valuecampaignfinance informationalmostasmuchasacandidate’sability tobepersuasive inpublic.Theexperimentalresultssuggestthataformof“meta”informationaboutcandidateandgroupdisclosurechangesvoterbehaviorinawaythatimprovesvotercompetence.24 We have muchmore to learn about the informational benefits of disclosure. Fornow, these results suggest that voters care enough about accessing campaign financeinformationtochangetheirvotesoverit.5.2FORCAMPAIGNSANDGROUPSTHATMAKEINDEPENDENTEXPENDITURES Thesurveyandexperimentalresultssuggestthatcampaignsandso-called“outsidegroups”standtogainfrom(1)pointingoutwhentheopposingsideissupportedbydark-moneygroups,(2)requestingthatanygroupswhowanttosupporttheirsidediscloseatleast some information about their donors, (3) over-complying with disclosurerequirementswheretheycanmeaningfullyinformvotersabouttheirsourcesofsupport.25 Thelargestbenefitismobilizingvotersentimentagainstanopponentsupportedbydarkmoney.Asforover-complying,theexperimentdidnotcompareparticularapproachesto over-compliance with disclosure regulations, but any move that enables a side tocredibly claim that they are more transparent than the law requires could result in anelectoral boost. Geo-coding donations by neighborhood, as described above, might be aparticularlyusefulandinformativewaytoover-comply.For501cgroups,disclosingabovea certain threshold can help the candidate the group supports. A press release aimed atorganizations like the Center for Responsive Politics (OpenSecrets.org) could helpcampaignsandgroupscapitalizeontheeffectsseenhere.

24Ididnotaskrespondentpreferencesaboutcandidatetransparencybeforerunningtheexperiment,soasnottoprimethem,whichmeansIamstrictlynotsurewhethertheyarevotingmoreinlinewiththeirinterests.Iamunawareofatheorythatmoreinformationhurtsvotercompetence,soitseemshighlylikelythattheywerebetterabletovoteinlinewiththeirpreferences.25Groupsmayhavegoodreasonstoorganizeasa501(c)4ratherthana527,PAC,orSuperPAC.The501(c)4groupscanconductany“socialwelfare”activity,andinsomeyearstheymaynotengagepoliticalcampaigningatall.Theotherorganizationalformsarecampaignfocused.


31

5.3FORLAWMAKERSANDREGULATORS Theexperimentalmanipulationherewasbasedonvoluntarydisclosure.Supportersofonecandidatechosenottovoluntarilydisclose,andsupportersoftheothercandidate–aswellasthecandidateherself–chosetovoluntarilydisclosemoreinformationthanthelawrequires.Theseeffectscanoccurwithoutfurtheractiononthepartoflawmakersandregulators. Nevertheless,thesurveyresultsrevealthatcampaignfinancedisclosureispopular,andmostvotersseektheinformationatleastsomeofthetime.Lawmakerslookingtoscorepointswiththeirconstituentscoulddosobyintroducingmoredisclosureintotheexistingframework.Moreover,despitethefactthatdisclosureisbecomingapolarizingissueamongpolitical elites, the subgroup analysis in Section 4 indicates that disclosure is not aparticularlypartisanissueforvoters. Asfordesigninganoptimalframework,policymakersmustconsiderthetwotypesof information maximized by mandatory and voluntary disclosure regimes. Mandatorydisclosure is thought to increase the information about the policies that candidateswillfavor once in office. But where policymakers and the court leave room for voluntarydisclosure, theresults fromthisexperimentsuggest thatvoters learnabouttransparencyandtrustworthinessofcandidatesandthegroupsthatsupportthem. Ifreformerserrtoomuchonthesideofvoluntarydisclosure,voterswillreceiveasuboptimal amount of information about policy. Despite the experimental findings here,dark money groups are balancing more considerations that “merely” gaining valencebenefits forthecandidates–namely,manyof theirdonorshavechosentogivetoa501crather than a SuperPAC or other disclosable avenue precisely so that they can remainanonymous. A purely voluntary disclosure regime will under-provide campaign financeinformation,becausemanydarkmoneygroupswill continue tochoosenot todisclose. Ifreformersinsteaderrtoomuchonthesideofmandatorydisclosure,voterswillmissoutoninformation about trustworthiness and transparency that comes when candidates andgroupsvoluntarilydisclosemorethanthelawrequires. Arguments for changing our disclosure regime have centered on disclosurethresholds, the type of donor information collected and disclosed, forcing mandatorydisclosureon501(c)4groups,andonlineadvertising.Thetradeoffsfacedbypolicymakersforeachofthesefourareasofdisclosureareworthabriefdiscussion. Modifying disclosure thresholds. Disclosure thresholds vary across jurisdictions.Somestatesrequiredisclosureofallcontributions,andothershavethresholdshigherthanthe threshold for federal candidates ($200). Some lawyers and scholars argue thatdisclosure thresholds shouldbe raised,meaning thatmore campaign financingwouldbesubjectonlytovoluntarydisclosure(Cain2010,Beckel2011),whileatleastonedisclosurescoring group rewards thresholds below $200 (Center for Public Integrity & GlobalIntegrity2015). In aworld inwhich individuals can give over $300,000 to “joint victoryfunds”andunlimitedamountstoSuperPACs,theidentityofdonorsof$200probablydoesaddmuch information to thevoter’s informationaboutpolicies the candidate is likely topursueonceinoffice.Raisingthethresholdwouldalsoallowmore“space”forcandidatestovoluntarilydisclose.


32

Byhowmuchcould reformers raise the thresholds?Voluntarydisclosureof smalldonors might not have salutary benefits for voters’ perceptions of the transparency ortrustworthiness of a candidate.26Suppose reformers raised threshold modestly, say to$500. Voters may not reward voluntary disclosure of donors between some smalleramountand thenew threshold.Relatively speaking, $500 is still notmuchmoney inourworld of big-money politics. Suppose now that policymakers raisemandatory disclosurethresholdsbyalot,say,tothefederalcontributionlimitof$2,700.Candidatesandgroupschoosing todisclose fairly largedonors, suchasalldonorsabove$1000or$2000,mightreap transparencyand trustworthinessbenefits.Futureresearchers–and thecampaignsthemselves– canestimatewhere thevalencebenefits kick in. In general, it seems that amodestraiseofdisclosurethresholdswouldnotdomuchtoaffectthemixofinformationavailabletothevoters,butalargerchangeindisclosurethresholdscould. ModifyingtheTypeofDonorInformationthatisCollectedorDisclosed.Policymakersmay consider changing the donor information gathered and reported by candidates inorder to optimize the informational tradeoffs. The FEC currently releases contributorname,city,state,zipcode,principalplaceofbusiness,occupation,dateofeachcontribution,andamountcontributedeachtime.Allofthisinformationcanbeinformativewithregardstothepoliciesthatadonormightsupport.

For example, a contributor’s city and state can help indicate the policies that thecandidate’ssupporter’sfavor.ConsidertwoCaliforniacontributors.AcontributorfromLosAngeles and a contributor fromSanDiego, despite living only 120miles apart, probablyhave different policy concerns. The Angeleno lives in a city culturally dominated by theentertainmentindustrybutalsodeeplyinvolvedininternationaltradeandpetroleum.TheSan Diego donor, on the other hand, lives on the Mexican border and near several U.S.militaryinstallations.Zipcodeswithinlargecitieslikethesecanhelpvoterstorefinetheirestimates of donor policy preferences even more closely. The geographic location of adonor, even just city and state, canhelp voters knowmore about thepolicy pressures acandidatewillface. Donors and voters alikewant to protect their economic interests,which includestheirjobs.Disclosingthesectorinwhichadonorisemployedisusefulforpredictingfuturepolicy responsiveness of an elected official supported by donors from that sector.27Similarly, a donor’s general occupation is useful forpredictingpolicypositions. Considerthehealthcare sector.Within the sector, or evenwithinonehospital, if nurses support acandidate but hospital management does not, voters can learn about pressures thecandidatemayfacewithregardstolaborissues. Similarly,thedateofthecontributionprovidesusefulinformation,insofarasitcanhelp voters (and campaigns) to understand which policy speeches and news stories26Contributorsmayfeel“protected”fromdisclosurewhentheygivebelowamandatorythreshold,andtheymayfeelunfairly“exposed”ifthecandidateorgroupdisclosestheiridentitywhenthelawdoesnotrequireit.Thishappensmoreoftenthanonemightexpect:instate-levelracesbetween2000and2008,17%ofallcontributionsthatcandidatesandcampaignsreportedwerebelowthestate’smandatorythreshold.Givingbelowthethresholdisnotasafeharborfrombeingidentified(Wood&Spencer2016).27Ofcourse,therearecorruption-basedreasonsforwantingtoknowthemorefine-grainedinformationoftheemployer,sothatthepubliccanholdelectedofficialsaccountableforworkingtobenefitdonorsvs.non-donors.


33

contributorsarereactingto.Famously, theObamacampaignraised$10million in the24hours following Sarah Palin’s speech at the 2008 Republican National Convention. Herspeech included policy mentions of energy, state budget surpluses, taxes, andmanufacturing.Policyinformationinthecampaignisrevealednotonlybygeography(city,state,zip)andsector,butalsoovertime.28 Foranyoneindividualdonation,publicdisclosureofthedonor’snameisprobablythe least informativepieceof informationavailabletovoters.However,nameisprobablythemost importantpieceofdonor information forpredicting the ideologies and roll callvotes of candidates across time. Donor name is the main piece of information that canreliablylinkcontributionsovertime.Donorideologycanbeestimatedwithsomeprecision.AdamBonicahasusedrepeatdonationsattostateandfederalcandidatestodoexactlythat(Bonica2014).And in turn,donor ideologies,whenaggregatedwithincandidates,canbeused to estimate the ideology ofmembers of Congress (Bonica, forthcoming). The sameeffectcouldbeachievedbyreplacingthedonor’snamewithadonorIDnumber,sothatthedonor’s name does not become public (Elmendorf & Wood, 2017). That said, there isprobablyathresholdabovewhichacontributorplayssuchanoutsizedroleinacampaignorgroupexpenditurethatvotersneedtoknowtheidentityofthecontributor.Changinganamelike“SheldonAdelson”or“GeorgeSoros”toanIDnumberwouldrepresentareallossofinformation,aboutpolicypressuresonandpreferencesofcandidates. Shouldpolicymakersmakeanyofthisdisclosurevoluntary,ratherthanmandatory?The two most sensitive items disclosed are contributor name and place ofemployment/occupation. Replacing name with an ID number would still allow theideology-predicting benefits of linked donations over time, so that experts could stillprovide estimates of candidate ideology. There could be some informational losses topeople searching the contributionsof respected friends,neighbors,or leaders, inmakingtheir voting decisions, but to my knowledge it has not been studied. On net, theinformational losses about candidate policy pressures that would result from replacingnames with ID numbers seem small, at least below a certain “mega donor” threshold.Candidatesmightbe able to gain valencebenefits by voluntarilydisclosing thenamesofdonors above a certain (very high) threshold, especially in states with no contributionlimits.Butat the federal level, it isunlikely thatmanycandidateswouldchoosetodoso.Thesystemishydraulic(Issacharoff&Karlan1999):especiallyduringtheprimaryseason,savvy large contributorsmight choose to give to candidateswith similar ideologieswhowillprotecttheiridentitiesbyusingonlyIDnumberstoreportcontributions. Thepublicdisclosureofplaceofemploymentandoccupationprovidesvoterswithpolicy information for the reasonsdescribedabove.Were thesedatano longer required,votersmightappreciateseeingthem,especiallyintheaggregate(eithercountsofamountscontributed by certain groups, for example). Such voluntary disclosure could promoteperceptions of trustworthiness and transparency. Policymakers might also choose tocontinue to require the information but de-link the information about employment andoccupation from the identity of the individual making the contribution when posting itonline. De-linking occupation and employment information from individual identifyinginformation would only marginally reduce voter information about policy – it rarely28Thebiggestbeneficiaryofsuchinformationisprobablythecandidate,asElmendorf&Wood(2017)argues.


34

mattersthatPersonXinJobYinCompanyZdonated.ItmightmatterthatPersonXgave(especiallyifthepersonisamegadonor),anditdoesmatterthatpeopledoingJobYlikethecandidateandthatpeopleatCompanyZarecontributing.Forcingmandatorydisclosureondonors501(c)groups Whataretheinformationaltradeoffsfacedbypolicymakersconsideringmandatorydisclosurefor501(c)groups?Infederalcampaigns,disclosurebythesegroupsiscurrentlyentirelyvoluntaryaslongastheyarenotrunningpoliticalads30daysbeforeaprimaryor60 days before a general election. As this experiment shows, respondents changed theirvotesbasedonknowledgethatoutsidegroupssupportingacandidatedisclosemorethanthe law requires, evaluating the candidates supported by these voluntarily-disclosinggroups as more trustworthy. Most politically active 501(c)4 groups do not voluntarilydisclosetheirdonors. Ifpolicymakersweretorequiredisclosureofatleastsomedonors(especiallylargedonors)to501(c)4or501(c)6groups,wouldvotersgainpolicy-relevantinformation?Onnet,theywould.Whilethepoliticalleaningsofthemostactivegroupsarealreadystronglyidentifiedwithpolicypositionsorideologies(considerNRAInstituteforLegislativeActionandUS Chamber of Commerce), other groups have names that are not identifiedwith aparticularideologyorpolicyplatform(Torres-Spelliscy2011).Consider“AmericanActionNetwork”,“OneNation”,“iAmericaAction”and“AmericaNext”,whichspentmillionsinthe2016elections,andwhosenamesdonot indicatetheirpolitical leanings.Thepublicdoesnot know who supports these groups, whether corporations, individuals, or otherorganizations. Ifdisclosureruleschange,anonymity-seekingdonorsmightstillfindnondisclosurewithin the organization due to the nature of 501(c)4 and 501(c)6 organizations. Theorganizations can engage in socialwelfare and policywork outside of election season.29Requiringmandatory disclosure of the donorswhosemoney goes to support campaign-related activities can therefore create the incentive of 501(c)4 groups to segregate theirdonors into “disclosable” and “non-disclosable” donors. Money from non-disclosabledonorscouldgotooverheadorsocial-welfare improvingactivitiesthatarenotcampaignrelated. Money from disclosable donors could go to campaign-related activities. Thedonation form can simply ask contributors whether they are comfortable with theirdonation being made public. Corporations, in particular, might opt for “nondisclosable”status. Insum,thehydraulicsofcampaignfinancemeanthatforcingmandatorydisclosureon these “dark money” groups might result in incomplete donor information, becausedonors who seek anonymity can still have it. The donor information that is revealed islikely to be the least “objectionable” set of donors, from a public opinion perspective –perhaps from themost politicallymoderate subset of the organization’s donors. Puttingaside the ban on coordination between candidates and groups, partial disclosurewouldprovide a skewed estimate of the policy pressures on the candidate as a result ofexpendituresmadebythegroup. Thepolicyinformationgainedbyincreaseddisclosuremustbeweighedagainstthetrustworthiness benefits lost by mandating disclosure. The amount of transparency29ThatiswhyOurRevolutionorganizedunderthisorganizationalform,ratherthanaPAC,inthefirstplace.


35

informationwewould sacrifice withmore disclosure depends onwhere themandatorythreshold is set. If policymakers require disclosure of only large donations, say above$10,000, disclosure-friendly groups could continue to voluntarily disclosure below thethresholdandvotersandthevoluntarilydisclosinggroupswillbenefitaccordingly. Forcingmandatorydisclosureonallonlineadvertising There are twoways inwhichonline advertising lacksdisclaimers. First, the smallads that appear next to search results or on Facebook page are exempt from federaldisclaimerrequirementsunderthe“smallitems”exemptionforadssosmallthatincludinga disclaimer on them is impractical. Second, the FEC exempts from disclaimers anyadvertisingplaced“fora feeonanother’swebsite”.ThissecondexemptionturnspoliticalcommunicationontheInternetintothe“wildwest”(Persily,2016).

Plenty of ads supporting or opposing candidates are placed for free and do notrequiredisclosure.Considerads(oreven“fakenews”)uploadedtoacandidateorgroup’swebpage,orsocialmediaprofile,ortoYouTube.Ifthoseadsgoviral,theirmessagereachesalotofpeople.Iftheylackdisclaimers,theycirculatewithnotransparencyattached.Inthe2016 election, most online video advertising seemed to be from campaigns and groupspostingadstheyhadmadeforTV,andmostseemedtoincludethedisclaimersthatwererequired for television. In otherwords,most online video advertising had at least somevoluntary disclosure attached, by default. We should not expect this to last – onlineadvertisingisalwaysevolving,eveninpoliticalcampaigns.

Especiallyinthecaseoffakenews,alackofdisclaimerswillleavevotersunabletoevaluatetheunderlyingpolicy leaningsofthecandidateswhoarethebeneficiariesofthemessages,muchlesstheveracityofthemessageoreventhenationalityofthespeaker.Theimbalance between the amount of politicalmessaging subject to disclosure online (verylittle) and the amount not subject to disclosure (almost all) is stark. More mandatorydisclosureontheInternetisademocraticnecessity.

Lawmakers and regulators could start with easy-to-implement reforms, likeMinnesota, Maine, and other states have done. Require that the “landing page” of smallonline ads disclose the identity of the group running the ad (at aminimum). Require adisclaimerattheendofallonlinevideoads,whetherplacedforafeeorforfree.Allofthiswould improve the amount of policy-relevant information that voters receive about thegroupsrunningtheadsandthecandidatesbenefitingfromthem.

Inthemeantime,canvoterslearnaboutcandidateandgrouptrustworthinessfromvoluntary disclaimers online? In theory, yes, they can. However, voluntary disclosure isonlyeffectivewhereauditsareavailable,anditisimpossibletoaudittheentireuniverseofonline advertising. Suppose a group called “AmericaPAC” runs an ad, with a voluntarydisclaimer,onitsownYouTubechannel.Thefactofthevoluntarydisclaimershouldworktothegroups,andthebenefittingcandidate’s,advantage.Butnoauditcouldeverconfirmthat the group is not also running other, disclaimer-free, ads from anonymous YouTubechannels.Therefore,auditreportsofgrouponlineactivityshouldusecareful language tospecify that of thematerial reviewedby the auditors, all of it containedmoredisclosurethanthelawrequires.Trustworthinessbenefitsshouldbemoderated,asaresult.


36

5.4FORFUTURERESEARCHERSThese results suggest that, in a voluntary disclosure regime, voters will punish

nondisclosureandrewarddisclosurewhentheyaremadeawareofthedisclosingactivitiesofcampaignsandoutsidegroups.Theeffectsizesarefairlylarge,butthemanipulationisquite strong, and thepolicy informationprovided is abstract. The study ismerely a firststep,andfutureresearchersshouldfocusonthisimportantandtimelyresearchquestion.

Futurestudiesshouldexplorewhethervotersassumecoordinationbetweenoutsidegroups and campaigns. The respondents in this study punished a candidate for thecompletely legal behavior of an uncoordinated group supporting her, but theywere nottold that the behavior of the independent groups complied with the law, nor did thepromptpresentdetailsaboutwhat“supportfromindependentgroups”mightmean.

Anotherfruitfulavenueforresearchwouldbetodisentanglewhetherrespondentshere are rewarding over-compliance with the law or transparency. The scores here are“transparency grades”, but Conley receives an “A” for her over-compliance with atransparencyrule,inparticular.Whatifthesettingwereadifferentvoluntarycompliancecontext,suchascorporatesocialresponsibilityinherprofessionallife?

Finally, theBuckley courtenvisions thedisclosed information–donor informationandamounts–asbeingparticularlyinformativetovoters.Moreresearchersshouldtacklethis question head-on. Lupia (1994) showed that knowledge of which policies industrypreferred improve voter competence, but the source of the knowledge, whetherendorsements, campaign finance information,or somethingelse,wasnot specified.Morework is needed to assist the courts in balancing the tradeoffs of campaign financedisclosure.6.CONCLUSION For40years,theSupremeCourthasassumedthatcampaignfinancedisclosurescanprovidevaluableinformationtovoters.Theassumptionhasrarelybeentested.Thesurveyandexperimentpresentedhereprovideevidencethatpeoplecareaboutcampaignfinancetransparency.Respondentsrewardedvoluntarydisclosureandpunishedacandidatesupportedbydarkmoneygroups.Informationaboutcampaignfinancedisclosuresprovidedvaluableinformationtovoters,whowereabletovotemorecompetentlyasaresultoflearningit.Respondentswerenotprovidedthecontentofdisclosures.Theinformationtheyreceivedassistedtheminevaluatingcandidatetrustworthiness,anditminimizeduncertaintyaroundpolicypositions. Theexperimentisdesignedcarefullytoreflectourexistingcampaignfinanceregime,forwhichagooddealofdisclosureremainsvoluntary.Werealldisclosuremandatory,theexperimentwouldneedtobedifferent.Wheredisclosureismandatory,over-complianceismorecostlyandshouldbemorerare.Under-complianceisaviolationofthelawinamandatoryregime.Voterreactionswouldlikelybedifferentiftheexperimentalcontextwereoneofmandatorydisclosure. Thesurveyresultsindicatethatvoterswouldwelcomemoremandatorydisclosure,ifCongressandtheFECweretobecomemoreopentoit.Thecurrentregulatoryframeworksuggestsvoterinformationcouldbeoptimizedbymakingsomeaspectsofcampaignfinancedisclosurevoluntary–perhapsbyraisingthemandatorydisclosurethreshold–andotheraspectsmandatory–like501cdisclosureandincreasingmandatorydisclosure


37

online.Inthemeantime,theresultsoftheprojectsuggestthattheSupremeCourt,inevaluatingthenextdisclosurechallenge,shouldtakeseriouslytheinformationbenefitsofcampaignfinancedisclosure.


38

REFERENCESDark Money. OpenSecrets Center for Responsive Politics.https://www.opensecrets.org/dark-money/.(July23,2017)ACLU. 2010. ACLU Urges No Vote on Disclose Act. 26 July.https://www.aclu.org/news/aclu-urges-no-vote-disclose-act.BeckelM.2011.“CourtBattlesoverCampaignDisclosureLoom,LegalExpertsPredict”.Open Secrets Blog. https://www.opensecrets.org/news/2011/03/court-battles-over-campaign-disclos/Bianco W. T. 1994. Trust: Representatives and Constituents, Ann Arbor, Michigan:UniversityofMichiganPress.Bonica,A.2014.MappingtheIdeologicalMarketplace.Am.J.Pol.Sci.,58,267.Bonica,A.Forthcoming. InferringRollCallScores fromCampaignContributionsUsingSupervised Machine Learning. Am. J. Pol. Sci. (manuscript available athttps://papers.ssrn.com/sol3/papers.cfm?abstract_id=2732913).BriffaultR.2010.CampaignFinanceDisclosure2.0.ElectionsL.J.,9,273.BriffaultR.2012.UpdatingDisclosurefortheNewEraofIndependentSpending.J.L.&Pol’y.,.27,708.Brown R. L. Martin A. D. 2015. Rhetoric and Reality: Testing the Harm of CampaignSpending.N.Y.U.L.Rev.,90,1066.Cain B. 2010. Shade from the Glare: The Case for Semi-Disclosure. CATOUnbound, 8November. http://www.cato-unbound.org/2010/11/08/bruce-cain/shade-glare-case-semi-disclosure.DowlingC.M.WichowskyA.2015.AttacksWithoutConsequences?Candidates,Parties,Groups,andtheChangingFaceofNegativeAdvertising.Am.J.Pol.Sci.,59,19.Downs.A.1957.AnEconomicTheoryofDemocracy,NewYorkCity,NewYork:HarperandRow.Dykhne,I.2017.PersuasiveorDeceptive?NativeAdvertisinginPoliticalCampaigns.S.Cal.L.R.90.EisenbergerR.2004.WhoTakestheMostRevenge?IndividualDifferencesinNegativeReciprocityNormEndorsement.Personality&Soc.Psychol.Bull.,30,787.ElmendorfC.S.2009.RefiningtheDemocracyCanon.CornellL.Rev.,95,1051.


39

Elmendorf C. S.Wood A. K. 2017. Elite Political Ignorance.Working Paper, availablefromauthor.EnelowJ.M.HinichM.J.1984.TheSpatialTheoryofVoting:AnIntroduction,Cambridge:CambridgeUniversityPress.Faturechi R. 2016. The Conservative Playbook for Keeping ‘Dark Money’ Dark.ProPublica, 4 February. https://www.propublica.org/article/the-conservative-playbook-for-keeping-dark-money-dark.FehrE.GachterS.2000.FairnessandRetaliation:TheEconomicsofReciprocity.J.Econ.Persp.,14,159.GarrettE.1999.TheLawandEconomicsof‘InformedVoter’BallotNotations.UniversityofChicagoPublicL.&LegalTheoryWorkingPaperNo.4.Garrett E. SmithD. A. 2005. Veiled Political Actors andCampaignDisclosure Laws inDirectDemocracy.ElectionL.J.,4,295.GaudianoN.2016.BernieSandersSpinoffGrouptoDiscloseBigDonors.USAToday,7September.https://www.usatoday.com/story/news/politics/onpolitics/2016/09/07/bernie-sanders-spinoff-group-disclose-big-donors/89974756/GerberA.S.HuberG.A.DohertyD.DowlingC.M.HillS.J.2013.DoPerceptionsofBallotSecrecyInfluenceTurnout?ResultsfromaFieldExperiment.Am.J.Pol.Sci.,57,537.GerkenH.K.GibsonW.LyonsW.2014.ReroutingtheFlowof‘DarkMoney’intoPoliticalCampaigns. Washington Post, 3 April.https://www.washingtonpost.com/opinions/rerouting-the-flow-of-dark-money-into-political-campaigns/2014/04/03/1517ac6e-b906-11e3-9a05-.Gilbert,M. 2013. Campaign FinanceDisclosure and the InformationTradeoff. Iowa L.Rev.98,1847.Gilbert,M.andAiken,B.2015.DisclosureandCorruption.ElectionL.J.14,148.GrosecloseT.2001.AModelofCandidateLocationWhenOneCandidateHasaValenceAdvantage.Am.J.Pol.Sci.,45,862.HartmanE.HidalgoD.F.WorkingPaper2016.What’stheAlternative?:AnEquivalenceApproachtoBalanceandPlaceboTests.


40

HasenR.2010.ASemi-ObjectiontoBruceCain’sSemi-CaseforSemi-Disclosure.CATOUnbound, 8 November. http://www.cato-unbound.org/2010/11/15/richard-hasen/semi-objection-bruce-cains-semi-case-semi-disclosure.HasenR.L.2009.TheDemocracyCanon.Stan.L.Rev.,62,69.Healy P.M. Paleu K. G. 2001. Information Asymmetry, Corporate Disclosure, and theCapital Markets: A Review of the Empirical Disclosure Literature. J. Accounting andEcon.,31,405.IssacharoffS.KarlanP.1999.TheHydraulicsofCampaignFinanceReform.TexasL.Rev.,77,705.Krosnick J. A. 1988. TheRole ofAttitude Importance in Social Evaluation:A Study ofPolicy Preference, Presidential Candidate Evaluations, and Voting Behavior. J.Personality&Soc.Psychol.,55,196.La Raja, R. 2014. Political Participation and Civic Courage: The Negative Effect ofTransparencyonMakingSmallCampaignContributions,36PoliticalBehavior753.LevinsonJ.2016.FullDisclosure:TheNextFrontierinCampaignFinanceLaw.Denv.L.Rev.,93,431.LowensteinD.H.HasenR.L.TokajiD.P.2016.ElectionLaw,CasesandMaterials, 5thedn.Durham,NorthCarolina:CarolinaAcademicPress.Lupia A. 1994. Shortcuts Versus Encyclopedias: Information and Voting Behavior inCaliforniaInsuranceReformElections.Am.Pol.Sci.Rev.,88,63.LupiaA.McCubbinsM.D.1998.TheDemocraticDilemma:CanCitizensLearnWhatTheyNeedtoKnow?,Cambridge,UniversityofCambridgePress.McCurley C. Mondak J. J. 1995. Inspected by #1184063113: The Influence ofIncumbents’CompetenceandIntegrityinU.S.HouseElections.Am.J.Pol.Sci.,39,864.McGehee M. 2015. Fix the FEC: Background Memorandum on New BipartisanLegislation to Address a Dysfunctional Agency. The Campaign Legal Center, 17September. http://www.campaignlegalcenter.org/news/publications-speeches/fix-fec-background-memorandum-new-bipartisan-legislation-address-0.Mondak J. J. 1995.Nothing toRead:Newspapers andElections in a Social Experiment,AnnArbor,Michigan:UniversityofMichiganPress.NyhanB.2014.ScandalPotential:HowPoliticalContextandNewsCongestionAffectsthePresident’sVulnerabilitytoMediaScandal.BritishJ.ofPol.Sci.,45,425.


41

OrtizD.2012.TheInformationalInterest.J.Law&Pol’y.,27,663.Overman T. 2002. Shame on You: Campaign Finance Reform Through Social Norms.Vand.L.Rev.,55,1243.PageB.I.JonesC.C.1979.ReciprocalEffectsofPolicyPreferences,PartyLoyaltiesandtheVote.Am.Pol.Sci.R.,73,1071.Persily,N.2015.TheCampaignRevolutionWillnotbeTelevised.TheAmericanInterest11(2). https://www.the-american-interest.com/2015/10/10/the-campaign-revolution-will-not-be-televised/Persily,N.2016.GoodLuckTryingtoRegulatetheNewWorldofOnlinePoliticalAds.N.Y. Post, Aug. 16, 2016. http://nypost.com/2016/08/16/good-luck-trying-to-regulate-the-new-world-of-online-political-ads/PiotrowskiS. J.VanRyzinG.G.2007.CitizenAttitudesTowardTransparency inLocalGovernment.Am.Rev.Pub.Admin.,37,306.PrimoD.M.2011.FullDisclosure:HowCampaignFinanceLawsFail to InformVotersand Stifle Public Debate. Institute for Justice, 11 October.http://www.ij.org/images/pdf_folder/other_pubs/fulldisclosure.pdf.PrimoD.M.2013.InformationattheMargin:CampaignFinanceDisclosureLaws,BallotIssues,andVoterKnowledge.ElectionL.J.,12,114.CenterforPublicIntegrity&GlobalIntegrity.2015.StateIntegrityInvestigation2015.https://www.publicintegrity.org/accountability/state-integrity-investigation/state-integrity-2015(selectanystatereport,clickon“PoliticalFinancing”,thenclickon“Inlaw,thenamesofcampaigncontributorsandamountsdonatedtocandidatesandpoliticalpartiesmustbedisclosedtothepublic.”)Rhodes, S.C., Franz, M.M., Franklin Fowler, E., Ridout T. N. 2017. The Role of DarkMoneyDisclosureonCandidateEvaluationsandViability.WorkingPaper,presentedatthe2017AnnualMeetingoftheAmericanPoliticalScienceAssociation.Ridout T. N., Shah D.V., Goldstein K.M., Franz M.M. 2004. Evaluating Measures ofCampaignAdvertisingExposureonPoliticalLearning.Pol.Behav.,26,201.Robertson, C., Winkelman, D. A., Bergstrand, K. and Modzelewski, D. 2016. TheAppearance and Reality ofQuid ProQuoCorruption: an Empirical Investigation. J. ofLegalAnalysis8,375-438.Sah S. 2016. Effect of Physician Disclosure of Specialty Bias on Patient Trust andTreatmentChoice.ProceedingsNat’lAcad.Sci.,113,7465.


42

SettleJ.E.BondR.M.CovielloL.FarrisC.J.FowlerJ.H.JonesJ.2016.FromPostingtoVoting: The Effects of Political Competition on Online Political Engagement. Pol. Sci.Research&Methods,4,361.ShawK.2016.TakingDisclosureSeriously.YaleL.&Pol’y.Rev.,34,18.Smith B. A. 1998. A Most Uncommon Cause: Some Thoughts on Campaign FinanceReformandaResponsetoProfessorPaul.Conn.L.Rev.,30,831.Spencer D. M. 2017. Campaign Finance and the Rhetoric of Corruption: A ConjointExperiment.Workingpaper.StokesD.E.1963.SpatialModelsofPartyCompetition.Am.Pol.Sci.Rev.,57,368.StoneW.J.SimasE.N.2010.CandidateValenceandIdeologicalPositionsinU.S.HouseElection.Am.J.Pol.Sci.Rev.,54,371.Torres-Spelliscy,C.2011.“HidingBehindtheTaxCode,theDarkElectionof2010andWhy Tax-Exempt Entities Should Be Subject to Robust Federal Campaign FinanceDisclosureLaws.”Nexus:Chapman’sJ.Law&Policy16:57-95.WangC.LeungA.K.2010.TheCulturalDynamicsofRewardingHonestyandPunishingDeception.Personality&Soc.Psychol.Bull.,36,1529.Weintraub,E.2017.HowOurBrokenCampaignFinanceSystemCouldAllowForeignGovernmentstoBuyInfluenceinOurElectionsandWhatWeCanDoAboutIt.TestimonyBeforetheSenateDemocraticPolicyandCommunicationsCommittee,July19,2017WesleyanMediaProjectandCenterforResponsivePolitics.2016.WMP/CRPSpecialReportOutsideGroupActivity,2000-2016.August24.http://mediaproject.wesleyan.edu/blog/disclosure-report/Wilson, M. R. 2016. FEC Deadlocked on ‘Dark Money’. The Hill. 11 February.http://thehill.com/homenews/campaign/232582-fec-deadlocked-on-dark-moneyWittman D. 2005. Valence Characteristics, Costly Policy, and the Median-CrossingProperty:ADiagrammaticExposition.PublicChoice,124,265.WoodA.K.SpencerD.M.2016.IntheShadowsofSunlight:TheEffectsofTransparencyonStatePoliticalCampaigns,InPress.ElectionL.J.Vol.4,15.WoodA.K.andGilbert,M.Disclosurecanencouragepoliticalspeech.TheHill,October21,2016.http://thehill.com/blogs/pundits-blog/campaign/302152-disclosure-can-encourage-political-speech


43

WrightM.LevyM.Citrin J.2016.PublicAttitudesTowards ImmigrationPolicyAcrosstheLegal/IllegalDivide:TheRoleofCategoricalandAttribute-BasedDecision-Making.Pol.Behavior,38,229.Zaller J. 1992. The Nature and Origins of Mass Opinion, Cambridge: University ofCambridgePress.


44

APPENDIX Randomlyassigningrespondents to treatmentconditionsshouldcreateconditionsthat are balanced on important characteristics, like partisan affiliation, age, or opinionstoward various policy areas. Figure A1 presents the distribution of key characteristicsacrosscategories.

Figure A1: Distribution of respondents’ answers to key survey questions on the verticalaxis.Treatment condition ison thehorizontal axis.C20andT20= respondents told thatConleywas20points from Johnson.Violinplotdisplays thedensityof responsesateachvalue,smoothingacrosslevels.Genderiscoded1forfemale,0formale.Ideologyiscoded1=“very liberal” to5=“veryconservative”.Levelofeducation iscodedfrom1=“Nohighschool” to 6 = “Post-graduate”. Party (Democrat) is coded 0 for Republicans, 1 forDemocrats.Finally,PoliticalInterestiscodedforthefrequencywithwhichtherespondentreadsthepoliticalnewsandiscoded1for“Hardlyatall”to4for“MostoftheTime”.Keyfeatures of respondents that could influence their reaction to the experiment are evenlydistributedamongthetreatmentconditionsbytherandomization. As is clear from Figure A1, features that could affect respondent reactions to theexperiment are randomly distributed across treatment conditions. In all treatment

Gender (female) Ideology (lib to cons)

Level of Education Party (Democrat)

Political Interest

0.00

0.25

0.50

0.75

1.00

1

2

3

4

5

2

4

6

0.00

0.25

0.50

0.75

1.00

1

2

3

4

C20 T20Treatment Condition

valu

e


45

conditions,inalmostallpanels,thedistributionofresponsesisclosetoidentical.Wecanbeconfident that the randomizationdistributedpotential confounding factors evenlyacrossthetreatmentandcontrolgroups,suchthatcomparisonsbetweenthemwillgiveusinsightintohowvotersreacttoinformationaboutcandidatecampaignfinancetransparency.Thetwopossiblevisualexceptions,political interestandideology,arestatistically identicalat0.23 and 0.25 standard deviations, respectively. They are not statistically different (p =0.36andp=0.23intwo-tailedt-test)(HartmanandHidalgo2017).


46

FigureA2presentsrandomizationchecksontheT10andC10data.

Figure A2: Distribution of respondents’ answers to key survey questions on the verticalaxis.Treatment condition ison thehorizontal axis.C10andT10= respondents told thatConleywas10points from Johnson.Violinplotdisplays thedensityof responsesateachvalue,smoothingacrosslevels.Genderiscoded1forfemale,0formale.Ideologyiscoded1=“very liberal” to5=“veryconservative”.Levelofeducation iscodedfrom1=“Nohighschool” to 6 = “Post-graduate”. Party (Democrat) is coded 0 for Republicans, 1 forDemocrats.Finally,PoliticalInterestiscodedforthefrequencywithwhichtherespondentreadsthepoliticalnewsandiscoded1for“Hardlyatall”to4for“MostoftheTime”.Keyfeatures of respondents that could influence their reaction to the experiment are evenlydistributedamongthetreatmentconditionsbytherandomization.

Gender (female) Ideology (lib to cons)

Level of Education Party (Democrat)

Political Interest

0.00

0.25

0.50

0.75

1.00

1

2

3

4

5

2

4

6

0.00

0.25

0.50

0.75

1.00

1

2

3

4

C10 T10Treatment Condition

valu

e


47

As in Figure A1, Figure A2 shows that features that could affect respondentreactions to the experiment are randomlydistributed across treatment conditions. In alltreatmentconditions,inalmostallpanels,thedistributionofresponsesisclosetoidentical.We can be confident that the randomization distributed potential confounding factorsevenlyacrossthetreatmentandcontrolgroups,suchthatcomparisonsbetweenthemwillgive us insight into how voters react to information about candidate campaign financetransparency.30 FiguresA3–A5replicate the findings inFigures7-9 in thepaper for theC10andT10data.

FigureA3:Thisfigurepresentsthedifferencesinthemeanfavorabilityandprobablevotechoiceamongthetreatmentgroup,minusthemeanfavorabilityandprobablevotechoiceamongthecontrolgroup,forJohnson(toptworows)andConley(middletworows).Theneteffectispresentedinthebottomtworows.Theverticallineisat0,wheretheestimateswould be if respondents were not affected by seeing an evaluation of the candidates’campaignfinance information.Thehorizontalaxismeasurespercentagepoints.Theopencirclesshowthedifferenceinmeans,andthep-valuesreportedareforatwo-sidedt-test.The black lines show a 95% confidence interval. All findings are statistically significant(andthesmallertheconfidenceinterval,themoreprecisetheestimate),exceptforConley’sfavorabilityboost,whichisnotstatisticallydistinguishablefromzero.As in themain text, Ialsoprovide themeantreatmentandcontrolvaluesunderlying theestimates. Johnson Favorability is 56.85 in control and drops to 45.46 in treatment.Johnson Probable vote is 56.25 in control and drops to 45.75 in treatment. Conleyfavorability is 57.00 in control and rises to 58.85 in treatment. Conley probable vote is

30 The possible visual exception is ideology. Respondent ideologies in C10 and T10 are statistically identical at 0.19 standard deviations. They are not statistically different (p = 0.64 in two-tailed t-test) (Hartman and Hidalgo 2017).

−20

−10 0 10 20

●

●

●

●

●

●Johnson − Favorability (T−C)

Johnson − Probable Vote (T−C)

Conley − Favorability (T−C)

Conley − Probable Vote (T−C)

Net Favorability (J−C)

Net Vote (J−C)

0.0000

0.0000

0.31

0.01

0.0000

0.0000

P ValueOutcome


48

57.55incontrolandrisesto62.75intreatment.Thenetfavorabilityis-0.67incontroland-13.79intreatment.Netvoteis-1.3incontroland-17.00intreatment.

Figure A4: This figure presents the differences in themean trustworthiness evaluationsamong the treatment group, minus the mean trustworthiness evaluations among thecontrolgroup,forJohnson(toprow)andConley(middlerow).Theneteffectispresentedin the bottom two row. That quantity is the effect for Conley (Treatment grouptrustworthiness–Controlgroup trustworthiness)subtracted fromtheeffect for Johnson.Theverticallineisat0,wheretheestimateswouldbeifrespondentswerenotaffectedbyseeinganevaluationofthecandidates’campaignfinanceinformation.Thehorizontalaxismeasures percentage points. As before, the open circles show the estimate, which ismeasuredasthedifferenceinmeansbetweenthecontrolandtreatmentgroups.P-valuesreportedareforatwo-sidedt-test.Theblacklinesshowa95%confidenceintervalaroundthe estimate. All findings are statistically significant, and the smaller the confidenceinterval,themoreprecisetheestimate.As in themain text, Ialsoprovide themeantreatmentandcontrolvaluesunderlying theestimates. JohnsonTrustworthiness is 56.52 in control and drops to 46.47 in treatment.ConleyTrustworthinessis56.5incontrol,risingto62.34intreatment.Nettrustis0.004incontrol,droppingto-15.87intreatment.

−20

−10 0 10 20

●

●

●Johnson − Trustworthiness (T−C)

Conley − Trustworthiness (T−C)

Net Trustworthiness (J−C)

0.0000

0.0000

0.0000

P ValueOutcome


49

FigureA5:Averagetreatmentandcontrolgroupevaluationsonfavorability(toprow),votechoice (center row) and trustworthiness (bottom row), for Johnson (left column) andConley(rightcolumn).Groupresponsesareshownbyideologyfrom1=veryliberalto7=very conservative, with 4 being “middle of the road”. Control group evaluations arepresented as the tail of the arrows. Treatment group evaluations are presented as theheadsofthearrows.

Ideology

Favo

rabi

lity −

John

son

1 2 3 4 5 6 7

40

50

60

70

80

Ideology

Favo

rabi

lity −

Con

ley

1 2 3 4 5 6 7

40

50

60

70

80

Ideology

Vot

e C

hoic

e −

John

son

1 2 3 4 5 6 7

40

50

60

70

80

Ideology

Vot

e C

hoic

e −

Con

ley

1 2 3 4 5 6 7

40

50

60

70

80

Ideology

Tru

stw

orth

ines

s −

Joh

nso

n

1 2 3 4 5 6 7

40

50

60

70

80

Ideology

Tru

stw

orth

ines

s −

Con

ley

1 2 3 4 5 6 7

40

50

60

70

80


University of Southern California Law School€¦ · Show Me the Money: “Dark Money” and the...

Documents

Transcript of University of Southern California Law School€¦ · Show Me the Money: “Dark Money” and the...