Hu e, Paul; Weishaa , Daniel
A icle — Published Ve sion
Jus cheap alk? In es iga ing ai ness p e e ences in
hypo he ical scena ios
The Jou nal o Economic Inequali y
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: Hu e, Paul; Weishaa , Daniel (2025) : Jus cheap alk? In es iga ing ai ness
p e e ences in hypo he ical scena ios, The Jou nal o Economic Inequali y, ISSN 1573-8701, Sp inge
US, New Yo k, NY, Vol. 23, Iss. 3, pp. 881-907,
h ps://doi.o g/10.1007/s10888-025-09700-w
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/330533
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/4.0/
h ps://doi.o g/10.1007/s10888-025-09700-w
RESEARCH
Jus cheap alk? In es iga ing ai ness p e e ences
in hypo he ical scena ios
Paul Hu e1
·Daniel Weishaa 2
Recei ed: 27 June 2025 / Accep ed: 8 July 2025
© The Au ho (s) 2025
Abs ac
The measu emen o p e e ences o en elies on su eys in which indi iduals e alua e hypo-
he ical scena ios. This pape p oposes and alida es a no el ac o ial su ey ool o measu e
ai ness p e e ences. We examine whe he a non-incen i ized su ey cap u es he same dis-
ibu ional p e e ences as an impa ial spec a o design, whe e choices may apply o a eal
pe son. In con as o p io s udies, ou design in ol es high s akes, wi h esponden s de e -
mining a eal pe son’s mon hly ea nings, anging om $500 - $5,700. We ind ha he
non-incen i ized su ey module yields nea ly iden ical esul s compa ed o he incen i ized
expe imen and eco e s ai ness p e e ences ha a e s able o e ime. Fu he mo e, we show
ha mos esponden s adop in e media e ai ness posi ions, wi h ewe exhibi ing s ic ly
egali a ian o libe a ian p e e ences. In sum, hese indings sugges ha high-s ake incen i es
do no signi ican ly impac he measu emen o ai ness p e e ences and ha non-incen i ized
su ey ques ions co e ing ealis ic scena ios o e aluable insigh s in o he na u e o hese
p e e ences.
Keywo ds Fai ness p e e ences ·Su ey expe imen ·Vigne e s udies
JEL Classi ica ion C90 ·D63 ·I39
This pape bene i ed s ongly om discussions wi h Ing ild Almås. I is pa o a la ge esea ch p ojec in
which we measu e ai ness p e e ences and belie s abou inequali y a ound he wo ld (see Riksbankens
Jubileums ond (P22-0564) and UKRI Fu u e Leade s Fellowship (MR/X033333/1): “(Un)Fai inequali y in
he labo ma ke : a global s udy”). We ecei ed use ul eedback om semina audiences a LMU Munich.
This esea ch was unded by he B i ish Academy (TDA21/210082) and by Deu sche
Fo schungsgemeinscha h ough CRC TRR 190 (No. 280092119) and unde Ge many’s Excellence S a egy
– EXC 2126/1-390838866. Hu e g a e ully acknowledges inancial suppo om UKRI (MR/X033333/1).
Weishaa g a e ully acknowledges inancial suppo om he Joachim He z Founda ion and he F i z Thyssen
Founda ion. The ques ionnai e and co e analysis we e p e- egis e ed ia he Open Science F amewo k
(OSF), No. DV3KP. We ob ained e hical app o al om he Ins i u ional Re iew Boa ds a he Uni e si y o
B is ol, LMU Munich, and NHH No wegian School o Economics. All emaining e o s a e ou own. The
online e sion con ains supplemen a y ma e ial a ailable a h ps://doi.o g/10.1007/s10888-025-09700-w.
BPaul Hu e
[email p o ec ed]
Daniel Weishaa
[email p o ec ed]
1Uni e si y o B is ol, B is ol, England
2Uni e si y o Cologne, Cologne, Ge many
123
The Jou nal o Economic Inequali y (2025) 23:881–907
/ Published online: 16 Sep embe 2025
P. Hu e, D. Weishaa
1 In oduc ion
The e is expanding li e a u e in economics and o he social sciences ha in es iga es which
inequali ies a e seen as un ai by people (Alesina e al. 2018; Almås e al. 2020; And e 2025;
Cappelen e al. 2007; Jasso and Webs e 1999; Konow 2000). In hese pape s, ai ness p e -
e ences a e ypically elici ed using incen i ized expe imen s o non-incen i ized su eys.
Resea che s conside ing he choice be ween hese wo esea ch designs ace a ade-o . On
he one hand, expe imen s combine s ylized ep esen a ions o eal-wo ld si ua ions wi h
payou - ele an choices o esponden s. On he o he hand, su ey ques ions o en mi o
eal-wo ld con ex s mo e closely; howe e , esponden s’ answe s ha e no consequences in
he eal wo ld. The e o e, su ey-based me hods a e o en conside ed un eliable p edic o s o
ac ual beha io . This aises he ques ion o whe he esea che s can employ non-incen i ized
su eys o analyze ai ness p e e ences o whe he such answe s mus be conside ed “cheap
alk.” In his pape , we add ess his ques ion by using a ep esen a i e sample o he US
adul popula ion o es whe he answe s o hypo he ical ques ions align wi h hose om an
incen i ized expe imen .
Ou su ey ool in eg a es co e unc ionali ies o impa ial spec a o expe imen s (Almås
e al. 2020; Almås e al. 2024a; And e 2025; Cappelen e al. 2013; Konow 2000; Konow e al.
2020) wi h he me hodological ad an ages o ac o ial su eys (Auspu g e al. 2017; Gae ne
and Schwe mann 2007; Jasso and Webs e 1999; Konow 1996).1The ques ions in ou su ey
ool show esponden s pai s o hypo he ical pe sons ha a e desc ibed in e ms o obse able
cha ac e is ics, i.e., hei gende , age, educa ional a ainmen , pa en al backg ound, wo king
hou s, and labo ma ke ea nings. Responden s hen dis ibu e ea nings be ween hese wo
pe sons based on he gi en in o ma ion.
In impa ial spec a o designs, choices a e no payo - ele an o he esponden s hem-
sel es. Ye , answe s in a hypo he ical ask migh s ill di e om hei incen i ized analogs.
Fi s , hypo he ical dis ibu ion asks a e p one o di e en biases, including expe imen e
demand e ec s and social desi abili y biases (S an che a 2023). Fo ins ance, in he absence
o eal-wo ld consequences, esponden s may end owa ds mo e equal income alloca ions
o comply wi h pe cei ed expec a ions om he esea che . Such demand e ec s a e less
o a conce n when eal money is a s ake (Haaland e al. 2023). Second, e en wi hou such
sys ema ic biases, hypo he ical dis ibu ion asks may be p one o measu emen e o , e.g.,
i non-incen i ized esponden s a e less a en i e when comple ing he asks.
Ou su ey ool di e s om p io li e a u e on ai ness p e e ences which has la gely
ocused on he ex en o which indi iduals ewa d a he abs ac concep s such as “luck,”
“p oduc i i y,” “ha d wo k,” and “ alen ” (e.g., Cappelen e al. 2010; Molle s om e al. 2015).
In con as o hese s udies, ou su ey ool allows us o elici ai ness p e e ences ha can
be di ec ly mapped o obse able labo ma ke inequali ies, e.g., gende gaps (Blau e al.
2017), e u ns o hou s (Kuhn and Lozano 2008), educa ion p emia (Ha mon e al. 2003),
and in e gene a ional pe sis ence (Roeme and T annoy 2016). Howe e , he ocus on hese
eal-wo ld inequali ies also makes i p ohibi i ely cos ly o elici he ele an p e e ences
in an expe imen al design whe e esponden s’ choices a e consequen ial o he ea nings
o ac ual pe sons. The e o e, i is impo an o es whe he hypo he ical dis ibu ion asks
deli e c edible esul s ha align wi h he “gold s anda d” o incen i ized expe imen s.
1Fo de ailed e iews on expe imen al and su ey-based e idence on ai ness p e e ences, see Almås e al.
(2023,2024b) and Gae ne and Schokkae (2012).
123
882
Jus cheap alk? In es iga ing ai ness p e e ences
To add ess his ques ion, we collec ed da a om 1,602 adul s in he Uni ed S a es be ween
Oc obe and No embe 2022. The sampling was designed o be ep esen a i e o a ious
demog aphic cha ac e is ics such as gende , age, educa ion, employmen s a us, and egion
o esidence. The su ey modules, as well as co e analyses, we e p e- egis e ed ia he Open
Science F amewo k (OSF), No. DV3KP.
We alida e ou su ey ool along h ee dimensions. Fi s , we es whe he he dis ibu ional
choices o esponden s a e di e en i hey a e payo - ele an . Fo his pu pose, we un an
expe imen wi h a be ween-subjec design. All esponden s answe a su ey whe e hey ace a
selec ion o asks om ou su ey ool. Responden s in he ea men g oup a e in o med ha
one o he pe sons shown o hem is a eal pe son and ha he decision made by a andomly
chosen esponden will de e mine he mon hly ea nings o his indi idual. Thus, in con as
o he con ol g oup, hey know ha each choice may ha e subs an ial inancial consequences
o a eal pe son. This design allows us o es whe he ai ness p e e ences in ou hypo he ical
ool a e consis en wi h he “gold s anda d” o an incen i ized expe imen (Baue e al. 2020;
Enke e al. 2022;Falke al.2023). Second, we es whe he he dis ibu ional choices a e s able
o e ime. Fo his pu pose, we employ a wi hin-subjec design and un an ob usca ed ollow-
up one week a e he baseline su ey.2In pa icula , we in i e esponden s o ano he su ey,
whe e hey again ace a selec ion o asks om ou su ey ool. Some asks a e epea ed om
he baseline wa e, allowing us o calcula e in e empo al co ela ions. This design enables us
o es he s abili y o ai ness p e e ences in ou hypo he ical ool and gi es c ucial in o ma-
ion on measu emen e o in he elici ed p e e ence da a (Gillen e al. 2019;S an che a2023).
Las ly, nex o he me hodological alida ion o he su ey design, we conduc a sugges i e
subs an i e analysis. Speci ically, we desc ibe he na u e o ai ness p e e ences iden i ied
h ough ou su ey and he he e ogenei y o ai ness iews wi hin he US popula ion. This
analysis has se e al ca ea s since he su ey design was p emised on me hodological ali-
da ion. Ne e heless, he subs an i e analysis p o ides an impo an c oss-check on whe he
ou hypo he ical su ey ool eco e s p e e ences ha a e consis en wi h p e ious s udies
on ai ness p e e ences in he US (Almås e al. 2020; Fisman e al. 2023;Konowe al.2020).
Ou esul s can be summa ized as ollows. Fi s , he dis ibu ional choices o esponden s
a e no a ec ed by making hem ele an o he ea nings o eal pe sons. The poin es ima es
o he “ eal-pe son ea men ” a e small and insigni ican a con en ional le els o s a is-
ical signi icance. This conclusion emains una ec ed when conside ing ea men e ec s
on he dis ibu ion o choices and ea men e ec s wi hin a ious popula ion subg oups.
Second, he dis ibu ional choices o esponden s a e ela i ely s able o e ime. The a e age
(in a- esponden ) in e empo al co ela ion o dis ibu ional choices is 0.56, which lies in
he ange o es - e es co ela ions o o he p e e ence su ey modules (Enke e al. 2022).
Fu he mo e, he in e empo al co ela ion is sligh ly highe in he incen i ized g oup, sug-
ges ing ha incen i es ha e a small posi i e e ec on educing measu emen e o in he
elici ed p e e ences. Thi d, we ind ha he na u e o he eco e ed ai ness p e e ences is
b oadly consis en wi h p e ious s udies on he US. Inequali y accep ance anges be ween
Gini coe icien s o 0.30 and 0.53 (e.g., Almås e al. 2020), he majo i y o esponden s
adop in e media e ai ness posi ions ha a e in luenced by disc e iona y a iables such as
educa ion and wo king hou s (e.g., Konow 2000), and he dis ibu ional choices o di e en
popula ion subg oups a e consis en wi h sel -se ing biases (Cos a-Fon and Cowell 2014).
2This choice o he ime lag is consis en wi h se e al ecen a icles ha alida e su ey-based measu emen
ools using ime-lags be ween one and wo weeks (Baue e al. 2020;Enkee al.2022; Falk e al. 2023;
Fallucchi e al. 2020).
123
883
P. Hu e, D. Weishaa
In summa y, he esul s om ou alida ion sugges ha he p oposed hypo he ical su ey
ool eco e s ai ness p e e ences ha a e consis en wi h incen i ized choices, s able, and
easonable in ligh o he exis ing li e a u e.
This pape con ibu es o wo s ands o he li e a u e. Fi s , we con ibu e o he li e a u e
on ai ness p e e ences. The e is a la ge li e a u e in economics and o he social sciences
ying o unde s and he na u e and ana omy o ai ness p e e ences in di e en popula ion
g oups (Almås e al. 2020; And e 2025; Cappelen e al. 2007; Gae ne and Schokkae 2012;
Jasso and Webs e 1999; Konow 2000; S a mans e al. 2017). In his pape , we alida e a
igne e-based su ey ool ha allows esea che s o in es iga e ai ness p e e ences in a
lexible and cos -e icien way. The e o e, his s udy p o ides a c ucial s ep o s eng hen he
me hodological oolki o in es iga ing ai ness p e e ences in applied esea ch. Fu he -
mo e, he e is a g owing li e a u e in es iga ing he ups eam de e minan s and downs eam
consequences o ai ness p e e ences (Ad iaans 2023; Alesina e al. 2018; Ande sen e al.
2023;Feh e al.2024). These s udies o en ely on su ey-based measu es o ai ness p e e -
ences. Ou esul s p o ide encou aging news o such esea ch designs as he consis ency o
hypo he ical and incen i ized choices sugges s ha su ey-based measu es a e no sys em-
a ically biased compa ed o hei incen i ized analogs. Second, we con ibu e o a g owing
me hodological li e a u e ha alida es su ey-based measu emen ools in a ious domains,
including isk, ime, compe i ion, and social p e e ences (Baue e al. 2020;Enkee al.2022;
Falk e al. 2023; Fallucchi e al. 2020). To he bes o ou knowledge, ou s udy is he i s
alida ion o an impa ial spec a o ask unde high mone a y s akes.
The emainde o he pape is o ganized as ollows. Sec ion 2desc ibes he su ey ool and
p o ides in o ma ion on he da a collec ion. In Sec ion 3, we p esen esul s o he e ec s o
he “ eal-pe son ea men .” Sec ion 4desc ibes he s abili y o ai ness p e e ences. Sec ion 5
p o ides a sugges i e compa ison o he eco e ed ai ness p e e ences ela i e o exis ing
li e a u e. Sec ion 6concludes he pape .
2 Su ey ool and da a collec ion
Su ey s uc u e Figu e 1p o ides an o e iew o he su ey used in ou analysis. The
su ey is s uc u ed in o wo wa es, each consis ing o mul iple modules. In he i s module
o he baseline wa e, we elici he demog aphic cha ac e is ics o esponden s. The second
and hi d modules measu e inequali y pe cep ions and ai ness p e e ences. The inal module
Baseline Wa e
Demog aphics Fai ness P e e ences Labo Ma ke s
Real Pe son
Hypo he ical
Real Pe son
Hypo he ical
5 Ques ions
5 Ques ions
7 Ques ions
7 Ques ions
Real Pe son
Hypo he ical
9 Ques ions
9 Ques ions
Follow-Up Wa e
Inequali y Pe cep ions
Financial Inc.
No Financial Inc.
Financial Inc.
No Financial Inc.
6 Ques ions
6 Ques ions
8 Ques ions
8 Ques ions
Financial Inc.
No Financial Inc.
10 Ques ions
10 Ques ions
Inequali y Pe cep ions Fai ness P e e ences
1week
Fig. 1 Su ey S uc u e. No e: This igu e isualizes he s uc u e o he su ey wi h wo wa es (baseline,
ollow-up). Each wa e consis s o mul iple modules. The modules on inequali y pe cep ions a e blu ed ou
since hey a e no co e ed in his pape . The main ea men g oup (con ol g oup) is highligh ed in ed (g ay)
123
884
Jus cheap alk? In es iga ing ai ness p e e ences
Fig. 2 Fai ness P e e ence Elici a ion, Exempla y Ques ion. No e: This igu e p o ides an example o a ques-
ion sc een in he ai ness p e e ence module. Each ques ion shows he cha ac e is ics o wo pe sons in six
dimensions (ea nings, gende , age, educa ion, pa en al educa ion, wo king hou s) in a able o ma . Each o he
wo pe sons has been alloca ed a andom pe son iden i ie om 1 o 9999. The o de ing o cha ac e is ics in he
able is andomized a he esponden -ques ion le el. A slide allows esponden s o selec hei p e e ed dis i-
bu ion o ea nings be ween he wo pe sons. The chosen alloca ion is also shown nume ically abo e he slide
con ains addi ional ques ions abou he labo ma ke . The wo modules o he ollow-up wa e
mi o he pe cep ions and p e e ence modules o he baseline wa e. In his pape , we ocus
exclusi ely on ai ness p e e ences.3
P e e ence module The module on ai ness p e e ences consis s o mul iple ques ions ha
ollow he design o ac o ial su ey expe imen s. Fac o ial su ey expe imen s a e well-
es ablished ools in he social sciences o assess p e e ences and belie s (Auspu g e al.
2015,2017;Fismane al.2020; Jasso and Webs e 1997,1999; Wiswall and Za a 2018). In
such expe imen s, esponden s e alua e mul iple hypo he ical scena ios ha a y a andom in
p e-de ined cha ac e is ics. The andom a ia ion o cha ac e is ics has wo main ad an ages.
Fi s , he design can eplica e he complexi ies o he eal wo ld. In pa icula , esponden s
a e o ced o make ade-o s and weigh he impo ance o di e en eal-wo ld a ibu es
agains each o he when making hei choices. Second, he simul aneous a ia ion o cha ac-
e is ics mi iga es expe imen e demand e ec s and social desi abili y biases—conce ns ha
a e pa icula ly ele an in he domain o ai ness p e e ences (Auspu g and Hinz 2014).
In ou su ey module, esponden s ecei e in o ma ion on he obse a ional cha ac e is ics
o wo pe sons—see Fig. 2 o an example.
We desc ibe hese pe sons in e ms o six cha ac e is ics, i.e., hei gende , age, own
educa ion, pa en al educa ion, wo king hou s, and labo ma ke ea nings. The choice o
hese cha ac e is ics was guided by he ollowing conside a ions. Fi s , we selec ed he
3The pe cep ions modules a e designed o assess esponden s’ pe cep ions o inequali y in he labo ma ke .
All ea men s in his module a e independen o he ea men s in he p e e ence module, allowing us o
analyze hese da a in isola ion.
123
885
P. Hu e, D. Weishaa
dimensionali y o he igne es by acknowledging he ade-o be ween p ecision and
complexi y in he design o ac o ial expe imen s. On he one hand, a highe numbe o
igne e cha ac e is ics has he bene i ha mo e con ounding ac o s can be held cons an .
On he o he hand, inc easing he numbe o cha ac e is ics also pu s a la ge cogni i e bu den
on pa icipan s when making dis ibu ional choices. Wi h six dimensions, we op ed o an
in e media e le el o complexi y (Auspu g and Hinz 2014). Second, we hen selec ed cha ac-
e is ics ha a e bo h ele an o unde s anding ea nings inequali y and widely a ailable in
s anda d household su ey da a (Bick e al. 2022; Goldin 2014;Lemieux2006; Magnac and
Roux 2021; Mazumde 2005). We a e awa e ha he esul ing selec ion o cha ac e is ics is
somewha disc e iona y. The e o e, we also alida e ou selec ion ex-pos by asking espon-
den s abou cha ac e is ics hey conside pa icula ly impo an when making dis ibu ional
choices. Appendix Figu e A1 shows ha i e ou o he six selec ed cha ac e is ics a e cho-
sen by a leas en pe cen o esponden s. Fu he mo e, many o ou chosen cha ac e is ics
co ela e s ongly wi h o he cha ac e is ics ha pa icipan s ind impo an , e.g., age (ou
selec ion) and wo k expe ience ( esponden s selec ion).
Each cha ac e is ic can ake mul iple exp essions. Fo ins ance, he cha ac e is ic o edu-
ca ion can ake h ee alues, i.e., High School D opou ,High School,o Uni e si y.The
po en ial exp essions o each cha ac e is ic a e shown in Table 1. The o de o cha ac e is-
ics is andomized a he esponden -ques ion le el o ensu e ha esul s a e no d i en by
o de e ec s (Day e al. 2012).
Combining all po en ial exp essions yields a se o 720 p o iles (2×2×3×3×4×5) and
a se o 258,840 (720×719
2) unique uno de ed p o ile pai s. In he ollowing, we will e e o
each unique p o ile as a igne e pe son and each unique p o ile pai as a igne e.Weshow
esponden s a selec ion o igne es de e mined ia a andom d aw om he ull se . Based
on he p esen ed in o ma ion, esponden s can use a slide o adjus he ini ial ea nings and o
implemen hei p e e ed ea nings dis ibu ion in he igne es. All esponden s answe ed he
same selec ion o igne es, wi h some a ia ion in he numbe o igne es pe esponden —
see also ou discussion on he “leng h ea men ” below.
T ea men 1: “Real-pe son ea men ” To assess whe he hypo he ical ques ions in ac-
o ial su eys eco e ai ness p e e ences consis en wi h incen i ized expe imen s, we
andomize esponden s in o wo g oups. Responden s in he con ol g oup comple e a se ies
o hypo he ical dis ibu ion asks. Responden s in he ea men g oup comple e he same
Table 1 Fai ness P e e ence Elici a ion, Cha ac e is ics and Exp essions
Cha ac e is ic Numbe Displayed Values o Exp ession
Gende 2 Male / Female
Age 2 (26 , 35 , 40) / (50, 55, 59)
Educa ion 3 High School D op-Ou / High School / Uni e si y
Pa en al Educa ion 3 High School D op-Ou / High School / Uni e si y
Wo king Hou s 4 (2, 5) / (20, 27, 31) / (39, 40) / (48, 51, 59)
Ea nings 5 $1,100 / $2,700 / $4,100 / $5,900 / $11,400
No e: This able shows he cha ac e is ics displayed in ques ions o he ai ness p e e ence module (column
1), he numbe o coa se exp essions wi hin each cha ac e is ic (column 2), and he displayed alues o each
exp ession (column 3). We employ a second andomiza ion o age and wo king hou s o display exac alues
ins ead o anges. Fo each o he wo age ange g oups (25-44, 45-65) and each o he ou wo king hou s
ange g oups (1-9, 10-34, 35-44, Mo e han 44), we d aw om in ege s in he espec i e ange
123
886
Jus cheap alk? In es iga ing ai ness p e e ences
se ies o asks. Howe e , hey a e in o med ha one o he igne e pe sons is a eal pe son.
Fu he mo e, hey a e in o med ha he choice o one esponden will be selec ed o de e -
mine he mon hly ea nings o his pe son. Responden s know ha he eal pe son is be ween
25 and 65 yea s old, is a esiden o he Uni ed S a es, and wo ks in a job o ea n money.
Impo an ly, his in o ma ion does no allow esponden s o dis inguish he eal pe son om
any o he igne e pe son. They also know ha he o al ea nings o he eal pe son consis
o wo pa s: (i) a ixed paymen o $500 and (ii) a lexible paymen ha can be changed by
he esponden .
The esea ch eam hi ed a eal pe son in Augus 2022—see Appendix Table A1 o he
cha ac e is ics o he eal pe son as displayed in hei igne e. The hi ed pe son was in o med
ha he ixed- e m con ac would ha e a du a ion o one mon h and ha he exac amoun
o hei ea nings would be de e mined by ano he indi idual; howe e , hey did no know he
exac p ocess o how his happens. The pe son only knew hei o al ea nings would be $500
o abo e. To de e mine he po en ial ea nings o he eal pe son, we p oceeded in wo s eps.
Fi s , we alloca ed he eal pe son a mon hly ea nings alue by andomly d awing om he
se o po en ial ea nings displayed in Table 1. Second, we andomly ma ched he eal pe son
wi h ano he (hypo he ical) igne e pe son. This wo-s ep p ocedu e ixed he olume o
ea nings in he igne e o he eal pe son a $5,200. The e o e, including he ixed paymen
o $500, he uppe bound o po en ial ea nings o he eal pe son was $5,700. This uppe
bound would be ealized i he decisi e esponden alloca ed all he igne e ea nings o he
eal pe son. Impo an ly, he igne e wi h he eal pe son was p esen ed alongside all o he
igne es, and he iden i y o he eal pe son was concealed om esponden s. Consequen ly,
esponden s also aced si ua ions in which he po en ial ea nings implica ions we e e en
highe . The a e age ea nings olume in he displayed igne es—and he e o e he a e age
uppe bound o po en ial paymen s o he eal pe son om he esponden s’ pe spec i e—
was $11,420 in he main igne es o he p e e ence module. Fu he mo e, we ensu ed he
salience o he eal-wo ld consequences h ough a aining ask. Speci ically, we ained
esponden s on an example igne e and highligh ed he po en ial ea nings consequences o
he eal pe son a e esponden s had made hei dis ibu ional choices.
Ou incen i iza ion is based on high s akes wi h a low p obabili y o implemen a ion.
Al e na i ely, we could ha e used an incen i e scheme based on lowe s akes bu a highe
p obabili y o implemen a ion. Due o ou ocus on eal-wo ld cha ac e is ics, he o me
pa h appea s mo e na u al since i allows us o ocus on payou s ha e lec ealis ic mon hly
wages d awn om he US wo king popula ion. Howe e , we con end ha a gene aliza ion
o ou esul s o al e na i e incen i e schemes is an in e es ing pa h o u u e esea ch.
T ea men 2: Leng h ea men To assess he sensi i i y o ou conclusions o he leng h
o he su ey, we a y he numbe o igne es in he su ey module. All esponden s made
a leas i e dis ibu ional choices; howe e , 1/3 o esponden s ecei ed 2 o 4 addi ional
igne es, espec i ely. Fo he main alida ion, we ocus on he i s i e ques ions answe ed
by all esponden s and use he a ia ion om he leng h ea men s in obus ness analyses.
The assignmen o he leng h ea men s was independen o he alloca ion o he “ eal-pe son
ea men .” The e o e, we ob ain six g oups o app oxima ely equal size ha a y in hei
exposu e o he “ eal-pe son ea men ” and he su ey leng h (Fig. 1).
Baseline wa e We adminis e ed he baseline wa e o he su ey o 1,602 adul ci izens o he
Uni ed S a es. Da a collec ion ook place be ween Oc obe and No embe 2022. Responden s
we e con ac ed h ough he su ey p o ide Dyna a and ecei ed a pa icipa ion paymen
123
887
P. Hu e, D. Weishaa
depending on he expec ed su ey leng h.4The mean (median) comple ion ime was 24 (19)
minu es o he baseline wa e (Appendix Figu e A2).
Responden s we e a ge ed o ma ch he popula ion along i e dimensions (gende , age,
educa ion, employmen s a us, and egion o esidence). In Panel A o Table 2, we compa e
ou sample o he Ame ican Communi y Su ey (ACS) ega ding he a ge ed cha ac e -
is ics. In gene al, we ma ch he da a well. Ou sample has a sligh unde ep esen a ion o
people wi h low educa ion. In addi ion, we ecei ed o e -p opo ional (unde -p opo ional)
esponses om mid-wes e n s a es (sou he n s a es). Panel B o Table 2 u he shows ha ou
sample is also b oadly ep esen a i e ega ding o he obse able cha ac e is ics like e hnici y
and income.
We ake a ious s eps o ensu e he quali y o su ey answe s. Fi s , we included an
a en ion check a he beginning o he p e e ence module. Responden s who ailed his
a en ion check we e sc eened ou di ec ly and we e no pa o he sample. Second, we
asked a aining ques ion a e explaining he asks in he p e e ence module. A ound 72%
o esponden s passed his ques ion on he i s y. In obus ness checks, we show ha ou
esul s a e no sensi i e o excluding esponden s who did no pass he aining ques ion on
hei i s a emp .
Follow-up wa e We in i ed esponden s o a ollow-up wa e one week a e hey comple ed
he baseline wa e. Among o he s, he ollow-up wa e consis s o a ai ness p e e ence module
wi h six ques ions. As in he baseline wa e, esponden s aced a selec ion o dis ibu ion asks
based on he igne es om ou su ey module. Th ee o he ques ions a e epe i ions om he
baseline wa e, which we e p esen ed o he esponden s in an ob usca ed way. This ea u e
allows us o assess he s abili y o ai ness p e e ences o e ime.
The ollow-up ob ained a esponse a e o a ound 44%, and a ound 90% o esponden s
answe ed wi hin wo weeks. The esul ing sample is sligh ly olde bu o he wise b oadly com-
pa able o ou baseline sample ega ding obse able demog aphics (Appendix Table A2). The
mean (median) comple ion o he ollow-up wa e was 14 (9) minu es (Appendix Figu e A2).
3 Effec s o he “ eal-pe son ea men ”
In his sec ion, we in es iga e whe he he po en ial o eal-wo ld implemen a ion a ec ed
he dis ibu ional choices o esponden s. Fi s , we p esen me hodological checks on he an-
domiza ion and he anonymi y o he eal pe son. Second, we p esen he ea men e ec s
on dis ibu ional choices. Thi d, we in es iga e po en ial he e ogenei ies by popula ion sub-
g oups. Fou h, we p esen obus ness analyses. All analyses in his sec ion a e p e- egis e ed
unless no ed o he wise.
Balancing To gi e ou es ima es a causal in e p e a ion, he ea men assignmen mus be
unco ela ed wi h any esponden cha ac e is ics ha may p edic hei dis ibu ional choices.
The e o e, we es he balance o esponden s’ socio-demog aphic cha ac e is ics be ween
he ea men and he con ol g oup. In pa icula , we eg ess he ea men s a us T ea
io
esponden ion Kp e-speci ied indi idual cha ac e is ics deno ed by xk
i:
xk
i=αk+βkT ea i+εk
i.(1)
4Fo he baseline wa e, esponden s we e able o ea n be ween $0.20 and $1.50. Fo he sho e ollow-up
wa e, he paymen a ied be ween $0.10 and $1.20. The a ying pa icipa ion paymen was used by Dyna a
o ob ain esponses om demog aphic segmen s o he popula ion ha a e mo e di icul o each.
123
888
Jus cheap alk? In es iga ing ai ness p e e ences
Table 6 Alloca ions, “Real-Pe son T ea men ”, Al e na i e Samples
Ques ion NPoin Es ima e Con ol Mean Model
p- alue
Resample
p- alue
Romano-Wol
p- alue
Panel A: Di ec ly Passed T aining Ques ion
Q1 1154 168.87 1308.25 0.237 0.244 0.686
Q2 1154 31.76 -2574.27 0.774 0.777 0.949
Q3 1154 112.27 3320.52 0.813 0.813 0.949
Q4 1154 -147.78 -4508.78 0.613 0.629 0.935
Q5 1154 -290.95 6884.88 0.361 0.339 0.774
Panel B: Exclude Response Time Ou lie s
Q1 924 162.99 1394.95 0.317 0.321 0.766
Q2 924 109.22 -2699.89 0.364 0.375 0.766
Q3 924 -270.11 4030.09 0.606 0.607 0.854
Q4 924 -30.23 -4623.23 0.925 0.933 0.933
Q5 924 -477.60 7159.91 0.169 0.177 0.572
No e: This able p esen s esul s o he eg ession analysis ou lined in Eq. 2 o di e en es ic ed samples.
Panel A displays esul s o he sample o esponden s ha passed he aining ques ion o he ai ness p e e ence
module a he i s y. Panel B p esen s esul s o a sample ha excludes esponden s wi h high (abo e
p90) and low (below p10) esponse imes o he ai ness p e e ence module. We p esen poin es ima es
o he coe icien o in e es βj, he mean o he con ol g oup, and he e oskedas ici y obus unco ec ed
analy ical p- alues (model p- alues), unco ec ed boo s apped p- alues ( esample p- alues), as well as p-
alues adjus ed o mul iple hypo hesis es ing using 1000 boo s ap eplica ions. See Appendix Tables B6-B7
o he co esponding balancing and Kolmogo o -Smi no es s
Sou ce: Own calcula ion based on su ey esponses
Resul s in Table 8show ha he e a e i ually no di e ences be ween ea men and
con ol g oups, sugges ing ha he e o o esponden s does no dec ease when acing
hypo he ical ins ead o incen i ized scena ios.8
In his sec ion, we ha e shown ha he e a e no sys ema ic di e ences in dis ibu ional
choices be ween hypo he ical and incen i ized scena ios. This conclusion holds o a e age
alloca ions, dis ibu ions o alloca ions, and wi hin a ious demog aphic subg oups. Fu he -
mo e, his conclusion is obus o a ious sensi i i y checks, such as excluding low-quali y
esponses. In summa y, he esul s sugges ha hypo he ical igne es cap u e he same ai -
ness p e e ences as hei incen i ized analogs.
4 S abili y o ai ness p e e ences
In his sec ion, we in es iga e whe he he su ey module cap u es genuine ai ness p e e -
ences ha a e s able o e ime. In pa icula , we use he longi udinal a ia ion be ween he
baseline wa e and he ollow-up wa e o he su ey. The ollow-up wa e consis s o a ai ness
8To limi he in luence o ex eme ou lie s, we ocus on esponden s whose esponse ime o a pa icula
ques ion is below he 99 h pe cen ile o he ques ion-speci ic esponse ime dis ibu ion. In Appendix Table 25,
we epea he exe cise wi hou his es ic ion. T ea men e ec s inc ease due o single ou lie s in he ea men
and con ol g oups. Howe e , none o he di e ences is s a is ically signi ican , and ou gene al conclusion
emains una ec ed.
123
895
P. Hu e, D. Weishaa
Table 7 Alloca ions, “Real-Pe son T ea men ”, by Su ey Leng h Subg oups
Rejec ed Hypo heses a 5% (10%)
Module Leng h NModel p- alue Resample p- alue RW p- alue
5 Ques ions 534 0 (1) 0 (1) 0 (0)
7 Ques ions 536 0 (0) 0 (0) 0 (0)
9 Ques ions 532 1 (1) 1 (1) 0 (0)
No e: This able p esen s esul s o he eg ession analysis ou lined in Eq. 2 o a subsample o esponden s
based on he su ey module leng h, i.e, whe he esponden s answe ed i e, se en, o nine ques ions in he
p e e ence module. We p esen he numbe o esponden s, he numbe o ejec ed hypo heses acco ding o
he e oskedas ici y obus model p- alues, esample p- alues, and p- alues adjus ed o mul iple hypo hesis
es ing. In o al, we es i e, se en, and nine hypo heses o each subsample depending on he numbe o
ques ions. De ailed in o ma ion on eg ession esul s a e shown in Appendix Tables B22 - B24
Sou ce: Own calcula ions based on su ey esponses
p e e ence module wi h six ques ions, h ee o which a e epe i ions om he baseline su ey.
To a oid esponden s ancho ing hei esponses on hei answe s in he baseline su ey, we
ob usca e he epea ed ques ions by mixing hem in andom o de wi h no el ques ions ha
ha e no been shown o esponden s p e iously. We p esen esul s in h ee s eps. Fi s , we
p esen in e empo al co ela ions based on he pooled ollow-up sample. Second, we in es i-
ga e whe he hese in e empo al co ela ions a y by ea men s a us in he baseline su ey.
Thi d, we p esen obus ness analyses. We egis e ed he ollow-up su ey in ou p e-analysis
plan. Howe e , since he su ey p o ide exp essed conside able unce ain y abou he likely
esponse a es, we did no p e-speci y he associa ed analyses p esen ed in his sec ion.
In e empo al co ela ions We es ima e in e empo al co ela ions h ough o dina y leas -
squa es using he ollowing model:
yij, =α+σyij, −1+εij,(3)
whe e yij is again he di e ence in alloca ions o igne e pe sons A and B by esponden
iin igne e jin he baseline wa e ( −1) and he ollow-up wa e ( ), espec i ely. In all
es ima ions, we s anda dize yij, and yij, −1on he es ima ion samples such ha hey
Table 8 P e e ences, Response Time (Min.), “Real-Pe son T ea men ”
Ques ion NPoin Es ima e Con ol Mean Model
p- alue
Resample
p- alue
Romano-Wol
p- alue
Q1 1585 0.00 1.00 0.892 0.904 0.989
Q2 1585 0.00 0.46 0.936 0.940 0.989
Q3 1585 0.00 0.50 0.837 0.839 0.989
Q4 1585 -0.01 0.37 0.501 0.555 0.936
Q5 1585 0.01 0.36 0.729 0.743 0.981
No e: This able p esen s esul s o he eg ession analysis ou lined in Eq. 2using he esponse ime in minu es
as he dependen a iable. Fo e e y ques ion, we ocus on esponse imes below he 99 h pe cen ile o he
ques ion-speci ic esponse ime dis ibu ion. We p esen poin es ima es o he coe icien o in e es βj, he
mean o he con ol g oup, and he e oskedas ici y obus unco ec ed analy ical p- alues (model p- alues),
unco ec ed boo s apped p- alues ( esample p- alues), as well as p- alues adjus ed o mul iple hypo hesis
es ing using 1000 boo s ap eplica ions
Sou ce: Own calcula ion based on su ey esponses
123
896
Jus cheap alk? In es iga ing ai ness p e e ences
Fig. 4 S abili y o Fai ness P e e ences, Co ela ion. No e: This igu e displays he in e empo al co ela ion
be ween he baseline and ollow-up wa e (Fig. 4a) and he cumula i e dis ibu ion unc ion o wi hin-
esponden co ela ions (Fig. 4b). In Fig. 4a, a iables a e s anda dized on he ull sample, and he line
indica es he line o bes i om a linea eg ession. In Fig. 4b, a iables a e s anda dized a he indi idual
le el, and he solid (dashed) line indica es he mean (median) co ela ion ac oss esponden s. Sou ce: Own
calcula ion based on su ey esponses
ha e a mean o ze o and a s anda d de ia ion o one. As a esul , es ima es o σcan be
in e p e ed as in e empo al co ela ion coe icien s.
Figu e 4a plo s he aw s anda dized da a o yij, agains yij, −1, wi h he i ed line
indica ing he poin es ima e o σ. The in e empo al co ela ion is es ima ed a 0.61, sug-
ges ing sizable s abili y o dis ibu ional choices o e ime.
Figu e 4b isualizes he cumula i e dis ibu ion o in e empo al co ela ions a he indi-
idual le el. Fo each esponden , es ima es o σa e based on he h ee epea ed ques ions
om baseline and ollow-up. Mo e han 80% o esponden s display a posi i e co ela ion,
and mo e han 65% ha e a co ela ion o 0.50 o highe . The mean (median) co ela ion
ac oss esponden s is a ound 0.56 (0.91). These high in a- esponden co ela ions ea i m
ou conclusion ha he eco e ed dis ibu ional choices a e ai ly s able o e ime o mos
esponden s in ou sample.
E ec o “ eal-pe son ea men ” The es ima ed in e empo al co ela ions in Eq. 3may
be a enua ed by measu emen e o in yij, −1, i.e., he dis ibu ional choices in he baseline
wa e. The e o e, we can use es ima es o σ o assess whe he incen i ized su ey ques ions
inc ease he signal- o-noise a io in he eco e ed ai ness p e e ences. I measu emen e o
in yij, −1was less p onounced in incen i ized scena ios, σwould be signi ican ly highe in
he ea men g oup han in he con ol g oup. Such a inding would sugges ha incen i ized
su ey modules yield less noisy es ima es o ai ness p e e ences.9
In Fig. 5a, we eplica e Fig. 4a by spli ing ou sample in o he ea men and con ol
g oups om he baseline wa e. Es ima es o σa e sligh ly highe in he ea men (0.66) han
in he con ol g oup (0.55). The di e ence o 0.10 is s a is ically signi ican a he i e pe cen
le el (p- alue=0.02). In Fig. 5b, we show ha he di e ence in s abili y is less p onounced
9In he ollow-up wa e, all ques ions we e hypo he ical. Since we use yij, as ou comes in equa ion (2),
he associa ed (classical) measu emen e o will no bias ou es ima es o σ.
123
897
P. Hu e, D. Weishaa
Fig. 5 S abili y o Fai ness P e e ences, “Real-Pe son T ea men ”. No e: This igu e displays he in e empo al
co ela ion be ween he baseline and ollow-up wa e (Fig. 5a) and he cumula i e dis ibu ion unc ion o
wi hin- esponden co ela ions (Fig. 5b) sepa a ely o ea men and con ol g oups. In Fig. 5a, a iables a e
s anda dized a he g oup le el, and solid lines indica e lines o bes i om a linea eg ession. In Fig. 5b,
a iables a e s anda dized a he indi idual le el, and solid lines indica e mean co ela ions ac oss esponden s.
Sou ce: Own calcula ion based on su ey esponses
when conside ing co ela ions a he indi idual le el. The a e age in a-indi idual co ela ion
is s ill sligh ly highe in he ea men (0.57) han in he con ol g oup (0.54). Howe e , he
di e ence o 0.03 is no s a is ically signi ican a con en ional le els o s a is ical signi icance
(p- alue=0.49).
These pa e ns sugges ha incen i ized su ey modules may yield sligh ly less noisy
es ima es o ai ness p e e ences. Howe e , hese gains a e ela i ely mode a e and may be
quickly ou weighed by he bene i s o an unincen i ized su ey, e.g., lowe cos , he po en ial
o a ge b oade popula ion samples, e c.
Robus ness We again implemen a se ies o obus ness checks o analyze he sensi i i y o
he p e ious indings. These obus ness checks a e summa ized in Table 9.
Fi s , we check whe he in e empo al co ela ions change when excluding low-quali y
answe s om ina en i e esponden s who do no pass he aining ques ion on he i s y.
In e empo al co ela ions inc ease sligh ly bu emain e y close o ou ull sample es ima e.
In an al e na i e es , we exclude esponden s in he ails o he esponse ime dis ibu ion.
This sample es ic ion has i ually no e ec on he es ima ed in e empo al co ela ions.
Second, we check whe he he es ima ed in e empo al co ela ions a e especially d i en
by indi iduals who always lea e he slide close o i s o iginal posi ion. In pa icula , we
exclude obse a ions whe e esponden s lea e he igne e slide wi hin a wo-sided i e
pe cen age poin band a ound he ini ial ea nings dis ibu ion in he baseline and he ollow-
up. Indeed, he e seems a sligh d op in in e empo al co ela ions when excluding hese
esponden s. Howe e , we also emphasize ha he implemen ed es is likely oo s ingen .
On he one hand, we exclude esponden s who lea e he slide unal e ed in bad ai h. On he
o he hand, we also exclude esponden s wi h genuine libe a ian p e e ences. The e o e, we
in e p e he s ill subs an ial in e empo al co ela ion as a posi i e signal ha we can eco e
s able p e e ences in a eas u he away om ini ial income posi ions.
123
898
Jus cheap alk? In es iga ing ai ness p e e ences
Table 9 S abili y o Fai ness P e e ences, Co ela ion, Robus ness
Full Sample Res ic ed Sample
T aining Ques ion Response Time No a S a us Quo
Agg ega e 0.605 (2127) 0.644 (1710) 0.609 (1707) 0.541 (1206)
Q1 0.375 (424) 0.389 (338) 0.391 (338) 0.413 (295)
Q2 0.328 (426) 0.335 (346) 0.298 (340) 0.436 (199)
Q3 0.407 (425) 0.390 (334) 0.399 (337) 0.356 (326)
Q4 0.332 (427) 0.337 (341) 0.312 (352) 0.317 (193)
Q5 0.382 (425) 0.377 (351) 0.335 (340) 0.497 (193)
No e: This able displays in e empo al co ela ions be ween he baseline and ollow-up wa es o he ull
sample and h ee es ic ed samples. The i s es iced sample ocuses on esponden s ha passed he aining
ques ion o he ai ness p e e ence module in he baseline wa e a he i s y. The second es ic ed sample
excludes esponden s wi h high (abo e p90) and low (below p10) esponse imes o he ai ness p e e ence
module in he baseline wa e. The hi d es ic ed sample excludes esponden s whose alloca ed sha es a e a
mos 5 pe cen age poin s away om he s a us quo dis ibu ion o ea nings. Va iables a e s anda dized on he
sample used in he co esponding eg ession. Sample sizes a e shown in pa en hesis
Sou ce: Own calcula ions based on su ey esponses
Thi d, all p e ious conclusions hold when calcula ing in e empo al co ela ions a he
le el o indi idual ques ions. In ou p e ious discussion, we especially ocused on in e em-
po al co ela ions a he indi idual le el. This is he app op ia e le el o analysis since
ac o ial su ey designs mos ly use in a- esponden a ia ion ac oss mul iple igne es o
iden i y he ele an p e e ences (Wiswall and Za a 2018). Howe e , depending on he
design, esea che s may wan o in e p e e ences om ewe igne es pe indi idual han
in ou se ing. In Table 9, we, he e o e, assess he ex eme case whe e p e e ences would be
iden i ied based on a single ques ion only. In his case, he p e e ence signal is mo e noisy,
ansla ing in o lowe in e empo al co ela ions. None heless, e en in he ex eme case o
using only one igne e, he co ela ions a e s ill subs an ial, anging om 0.33 o 0.41 in he
ull sample.
In his sec ion, we ha e shown ha he dis ibu ional choices a e ela i ely s able o e ime.
This conclusion is obus o a ious sensi i i y checks, among o he s, excluding low-quali y
esponses. The p esence o incen i es sligh ly dec eases he noise in elici ed p e e ences.
This dec ease in noise, howe e , is ai ly mode a e and may be quickly ou weighed by he
po en ial bene i s o unning an unincen i ized su ey. In summa y, he esul s sugges ha
hypo he ical dis ibu ion asks can yield high-quali y da a on s able ai ness p e e ences.
5 Na u e o ai ness p e e ences
In his sec ion, we accompany he main me hodological alida ion o he p e ious sec ions
by gi ing some sugges i e insigh s in o he na u e o elici ed ai ness p e e ences. To be su e,
his analysis comes wi h ca ea s. The p ima y pu pose o his pape is o assess he measu e-
men o ai ness p e e ences in hypo he ical se ings as compa ed o he “gold s anda d” o
incen i ized expe imen s. The e o e, we made se e al me hodological choices ha p e en
a ull subs an i e analysis o he eco e ed p e e ences. Fo example, o maximize s a is ical
powe o de ec di e ences be ween he ea men and con ol g oup, we show all esponden s
he same andomly selec ed subse o igne es. Consequen ly, igne e cha ac e is ics a e
123
899
P. Hu e, D. Weishaa
no equally ep esen ed, and co ela ions may exis among hem. These ea u es may a ec
esponden s’ willingness o ole a e inequali y and how hey inco po a e di e en igne e
cha ac e is ics in o hei choices. The e o e, we iew he ollowing analysis as a sugges-
i e es o whe he he eco e ed p e e ences a e b oadly consis en wi h indings om he
exis ing li e a u e.
Wi h hese ca ea s in mind, we will p esen he esul s o his sec ion as ollows. Fi s ,
we analyze he le el o inequali y implemen ed by esponden s. Second, we will analyze he
p e alence o di e en ai ness ypes in ou sample. Las ly, we show he sensi i i y o ai ness
p e e ences o di e en cha ac e is ics o he e alua ed igne e pe sons. In all analyses, we
will ocus on unincen i ized scena ios om he con ol g oup. Howe e , ou conclusions
emain una ec ed when ocusing on he incen i ized sample—see Appendix Figu es C4, C6
and Appendix Table C26 o eplica ions o he main exhibi s o his sec ion based on he
ea men g oup. The analyses o his sec ion a e explo a o y. The e o e, hey ha e no been
egis e ed in ou p e-analysis plan.
Implemen ed inequali y Figu e 6compa es implemen ed inequali y by esponden s o he
ini ial inequali y sepa a ely o each igne e.
The implemen ed Gini coe icien s show subs an ial a ia ion ac oss igne es (0.30–0.53).
Fo compa ison, Almås e al. (2020) use a ep esen a i e sample o Ame ican esponden s o
show ha hey would implemen a Gini coe icien o 0.35 (0.54) i he income-gene a ing
p ocess we e pu ely based on luck (me i ). This sugges s he ange o implemen ed inequali y
ac oss di e en scena ios in ou se ing is plausible.
Fig. 6 Gini Coe icien . No e: This igu e compa es implemen ed Gini coe icien s (g ay do s) o ini ial Gini
coe icien s (g een c osses) in each igne e. G ay ba s indica e 95 pe cen con idence in e als. Gini coe i-
cien s a e calcula ed a he esponden le el as |x−y|
x+ywhe e x(y) is he amoun alloca ed o igne e pe son A
(B). We ocus exclusi ely on esponden s in he con ol g oup, i.e., hose esponden s who aced hypo he ical
scena ios. Appendix Figu e C4 eplica es he analysis o esponden s in he ea men g oup. Sou ce: Own
calcula ions based on su ey esponses
123
900
Jus cheap alk? In es iga ing ai ness p e e ences
In Appendix Figu e C5, we u he mo e illus a e how inequali y accep ance a ies ac oss
esponden s wi h di e en socio-demog aphic cha ac e is ics. Responden s who a e mo e
inequali y-accep ing end o be olde , mo e educa ed, and wo k longe hou s. Those wi h a
lowe inequali y ole ance end o be emale and non-whi e. Again, hese pa e ns a e b oadly
consis en wi h exis ing li e a u e. Fo ins ance, he indings o Almås e al. (2020) indica e
ha women and indi iduals wi h lowe educa ional a ainmen a e less accep ing o inequali y
compa ed o men and hose wi h highe educa ion. The au ho s in e p e hese pa e ns in he
ligh o po en ial sel -se ing biases in ai ness p e e ences. I is eassu ing ha ou su ey
eplica es hese pa e ns as well.
Fai ness ypes Expe imen al li e a u e has ocused on es ima ing he p e alence o di e en
ai ness ypes ha can be mapped o ai ness p inciples in he philosophical li e a u e—see
Almås e al. (2024b) o a ecen o e iew. On one end o he spec um is he egali a ian posi-
ion. Egali a ians conside all inequali ies un ai , ega dless o how hese inequali ies come
abou . The e o e, he egali a ian posi ion p esc ibes an equal income dis ibu ion in any dis-
ibu i e si ua ion. A he opposi e end o he spec um is he libe a ian posi ion. Libe a ians
conside all inequali ies ai ega dless o how hese inequali ies come abou (Nozick 1974).
The e o e, he libe a ian posi ion p esc ibes a dis ibu ion o income ha co esponds o he
ini ial dis ibu ion in any dis ibu i e si ua ion. Be ween hese wo ex eme posi ions, he e
a e se e al in e media e posi ions, such as he esponsibili y-sensi i e posi ions p oposed by
A neson (1989); Cohen (1989); Dwo kin (1981a,b). These in e media e posi ions ad oca e
o dis inguishing be ween di e en sou ces o inequali y, such as disc e iona y choices,
abili y, p e e ences, o ci cums an ial ac o s.
We es ima e he p e alence o he egali a ian posi ion by calcula ing he sha e o espon-
den s who implemen equal spli s in all igne es. Simila ly, we es ima e he p e alence o he
libe a ian posi ion by calcula ing he sha e o esponden s who accep ini ial inequali y in
all igne es. When calcula ing hese sha es, we allow o “ embling hand” mis akes (Choi
e al. 2007). Fo ou baseline es ima es, we use wo-sided i e pe cen age poin bands a ound
he egali a ian and libe a ian answe s o a igne e and allow esponden s o be ou side o
he co esponding band o a mos one igne e wi hou epe cussions on hei classi ica ion
as egali a ians o libe a ians. We es ima e he p e alence o he in e media e posi ion as he
emaining sha e o esponden s who a e no classi ied as egali a ians o libe a ians.
Table 10 shows he esul s, whe e he highligh ed a eas ep esen ou baseline es ima es.
A ound wo pe cen o esponden s a e classi ied as egali a ian, whe eas a ound nine pe cen
Table 10 P e e ence Types
Sha e Egali a ians (%) Sha e Libe a ians (%)
Max. abs. di e ence (pp) 2 5 10 2 5 10
Allow o 0 inconsis en answe s 0.00 0.51 1.73 5.68 6.41 7.65
Allow o 1 inconsis en answe s 0.00 1.46 3.75 7.89 9.47 13.84
Allow o 2 inconsis en answe s 0.89 2.71 7.69 10.75 13.90 22.22
No e: This able p esen s sha es o egali a ians (libe a ians) acco ding o consis en choices in all ques ions
o he baseline wa e. We also a y he leniency o he classi ica ion by allowing o 0, 1, 2 answe s ha
a e inconsis en wi h egali a ian (libe a ian) choices. In he baseline (highligh ed es ima es), we allow o a
de ia ion o +/-5 pe cen age poin s and inconsis en choices in one ques ion only. We ocus exclusi ely on
esponden s in he con ol g oup, i.e., hose esponden s who aced hypo he ical scena ios. Appendix Table
C26 eplica es he analysis o esponden s in he ea men g oup
Sou ce: Own calcula ions based on su ey esponses
123
901
P. Hu e, D. Weishaa
a e classi ied as libe a ians. The emaining 89% pe cen o esponden s adop in e media e
posi ions. The e o e, mos esponden s adop ai ness posi ions ha a y wi h he cha ac e -
is ics o he espec i e igne e. We no e ha his conclusion does no a y wi h he leniency
wi h which we accep “ embling hand” mis akes. E en in he mos lenien speci ica ions
whe e we allow o wo-sided en pe cen age poin bands and wo inconsis en answe s, he
sha e o esponden s adop ing in e media e posi ions is s ill 70%. Fu he mo e, we no e ha
condi ional on he adop ed ule o “ embling hand” mis akes, he p esen ed es ima es o
he p e alence o egali a ian and libe a ian posi ions should be in e p e ed as uppe bounds.
We only p esen ed esponden s wi h a limi ed selec ion o i e o nine igne es. The e o e, in
addi ional ques ions, he numbe o di e gences om he egali a ian and libe a ian posi ions
can s ay cons an a bes bu no dec ease. The es ima ed sha es o egali a ians and libe a -
ians in he US a e smalle han he co esponding sha es es ima ed in Almås e al. (2020).
Thei es ima es classi y 15% and 29% o he US popula ion as egali a ians and libe a ians,
espec i ely. This di e ence may be a ionalized by a ia ions in how di e en p e e ence
ypes a e iden i ied o by he inc eased ichness o he dis ibu ional scena ios in ou se ing.10
Since he igne es p o ide mul idimensional in o ma ion on he ea nings- ele an cha ac-
e is ics o he ecipien s, esponden s can exp ess posi ions ha de ia e om he pola cases
o egali a ian/libe a ian ai ness p e e ences in mo e nuanced ways.
Impo ance o igne e pe son cha ac e is ics In he las s ep, we analyze he impac o
pa icula ea nings- ele an cha ac e is ics on he ai ness p e e ences o esponden s. To his
end, we ans o m ou da a as ollows. We c ea e a da a se whe e each ow ep esen s one
pe son m om igne e j. Then, we eplica e hese da a o each esponden iwho made a
dis ibu ional choice o igne e jand include he co esponding income alloca ions yim(j)
as he ou come a iable o in e es . S acking hese da a, we ob ain a panel da a se wi h
mul iple obse a ions o each igne e pe son m(j)and each esponden i.
We hen es ima e he ollowing model ia o dina y leas -squa es:
ln yim(j)=β1gende m(j)+β2agem(j)+β3educm(j)
+β4educpa m(j)+β5hou sm(j)+β6ln ea nm(j)
+θ[ea nA(j)+ea nB(j)]+im(j).
(4)
The igh -hand side a iables in he i s wo lines o Eq. 4 ep esen he six igne e
cha ac e is ics conside ed in ou ai ness p e e ence module. The associa ed coe icien s
β1−β6cap u e he linea e ec o pe sonal cha ac e is ics on ai ea nings condi ional on
he emaining igne e cha ac e is ics. We con ol non-pa ame ically o he o al sum o
igne e ea nings ea nA(j)+ea nB(j)by including co esponding ixed e ec s. The eby,
we accoun o he ac ha igne e pe sons who a e pai ed wi h high-ea ning pe sons in hei
igne e may mechanically ecei e highe income alloca ions. S anda d e o s a e clus e ed
a he esponden le el.
The in e p e a ion o he es ima ed coe icien s comes wi h wo impo an ca ea s. Fi s ,
we canno con ol how esponden s associa e he displayed igne e cha ac e is ics wi h o he
unobse ed cha ac e is ics, such as p oduc i i y o job pe o mance. The e o e, β1−β6a e
composi e pa ame e s ha cap u e he ewa d o a ce ain cha ac e is ic, and he ewa d
o unobse ed ac o s co ela ed wi h i while holding all o he obse ed cha ac e is ics
10 Almås e al. (2020) iden i y egali a ians as esponden s who dis ibu e esou ces equally in a si ua ion whe e
ini ial inequali y is d i en by p oduc i i y. They iden i y libe a ians as esponden s who do no edis ibu e
a all when ini ial inequali y is pu ely based on luck.
123
902
Jus cheap alk? In es iga ing ai ness p e e ences
(a) Reg ession Analysis (b) Tex Analysis
Fig. 7 Fai ness P e e ences, Impo ance o P o ile Cha ac e is ics. No e: This igu e displays how esponden s
ake he igne e cha ac e is ics in o accoun o hei dis ibu ional choices. Figu e 7a displays he poin
es ima es and 95 pe cen con idence in e als om Eq. 4. Figu e 7b displays a wo d cloud om a ex analysis
o he open-ended ques ion in he ai ness p e e ence module. A he end o he module, people we e asked abou
how hey came up wi h hei dis ibu ional choices and o desc ibe hei easoning in hei own wo ds. Based
on he ex co pus om he open-ended answe s, we used na u al language p ocessing echniques o ank he
equency o speci ic e ms. We ocus exclusi ely on esponden s in he con ol g oup, i.e., hose esponden s
who aced hypo he ical scena ios. Appendix Table C27 eplica es he analysis o di e en ans o ma ions o
he ou come a iable, and unde he inclusion o esponden ixed e ec s. Appendix Figu e C6 eplica es he
analysis o esponden s in he ea men g oup. Sou ce: Own calcula ion based on su ey esponses
cons an . Second, we dis ega d non-linea e ec s ac oss he o dinal igne e cha ac e is ics.
The me hodological choices men ioned a he beginning o his sec ion limi he a ailable
a ia ion in ou da a and p e en us om elaxing his s ingen unc ional o m assump ion.
Figu e 7a displays he coe icien s om eg ession Eq. 4along wi h he co esponding
95% con idence in e als.
On he one hand, he esul s indica e ha ai labo ma ke ea nings inc ease wi h igne e
cha ac e is ics ele an o labo ma ke pe o mance and ha a e (pa ially) unde he con ol
o indi iduals. Fo example, he e a e s ong posi i e e ec s o educa ion and weekly wo king
hou s on ai income alloca ions. This esul sugges s ha ai ness p e e ences in he US a e
a leas pa ially consis en wi h no ma i e heo ies ha emphasize he ole o disc e iona y
choices in de e mining ai income sha es, (e.g., Konow 2000). Simila conclusions can be
d awn o age and ini ial ea nings, i in e p e ed as p oxy indica o s o ele an labo ma ke
expe ience and on- he-job pe o mance, espec i ely.11
On he o he hand, he esul s show ha non-disc e iona y igne e cha ac e is ics a e no
ewa ded in ai income alloca ions. Fo example, condi ional on he o he igne e cha ac-
e is ics, he poin es ima e o gende canno be dis inguished om ze o, sugges ing ha US
esiden s pe cei e adjus ed gende gaps in labo ma ke ea nings as un ai . Fu he mo e, he
11 By con olling o ini ial labo ma ke ea nings, we ensu e ha we measu e he desi ed ewa d o all o he
cha ac e is ics, holding ini ial labo ma ke ea nings cons an . Fo ins ance, a posi i e coe icien on educa ion
implies ha condi ional on ini ial ma ke ea nings, highe -educa ed people a e conside ed mo e dese ing
han lowe -educa ed people. In a ull oll-ou o ou ac o ial su ey, ini ial labo ma ke ea nings would be
unco ela ed wi h all o he igne e cha ac e is ics and we could eplica e ou analyses excluding ini ial labo
ma ke ea nings. Howe e , due o he me hodological ocus o ou da a collec ion, we ope a e wi h a limi ed
numbe o igne es whe e he cha ac e is ics o in e es a e co ela ed wi h each o he . To accoun o such
co ela ions, we include ini ial labo ma ke ea nings in all ou es ima ions.
123
903
P. Hu e, D. Weishaa
esponden s in ou sample assign highe (lowe ) ea nings o indi iduals wi h lowe (highe )
pa en al educa ion. This inding could be a ionalized by he ac ha esponden s a e willing
o compensa e people o a disad an aged socio-economic amily backg ound. No e, how-
e e , ha ou analysis only p esen s ela i e e ec s o igne e cha ac e is ics. The e o e, we
canno dis inguish whe he esponden s compensa e indi iduals o a disad an aged back-
g ound (low pa en al educa ion) o penalize indi iduals coming om an ad an aged amily
backg ound (high pa en al educa ion).
Gi en he abo emen ioned ca ea s, we in e p e hese indings as sugges i e. Howe e ,
in Appendix Table C27, we show ha all p e ious conclusions a e obus o al e na i e
speci ica ions. In pa icula , we use p e e ed income sha es o p e e ed income a ios as
ou come a iables o in e es . While he magni ude o coe icien s changes, he di ec ion o
e ec s emains unal e ed. Also, ou esul s do no change quali a i ely when con olling o
esponden ixed e ec s, i.e., when we only use wi hin- esponden a ia ion o es ima e p e-
e ed ewa ds. In addi ion, while ou baseline analysis only uses esponses om esponden s
wi h hypo he ical asks, he measu ed ewa ds a e simila in he eal-pe son ea men g oup
(Appendix Figu e C6).
To subs an ia e he quan i a i e e idence, we also use na u al language p ocessing ech-
niques o p esen indings om a ex analysis (Fe a io and S an che a 2022). A he end
o he baseline su ey, we asked esponden s how hey made hei dis ibu ional choices and
allowed hem o desc ibe hei easoning in an open- ex ield. Figu e 7b isualizes he ex
analysis in a wo d cloud, highligh ing he equency o obse ed e ms. The wo d cloud
shows ha esponden s pu a s ong emphasis on wo king hou s, ea nings, and he educa ion
o he igne e pe sons when making hei dis ibu ional choices. The emphasis on hese
cha ac e is ics, he e o e, echoes he esul s om ou quan i a i e analysis.
In his sec ion, we ha e shown ha ou su ey module eco e s ai ness p e e ences ha
a e b oadly consis en wi h he exis ing li e a u e. This conclusion holds o he deg ee o
inequali y accep ance, he p e alence o ai ness ypes, and he cha ac e is ics de e mining he
ex en o ai income alloca ions. Since he da a collec ion was designed o me hodological
alida ion, we u ge eade s o ea hese subs an i e esul s cau iously. Howe e , he esul s
poin o he abili y o ou su ey module o unco e nuanced ai ness posi ions and o desc ibe
ai ness p e e ences in socie ies mo e b oadly.
6 Conclusion
This s udy alida es a no el su ey ool designed o measu e ai ness p e e ences using
ealis ic ye hypo he ical scena ios.
We conduc his alida ion using a wo-wa e su ey co e ing a ep esen a i e sample o
he US popula ion. Ou esul s demons a e ha ai ness p e e ences a e no in luenced by he
p ospec o eal-wo ld implemen a ion, e en when mone a y s akes a e high. This conclusion
holds ue o bo h he gene al popula ion and ac oss a ious demog aphic subg oups. Mo e-
o e , compa ing indi idual esponses ac oss he wo wa es e eals ha ai ness p e e ences
a e s able o e ime, ega dless o whe he hey o igina e om hypo he ical o incen i ized
scena ios. We u he mo e p o ide sugges i e e idence ha he elici ed p e e ences a e con-
sis en wi h es ablished indings on ai ness p e e ences in he US.
The e o e, ou alida ion p o ides compelling e idence ha ai ness p e e ences om
hypo he ical su eys a e no “jus cheap alk.” Ins ead, hey can yield c edible insigh s in o he
na u e and ana omy o hese p e e ences. We emphasize ha hese conclusions a e con ex -
123
904
