Elite and Mass Support for Foreign Aid versus Government Programs

ELITE AND MASS SUPPORT FOR FOREIGN AID VERSUS GOVERNMENT PROGRAMS: EXPERIMENTAL EVIDENCE FROM UGANDA MichaelG.Findley
DepartmentofGovernment
UniversityofTexasatAustin
[email protected]
AdamS.Harris
DepartmentofPoliticalScience
NewYorkUniversity
[email protected]
HelenV.Milner
DepartmentofPolitics
PrincetonUniversity
[email protected]
DanielNielson
DepartmentofPoliticalScience
BrighamYoungUniversity
[email protected]
January31,2015
Abstract Wecompareeliteandmasssupportforforeignaidversusgovernmentspendingondevelopment
projects.WeperformedanexperimentandsurveyonmembersoftheUgandannationalparliament
andaparallelstudyonarepresentativesampleofroughly3,600Ugandancitizens.Fortwoactual
aidprojectsinthepipeline,werandomlyassignedexposuretotheprojects’differentfunders.Sig‐
nificant treatment effects on attitudes and behaviors reveal that members of parliament support
government programs over foreign aid, whereas citizens prefer aid over government. Using sub‐
group analysis, we explore several mechanisms that might explain this pattern: partisanship, co‐
ethnic bias, nationalism, incumbency, a foreign media effect, and corruption. Effects are most ap‐
parent for members of parliament and citizens who perceive significant government corruption,
suggestingthatcitizensseeforeignaidasanescapefromcorruption,butelitesperceivemoreave‐
nuesforthecaptureofgovernmentresourcescomparedtoaid.
Prepared for the 8th annual Political Economy of International Organizations (PEIO) con‐
ference,HertieSchoolofGovernance,Berlin,Germany,Feb.12‐14,2015.
1
Introduction Academic disputes rage over what foreign aid allows politicians to do. For some
scholars,aidenablespoliticalelitestobuyvotes,buildmilitariesforrepression,andenrich
themselvesthroughcorruption(Svensson,2000;AlesinaandWeder,2002;Bräutigamand
Knack,2004;Easterly,2006;Morrison,2009;Moyo,2009;Morrison,2012).Aidmaythus
underminecitizens’abilitytoholdpoliticalelitesaccountableforhowpublicrevenuesare
spent(Ross,2004;Knack,2009;Morrison,2009;Moyo,2009).
Alternatively,othersarguethatforeignaidcanbypasscorruptpoliticiansandthus
deliverneededpublicgoodsdirectlytorecipientsevenasitpromotescivilsocietyactors
thatcandemandbettergovernance(Finkeletal.,2007;WrightandWinters,2010;Bermeo,
2014; McLean and Schneider, 2014; Mosley, forthcoming). Additionally, much aid targets
government capacity building and might contribute to the development of better‐
functioninginstitutionsandthusconstrainpoliticianstocleanupcorruptionandmisman‐
agement(Riddell,2007;BaserandMorgan,2008).
Missingfromthedebate,however,aretwoimportantelements.First,greateratten‐
tionneedstobefocusedonthesupportforaidofbothpoliticalelitesandcitizensinrecipi‐
entcountries,whoseoutlookislittleunderstooddespitebeingaid’scriticalintermediaries
andultimatebeneficiaries,respectively.Second,aiddoesnotoccurinaninstitutionalvacu‐
um, so we need a meaningful baseline to which to compare it, and government projects
providethemostrelevantalternative.Whatmightbeinferredfromtheoreticalmodelsand
cross‐national statistical studies based on observational evidence may be different from
therealityontheground.Mosttheoriesshareexpectationsaboutdomesticelites,arguing
2
thattheyplayasignificantroleinshapinghowforeignaidaffectstheircountry.Yet,toour
knowledge,nodirectandsystematicevidencehasbeengatheredthatemployspoliticians
asrespondentsinstudieswheretheyreflectonthedispositionofaid.Studiesofrecipient
citizens’supportforaidarelikewiserare.Moreover,thecurrentevidencelacksidentifica‐
tionofcausaleffectsthatmightbebettersuppliedbyexperimentalmethodsthatcompare
attitudinal and behavioral support for aid versus government spending within and be‐
tweenpoliticiansandcitizens.
Wereporttwoparallelexperimentsperformedin2012inUgandathatcontrastelite
andcitizensupportfordevelopmentprojectsintreatmentconditionsattributedtoforeign
donorscomparedtoidenticalprojectsinacontrolconditioninwhichnodonorwasexplic‐
itly mentioned and that most sampled Ugandans took to be the domestic government. A
sizableminorityofsubjectsbelievedtheunnamedfunderinthecontrolconditionwasac‐
tuallyaforeigndonor,butthisworksinfavorofthenullhypothesisofnosignificantdiffer‐
ence between treatment conditions and control. The differences reported below thus
understateelites’andcitizens’contrastingpreferencesforaidversusgovernmentprojects
giventheinformationwehave,aresultweshowindetailintherobustnesssection.
While our experiments cannot settle the broader debate on aid effectiveness, they
can provide direct causal evidence about the effects of different sources of development
fundingontheattitudesandbehaviorofelitesandcitizens.Theseattitudesandbehaviors
areimportantandworthyofstudy.Buttheymayalsoreflectatleastindirectlyonanim‐
portantlinkinthechainconnectingrevenuesourcestogovernanceandpolicyoutcomes.
3
In all, we were able to conduct nearly hour‐long interviews with more than two
thirds of the sitting Ninth Ugandan Parliament (276 out of 375 members of parliament).
Wealsoconductedthestudywith78formerMPsfromtheEighthParliament.(Totalcur‐
rentandformerMPssurveyedis354.)Moreover,werandomlysampledanationallyrepre‐
sentative group of roughly 3,600 Ugandan citizens for comparisons. The parallel
experiments provided an opportunity for each set of subjects to demonstrate individual‐
levelsupportforforeignaidorgovernmentfundsthroughbehavioralactionsthatimposed
personalcostsaswellasthroughresponsestoattitudinalsurveyquestions.
Interestingly,wefindthatthemembersofparliamentaresignificantlymorelikely
to support projects in the control condition that most took to be the government rather
than treatment projects identifying foreign donors. However, citizens are significantly
morelikelytosupportforeigndonorsoverthecontrolcondition,preciselytheoppositeof
theelites.Effectsizesaregenerallymodestandapproachaceiling,butthedifferencesare
significantandrobustacrossavarietyofspecifications.Thesedifferencesbetweenthepub‐
licandeliteinUgandaformaninterestingandnovelpuzzle.
Employing subgroup analysis, we explore the leading candidates for the mecha‐
nisms that might underlie these differences: partisanship, ethnicity, nationalism, incum‐
bency,aforeignmediaeffect,andcorruption.Ingeneral,thesubgroupresultsprovidenull
or weak evidence for all of the possible mechanisms, except one: the corruption mecha‐
nism.MPswhoperceivegreatergovernmentcorruptionareespeciallylikelytopreferthe
governmentprojectsoverforeignaid,whereascitizensperceivinggovernmentcorruption
were significantly more likely to support the aid projects. Subjects who did not perceive
4
corruptionappearedindifferentbetweenaidandgovernmentprojects.Effectsizesforcor‐
ruptionsubgroupanalyseswereconsiderablylargerthaninthemainanalysis.
Thisstudymakesatleastthreecontributionsthatdirectlyaddresskeygapsinour
understandingofthepoliticaleconomyofdevelopment.Itistheonlyexperimentalstudy
randomlyassigningforeignversusdomesticsourcesofprojectfundingandthusabletore‐
vealcausaleffectsonboththeattitudesandbehaviorofrecipients.Second,itisthefirstde‐
tailed study of foreign aid versus domestic government funding to employ members of
parliamentinarecipientgovernmentasrespondents.Third,itprovidesaninvestigationof
sixalternativemechanismsthatmightexplainthesignificantdifferencesbetweeneliteand
masssupportfordevelopmentprojectsfromcompetingsourcesoffunding.
In the next section, we review the debate over how political elites might use aid
comparedtogovernmentresourcestopromotetheirownendsontheonehandorserve
publicgoalsontheother,andwefurtherdelineatetheempiricalexpectationsoftheoppos‐
ingarguments.Wethendiscusstheevidenceforthecompetingargumentsdrawnfromour
experiment on members of parliament and citizens. The results tell an interesting story
aboutmassandelitebeliefs,preferences,andbehavior.Inparticulartheysuggestthat–at
leastinthemindsofthosewithdirectexperience–aidmaybelesssusceptibletopolitical
capturethangovernmentresources.Inthenextsection,welookatthemainalternativeex‐
planationsforourfindingsandevaluatewhetherourdatasupporttheseinterpretations.In
therobustnesssection,wediscusstwoissuesthatarisefromourexperimentalprocedures
andshowthatneitherofthemshouldinvalidateourresults.Weconcludebyreturningto
the main themes about attitudes toward aid relative to government funding of develop‐
5
ment projects. And we note how bringing together the comparative politics research on
corruptionandclientelismwiththeliteratureonforeignaidmaybefruitful.
The Debate Strongargumentsexistonbothsidesofthedebateoverhowelitesmightemployaid
versusgovernmentrevenuestopromotetheirpoliticalcareers.Ontheonehand,analysts
have likened aid to natural resources in the way they “curse” developing countries with
conflict, autocracy, and poor governance (Morrison, 2009; Moyo, 2009, p. 59; Morrison,
2012). Prominent studies have held that external sources of money, such as natural re‐
sourcesandaid,enablepoliticianstoentrenchthemselvesratherthanbeheldaccountable
for the use of tax revenues (Bräutigam, 2000; Smith, 2008; Morrison, 2009; Gervasoni,
2010).Citizens,itisclaimed,demandmoreaccountabilityandbetteroutcomeswhentheir
taxdollarsareatplay(Schumpeter,1954;NorthandWeingast,1989;Ross,2004).
Prominentworkimpliesthattheremightexistadifferencebetweenpoliticalelites
and the mass public in attitudes toward aid. For example, Bueno de Mesquita and Smith
(2007,2009)arguethatpoliticalelitesinthedonorandrecipientcountriesarethemajor
beneficiariesofaid,soaidfunctionstoincreasethepoliticallongevityofelitesinrecipient
countries.Theyalsoarguethatthebiggestlosersfromaidarethemasspublicsinpoorna‐
tionswhoareforcedtoacceptpolicyconcessionstheyopposeandtoenduremorecorrup‐
tionfromtheirownleaders(BuenodeMesquitaandSmith,2009,p.311).
6
Alternatively,otherscholarsarguethatdonorsunderstandmanyofthechallenges
they face in developing countries and try to act strategically to advance their goals.1For‐
eignaid–especiallyifitismonitored,hasconditionsattachedtoit,orislessfungible–may
servemoreasapublicgoodthatpoliticiansstruggletodiverttothemselvesandtheirallies
(MavrotasandOuattara,2006).Often,aidexplicitlytargetsimprovementsingovernment
capacity.Forexample,Bermeo(2014,p.4)presentsfindingsthataiddoesnotinhibitde‐
mocratization,notingthat“aidisnotoil.Foreignaidcomesfromdonorsanddonorshave
preferences.Theyalsohavetoolstoprovideaheterogeneousbasketofaidwhichcanlook
verydifferentfromtherevenuestreamattachedtoastate‐ownedenterprise.”
Evenif aidfails to strengtheninstitutions orbuild capacity directly, an alternative
literature claims that aid – especially for democracy and governance – can find ways
aroundincumbentpoliticians.Dietrich(2013)showsthatdonorsstrategicallydecidehow
muchaidtoprovidedirectlytogovernmentsandhowmuchtobypassthem.This“circum‐
vention” aid can support opposition parties, watchdog media, and civil society organiza‐
tionsthatmighteffectivelydemandmoreaccountability.Indeed,someevidencesuggests
thataidcontributessignificantlytodemocratizationinrecipientgovernments(Finkeletal.,
2007;ScottandSteele,2011;Aronowetal.,2012).
Thedebateremainsunresolved.Extantstudiesaddresstheproblemfrommultiple
methodologicalperspectives,butthusfarhavenottakenthestudydirectlytothepolitical
elites who may use (or abuse) foreign aid nor to the masses who may benefit or suffer.
1Bermeo,2010;Bermeo,2011.showsthatdonorsgoalsseemtohavechangedsincetheendoftheColdWar
andthattheyaremorefocusednowondevelopmentanddemocracypromotion.Shedemonstratesthatdo‐
norspursue“strategicdevelopment,”targetingandstrategicallyallocatingdifferenttypesofaidtodifferent
typesofcountries.
7
Whilenomethodologyisperfect,andourexperimentalapproachdoesnotresolvetheaid
effectiveness debate, focused experiments using elites and citizens as subjects may con‐
tributetothisdebatebyprovidingevidenceaboutwhosupportsforeignaidorgovernment
funding for development projects, and their reasons for doing so. Support for different
sourcesofrevenueamongmassesandelites–particularlytheirbehavioralsupport–ought
toshedsomelight,evenifindirect,onakeylinkinthecausalchainconnectingfundingto
outcomes.
Iftheliteratureclaimingthataidenablescorruptpoliticianstoentrenchthemselves
is correct, politicians should prefer foreign aid over government‐funded programs since
theywillbefreertousetheaidfortheirownpurposescomparedtotaxrevenuesforwhich
they must answer to the public. If aid is more accountable to foreign agencies, however,
anddonorshavethecapacitytoaudit,constrain,andpunishpoliticianswhotrytouseaid
fortheirownpoliticalends,thenweexpectadifferentresult:elitesshouldprefergovern‐
mentprogramstoforeignaidprojects.MPsshouldbeespeciallykeentouseavailablere‐
sources to maintain their privileged position through clientelism and should therefore
preferthefundingsourcewithfewerconstraints(seevandeWalle,2003,p.313).2
Preferencesofcitizensshouldmoveintheoppositedirection.Ifcitizensbelievethey
canbetterholdtheirrepresentativesaccountablefortaxrevenue,theyshouldmorestrong‐
lysupportgovernmentspendingoveraid.Butifcitizensperceiveforeigndonorsandaidas
2Similarly,whatRothchild,1986.called“hegemonialexchange”andBayart,1993.“reciprocalassimilationof
elites,” clientelism pervades Africa since political stability there has often been constructed by using state
resourcesto forge alliances across different social elites, often in the formofovertpower‐sharingarrange‐
mentsvandeWalle,2003..
8
betteratproducingpublicgoodsthangovernmentspending,theyshouldpreferaid.3Along
nearlyeverymajorbywayinUganda,asinmanydevelopingcountries,signstyingprojects
toforeignordomesticdonorscrowdtheroadsideandthereforewouldmaketheconnec‐
tionofaidtooutcomespossible.Ofcourse,citizens’perceptionsmaybemistakenaboutthe
effectsofaid,andweareopentointerpretingourdatainthislight.However,evenifciti‐
zensaremisinformed,politicians,wewouldargue,haveabettergraspoftheeffectsofdif‐
ferentsourcesofrevenueontheirpoliticalcareers.Andsocombiningstudiesofthetwo
subjectpoolsandcomparingtheirattitudesandbehaviorstowardthesameexperimental
conditionsisparticularlyusefulandnovel.
Ourexperimentscannotresolvetheaideffectivenessdebate.However,wedocon‐
tendthattheresults–especiallyfortheMPexperiment–mayreflectindirectlyonkeylinks
in the causal chain connecting aid to outcomes. Nevertheless, we stress that the disposi‐
tions–andespeciallythebehavioralactionstaken–towardaidbycitizensandespecially
MPsareinterestingintheownright.Citizensaretheultimatebeneficiariesofaid,andthey
notoriously suffer from the results of “broken feedback loop” underscoring the missing
communicationbetweendonorsandrecipients(Martensetal.2002).Thebrokenfeedback
looprequiresthatcitizensbeaskedaboutaid,whichiswhatwedohere.
Moreover,membersofparliamentofteninfluencehowaidisdistributedinrecipient
countries. Learning MPs’ disposition and behavior toward aid, especially as it compares
with government funding, appears important to understanding how aid might be chan‐
3Wealsoinvestigatedpreferencesoflocalgovernmentofficials(similartostateandcitylevelofficialsinthe
US),butfoundnosignificantpreferenceforgovernmentoraid‐fundedprojects.Thisislikelythecasebecause
theseofficialsmostlyreceivefundsfromthecentralgovernmentthatareearmarkedalready.
9
neledthroughdomesticinstitutions.Itmayalsoreflectonthewaysthataidmightbedif‐
ferentfromgovernmentmoneyingarneringparliamentarysupportforprogramsthatfor‐
eigngovernmentsandinternationalorganizationswantrecipientstopursue.
The Ugandan Context Uganda currently has a semi‐authoritarian regime in which the government of
YoweriMuseveni’sNationalResistanceMovement(NRM)hasretainedpowerfornearly30
years(vandeWalle,2007;Greene,2010).In2006,Ugandabeganholdingmultipartyelec‐
tions;yettheyhavenotbeenfullyfreeandfair(Cheibubetal.,2010;HydeandMarinov,
2012). Scholars describe the party’s ruling methods as relying heavily on patronage and
clientelismtoretainitscontrol(vandeWalle,2003,2007;Muhumuza,2009;Green,2010;
Tripp,2010).Asonerecentstudypointsout,“InUganda,therulingNRMhasestablished
patronagenetworksthroughoutthecountrythroughtheuseoflocalgovernment.Thecivil
serviceisanothersuchnetworkofpatronage,andperhapsthemostimportantisthemili‐
tary. These clientelist networks, while consolidating key sources of support, at the same
timeunderminegovernanceanderodetheviabilityofinstitutionsandleadership”(Tripp,
2010,p.25).Asof2010,Ugandaranksonthehigherendofcorruptionscales,scoringin
the 72nd percentile (129th out of 178) onTransparency International’s CorruptionPercep‐
tionsIndex.
Partisanship and ethnic attachments have also played important roles in Ugandan
politics at both the citizen and elite levels, revolving around the NRM‐opposition split.
However,recently,theNRMhasfacedimportantoppositionnotjustfromoppositionpar‐
10
ties,whoarefracturedandcurrentlyonlyhold16%oftheseatsinparliament,butmainly
fromwithintheNRM’sownranks.Intherun‐uptothe2011parliamentaryelections,for
example,theNRMprimarieswerehotlycontested(inmanyinstancesmorecontestedthan
thegeneralelections).Ballotboxeswerestuffedandelectionsriggedtoensurethatparty‐
leaderfavoriteswontheNRMpartynomination(Malinga,2010).Thisledtowidediscon‐
tentamongthelosersoftheprimaries.Manyoftheseindividualsthenranasindependents
in2011,andnowthecurrentparliamenthasmoreindependentsthananysingleopposi‐
tionparty,whichmakestheroleofpartisanshipmorecomplicatedandlesspredictable.4
EthnicityisalsoimportantintheUgandanpoliticalcontext.Critically,sharedethnic‐
itywiththechiefexecutiveofthecountryhasimportantpoliticalanddevelopmentalcon‐
sequencesinthattheleader’sco‐ethnicsarelikelytobenefit(FranckandRainer,2012).In
Uganda,itiscommonlyunderstoodthatwhenNorthernerssuchasPresidentMiltonObote
wereinpower,theNorthernregionofthecountryreceivedthemostbenefitsintermsof
development, government employment, and other material goods. Now, under President
Museveni,manycitizensarguethattheWesternregion,especiallythoseareasinwhichhis
fellowMuyankolearedominant,receivethenewroads,schools,andclinics.Whetherornot
thisisinfacttrue,citizenstendtooperateundertheseassumptions(seePosner,2005).
4ThecurrentUgandanparliamenthas375membersrepresenting7politicalparties:238ConstituencyMPs,
112WomanMPs,10UgandanPeople’sDefenseForce(UPDF)representative,and5representativesforeach
ofthefollowingspecialinterestgroups:peoplewithdisabilities(PWD),workers,andyouth.Thevastmajority
ofMPsareelectedunderpluralityrulesinsingle‐memberconstituencies.Eachdistrictelectsonefemalerep‐
resentative and eachconstituency (usually two or three constituencies are contained in one district) elects
oneConstituencyMP.Elevenex‐officiomembersareappointedinaddition,andeachofthefivegeographical
regionselectsonePWD,Worker,andYouthrepresentativeandtwoUPDFrepresentatives.
11
Ugandaprovidesausefulsettingfortheexperimentbecauseitreceivessubstantial
amounts of foreign development assistance. Since the 1990s, aid including off‐budget
sources equals approximately 70 percent of government expenditures. Moreover, aid en‐
compassedabout15percentoftotalGDPformuchofthatperiod,thoughthesharehasde‐
clinedtosomeextentinthelastfewyearsaseconomicgrowthhasincreased.Somebasic
informationgarneredfromoursurveyconfirmedthatcitizenshadgeneralawarenessboth
of foreign aid and their parliamentary representatives. More than two thirds of subjects
knewthatmorethan30percentoftheUgandanbudgetcomesfromforeignaid;thevast
majority(66percent)couldnameboththeirConstituencymemberofparliamentandDis‐
trictWomanmemberofparliament;andthemajorityofsubjectswereawareofforeignaid
flowingtotheirlocalareas.UgandaisalsotypicalofAfricancountriesintermsofitsde‐
mocratizationprocesses,currentlevelofdemocratization,andexecutivedominance(Brat‐
tonandvandeWalle1997,ResnickandvandeWalle2013).Likewise,muchlikemanyof
the countries across the continent, the Ugandan parliament sees quite competitive elec‐
tionsandwhileitismuchweakerthantheexecutiveitismuchmorethansimplyarubber
stampandisavenueforimportantandlivelydebates(HumphreysandWeinstein2013).If
we consider Uganda’s level of democracy (Polity IV score) and degree of aid dependence
(World Bank’s World Development Indicator of net official development assistance per
capita),UgandaisverysimilartoEthiopia,Guinea‐Bissau,Togo,Chad,andtheCentralAfri‐
canRepublic.Whiletherearevariousotherconsiderationssuchasethnicdiversity,coloni‐
alhistory,levelsofeconomicdevelopment,etc.similarityonthesetwokeymeasuresare
importantforunderstandingtowhichcasestheseresultsmostlikelygeneralize.
12
ResearchDesign
Toinvestigatecompetingexpectationsregardingeliteandmasspreferencesforfor‐
eign aid compared to a control condition implying government programs, we conducted
twodifferentexperimentsinthefield,eachwithcompanionsurveys.First,wecarriedout
anexperimentonaconveniencesampleof276ofthe375Membersofthe9thUgandanPar‐
liament(thesittinglegislature)and78formerMPsfromthe8thParliament(totalcurrent
andformerMPssurveyedis354).AlthoughwesampledMPsbyconvenience,thedistribu‐
tionisstrikinglysimilartotheactualparliamentatthattime,whichwediscussbelow(See
Table1).Second,weconductedanationallyrepresentativeexperimentonnearly3,600cit‐
izens in 42 of Uganda’s 112 districts. We used a clustered random sample for the citizen
surveytoensureregionalandpoliticalrepresentativeness.Bothexperimentsweresimilar,
butnotidentical.TheywereperformedbetweenJuneandOctober2012bylocalUgandan
enumerators.5
[TABLE1ABOUTHERE]
TomaximizethenumberofresponsesintheMPsurvey,weattemptedtoconducta
censusofallcurrentMPsandachieveda72percentresponserate.Inaddition,wealsocon‐
tactedasmanyformerMPsaspossible(fromthepreviousparliament)andobtaineda55
5ItispossiblethatsubjectsbelievedthelocalUgandanenumeratorsrepresentedaforeignaiddonororthe
governmentratherthanacademicresearchers.Indeed,datafromtheAfrobarometersuggeststhatmostre‐
spondentsthinkthegovernmentistheonedoingsurveys.WeexaminedtheAfrobarometerquestionsabout
subjects’perceptionsofwhosenttheenumerators.Inthe2012roundofUgandaAB(Round5),56%ofre‐
spondentsthoughtthegovernmentsenttheinterviewer.Inthe2008roundfor20differentcountriespooled
together,58%ofrespondentsthoughtthegovernmentsenttheinterviewer.Thisshouldbiasagainstthere‐
sultswefind.
13
percentresponserate.6Whilekeyaspectsoftheexperimentalinstrumentswereidentical
foreachpopulationgroupinordertofacilitatecomparisons,thecitizensurveywaslength‐
ierandthusgeneratedricherinformation.
Thesamplesofrespondentsreflecttheunderlyingpopulationswell,andassignment
to treatment conditions is not predicted by available observables, providing evidence of
randomassignment.FortheMPsurvey,wedonothavedataonindividualMPcharacteris‐
ticssuchasreligionoreducationlevels.7Table1,however,presentsdescriptivestatistics
from our sample and from Parliament as a whole for gender, party, region and MP type,
whichgenerallymatchesthe9thParliamentasawhole.ThedistributionofMPsbyregionis
largely representative, though it slightly oversamples those from the Central region and
undersamplesthosefromtheNorthernregion.Andfinally,assignmenttotreatmentcondi‐
tionsamongMPsisnotsignificantlyrelatedtoparty,gender,MPtype,orregion,sothereis
good covariate balance across experimental conditions. For the citizen survey, balancing
andrandomizationproceduresalsoworkedwell.Keyvariables,suchaseducation,gender,
age,party,religion,andregion,werenotsignificantlyrelatedtowhethercitizenswereas‐
signedtoagivenexperimentalcondition.
6TheformerMPresponserateislikelylowerbecausemanyformerMPsarescatteredthroughthecountry
andnotaseasilyaccessibletoourresearchteamwhencomparedtocurrentMPswhoareusuallyinthecapi‐
talofKampala(ourresearchheadquarters).
7GiventimeconstraintsintheMPsurvey,wewereunabletoobtainmuchdemographicdataonMPstocom‐
parewiththemasssample.Beyondthecomparisonsinoutcomedataexploredbelow,wecanidentifysome
comparisons. For example, the MP sample has more men, which is understandable given the parliament is
disproportionatelymaleandtheMPsamplealsohasahigherproportionofNRMthanthemasssample.
14
Interventions The experimental manipulation presented each subject with a randomly assigned
projectdescriptionandarandomlyassignedfunderforthatproject.Thisbetween‐subjects
designisimportantforelicitingcomparisonsbetweengovernmentandforeigndonorpro‐
jectswheredirectcomparisonsmightbetoosensitive.Werandomlyassignedthemanipu‐
lationforactualpipelineprojectsforthcomingfromtheWorldBankanditsco‐financers.As
such,thestudyavoidedactivedeception.Theprojectswereco‐financedbytheWorldBank
andmultipleagencies,whichallowedustomanipulatewhichof themultipledonorswas
presentedtothesubjectsasfunderoftheproject.Wealsorandomlyassignedthetypeof
project: an infrastructure project (electricity) and an education project. Multilateral and
bilateral donors jointly funded the projects. As is true with all World Bank projects, the
Ugandangovernmentwasalsoinvolvedinthefundingandimplementation.MPs,inorder
to increase the number of observations, were presented with and asked to express their
supportinvariouswaysforboththeelectricityandeducationprojectsindividually(andin
randomorder)butonlyonedonor.Citizensreceivedonlyoneofthetwopossibleprojects.
Wechosetheelectricityandeducationprojectsbecausetheyrepresentthetypesof
projects that can be given selectively to constituencies that support politicians. For the
masssurvey,werandomlyassignedthedonorandtheprojecttype.Neitherprojecttypein
themasssurveywassignificantlypreferredovertheotherinthebetween‐subjectsdesign,
whichmayreflectthefactthatbothtypesofprojectsaredesperatelysoughtafterinUgan‐
da.Becausetherewerenosignificantdifferencesbetweenprojecttypes,wefocusourdis‐
cussionsonthedifferenceacrossfundertypes.
15
Our framing question read, “The Electricity Sector Development Project will im‐
provethereliabilityofandincreaseaccesstoelectricity.Onemajoraspectoftheprojectis
toextendelectricitytothosewhodonotyethaveaccesstoit.Theprojectmayrequireyour
communitytoprovidefundingformaintenanceinthefuture.[Thisprojectwillbefunded
bythe{RANDOMLYASSIGNEDFUNDER}.]Howmuchwouldyousupportthisproject?”We
includethetextfortheeducationprojectintheappendix.
Weincludedthesentenceaboutfutureexpenses(“mayrequireyourcommunityto
providefunding…”)toincreasetherespondents’sensethatthisprojectmightcostthemin
themediumandlongtermtosupportit.Giventhataidmaybeperceivedas“freemoney”
whereasgovernmentprogramsmayimplyincreasedtaxes,wewereconcernedthatoffer‐
ingaprojectwithoutanynotedcostsmightleadallsubjectstosupportit.Askepticmight
worrythattheaddedcostconditionisnotsufficienttoovercomeabiastoward“free”re‐
sourcesamongsubjects.Aidmayfeellikeawindfall,butgovernmentprogramsmayseem
to cost something tangible. This is a reasonable concern, and we took some measures to
addressit.Wepresentthesedetailsintherobustnesssectionbelow.
The funding organizations we randomly assigned in the MP experiment were the
WorldBank,theGovernmentoftheUnitedStates,agenericmultilateralinstitution(“anin‐
ternationalorganizationfundedbymanycountries”),agenericbilateralagency(“asingle
foreigncountry”),andNoDonor,inwhichweomittedthesentenceindicatingwhichagency
was funding the project and served as the control condition. In the mass experiment, we
also included the African Development Bank and the Government of China because the
16
largersubjectpoolenabledgreaterpowerandthereforethepossibilityformoretreatment
conditions.8
Inthecaseofthecontrolcondition,weassumedthatrecipientswouldassociatethis
casewithdomesticgovernmentspending.Wementionednothingaboutforeignaidorfor‐
eigndonorsinthisversionofthesurvey.Weelectednottonamethegovernmentexplicitly
forthecitizensurveyoutoffearthatgeneralizedparanoiatowardgovernmentorassocia‐
tionswiththerulingpartymightbiasresponses.9WedidthesamefortheMPstoavoidso‐
cialdesirabilitybias(i.e.,governmentMPsmightfeelthey shouldsupportprojectsbythe
government,andthuswhenthegovernmentisexplicitlynamed,wewouldreceiveinaccu‐
rateresponses).
Althoughwemadethisdesignchoiceingoodfaithatthetime,inretrospectthisde‐
signchoicemaynothavebeenoptimal:explicitidentificationoftheUgandangovernment
wouldhavepresentedalessambiguouscontrolcondition.Butaswedescribebelow,this
design choice actually works in favor of the null hypothesis of no difference between
treatmentandcontrol;andthereforeourresultslikelyunderstatethefullextentoftreat‐
menteffects.Moreover,giventheinformationwehave,weshowthatourresultsholdeven
8Inadditiontotestingthetreatmenteffectofreceivinganaiddonorrelativetothegovernmentcontrol,we
alsotestedtheeffectofindividualdonorsacrossgroups.BecauseelitesdidnotreceivetheAfricanDevelop‐
mentBankandGovernmentofChinatreatments,weestimateddifferenceinmeansteststodetecttheeffect
ofthevarioustreatmentsrelativetothecontrolandtheothertreatmentconditions.Acrossallgroups,there
isneveraconsistentlysignificanteffectforanyoftheindividualdonors.TheseresultsarereportedinAppen‐
dixTablesA1andA2.
9Concernforbiasedresponsesoutoffearaboutthegovernmentseemsfairlyreasonableinanon‐democratic
contextlikeUganda.Inround5ofAfrobarometerdonein2012,thesameyearasourstudy,50%ofthere‐
spondents said they did not feel completely free to say whatever they believe, and 1/3 said they felt some
kindofpressureaboutwhomtovotefor.Moreover,63%admittedfearofbeingintimidatedinelectioncam‐
paigns.
17
ifsomesubjectsmisinterpretedthecontrolcondition;seethedetaileddiscussioninthero‐
bustnesssectionbelow.
Further, our intervention focuseson one type of aid: project aid.Thus, our results
may not apply to general budget support. However, we chose to focus on project aid be‐
causeitisthemostcommontype,itconstitutestheoverwhelmingmonetaryshare,anditis
themostvisibletocitizensandthuswouldmaximizeourabilitytoobtaininformedprefer‐
encesregardingaid.AccordingtotheAidDatainformationbase,whichisthelargestreposi‐
toryofaidstatistics,between2000and2012Ugandareceived157budgetsupportgrants
andloanssummingto$3.2billion.Overthesameperiod,thecountrywashostto16,019
aidprojectssummingto$24.5billionintotalaid.Thissuggeststhatbudget‐supportaidin
Uganda constitutes 1 percent of the count but 13 percent of total Ugandan aid. This is
roughlyonparwiththerestofSub‐SaharanAfrica,whichreceived3,811budgetsupport
grantsandloansfor$57.5billioninrelationto352,839projectsthattotaled$615billion.
Budget support in the region thus comprised 1 percent of the count but 9 percent of the
money(Tierneyetal.2011).
Weacknowledgethatprojectaidandbudget‐supportaidmighthavedifferentpolit‐
icaleffects.Indeed,givenourfocusonprojectaid,pastfindingswouldsuggestthatourre‐
sults regarding corruption are actually likely to be conservative. The findings of Tripp
(2013)andGazibo(2013)inTanzaniaandBenin,respectively,suggestthatbudgetsupport
aidismorecorruptiblethanprojectaid.Infact,perceptionsofincreasedcorruptionhave
leddonorstoreducebudgetsupportinBeninandUgandainfavorofprojectaid.
18
Outcomes Eachsurveyaccompanyingtheexperimentaskedavarietyofdemographic,political,
andaid‐relatedquestions.Toavoidanyprimingeffects,weposedallaidquestionsafterthe
experimentalportionofthesurvey.Tomeasuretheoutcomeofsupportoroppositionfor
theforeignordomesticfundedprojects,weaskedallrespondentstofirstexpresstheirlev‐
elofsupport,thentoreporttoustheirwillingnesstotellahigherauthority(Partyleader
forMPs,andLocalCouncilofficialforcitizens)oftheirsupport(ornot)fortheproject,their
willingnesstosignapetitionvoicingtheirsupport,andtoactuallysignthepetition.10
MPswereaskedtoexpresstheirwillingnesstocoordinatewithpeersinsupportof
(orinoppositionto)theproject,tellconstituentsabouttheproject,rallylocalsinsupport
of(orinoppositionto)theproject,andsignalettertothePresidentinsupportoforoppo‐
sitiontotheprojects.Citizens,butnotelites,werealsoaskediftheywerewillingtosenda
textmessage(SMS)andtoactuallysendtheSMSinsupport(ornot)oftheproject.
BecausetheMPswerepresentedwithbothprojects,wehavetwoobservationsfor
eachonalloftheseoutcomes,exceptthepetitiontothepresident.EachMPwasaskedto
signasinglepetitionthatreportedtheirlevelofsupportforbothprojectstothePresident,
thuswehaveoneobservationforeachMPonthisoutcome.Thisdesignchoicewasmadeto
reduce the burdenon the MPs and toreduce redundancyof sendingtwonearlyidentical
letterstothepresident.BecausetheMPsreceivedthesamedonoracrossthetwoprojects
10Full text of the petition language is included in the Appendix. Note that the language in the petition only
asksthemtosignwithoutspecifyingaforeigndonororgovernment.Ifadonorwouldhavebeennamedinthe
petition language, then the treatment condition (with a donor named) would not have been comparable to
thecontrolcondition(wherenoonewasnamed).Instead,weoptedtosimplyaskthemtosignapetitionin
supportoropposition.
19
thisshouldnotaffecttheresultswereportherebecausewearecomparingdifferencesin
donors and not sectors (given that there was no meaningful difference between project
types).Thesevariousmeasuresofsupportpresenttherespondentswithvaryinglevelsof
cost(attitudinalvs.behavioralresponses)andwillbeusedasthekeyoutcomevariablesto
gaugesupportforprojectsacrosstreatmentarms.
Results Inthissectionweexaminethesurveyandexperimentaldatatounderstandourre‐
spondents’ views. First, we ask whether MPs are more supportive of government pro‐
gramsorforeignaidprojects.Second,weaskwhetherthemasspublicismoresupportive
ofaidorgovernmentprogramsandthencomparethemtoUgandan MPs.Andfinally,we
considerpossiblemechanismsthatcouldexplaintheoveralltrendsinpreferences.
Differences across groups Table2reportsresultsfromdifference‐in‐meanstestscomparinglevelsofsupport
underalloftheaidtreatmentconditionscomparedtothegovernmentcontrolconditionfor
MPsandmasses.PanelAreportsoutcomesthatweremeasuredforallrespondents(plus
theSMSandPresidentialLetteroutcomesforcitizensandMPs,respectively),andPanelB
reports outcomes for those only measured for MPs. These overall results show that with
onlyoneexception,MPsareconsistentlymoresupportiveofgovernmentprojectsthanfor‐
eignaid.Thisdifferenceinsupportissignificantin3ofthe9outcomes,andtreatmentef‐
20
fectsrangefromlessthan1to12percentagepoints.11SeeFigureA1intheAppendixfora
graphicrepresentationofthetreatmenteffectsandsignificancelevelsforasubsetofout‐
comes.
The results in Table 2 also show that citizens consistently prefer aid over govern‐
mentprojects;thisdifferenceinsupportissignificantin5ofthe6outcomes,includingthe
behavioraloutcomes.Thetreatmenteffectsrangefrom2to4percentagepoints,whichare
notlargebutnonethelesssignificantstatistically.Themodestsubstantivedifferencesmay
resultfromstrongceilingeffectsgiventhattheprojectsareextremelypopularandthere‐
foreclusteredneartheupperboundof100percentsupport.Despitethesesmallsubstan‐
tive differences in the overall analysis, when we subgroup on the corruption/clientelism
mechanismweobservequitelargesubstantivedifferencesbetweenelitesandmasses.As
detailedbelow,foreliteswhobelievegovernmentfundsareusedforcorruptionthetreat‐
ment effects on four of the most important outcomes range from 10 to 19 percentage
points.
[TABLE2ABOUTHERE]
11NotethatTable2reportsintent‐to‐treateffects.Forthemasses,weaskedamanipulationcheckthatallows
ustodeterminethelevelofcompliance.Thoseresultsarequalitativelythesameastheintent‐to‐treateffects.
Moreover,ifwerestricttheanalysistothosesubjectspassingthemanipulationcheck,theresultsshowstrong
differences in favor of masses supporting aid in every possible outcome category. Because of the status of
membersofparliament,weoptednottoaskmanipulationcheckquestions.Wethusmustrelyontheintent‐
to‐treateffectsalone.
21
Why do these differences appear? These results present some interesting and counterintuitive findings. We explore
thesixpossibilitiesthattousappearedthemostprobableinaccountingforthisdifference.
To provide a plausible explanation, a subgroup mechanism needs to differ substantively
betweenthemassesandMPs,toexplainthedifferencesbetweenthetreatmentandcontrol
withineachsubgroup,andmostofalltoaccountforthedifferencesbetweentreatmentand
controlacrosstheelitesandmasses.Sixviablecontendersarepartisanship,ethnicity,na‐
tionalism, government incumbency bias, a bias due to foreign media, and corruption and
clientelism. We present the basic argument for each mechanism and the difference‐in‐
meansresultsofteststhatevaluatethem.WesplitthemassesandMPsalongtherelevant
dimension (e.g., those who do and do not perceive corruption) and compare these sub‐
groups to determine if that dimension produces separation on the post‐experiment out‐
comes.Forthemechanismthatwasbestsupportedbyevidence,thecorruptionandclien‐
telismcondition,wereportresultsbelow;therestoftheresultsarereportedgraphicallyin
theAppendix(FiguresA2‐A13).12
Beforeweproceedtothespecificmechanisms,itisimportanttofirstestablishthat
MPs do not simply think that government projects are more effective or superior in any
way simply because the Ugandan government is involved. If that were the case, then our
12Wedonotdiscusstheforeignmediaeffectindetail,butnotethatmassrespondentscouldbemorelikelyto
voicetheirsupportforaprojectwhenitisassociatedwithaforeigndonorratherthanwhenitisadomestic
source.Citizensandelitesthatpreferforeignmediaaremorelikelytobebiasedinfavorofforeignprojects
becausethoseprojectsarereferencedfavorablyintheforeignmediaWethusseparatedmassesandMPsby
theextenttowhichtheypreferforeignmediaoverUgandanmedia.SeeFiguresA2andA3inappendix.The
resultsgenerallyareinconsistentandweak.
22
storywouldbesimple:MPsprefergovernmentprojectsbecausetheyviewthegovernment,
ofwhichtheyareapart,asapreferablemanagerofaidfundsandprojects.
However,MPsonaveragedonotholdtheviewthatgovernment‐fundedprojectsare
superior.Only32percentofMPsbelievegovernmentfundsaremorelikelytogotothose
mostinneedcomparedtoforeignaidfunds,which59percentbelievegomoretotheneed‐
iest.Inaddition,only34percentofMPsbelievethatgovernmentfundsaremoreeffective
andlesswastefulcomparedtoforeignaidfunds.Only39percentofMPsbelievethatgov‐
ernment‐funded projects better meet the needs of their constituents than do foreign‐
fundedprojects.Andonly31percentbelievethatgovernment‐fundedprogramsaremore
transparent than foreign aid projects. Finally, when asked who they think would be the
mosteffectiveincarryingouttheelectricityoreducationproject,only23percentofMPs
namedtheUgandangovernment.Moreover,roughly80percentofMPsthoughtforeignaid
had a positive effect on the government and their constituents. Therefore, MPs actually
tendtohavelessconfidenceingovernment‐fundedprojectscomparedtoforeignaideven
thoughtheytendtomorereadilysupportgovernmentprograms.Sowhyshouldtheythen
choosetosupportgovernmentprojectsmorethanforeignaid?
Partisanship One might expect that ruling‐party NRM MPs and mass NRM partisans would be
strongsupportersoftheirgovernment’sownprojects,butitisalsopossiblethatNRMMPs
and supporters favor foreign aid because it shows that the government is capable of at‐
tractingfundsfromabroad.OnemightalsoexpectMPsandmasseswhosupporttheoppo‐
sition to favor foreign projects simply because they are not run completely by the NRM.
23
Thus,forpartisanshiptoexplainthedivergentfindingsacrossMPsandvoters,theremust
besignificantdifferencesbetweenNRMandoppositionsupportforgovernmentvs.aidpro‐
jects,andthedifferenceindifferencesmustfavorgovernmentprojectsforMPsandaidpro‐
jectsforcitizens.Inotherwords,NRMMPsmustprefergovernmentmorethanopposition
MPspreferaid,andoppositionsupportersamongthecitizenrymustpreferaidmorethan
NRMsupportersprefergovernment.
Thedifferenceinmeanstestscomparingpreferencesforforeignaidversusgovern‐
mentfundsamongbothMPsandmasseswhoareandarenotmembersoftheNRMarere‐
ported in Figures A2‐A3 in the Appendix. Figure A2 shows that among the public,
oppositionsupportershavestrongerpreferencesforaid‐fundedprojects(in3of6condi‐
tions),whereasoppositionMPshavestrongerpreferencesforgovernmentfundedprojects
(in4of9conditions)althoughthenumberofMPsinthatcellisverysmall.FigureA3shows
thatNRMsupportersinthecitizenryarenotsignificantlymorelikelytopreferonetypeof
project over another. NRM MPs are likewise not significantly more likely to prefer either
typeoffunding.ItispuzzlingthatNRMMPsandmassesdonotsupporttheirowngovern‐
mentprojectsmostofall,butgiventhedivisionswithintheNRMthismaybeunderstanda‐
ble. Overall, partisanship does not seem to explain the main results about differences
betweenelitesandmasses.
Ethnicity AnotherpossibleexplanationforthedivergentpreferencesbetweenMPsandmass‐
esisco‐ethnicidentitywiththepresident.Ontheonehand,ethnicityisoftenunderstoodto
beavehicleforclientelismorpatronage(Posner2005,FrankandRainer2012)andmay
24
thereforebeassociatedwithmassandMPsupportforthegovernment.Accordingtothis
logic, masses and MPs who are of the same ethnicity as the President should prefer gov‐
ernment spending over aid. But this would not present a possible explanation for the di‐
vergentpreferencesthatweactuallyobserveinthedata.Instead,itwouldhavetobethe
case that co‐ethnic masses had a different set of preferences relative to co‐ethnic MPs.
Whilewequestionwhetherthiscouldbethecase,wewanttoconjectureaboutethnicex‐
planationsthatcanaccountforthedivergentpreferences.
Onepossibilitywouldbetotieintothelogicofdescriptiverepresentationbenefits
forvoters(BoboandGilliam,1990;Barretoetal.,2004).Bythisargument,Musveni’sfel‐
low Munyankole citizens13receive sufficient benefits from having one of their own inthe
highestofficeinthecountrysuchthattheydonotnecessarilyexpectmaterialbenefitsfrom
thegovernmentandthereforemightpreferaidprojects.Co‐ethnicMPs,ontheotherhand,
expect greater favor, access to more funds, and cabinet positions from the president be‐
cause they are from the same ethnic group as the president. Thus, these MPs should be
more likely to support government funds because they should be the ones most likely to
directlybenefit.
Ifco‐ethnicityprovidesdescriptivebenefitstothemassesandfinancialbenefitsto
MPs, then co‐ethnic masses should have a stronger preference for aid and co‐ethnic MPs
shouldprefergovernmentfunds.Theresultsofthedifferenceofmeanstestcomparingco‐
ethnicandnon‐co‐ethnicmassesaswellasMPsarereportedinFiguresA4andA5.There‐
sultsshowthatneithernon‐co‐ethnicmassesnorMPshaveaclearandconsistentprefer‐
13Runyankole
is the language that Munyankole speak, which are the President’s ethnic group and language.
25
enceforonefundingsourceoveranother(FigureA4–non‐coethnic),whichatbestpartial‐
lysupportstheargument.However,FigureA5(co‐ethnic)showsthissametrend:neither
co‐ethnicmassesnorMPshaveaclearpreferenceforeithersourceoffunding.Thus,there
isnoclearevidencethatethnicityisdrivingthemaineffects.
Nationalism Relatedtotheethnicityargument,analternativeexplanationfortheresultscouldbe
thatMPsarenationalisticandresentrelyingonoutsidedonorsfordevelopmentinterven‐
tions.Theyshouldthusbeaversetosupportingwhatappearstobecharityfromoutsiders.
Whilesomecitizensmayalsoholdnationalisticviews,theymaybelessnationalisticthan
MPswhoserveinnationaloffice.Wespecificallyquestionwhetherindividualswhofeela
strongsenseofcommitmenttoanationalorstateidentityasopposedtoaparticularethnic
groupviewforeigninvolvementdifferently.14
Followingothersurveys,wemeasuredattachmenttonationvs.ethnicgroupbyask‐
ing individuals to respond to the following: “Let us suppose that you had to choose be‐
tween being Ugandanand being [insert respondent’sethnic identity].” Response options
range from feeling exclusively Ugandan, to mostly Ugandan, to equally Ugandan and a
member of one’ ethnic group, to mostly ethnic, to exclusively ethnic. We then measured
whetherthosewhofeelmoreUgandanbehavedifferentlyfromthosewhofeelgreaterat‐
tachmenttotheirownethnicgroup.Inourdata,relativetothemasses,UgandanMPsre‐
14Nationalismanditsrelationshiptoethnicityareoftendebatedandafulldiscussionisbeyondthescopeof
thispaper.SeefullertreatmentselsewhereincludingCalhoun,1993.andChandra,2006..
26
porthigherlevelsofattachmenttoUgandaasanationalidentitythantheydototheirpar‐
ticularethniccategory.
Forindividualswhodonotconsiderthemselvesnationalist(seeFigureA6),masses
prefer foreign aid whereas MPs prefers government spending, though the results are not
consistentlysignificant.Forthoseindividualswhoconsideredthemselvesmorenationalist
(seeFigureA7),therearenoconsistentpatternsofsupportforaidorgovernmentspend‐
ing.Moreover,therearenoconsistentdifferencesbetweenmassesandMPsamongnation‐
alistrespondents.Theevidenceforanationalistexplanationisthusweakatbest.
Government Incumbency Bias Relatedtothenationalismclaim,MPsmaysimplybemorelikelytoprefergovern‐
mentprogramsbecausetheyarepartofthegovernment.Totestthis,wetookadvantageof
auniqueaspectofourstudy:wesurveyed78formerMPs.Ifbeinginthegovernmentmat‐
ters,thenweshouldseegreaterlevelsofsupportforgovernment‐fundedprojectsamong
currentMPscomparedtoformerMPs.Thedifference‐in‐meansteststhatcompareformer
andcurrentMPsarereportedinFiguresA8andA9.Thesefiguresshowthatthereisalmost
nodifferencebetweencurrentandformerMPs,thuscastingdoubtonthisalternativeclaim
thatactualpresenceingovernmentdrivesthepro‐governmentbias.
Corruption and Clientelism IsthedifferencebetweenMPsandthepublicproducedbydifferentviewsofcorrup‐
tionandclientelism?Evidenceinfavorofthecorruptionandclientelismmechanismwould
indicate that citizens who believe that the government is corrupt and clientelistic would
preferforeignaidprojects.MPsshouldhavetheoppositepreference.Whenpoliticianssee
27
corruption in government it may be a boon to them personally or electorally, and hence
theymayprefergovernmentprojectsbecausetheyprovideaneasierwaytoaccessmoney
fortheirownpersonalgains.
Inthesurvey,weaskedbothMPsandcitizenswhethergovernmentfundsaremost
likely“tobenefitgovernmentofficialsandtheirpoliticalallies”or“helpthosemostinneed”
to capture aspects of both clientelism and corruption (using money to help friends and
themselves).BothareintimatelylinkedconceptsinAfricabecausecorruptionlargelysus‐
tainsclientelism(Szeftel,2000).WeseeaverylargedifferencebetweenthepublicandMPs
in their perceptions of corruption and clientelism: 75 percent of the public believes that
current government leaders take government money to benefit themselves and their
friendsratherthaneveryoneinthecountry,whileonly35percentoftheMPsagreewith
thisstatement.Wethereforeusethisquestiontodividethesampleintothosewhoseegov‐
ernmentfundsasmoresusceptibletocaptureandabuseandthosewhodonot.
Weacknowledgethefactthatthecorruptionquestionmaybesusceptibletosocial
desirabilitybias;however,furtheranalysismitigatestheconcern.First,whileMPswhoare
morecorruptmaybelesslikelytoadmittocorruptioningovernment,onecouldeasilyar‐
guethecontrary.ItmayjustaswellbethecasethatMPswhoareinvolvedincorruption
aremorelikelytoreportthatcorruptionisaprobleminordertopresentanevenstronger
signal that they themselves are not corrupt. So, on average, the social desirability effects
mightplausiblywashout.
However,wecanleveragesomeevidencefromoursurveytoseeifsocialdesirabil‐
ity is at play. If some MPs are more susceptible to social desirability pressures, then we
28
shouldexpectastrongpositivecorrelationbetweenlowresponsestothecorruptionques‐
tionandavarietyofotherquestionsreflectingsociallydesirableresponses,suchasreport‐
ingmorevisitstotheMP’sconstituency,perceivingagoodeconomy,andclaimingbetter
attendance at plenary sessions of parliament. However, answers to these questions are
neverstronglypositivelycorrelatedwithlowcorruptionperceptions.Thereisaweakneg‐
ative correlation (Pearson’s r) between low corruption reports and claiming more days
spentintheMP’sconstituencyeachmonth(‐0.1189).Thereisaweakpositivecorrelation
betweenlowcorruptionreportsandmaintainingthatthenationaleconomyisinatleasta
“good” condition (0.1017). Andfinally thereis a weaknegative correlations between low
corruptionreportsanddeclaringhigherratesofattendanceatplenarysessionsofparlia‐
ment(‐0.0183).Thereseemstobenoconsistentsocialdesirabilitybias.Whilewecannot
ruleoutthepossibility,ourdatasuggestthatMPswhoarecorruptarenotnecessarilyun‐
der‐reportingcorruption.
TheanalysisforMPs(reportedinTable3)showsthatMPswhobelievegovernment
fundsaremorelikelytobeusedforcorruptionandclientelismaresignificantlymorelikely
toprefergovernment‐fundedprojects.For5ofthe9outcomes,MPswhoseegovernment
funds as more corrupt and clientelist are significantly more likely to prefer government
funds.Importantly,theseeffectsarestronglysignificantforthebehavioraloutcomes.The
treatmenteffectsrangefrom4%to19%,thusindicatingmoremeaningfulsubstantiveef‐
fectsizes.
[TABLE3ABOUTHERE]
29
Conversely,thedifferenceinsupportforaidandgovernmentprojectsisnotstatisti‐
callysignificantforMPswhodonotperceivesignificantcorruptionandclientelism(thisis
trueforalloutcomemeasures;seeTable3).ThissuggeststhattheMPswhoseefewave‐
nuesforcorruptionandclientelismexpressnopreferenceforgovernment‐fundedprojects
overaid.OneplausibleinterpretationoftheseresultssuggeststhatiftheMPcannotcap‐
turesomeofthefunding,thens/hedoesnotmanifestaclearpreferencetowardsuchpro‐
jects.
The results in Table 4 report the difference‐in‐means tests and support the claim
thatcitizensupportforaidisalsoconditionalontheirperceptionsofcorruptionandclien‐
telism,but intheoppositedirection.Thecitizenswhobelievethatgovernmentfundsare
used for corruption and clientelism are significantly more likely to support aid over gov‐
ernmentprojectsfor3ofthe6outcomes(4ofthe6atthe0.1level).Amongsubjectswho
donotperceivethecorruptuseofgovernmentfunds,therearenosignificantdifferences,
butmostofthenegativesignssuggestaslightpreferenceforgovernmentprojectsoveraid.
(TheresultsforMPsandmassesarealsoplottedinFiguresA12andA13.)
[TABLE4ABOUTHERE]
Wealsoranadditionalteststodetermineiftheeffectofcorruptionandclientelism
ismediatedbyethnicity,partisanship,orregionalidentities.Eventhoughmanystudiesar‐
gue that clientelism operates along ethnic or regional channels or through partisan net‐
works(i.e.,Wantchekon,2003;Stokes,2005),wefindnoclearevidencethattheeffectof
clientelism found here is mediated by any of these variables. These null results could in
30
partbeduetothefactthatwemeasureperceptionsof,ratherthaninvolvementin,corrup‐
tionandclientelism.ThemediationanalysisisreportedinAppendixFiguresA14andA15.
Thisanalysisprovidessupportfortheargumentthatcitizenssupportaidovergov‐
ernmentprogramsconditionalontheirperceptionsofcorruptionandclientelism.Further,
wefindthatcitizensdoinfactconsidercorruptiontobeabadthing:peoplewhoperceive
theretobemorecorruptionaresignificantlylesslikelytotrustparliament,theirMPs,and
thepresident(effectsaresignificantatthe0.01levels).Takentogether,thisanalysisoffers
someevidencethatpoliticalelitesmaybelievethatgovernmentfundsaremoresusceptible
to clientelism and corruption. In addition, ordinary citizens who perceive corruption and
clientelism in government behave in ways that suggest they see aid‐funded projects as a
more preferred mechanism than government action to obtain the public goods that they
expresstheysodesperatelyneed.
Discussion of Robustness Aswithanyexperimentnumerousdesignchoiceswererequired,allofwhichpre‐
sented difficult tradeoffs. In this section, wediscuss two importantaspects of the experi‐
mentaldesignthatatfirstglancemayseemtoraisequestionsaboutourresults.First,we
discussthecostcondition,whichtriestoaddressthedifferencebetween“tax‐based”gov‐
ernmentprojectsand“free”aidprojectseventhough,aswearguebelow,thischaracteriza‐
tionisnotaccurateinthecontextofthestudy.Second,wediscussthefactthatthecontrol
conditiondoesnotexplicitlynamethegovernmentasthefunderbutisneverthelessinter‐
pretedasthegovernment.
31
Taxed Government Projects Versus Free Aid Projects: A False Dichotomy Afirstdesignobjectionmightbethatthepublicmaypreferaidbecauseitisviewed
asfree,whereasgovernmentprojectsrequirecitizenstopaytaxes.Wedonotbelievethis
isthefactordrivingourresultsforseveralreasons.First,weaddedthecoststatementto
both the treatment and control conditions, so that individuals are awarethat any project
mayrequirelocalfundsandsupport.
Second,weundertookafollow‐upstudywherewerecruitedanadditional460sub‐
jectsandrandomlyassignedthemtoreceivethecoststatement(ornot)inassociationwith
oneofthetworandomlyassignedprojectdescriptions.Thecoststatementhadnosignifi‐
canteffectonsubjects’supportfortheproject.Thismaybeeitherbecausethecoststate‐
ment was too weak to produce treatment effects or because subjects were indifferent to
costsforprojectstheyfeeltheydesperatelyneed.Whilethecoststatementmaybeweak,
multiplereasonsleadustobelievethatcitizensarerelativelyindifferenttocostsforpublic
goods.
First, subjects likely do not see government projects as costly to themselves any
morethanforeignaidiscostly.ThevastmajorityofUgandans–86percentinournational‐
ly representative subject pool – fall below the earnings threshold for paying income tax,
whichisroughly600dollarsperyear.AndasMartin(2013)notes,taxrateshaveactually
falleninthecountryrecently.
Second,Ugandancitizensreceiveveryfewpublicgoods,beitfromforeignorgov‐
ernmentsources.Ugandaranksamongthepoorestcountriesintheworld,andpublicser‐
vicesdonotextendbroadlytothegeneralpopulation.EightypercentofUgandanslivein
32
rural areas and more than ninety percent of our subjects reported earning less than two
dollarsperdayinincome.FormostUgandans,publicservices(healthcare,education,in‐
frastructure)areweakandunderdeveloped,whichsuggeststhatvotersshouldhavestrong
preferencesforpublicgoodsregardlessofsource,whichwefind.Further,thegovernment
raiseslittlerevenuefromtaxesandwhatrevenuesexistareoftentransformedintoprivate
goodsand/ordirectedtopoliticalalliesforthepurposesofcorruptionorclientelism(see
CoxandMcCubbins,2001;Martin,2013).
The Government Control Condition Aseconddesignobjectionmightbethatthegovernmentwasnotnamedinthecon‐
trolcondition.Wewereconcernedaboutsocialdesirabilityinresponsesifweactuallyla‐
beled the control as the government. This is not a trivial concern in this context as the
Afrobarometer data show (see footnote 9). For the citizens, we worried that they might
feargovernmentreactionsandsoalwaysrankthegovernmentprojectsfirst.FortheMPs,
wefearedsocialdesirabilitybiasinwhichtheyalwayssaidtheypreferredthegovernment
projectssincetheywerepartofthegovernmentandwouldwanttoavoidbeingseenasnot
supportinggovernmentdevelopmentprojects.
Toassesswhatcitizensperceivedwhentheyviewedthecontrolcondition,aswell
aswhattheimplicationsofthisare,weconductedafollow‐upsurveytoensurethatsub‐
jectsinthemassexperimentdidinfactinterpretthecontrolconditionasthegovernment,
andwefoundthatthemajorityofsubjectsdidso(52%and51%fortheeducationandelec‐
tricityprojects,respectively)(Author2013).Morethanonethirdofsubjectsinthefollow‐
upstudy,however,attributedthecontrolconditiontoaforeigndonor.Whileamoredirect
33
comparison could have been preferable, attribution of the control projects to foreign do‐
norsworksinfavorofthenullhypothesisofnotreatmenteffects.
Theconcernisthatthecontrolconditionrepresentedacombinationofpeoplewho
believeitimpliedeitherthegovernmentoraforeignaiddonor;thatis,supportforthecon‐
trolisequaltosomeaverageofsupportforforeignaidprojectscombinedwithsupportfor
government projects. Because we know two of these three values—the outcome in the
controlconditionoverallandtheoutcomeintheforeignaidcondition,wecancalculatethe
third:thelevelofsupportthatsubjectswouldprovidehadtheybeengiventhegovernment
controlconditionexplicitly.
First,weknowtheaveragevaluethatmassandMPrespondentsgaveinsupportof
theprojectsiftheywereassignedaforeigndonor.Inthetwosurveysweaskedaboutsup‐
portfortheprojectsusing6differentaiddonorsforthemassesand4fortheMPs,assign‐
ingeachsubjectadonoratrandom.15Ourdatashowthatacrossalltheseforeigndonors,
themassrespondentsdidnotdifferentiatesignificantlybetweenthem,butonaveragethey
supportedtheforeign‐fundedprojectsatahigherlevelthandidthecontrolgroup.Wehave
similarevidencefortheMPs,excepttheMPs,ontheotherhand,supportedalltheaidpro‐
jectsonaveragelessthanthecontrolcondition.Thisimpliesthatwecancalculateanaver‐
agevalueofsupportamongthemasspublicandMPsforprojectsledbyanyforeigndonor.
Second,onaveragethecontrolgroup'slevelofsupportfortheprojectswaslower
thantheaverageforalltheforeign‐donortreatmentgroupsforthemassexperiment.For
15Respondentswererandomlyassignedoneof6donors:US,China,WorldBank,AfricanDevelopmentBank,
genericbilateralandmultilateraldonor.MPsdidnotseeChinaortheAfDB.
34
theMPs,thecontrolgroups’supportwashigherthanforalltheforeignaidprojects.Third,
ourpost‐surveydatashowthat51or52%,dependingontherandomassignmentofelec‐
tricityoreducationproject,respectively,believethatthecontrolwasagovernmentproject
andmostoftheremainingbelieveditwasaforeigndonor.Sotheactualvalueofsupport
forthecontrolgroupforthosewhothoughtitwasthegovernmentcanbededucedfrom
this information. In the mass experiment it must necessarily be lower than that for the
group that was given the foreign aid conditions, while in the MP experiment it must be
higher.
Wecanusethesethreepiecesofinformationtocalculatethemeanandstandarder‐
rorsofthemassrespondentswhoattributedthecontrolconditiontothegovernment.We
canonlyobtainanestimatefortheMPssincewedidnotaskthemwhotheythoughtwas
funderinthecontrolcondition,butthisstillimplieswhatthecontrolgroupwhoattributed
ittothegovernmentwouldhavescored.Calculatingthemeanisstraightforward.Weknow
thatthemeanofthecontrolgroupismadeupoftherespondentswhothoughtthatthecon‐
trolwasaforeigndonorandthosewhothoughtthecontrolwasthegovernment:
∗
where
and
1
∗
aretheaveragelevelsofsupportforthedevelopmentprojects
underthecontrolandtreatmentconditions,respectively.Thesevaluesareknownfromthe
data and is the percentage identifying the control as the government. Rearranging to
solvefor
wederive:
1
∗
35
Calculatingthestandarderrortocreatetheconfidenceintervalsisalittlemoredifficultand
we describe the procedure in the following footnote.16Using these calculated means and
standarddeviations,wecanthencomparethosereceivingtheforeigntreatmenttothecon‐
trolconditionasreportedinthepapertotheportionofcontrolrespondentswhothought
theconditionwasthegovernment.Figure1demonstratestherelativedifferencesforthe
strongsupportconditionamongthemasses.Forallotheroutcomeconditionsformasses
and MPs, this relative ordering holds and so we do not display them here. As the figure
shows, the difference between those receiving the explicit foreign condition and those
thinkingthecontrolrepresentedthegovernmentismuchlargerthanbetweentheexplicit‐
lyforeignconditionandtheundifferentiatedcontrol.Thedirectionoftheeffectisopposite
intheMPcase.Thus,theresultswereportinthepaperworkagainstourstatedhypotheses
and therefore provide the most conservative test; that is, had we named the government
explicitly in the control condition,we would have observed a much larger difference and
thereforetheresultsreportedwouldbeevenstronger.
[FIGURE1ABOUTHERE]
16WeknowthatthestandarderroristhestandarddeviationdividedbythesquarerootofN.Thestandard
errorforthecontrolwillbecomprisedofthestandarderroroftherespondentswhobelievedthegovernment
wasthedonorandthosewhobelievedthatitwasaforeigndonor:
2
Againwecanrearrangetheformulatocalculatethestandarddeviationofthegovernmentrespondents:
2∗
∗
SincethestandarderrorisjustthestandarddeviationdividedbythesquarerootofN,wecancalculate
togetthestandarderrorsofthegovernmentrespondents.
36
Whilewehadlegitimatereasonsfornotnamingthegovernmentinthecontrolcon‐
dition (because of the generalized paranoia toward the regime among citizens and social
desirability among MPs), in retrospect doing so would have made the analysis clearer.
Nonetheless,wedonotbelievethatthewaythecontrolconditionwasstatedvitiatesour
results; indeed, it may have led us to understate them. Thus for evaluating the different
massandelitepreferencesforforeignvs.governmentprograms,thebasicresultsalready
providereasonablystrongsupportforourclaims,andwiththeseadjustmentstheevidence
insupportwouldbeevenstronger.
ForcostandlogisticalreasonswedidnotperformthesamefollowupstudyonMPs,
butweexpectthatMPsmighthavevoicedsimilarperceptionstootherUgandans.Thisof
coursemeansthatsomeMPs,likesomecitizens,probablyperceivedthecontrolcondition
assponsoredbyforeigndonors.Again,thiswouldhaveledtoanunderstatementofthedif‐
ferencebetweentreatmentandcontrolintheMPexperiment.
Ontheotherhand,itisalsopossiblethatMPsmaymerelyhaveapreferenceforun‐
specifiedoverspecifiedfunding.Weemphasizethattheprojectdescriptionswereidentical
acrossconditionssavethestatementofthefundingsource,sothedetailsoftheeducation
and electricity projects were equally specified between treatment and control, which di‐
minishes our concern on this count. Moreover, two of the foreign donor conditions were
deliberatelygenericinthattheyattributedtheprojectstoeitheranunspecifiedmultilateral
orbilateraldonor.MPsdidnotsignificantlypreferthesegenericconditionstoconditionsin
whichtheWorldBankortheUnitedStateswerenamed,whichmovesustowarddiscount‐
ingthepossibilitythatMPssimplypreferprojectswithunspecifieddonors.Rather,itap‐
37
pearsmorelikelythattheirattributionofthecontrolprojectstothegovernmentprompted
thetreatmenteffects.
Conclusion Thispaperprovideswhatis,toourknowledge,thefirstexperimentalstudytocom‐
pareaidpreferencesandactionsformembersofparliamentandanationallyrepresenta‐
tive, random sample of ordinary citizens in a prominent developing country. We were
specifically concerned with preferences towards foreign vs. domestic development pro‐
jects. Citizens preferred aid over government programs consistently, and with most de‐
pendentvariablestostatisticallysignificantdegrees,especiallyinthebehavioraloutcomes.
Thiswasparticularlysoamongtherespondentswhoperceivedproblemswithgovernment
corruptionandclientelism,thusprovidingevidenceconsistentwiththeargumentthataid
can help overcome governance problems. Likewise, members of parliament consistently
preferredgovernmentprogramsoveraid.
Lookingatsuchindividual‐levelevidence–forpoliticalelitesandcitizens–givesus
somesenseofhowaidmightfunctionatthegroundlevel.Ourmicro‐levelevidencepro‐
videssomesupportfortheargumentthatcitizensmayseeforeignaidasawayofpromot‐
ingpublicgoodswithoutstrengtheningcorruption.Citizensaremorewillingtosupportaid
bytakingbehavioralactionimposingpersonalcoststhroughsigningapetitionandsending
anSMS.Theyviewaidaslesspoliticizedthangovernmentprograms.
Forthesesamereasons,itperhapsmakessensethatpoliticaleliteswerelessenthu‐
siasticaboutaidthantheywereaboutgovernment‐fundedprojects.MPs’likelyfacefewer
38
constraintsoverhowtheymightutilizethesedomesticgovernmentresources.Highlevels
ofcorruptionandclientelismexistindevelopingcountriesevenintheabsenceofforeign
aid.Anddomesticresourcesmaybeeveneasierforgovernmentstodiverttothesepurpos‐
essincethereareoftennostrongaccountabilitymechanismsatworkinpoordeveloping
countries. But citizens’ significant preference for aid over government programs should
give readers pause. Perhaps citizens are merely ill informed about aidand donot under‐
standhowitharmsthem.Alternatively,perhapsthroughaidtheyseeapotential,ifpartial,
meansofescapefromasystemofgovernancethatpreventsthemfromreceivingthepublic
goodstheystronglydesire.Ourdataseemtosupportthislatterclaim.
Our study brings together two complimentary literatures. The large literature on
clientelismandcorruptionindevelopingcountriesstronglyimpliesthatgovernmentshave
the desire and will to use their funds to promote their own political purposes first and
foremost.Stayinginofficeiscriticalandusinggovernmentprojectstobuildsupportisone
way to do this. Uganda’s government is no exception. However, aid scholars often assess
foreignassistancewithoutanydirectcomparisontothemostrealisticalternative,whichis
governmentfunding.Ourstudyexaminesthebeliefsandactionsofbothelitesandcitizens
bycomparingtheirsupportforthesetwodifferentdevelopmentmechanisms.Theseforms
ofevidenceshednewlightontwoveryprominentliteraturesbymakingmorecentralthe
preferencesofpoliticalelitesandcitizens.Weexpectthatmuchistobegainedbycomple‐
mentingexistingmacro‐levelstatisticalapproacheswithmicro‐levelexperimentaldataon
politiciansandbeneficiariesofaidindevelopingcountries.
39
References: Alesina, Alberto, and Beatrice Weder (2002). Do Corrupt Governments Receive Less
ForeignAid?AmericanEconomicReview.92(4):1126‐1137.
Aronow, Peter M., Allison Carnegie, and Nikolay Marinov (2012). The Effects of Aid on
Rights and Governance: Evidence from a Natural Experiment. Working Paper. New
Haven:
Yale
University.
August
2,
2012.
Retrieved
from
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2199131
Barreto, Matt A., Gary M. Segura, and Nathan D. Woods (2004). The Mobilizing Effect of
Majority–Minority Districts on Latino Turnout. American Political Science Review.
98(01):65‐75.
Baser, Heather, and Peter Morgan (2008). Capacity, Change, and Performance. ECDPM
Study Report. Maastricht: European Centre for Development Policy Management.
April2008.
Bayart, Jean‐François (1993). The State in Africa: The Politics of the Belly. New York:
Longman.
Bermeo, Sarah Blodgett (2010). Development and Strategy: Aid Allocation in an
InterdependentWorld.Durham,NC:DukeUniversity.September22,2010.Retrieved
fromhttp://papers.ssrn.com/sol3/papers.cfm?abstract_id=1681104
‐‐‐ (2011). Foreign Aid and Regime Change: A Role for Donor Intent. WorldDevelopment.
39(11):2021‐2031.
‐‐‐(2014).AidisNotOil:DonorPreferences,HeterogeneousAid,andtheAid‐Democratization
Relationship.WorkingPaper.Durham,NC:DukeUniversity.July2014.
Bobo,Lawrence,andFranklinD.Gilliam,Jr.(1990).Race,SociopoliticalParticipation,and
BlackEmpowerment.AmericanPoliticalScienceReview.84(2):377‐393.
Bräutigam, Deborah (2000). Aid Dependence and Governance. Stockholm: Almqvist &
WiksellInternational.
Bräutigam,Deborah,andStephenKnack(2004).ForeignAid,Institutions,andGovernance
inSub‐SaharanAfrica.EconomicDevelopmentandCulturalChange.52(2):255‐285.
BuenodeMesquita,Bruce,andAlastairSmith(2007).ForeignAidandPolicyConcessions.
JournalofConflictResolution.51(2):251‐284.
‐‐‐(2009).APoliticalEconomyofAid.InternationalOrganization.63(2):309‐340.
Calhoun,Craig(1993).NationalismandEthnicity.AnnualReviewofSociology.19:211‐239.
Chandra, Kanchan (2006). What Is Ethnic Identity and Does It Matter? AnnualReviewof
PoliticalScience.9(1):397‐424.
Cheibub,JoséAntonio,JenniferGandhi,andJamesRaymondVreeland(2010).Democracy
andDictatorshipRevisited.PublicChoice.143(1‐2):67‐101.
Cox, Gary W., and Mathew D. McCubbins (2001). The Institutional Determinants of
Economic Policy Outcomes. In Stephan Haggard and Mathew D. McCubbins (Eds.),
Presidents, Parliaments, and Policy (pp. 21‐63). Cambridge (UK): Cambridge
UniversityPress.
Dietrich,Simone(2013).BypassorEngage?ExplainingDonorDeliveryTacticsinForeign
AidAllocation*.InternationalStudiesQuarterly.57(4):698‐712.
40
Easterly, William (2006). The White Man's Burden: Why the West's Efforts to Aid the Rest
HaveDoneSoMuchIllandSoLittleGood.Oxford;NewYork:OxfordUniversityPress.
Finkel, Steven E., Aníbal Pérez‐Liñán, and Mitchell A. Seligson (2007). The Effects Of U.S.
Foreign Assistance On Democracy Building, 1990‐2003.WorldPolitics.59(3): 404‐
439.
Franck, Raphaёl, and Ilia Rainer (2012). Does the Leader's Ethnicity Matter? Ethnic
Favoritism,Education,andHealthinSub‐SaharanAfrica.AmericanPoliticalScience
Review.106(2):294‐325.
Gazibo,Mamoudou(2013).BeyondElectoralDemocracy:ForeignAidandtheChallengeof
DeepeningDemocracyinBenin.InDanielleResnickandNicolasvandeWalle(Eds.),
Democratic Trajectories in Africa: Unravelling the Impact of Foreign Aid (pp. 228‐
255).Oxford:OxfordUniversityPress.
Gervasoni, Carlos (2010). A Rentier Theory of Subnational Regimes: Fiscal Federalism,
Democracy,andAuthoritarianismintheArgentineProvinces.WorldPolitics.62(2):
302‐340.
Green, Elliott D. (2010). Patronage, District Creation, and Reform in Uganda. Studies in
ComparativeInternationalDevelopment(SCID).45(1):83‐103.
Greene, Kenneth F. (2010). The Political Economy of Authoritarian Single‐Party
Dominance.ComparativePoliticalStudies.43(7):807‐834.
Hyde, Susan D., and Nikolay Marinov (2012). Which Elections Can Be Lost? Political
Analysis.20(2):191‐210.
Knack, Stephen (2009). Sovereign Rents and Quality of Tax Policy and Administration.
JournalofComparativeEconomics.37(3):359‐371.
Malinga, Joseph (2010). Vote Rigging and Violence in Primary Elections Sets Worrying
Trend
in
Uganda.
The
Guardian,
London.
Retrieved
from
http://www.theguardian.com/katine/2010/sep/07/katine‐election‐malpractice
Martin, Lucy (2013). Taxation and Accountability: Experimental Evidence for Taxation's
EffectonCitizenBehavior.PreparedfortheWorkshoponAfricanPoliticalEconomy,
Washington,D.C.,May20,2013.
Mavrotas, George, and Bazoumana Ouattara (2006). Aid Disaggregation and the Public
Sector in Aid‐Recipient Economies: Some Evidence from Cote D’Ivoire. Review of
DevelopmentEconomics.10(3):434‐451.
McLean, Elena V., and Christina J. Schneider (2014). Limits of Informal Governance? The
ScopeofConditionalityintheWorldBank. Prepared for the Annual Meeting on the
PoliticalEconomyofInternationalOrganizations,Princeton,January16‐18,2014.
Morrison, Kevin M. (2009). Oil, Nontax Revenue, and the Redistributional Foundations of
RegimeStability.InternationalOrganization.63(1):107‐138.
‐‐‐(2012).WhatCanWeLearnaboutthe“ResourceCurse”fromForeignAid?WorldBank
ResearchObserver.27(1):52‐73.
Mosley,Paul(forthcoming).FiscalCompositionandAidEffectiveness:APoliticalEconomy
Model.WorldDevelopment.
Moyo, Dambisa (2009). DeadAid:WhyAidIsNotWorkingandHowThereIsaBetterWay
forAfrica(1stAmericaned.).NewYork:Farrar,StrausandGiroux.
41
Muhumuza, William (2009). From Fundamental Change to No Change: The National
ResistanceMovement(NRM)andDemocratizationinUganda.LesCahiersd'Afrique
del'Est.41:21‐42.
North, Douglass C., and Barry R. Weingast (1989). Constitutions and Commitment: The
EvolutionofInstitutionsGoverningPublicChoiceinSeventeenth‐CenturyEngland.
JournalofEconomicHistory.49(4):803‐832.
Posner, Daniel N. (2005). InstitutionsandEthnicPoliticsinAfrica. Cambridge; New York:
CambridgeUniversityPress.
Riddell,RogerC.(2007).DoesForeignAidReallyWork?Oxford:OxfordUniversityPress.
Ross,MichaelL.(2004).DoesTaxationLeadtoRepresentation?BritishJournalofPolitical
Science.34(2):229‐249.
Rothchild, Donald S. (1986). Hegemonial Exchange: An Alternative Model for Managing
Conflict in Middle Africa. In Dennis L. Thompson and Dov Ronen (Eds.), Ethnicity,
Politics,andDevelopment(pp.65‐104).Boulder,CO:LynneRienner.
Schumpeter,JosephA.(1954).TheCrisisoftheTaxState.InternationalEconomicPapers.4:
5‐38.
Scott,JamesM.,andCarieA.Steele(2011).SponsoringDemocracy:TheUnitedStatesand
DemocracyAidtotheDevelopingWorld,1988‐2001.InternationalStudiesQuarterly.
55(1):47‐69.
Smith,Alastair(2008).ThePerilsofUnearnedIncome.JournalofPolitics.70(3):780‐793.
Stokes,SusanC.(2005).PerverseAccountability:AFormalModelofMachinePoliticswith
EvidencefromArgentina.AmericanPoliticalScienceReview.99(3):315‐325.
Svensson,Jakob(2000).ForeignAidandRent‐Seeking.JournalofInternationalEconomics.
51(2):437‐461.
Szeftel, Morris (2000). Clientelism, Corruption & Catastrophe. Review of African Political
Economy.27(85):427‐441.
Tripp, Aili Mari (2010). Museveni's Uganda: Paradoxes of Power in a Hybrid Regime.
Boulder:LynneRiennerPublishers.
‐‐‐ (2013). Donor Assistance and Political Reform in Tanzania. In Danielle Resnick and
NicolasvandeWalle(Eds.),DemocraticTrajectoriesinAfrica:UnravellingtheImpact
ofForeignAid(pp.170‐199).Oxford:OxfordUniversityPress.
van de Walle, Nicolas (2003). Presidentialism and Clientelism in Africa's Emerging Party
Systems.JournalofModernAfricanStudies.41(2):297‐321.
‐‐‐(2007).MeettheNewBoss,SameAstheOldBoss?TheEvolutionofPoliticalClientelism
in Africa. In Herbert Kitschelt and Steven Wilkinson (Eds.), Patrons, Clients and
Policies:PatternsofDemocraticAccountabilityandPoliticalCompetition (pp. 50‐67).
Cambridge,UK;NewYork:CambridgeUniversityPress.
Wantchekon, Leonard (2003). Clientelism and Voting Behavior: Evidence from a Field
ExperimentinBenin.WorldPolitics.55(3):399‐422.
Wright,Joseph,andMatthewWinters(2010).ThePoliticsofEffectiveForeignAid.Annual
ReviewofPoliticalScience.13(1):61‐80.
42
43
Table 1: Comparison of MP Sample to the Actual 9th Parliament Gender
%Male
%Female
Party
%NRM
%Independents
%FDC
%DP
%UPC
%CP
%JEEMA
Region
%fromCentral
%fromEastern
%fromNorthern
%fromWestern
MPType
%ConstituencyMPs
%DistrictWomenMPs
%SpecialInterestMPs
%Ex‐OfficioMPs
Sample
67
33
74.6
10.2
8.5
3.1
3.1
0.25
0.25
28
28
18
26
59
28
6
8
9thParliament
65
35
73.5
11.2
8.8
3.4
2.6
0.25
0.25
25
27
22
26
62
29
7
2
44
Table 2: Citizen and MP Preferences for Government versus Aid Projects MPs
Govt
N
Aid
N
Difference
Masses
Govt
N
Aid
N
Difference
MPs
Govt
N
Aid
N
Difference
PanelA:MPandCitizenOutcomes
Strong
Tell
Willing to Signed
Willing to Signed
Support
sign
SignPres. Pres.
0.84
0.97
.89
.78
.86
.75
136
136
136
138
59
59
0.83
0.99
.82
.75
.75
.68
567
567
567
570
292
292
‐0.01
0.02
‐0.07**
‐0.04
‐0.12**
‐0.06
Strong
Tell
Willing to Signed
Willing to SentSMS
Support
sign
SMS
0.73
0.91
0.82
0.77
0.59
0.02
528
520
528
538
538
202
0.77
0.94
0.83
0.80
0.64
0.05
3007
2967
3008
3017
3017
1143
0.03*
0.03**
0.02
0.04*
0.04*
0.02*
PanelB:EliteOnlyOutcomes
TellConstituents
RallyLocalOfficials Coordinatewith
Peers
0.99
0.98
0.99
136
123
136
0.98
0.97
0.97
567
501
567
‐0.00
‐0.01
‐0.02**
Anegativedifferencemeansthattheproportionofsupportforprojectsinthecontrolcondition(government)
islargerthantheproportionunderthetreatmentcondition(aid),implyingthegovernmentconditionispre‐
ferredtotheaidone.Notethatifasubjectstateds/hedidnotwanttosignthepetition(thirdcolumn)westill
presentedthemthepossibilityofsigningthepetition(fourthcolumn).ThehigherNsforwillingnesstoSMSin
thefifthcolumn(e.g.,538and3017)arearesultofsubjectrefusalstoanswerthepetitionquestions(where
correspondingNsarelower:528and3008).Thatis,ifasubjectrefusedtoanswerpetitionquestions,westill
asked about SMS and fewer subjects declined to answer SMS questions. Also, the Ns decrease in the “Sent
SMS”condition(relativeto“WillingtoSMS”)becauseweonlycalculateSentSMSforsubjectswhoowneda
phone.
45
Table 3: Testing the Corruption Mechanism (MPs) MPSupportConditionalonPerceptionsofCorruption
Strong
Tell
Willing to Signed
Willing to
Support
Sign
SignPres.
Yes,GovernmentFundsusedforCorruption
Govt
0.86
0.98
0.95
0.89
0.94
N
44
44
44
45
18
Aid
0.82
0.98
0.85
0.78
0.77
N
195
195
195
197
103
Difference
‐0.05
0.01
‐0.10**
‐0.11**
‐0.18**
No,GovernmentFundsnotusedforCorruption
Govt
0.82
0.97
0.86
0.73
0.83
N
90
90
90
91
41
Aid
0.83
0.99
0.81
0.73
0.74
N
366
366
366
367
188
Difference
0.01
0.02
‐0.05
0.00
‐0.09
Signed
Pres.Pet.
0.89
18
0.70
103
‐0.19**
0.68
41
0.68
188
‐0.01
MPSupportConditionalonPerceptionsofCorruption
Tell
RallyLocals
Coordinate
Constituents
WithPeers
Yes,GovernmentFundsusedforClientelism
Govt
0.98
0.97
1.00
N
44
39
44
Aid
0.98
0.96
0.96
N
195
171
195
Difference
‐0.01
‐0.01
‐0.04***
No,GovernmentFundsnotusedforClientelism
Govt
0.99
0.98
0.99
N
90
82
90
Aid
0.98
0.97
0.97
N
366
324
366
Difference
‐0.01
‐0.00
‐0.02
Anegativedifferencemeansthattheproportionofsupportforprojectsinthecontrolcondition(government)
islargerthantheproportionunderthetreatmentcondition(aid),implyingthegovernmentconditionispre‐
ferredtotheaidone.Apositivedifferenceimpliesthattheaidconditionispreferredtothegovernmentcon‐
dition.Notethatifasubjectstateds/hedidnotwanttosignthepetition(thirdcolumn)westillpresented
themthepossibilityofsigningthepetition(fourthcolumn).
46
Table 4: Testing the Corruption Mechanism (Masses) MassSupportConditionalonPerceptionsofCorruption
Strong
Tell
Willing to Signed
Willing to SentSMS
Support
sign
SMS
Yes,GovernmentFundsusedforCorruption
Govt
0.71
0.90
0.80
0.74
0.69
0.03
N
393
386
393
402
157
157
Aid
0.77
0.94
0.83
0.80
0.71
0.05
N
2274
2241
2274
2279
894
894
Difference
0.06**
0.04**
0.03
0.05**
0.02
0.03*
No,GovernmentFundsnotusedforCorruption
Govt
0.82
0.93
0.87
0.85
0.86
0.00
N
126
126
126
176
42
42
Aid
0.76
0.94
0.84
0.82
0.78
0.03
N
695
688
696
699
236
236
Difference
‐0.06
0.01
‐0.04
‐0.03
‐0.08
0.03***
Anegativedifferencemeansthattheproportionofsupportforprojectsinthecontrolcondition(government)
islargerthantheproportionunderthetreatmentcondition(aid),implyingthegovernmentconditionispre‐
ferredtotheaidone.Apositivedifferenceimpliesthattheaidconditionispreferredtothegovernmentcon‐
dition.Notethatifasubjectstateds/hedidnotwanttosignthepetition(thirdcolumn)westillpresented
themthepossibilityofsigningthepetition(fourthcolumn).ThehigherNsforwillingnesstoSMSinthefifth
column(e.g., 538 and 3017) are a resultofsubject refusalstoanswer the petition questions(where corre‐
spondingNsarelower:528and3008).Thatis,ifasubjectrefusedtoanswerpetitionquestions,westillasked
aboutSMSandfewersubjectsdeclinedtoanswerSMSquestions.Also,theNsdecreaseinthe“SentSMS”con‐
dition(relativeto“WillingtoSMS”)becauseweonlycalculateSentSMSforsubjectswhoownedaphone.
47
Figure 1: The Value of the Control if Government was Named .7
.75
% Supporting
.8
.85
.9
.95
% Strongly Supporting project by treatment
Foreign
Control
Mass
Gov (calc.)
MP
Foreign and control show actual values; gov (calc.) estimates the mean
for respondents believing control was government
Calculationofcontrolvalueifithadspecifiedgovernmentusinginformationfromsurveys.
48