Does poverty negate the impact of social norms on cheating?

Boonmanun , Supa ee; Kajackai e, Agne; Meie , S ephan A icle — Published Ve sion Does po e y nega e he impac o social no ms on chea ing? Games and Economic Beha io P o ided in Coope a ion wi h: WZB Be lin Social Science Cen e Sugges ed Ci a ion: Boonmanun , Supa ee; Kajackai e, Agne; Meie , S ephan (2020) : Does po e y nega e he impac o social no ms on chea ing?, Games and Economic Beha io , ISSN 1090-2473, Else ie , Ams e dam, Vol. 124, pp. 569-578, h ps://doi.o g/10.1016/j.geb.2020.09.009 This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/228485 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-nc-nd/4.0/ Games and Economic Beha io 124 (2020) 569–578 Con en s lis s a ailable a ScienceDi ec Games and Economic Beha io www.else ie .com/loca e/geb Does po e y nega e he impac o social no ms on chea ing? Supa ee Boonmanun a,b,∗, Agne Kajackai e c, S ephan Meie d aDepa men o Clinical Epidemiology and Bios a is ics, Facul y o Medicine Rama hibodi Hospi al, Mahidol Uni e si y, Rachahe i, Bangkok 10400, Thailand bFacul y o En i onmen and Resou ce S udies, Mahidol Uni e si y, Phu hamon hon, Nakhon Pa hom 73170, Thailand cWZB Be lin Social Science Cen e , Reichpie schu e 50, 10785 Be lin, Ge many dColumbia Business School, 3022 B oadway, New Yo k, NY 10027, USA a i c l e i n o a b s a c A icle his o y: Recei ed 15 Janua y 2020 A ailable online 13 Oc obe 2020 JEL classifica ion: C91 C93 D82 D91 Keywo ds: Chea ing Lying Po e y Social no ms In e en ions Lab-in- he-field expe imen Chea ing such as co up ion and ax e asion is p e alen in he de eloping wo ld; he e o e, many in e en ions ha e been unde aken o educe chea ing in de eloping coun ies. Al hough some field e idence shows ha po e y is co ela ed wi h chea ing, he causal e ec o po e y on chea ing in he field and he e ec i eness o in e en ions o financially cons ained people emain an open ques ion. We p esen esul s om a lab-in- he-field expe imen wi h low-income ice a me s in Thailand (N= 568), in which we, fi s , in es iga e he causal e ec o po e y on chea ing and, second, es whe he po e y a ec s he e ec i eness o a social-no m in e en ion o educe chea ing. We show po e y i sel does no a ec willingness o chea . Howe e , al hough a social-no m- eminde in e en ion educed chea ing when he popula ion was iche (a e ha es ), i had no e ec when he popula ion was poo e (be o e ha es ). Ou esul s sugges ha he iming o in e en ions o change beha io migh ma e . ©2020 The Au ho (s). Published by Else ie Inc. This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/). 1. In oduc ion The e has long been deba e in economics on whe he po e y os e s o dampens p osocial beha io . Empi ically, he ques ion emains open. One s and o empi ical li e a u e using field da a showed ha poo o lowe -social-class indi iduals a e mo e gene ous, cha i able, us ing, willing o help, and less likely o chea (Pi e al., 2010, 2012). O he schola s ound ha , in con as , poo o lowe -social-class indi iduals a e less us ing, less us wo hy, less in insically mo i a ed, mo e likely o beha e an isocially, and less likely o en o ce sha ing (Ba os, 2016; Glaese e al., 2000; Hausho e , 2013; Hausho e and Feh , 2014; Jiang and Lim, 2016; P edige e al., 2013, 2014; Shal i, 2016). Finally, some schola s ound no e ec o po e y o lowe social class on chea ing, gi ing, o coope a ing (Aksoy and Palma, 2019; And eoni e al., 2017; Ba os, 2016; Gäch e and Schulz, 2016; P edige e al., 2013). The mixed empi ical e idence may be due o ac o s such as di e ences in expe imen al design and p ocedu e as well as di e en ways o looking a po e y ( ela i e s. absolu e).1Impo an ly, a la ge ac ion o he s udies a e co ela ional (And eoni e al., 2017; Gäch e and Schulz, 2016; Glaese e al., 2000; Hausho e , 2013; Hausho e and Feh , 2014; Jiang and Lim, 2016; Pi e al., 2010, 2012; Shal i, 2016), and only a ew s udies used an exogenous income shock o es ablish a causal *Co esponding au ho . E-mail add esses: supa [email p o ec ed] (S. Boonmanun ), agne.kajackai [email p o ec ed] (A. Kajackai e), [email p o ec ed] (S. Meie ). 1Some o he pape s men ioned abo e (e.g., Pi e al., 2012) looked a ela i e po e y; o he s (e.g., Ba os, 2016) also conside ed absolu e po e y. h ps://doi.o g/10.1016/j.geb.2020.09.009 0899-8256/©2020 The Au ho (s). Published by Else ie Inc. This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/). S. Boonmanun , A. Kajackai e and S. Meie Games and Economic Beha io 124 (2020) 569–578 link be ween po e y and p osocial/an isocial beha io (Aksoy and Palma, 2019; Ba os, 2016; P edige e al., 2013, 2014; see Sha ma e al., 2014, o a lab s udy). In o he wo ds, mos o he field s udies canno show a causal e ec o po e y on p osocial beha io because hey canno exclude e e se causali y, ha is, ha p osocial p e e ences p edic income and no he o he way a ound, o because hey canno exclude omi ed a iable bias, ha is, a hi d unobse able ac o a ec ing bo h income and p osocial beha io . Ou in es iga ion ocused on p o iding a causal link be ween po e y and beha io in a field se ing. In pa icula , we in es iga ed how po e y a ec s chea ing beha io . Many e e yday in e ac ions a e based on asymme ic in o ma ion (i.e., I know some hing ha you do no know), which emp s some people o chea o mone a y benefi , such as solici ing b ibes o e ading axes. The economic and social consequences o chea ing a e subs an ial (Mau o, 1995; Olken and Pande, 2012; P anab, 1997); hus, unde s anding he ac o s ha influence chea ing decisions is impo an . Al hough he e is obus expe imen al e idence abou he e ec o he cos o lying, gende , cul u e, o p obabili y o de ec ion on chea ing in he lab (Abele e al., 2019; Fischbache and Föllmi-Heusi, 2013; Gäch e and Schulz, 2016; Gneezy e al., 2013, 2018; Kajackai e and Gneezy, 2017; Maza e al., 2008), he e is limi ed e idence abou whe he po e y causally a ec s he decision o chea in he field. The e is also no e idence abou whe he po e y a ec s he e ec i eness o in e en ions o educe chea ing— despi e he conside able unds spen c ea ing and implemen ing such in e en ions. In a numbe o s udies, po e y (o sca ci y o ano he kind) was shown o educe cogni i e capaci y and cogni i e pe o mance. The main e idence showing ha sca ci y educes cogni i e capaci y is he expe imen al e idence ha cogni i e unc ion sco es a e lowe in a sca ci y pe iod han in an abundance pe iod o he same people (Mani e al., 2013; Mullaina han and Shafi , 2013). The e o e, po e y can easily educe he e ec i eness o in e en ions equi ing cogni i e capaci y. The goal o his s udy was he e o e wo old. Fi s , we we e in e es ed in he causal e ec o po e y on chea ing beha io in he field, in gene al. Fo his pu pose, we ec ui ed a unique popula ion o Thai ice a me s, le hem play a simple chea ing game (Gneezy e al., 2018), and exploi ed he di e ences in financial cons ain s hey ace be o e ha es (when hey a e poo ) and a e ha es (when hey a e iche ; simila o Ba os, 2016; Mani e al., 2013). Second, we explo ed he e ec i eness o an in e en ion o educe chea ing when ou pa icipan s ace di e en le els o financial cons ain s. Fo his, we used a popula ins umen —a social-no m eminde . We belie e ha i is impo an no only o analyze he e ec o po e y on he gene al endency o e hical and p osocial beha io bu also o conside how in e en ions can achie e desi able beha io when he a ge popula ion aces (o does no ace) po e y. Whe he po e y causes mo e chea ing is ul ima ely an empi ical ques ion, as he e a e a gumen s o why being fi- nancially cons ained can inc ease he willingness o chea o lea e i unchanged. On one hand, clea ly, being financially cons ained inc eases he immedia e need o money as well as i s ma ginal u ili y (Ca alho e al., 2016). As a esul , he immedia e need o money can lead o mo e chea ing han when one is less financially cons ained. On he o he hand, he e is as expe imen al lab e idence ha chea ing can be insensi i e o many pa ame e s such as s akes (Abele e al., 2019; Kajackai e and Gneezy, 2017). I expe imen al pa icipan s do no eac o changes in s akes while chea ing, hen hey migh no eac o changes in income. In e ms o he e ec i eness o a social-no m- eminde in e en ion, he social-no m eminde has p o ed o wo k well and is a powe ul ool in many con ex s (F ey and Meie , 2004; Golds ein e al., 2008; Schul z e al., 2007; Thale and Suns ein, 2008), including ax compliance (e.g., Hallswo h e al., 2017 [conduc ed in he UK]; Del Ca pio, 2014 [conduc ed in Pe u]). We he e o e expec he no m- eminde in e en ion o wo k well in ou sample, especially when people a e no financially cons ained.2Howe e , po e y migh influence he in e en ion’s e ec i eness. P e ious expe imen al e idence showed ha sca ci y educes one’s cogni i e capaci y because i cap u es one’s a en ion (Mani e al., 2013; Mullaina han and Shafi , 2013; Shah e al., 2012). In o he wo ds, sca ci y leads o “ unneling”: being (financially) cons ained consumes many cogni i e esou ces i sel and lea es one wi h ewe cogni i e esou ces o o he domains (because o he human cogni i e sys em ha ing limi ed capaci y; Baddeley and Hi ch, 1974; Mille , 1956; Neisse , 1976). Following on his ex- pe imen al e idence, we a gue ha in ou expe imen , indi iduals in po e y may lack he cogni i e bandwid h o diges he no m eminde and may be less likely o eac o i . The easoning is simple: he mo e cogni i e capaci y is used o managing po e y, he ewe cogni i e esou ces can be used o managing he social-no m eminde . In addi ion, e en i ou expe imen al indi iduals in po e y ha e sufficien cogni i e bandwid h o in e nalize he no m eminde , hey ha e ewe financial possibili ies o adop ing he desi ed beha io han iche pa icipan s because hey need he cash mo e. In o he wo ds, poo pa icipan s ha e less space o “maneu e ing” han iche pa icipan s. Pu oge he , because o ei he “ unneling” o a limi ed financial possibili y o ollowing he no m, o bo h, we expec no m in e en ions o wo k wo se when ou pa icipan s ace po e y. We conduc ed a la ge-scale lab-in- he-field expe imen wi h ice a me s in u al Thailand o es he hypo heses de- sc ibed abo e. We used a simple lying game and social-no m eminde s in a be ween-subjec s expe imen al design o 2Despi e hei popula i y among policymake s and he many success s o ies, he e a e some cases whe e social-no m in e en ions we e unsuccess ul in changing beha io such as chea ing o ax noncompliance (e.g., Cas o and Sca ascini, 2015 [A gen inian sample]; Fellne e al., 2013 [Aus ian sample]). In hese s udies, social-no m eminde s had no e ec o e all bu wo ked depending on belie s abou e asion beha io o o he s (Fellne e al., 2013) o depending on pas compliance beha io s (Cas o and Sca ascini, 2015). Howe e , because Thailand is a highly collec i is coun y acco ding o he Ho s ede model o na ional cul u e (Ho s ede, 1984; see www.ho s ede -insigh s .com o cu en in o ma ion on na ional cul u e o o e 100 coun ies) and because people in collec i is socie ies end o con o m mo e o social no ms han people in indi idualis ic cul u es (Bond and Smi h, 1996; Cialdini e al., 1999), we expec he social-no m eminde o wo k be e in ou sample han in he indi idualis ic Aus ian and A gen inian samples used in he examples abo e. 570 S. Boonmanun , A. Kajackai e and S. Meie Games and Economic Beha io 124 (2020) 569–578 measu e chea ing beha io and he e ec i eness o a social-no m in e en ion o educe chea ing be o e and a e ha es . We ound ha po e y i sel does no a ec willingness o chea — ha is, pa icipan s chea ed o a simila ex en be o e and a e ha es . When eminded o he social no m, hey did no chea less when hey we e poo e (be o e ha es ), bu he social-no m- eminde in e en ion educed chea ing when hey we e iche (a e ha es ). This esul sugges s ha he iming o in e en ions o change beha io migh ma e . Howe e , no e ha e en hough in e ac ion-e ec s analysis shows ha no ms wo k mo e han wo imes be e a e ha es han be o e ha es , which we pe cei e o be economically significan , he in e ac ion e ec is no s a is ically significan . The emainde o he pape is o ganized as ollows. In Sec ion 2, we desc ibe he unique financial si ua ion o ou pa icipan s, he expe imen al design and p ocedu e. Sec ion 3p o ides he expe imen al esul s. In Sec ion 4, we discuss he esul s and p o ide policy implica ions. 2. Expe imen al design and p ocedu e In he expe imen , we use a 2×2 be ween-subjec s design wi h ou ea men s, a ying he financial si ua ion o he pa icipan s (be o e ha es s. a e ha es ) and he exis ence o a social-no m eminde (baseline chea ing game [no no m- eminde included] s. no m- eminde game). In his sec ion, we fi s demons a e he financial di e ences be ween he pa icipan s be o e and a e ha es and hen desc ibe he expe imen al ea men s and p ocedu e. 2.1. Pa icipan s and hei financial si ua ion be o e and a e ha es Fo ou expe imen , we ec ui ed 568 ice a me s om 48 illages in u al Thailand.3These ice a me s ha e gene ally a low income (93% a e eligible o apply o a go e nmen subsidy o g oce ies; see Table A.1 in Appendix A in Supplemen a y Ma e ial o mo e sociodemog aphic cha ac e is ics o ou sample). Rice is impo an o ou pa icipan s. They use i o subsis ence consump ion; i cons i u es hei main sou ce o nu i ion. Also, mos (65% o all pa icipan s) sell hei ice on he ma ke , a la ge and impo an addi ional sou ce o income. The income om o he ac i i ies h oughou he yea is no high, and he in es men in ice p oduc ion is subs an ial (18% o o al annual expendi u es). Impo an ly o ou s udy, ice cul i a ion happens jus once a yea , a he beginning o he ainy season, because o wa e a ailabili y. As a esul , he a me s ha e difficul ies in smoo hening hei consump ion o e he yea . The e o e, hey a e ela i ely poo be o e ha es bu ela i ely ich a e ha es . This exogenous income gene a ion caused by he ha es allows us o in es iga e a causal e ec o po e y on chea ing and is why we chose he Thai ice a me s as he pa icipan pool o ou expe imen . The expe imen ook place in 48 illages in Ubon Ra cha hani, no heas Thailand. 283 a me s pa icipa ed in ou expe - imen be o e ha es (Sep embe 2017) and ano he 285 a me s pa icipa ed a e ha es (Decembe 2017).4The be o e- and a e -ha es a me s a e om he same subdis ic s bu di e en illages (24 illages be o e ha es and ano he 24 illages a e ha es ). The illages we e andomly assigned o ea men s and expe imen al sessions, wi h each illage pa - icipa ing in only one expe imen al session (and one ea men ).5The be o e- and a e -ha es a me s do no di e in sociodemog aphic cha ac e is ics such as age, gende , educa ion, o amily size (see Table A.1 in Supplemen a y Ma e ial), which shows ha ou andomiza ion wo ked. The da a collec ed in a pos -expe imen al ques ionnai e showed ha he ice a me s a e indeed much poo e be o e ha es han a e ha es —confi ming ou empi ical s a egy (see Table 1). On a e age, he be o e-ha es pa icipan s epo ed 73% lowe household income in he ele an mon h han he a e -ha es pa icipan s (mean income Mbe o e = ฿11,533 s. Ma e = ฿42,442, p< 0.001, - es ; all es s in he pape a e wo-sided, i no no ed o he wise). I ollows ha he e ec i e income—compu ed by di iding household income by he squa e oo o household size—is also 73% lowe be o e ha es han a e ha es (Mbe o e = ฿5,467 s. Ma e = ฿20,217, p< 0.001, - es ). Also, household expendi u es o be o e-ha es pa icipan s we e 38% lowe han hose o a e -ha es pa icipan s (Mbe o e = ฿11,269 s. Ma e = ฿18,238, p< 0.01, - es ). Fu he mo e, he epo ed household deb is 28% highe be o e ha es (Mbe o e = ฿192,086 s. Ma e = ฿138,338, p< 0.01, - es ), and he be o e-ha es pa icipan s a e less likely o ha e sa ings o any kind, including li es ock (76% s. 95%, p< 0.001, es o p opo ion). In Table 1, we p esen o dina y leas squa es (OLS) eg ession esul s, in which we use he ollowing specifica ion: Yi=α+γ0Ha es i+γ1No.o HH membe si+ei, whe e he dependen a iable Yiis ei he household (HH) Income, HH Expendi u e, Amoun o HH Deb s anding o Sa ings ( his is a dummy, which is equal o 1 i he household has any sa ings, and 0 o he wise). The indica o a iable Ha es i is equal o 1 i he indi idual is in an a e -ha es g oup and 0 i he indi idual is in a be o e-ha es g oup. All s anda d e o s a e clus e ed a he illage le el. In line wi h he p e ious analysis, we find ha a e he ha es , households ha e a significan ly highe income and expendi u es, ha e a lowe amoun o deb s anding, and a e mo e likely o ha e sa ings. 3We excluded nine a me s because hey did no cul i a e and ha es ice ha yea . This yielded a final sample o 559 a me s. 4Since he e we e ou expe imen al condi ions in he expe imen , we ob ained a ound 140 independen obse a ions pe cell. This sample is la ge han hose in o he li e a u e using simila andom-d aw games. We de e mined he sample size based on ou financial limi a ions, and decided be o e conduc ing he s udy ha we would end da a collec ion upon comple ing ou isi s o he 48 illages selec ed. 5Assignmen o a be o e- o a e -ha es condi ion and o a baseline chea ing game o a no m- eminde game was andom. 571 S. Boonmanun , A. Kajackai e and S. Meie Games and Economic Beha io 124 (2020) 569–578 Table 1 Be o e- and a e -ha es di e ences in household (HH) financial si ua ion. HH Income HH Expendi u es Amoun o HH Deb S anding Sa ings (dummy) {A e -ha es } ฿31,241 (5,719)*** ฿7,111 (1,377)*** -฿52,698 (24,169)** 0.19 (0.04)*** No. o HH membe s ฿4,111 (1,116)*** ฿1,762 (427)*** ฿13,022 (4,420)*** - Cons an -฿7,296 (5,023) ฿3,196 (1,649)* ฿132,436 (28,975)*** 0.76 (0.04)*** p- alue Wilcoxon es equali y o dis ibu ion <0.001 <0.001 <0.01 <0.001 Obse a ions 559 559 559 559 No es: This able epo s esul s om OLS eg essions o he dependen a iables shown in he column headings on an indica o a iable iden i ying pa icipan s assigned o he a e -ha es g oups and a cons an con olling o HH size when a iables a e a he HH le el ( he fi s h ee). S anda d e o s, clus e ed a he illage le el, a e in pa en heses. The ou h ow shows he p- alue o a Wilcoxon ank-sum es . HH expendi u es include expendi u es on ag icul u al ac i i ies. ฿= Thai Bha . * p< 0.10, ** p< 0.05, *** p< 0.01. O he ac o s such as nu i ion, s ess, o subjec i e well-being migh also be di e en be o e and a e ha es and hus a ec beha io in expe imen s. Howe e , be o e- and a e -ha es a me s epo ed simila le els o s ess (Mbe o e = 6.09 s. Ma e = 5.80, p= 0.19, - es ) and subjec i e well-being (Mbe o e = 8.27 s. Ma e = 8.23, p= 0.81, - es ).6Rega ding nu i ion, only one pa icipan (in he be o e-ha es g oup) indica ed he mon hs when we conduc ed ou expe imen as being ha des in e ms o ood. Thus, we concluded ha nu i ion is good bo h be o e and a e ha es . The eg ession esul s in Table 2confi m hese esul s. In Table 2, we use he ollowing eg ession specifica ion: Yi=α+γ0Ha es i+ei, whe e he dependen a iable Yiis ei he S ess le el, Subjec i e well-being, o Cu en -mon h-is-ha des -in- e ms-o - ood (a dummy equal o 1 i he cu en mon h is he ha des o he pa icipan in e ms o ood). The indica o a iable Ha es iis equal o 1 i he indi idual is in an a e -ha es g oup and 0 i hey a e in a be o e-ha es g oup. All s anda d e o s a e clus e ed a he illage le el. In line wi h he p e ious analysis, we find ha all h ee dependen a iables a e no di e en be o e and a e ha es . Tha is, we ound clea e idence ha only financial si ua ion, and no he o he ac o s such as subjec i e well-being, s ess, o nu i ion, is significan ly di e en be o e and a e ha es . Na u ally, he e migh also be o he ac o s ha di e be o e and a e ha es and which we did no con ol o . Al hough we canno be su e o his, we used answe s o an ex ensi e pos -expe imen ques ionnai e o disco e as much abou ou pa icipan s as possible (see Appendix B.2 in Sup- plemen a y Ma e ial) and ound ha he only a iables ha we con olled o and ha di e ed be o e and a e ha es a e hose ela ed o financial si ua ion. 2.2. Expe imen al ea men s We used a simple game o measu e chea ing beha io o he ice a me s (Fischbache and Föllmi-Heusi, 2013; Gneezy e al., 2018). In he fi s expe imen — he baseline chea ing game—we gi e each pa icipan a sealed en elope con aining 10 olded pieces o pape bea ing he numbe s om 1 o 10. We asked pa icipan s o blindly ake ou one piece o pape , obse e he numbe , e u n i o he en elope, seal he en elope, and hen epo he obse ed numbe on a epo ing shee . The payo was 10 Thai Bah imes he numbe epo ed. This c ea ed an incen i e o chea o mone a y benefi . As esea che s, we did no know exac ly which indi iduals chea ed o by how much, bu we could in e he app oxima e le el o chea ing by compa ing he expec ed heo e ical dis ibu ion o epo s (i.e., uni o m dis ibu ion o he numbe s be ween 1 and 10) wi h he ac ual epo ed dis ibu ion o numbe s. Tha is, in he absence o chea ing, we should obse e ha e e y numbe be ween 1 and 10 occu s app oxima ely 10% o he ime and ha he a e age epo ed numbe is no s a is ically significan ly di e en om 5.5 ( he sum o 1 h ough 10, di ided by 10). By con as , i epo ed numbe s we e highe han expec ed, his would indica e ha pa icipan s chea ed. 6A fi s sigh , i is a he su p ising ha pe cei ed happiness and s ess do no di e be o e and a e ha es . Ou specula i e in e p e a ion is ha he mainly Buddhis Thai a me s see “s ess” o “happiness” as some hing bigge han being financially cons ained / no cons ained (see Ekman e al., 2005, o a discussion on Buddhism and happiness). O e all, ou pa icipan s epo ed low s ess le els and high happiness le els bo h be o e (means o s ess and happiness a e 6.09 [ou o 14] and 8.27 [ou o 10], espec i ely) and a e ha es (means o s ess and happiness a e 5.80 [ou o 14] and 8.23 [ou o 10], espec i ely). Impo an ly, we ollowed Mani e al. (2013)and Mullaina han and Shafi (2013)when dis inguishing be ween sca ci y “ educing cogni i e capaci y” and sca ci y “causing s ess.” O cou se, one migh eel s essed when acing po e y (which does no show in ou pos -ques ionnai e da a bu , o example, is he case in da a om Mani e al., 2013). Howe e , acco ding o he as sca ci y li e a u e, expe iencing biological s ess is no necessa y o expe iencing a axed cogni i e capaci y. Sca ci y will p eoccupy a pe son and consume hei cogni i e capaci y, which in u n will ha e nega i e e ec s on hei cogni i e unc ioning; biological s ess can accompany i , bu no necessa ily. 572 S. Boonmanun , A. Kajackai e and S. Meie Games and Economic Beha io 124 (2020) 569–578 Table 2 Be o e- and a e -ha es di e ences in s ess le el, subjec i e well-being, and nu- i ion. S ess Le el Subjec i e Well-Being Cu en Mon h Is Ha des o Food (dummy) {A e -ha es } −0.29 [0.24] −0.04 [0.21] −0.004 [0.004] Cons an 6.09 [0.17]*** 8.27 [0.17]*** 0.004 [0.004] p- alue Wilcoxon es equali y o dis ibu ion 0.37 0.46 0.32 Obse a ions 559 559 559 No es: This able epo s esul s om OLS eg essions o he dependen a iables shown in he column headings on an indica o a iable iden i ying pa icipan s as- signed o he a e -ha es g oups and a cons an . S anda d e o s, clus e ed a he illage le el, a e in b acke s. The hi d ow shows he p- alue o a Wilcoxon ank- sum es . S ess le el anges om 1 o 14. Subjec i e well-being anges om 1 o 10. * p< 0.10, ** p< 0.05, *** p< 0.01. In he second expe imen — he no m- eminde game—we es ed o he e ec i eness o a social-no m- eminde in e - en ion o educe chea ing. In his expe imen , be o e playing he chea ing game, pa icipan s we e in o med ha mos ice a me s in hei p o ince find chea ing o one’s own benefi unaccep able. The exac wo ding o he p omp was “We an a su ey on a me s in his p o ince and he majo i y finds chea ing o one’s own benefi unaccep able.” We in ac asked pa icipan s, a he end o he fi s expe imen al sessions o he baseline chea ing game, o a e how accep able chea ing o one’s own benefi is, on a scale om 1 o 10. Mos pa icipan s ound such chea ing e y unaccep able.7 We conduc ed bo h he baseline chea ing game and he no m- eminde game be o e and a e ha es , wi h di e en pa icipan s, which leads us o a 2×2 be ween-subjec s ac o ial design. Table 3p esen s he ea men s and he numbe o independen obse a ions in each ea men . Table 3 T ea men s. 2×2 Be ween-Subjec s Design Be o e ha es (N) A e ha es (N) Baseline chea ing game 140 144 No m- eminde game 139 136 2.3. P ocedu e We conduc ed pape -and-pencil expe imen s in all sessions (see Appendix B.1 in Supplemen a y Ma e ial o he exac ins uc ions and Appendix C in Supplemen a y Ma e ial o he de ailed expe imen al p ocedu e). In his pape , we use da a om he chea ing expe imen ,8 o which we used he ollowing p o ocol. Fi s , pa icipan s d ew a andom sea numbe ( om 1 o 12) and ook he co esponding sea . A e hey signed he consen o m, we explained he expe imen s o hem. They we e in o med ha he e would be ou decision ac i i ies and ha we would explain he ules o each game a he beginning o ha pa icula game. A e we desc ibed he game h ough a w i en sc ip and p esen a ion ma e ials, pa icipan s had o answe es ques ions co ec ly. Only hen did we p oceed wi h he expe imen . The en i e expe imen al session ook on a e age 74 minu es, o which he Chea ing expe imen ook only 7 minu es. The pos -expe imen al ques ionnai e sec ion las ed an addi ional 96 minu es. I ook abou 20 minu es o in e iew each pe son, bu because he e we e only h ee o ou assis an s in each session, mos pa icipan s had o wai o be in e iewed. The a e age expe imen al ea nings we e 279 Bah (8.45 USD, equi alen o he pu chasing powe o 22.63 USD), wi h a e age ea nings o 64 Bah in he Chea ing expe imen (1.94 USD, equi alen o he pu chasing powe o 5.19 USD).9In addi ion, 7We used he fi s wo expe imen al sessions o de e mine he social-no m eminde . These sessions we e baseline chea ing-game sessions wi h 24 pa icipan s all oge he . On a e age, he pa icipan s a ed ha lying is accep able a a le el o 1.63 (wi h 1 being e y unaccep able and 10 e y accep able), and 79.17% pa icipan s chose “ e y unaccep able” (ma king 1) as hei answe . We ask pa icipan s how accep able lying is, a he end o each expe imen al session. The a e age a e o accep ance amoun s o 1.70 o e all pa icipan s, wi h 79.43% o pa icipan s choosing “ e y unaccep able” as hei answe . Tha is, lying o one’s own benefi is pe cei ed as being e y unaccep able in ou sample. 8Pa icipan s pe o med h ee expe imen s p io o he Chea ing expe imen : P isone ’s Dilemma, P isone ’s Dilemma wi h hi d-pa y punishmen , and he Dic a o game. All decisions we e made in p i a e. The o de o he expe imen s emained he same o all pa icipan s in all sessions. No eedback abou expe imen al ea nings om each game was p o ided du ing he expe imen . We checked whe he a me s’ ea nings and decisions in he p e ious games had an e ec on chea ing beha io and ound no significan e ec using eg ession analyses. 9The exchange a e o 1 USD was 33 Thai Bah on expe imen al days. Howe e , he pu chasing powe pa i y (PPP) con e sion ac o (GDP) was 12.33 Thai Bah pe 1 USD in 2015 (Wo ld Bank, 2018). 573 S. Boonmanun , A. Kajackai e and S. Meie Games and Economic Beha io 124 (2020) 569–578 Fig. 1. A e age epo ed numbe s in he baseline chea ing game and he no m- eminde game. pa icipan s ecei ed 100 Bah o showing up and 100 Bah o he in e iew. The addi ional paymen o he in e iew was announced a e he expe imen had ended. 3. Resul s Panel A in Fig. 1shows he a e age epo ed numbe , and Fig. 2shows he dis ibu ion o he epo ed numbe s in he wo games, be o e and a e ha es . In he baseline chea ing game, a me s chea ed s a is ically significan ly. They epo ed, on a e age, 6.62. Impo an ly, he ex en o chea ing was he same be o e and a e ha es (Mbase_be o e = 6.61 s. Mbase_a e = 6.64, p= 0.92, - es ). Be o e ha es , in expec a ion, pa icipan s o e epo ed by 20.18% (chea ing was s a is ically sig- nifican , p< 0.001, Chi-squa e- es ) and a e ha es by 20.73% (chea ing was s a is ically significan , p< 0.001, Chi-squa e es ). Tha is, in ou se ing, po e y i sel does no change willingness o chea . In he no m- eminde game, be o e ha es , pa icipan s o e epo ed by 14.73% (chea ing was s a is ically significan , p = 0.005, Chi-squa e- es ). On a e age, pa icipan s epo ed 6.31, which is no significan ly less han in he baseline chea ing game (Mbase_be o e = 6.61 s. Mno m_be o e = 6.31, p= 0.38, - es ). In con as , he no m- eminde ool was e ec i e a e he ha es : i educed chea ing significan ly ela i e o he baseline chea ing game (Mbase_a e = 6.64 s. Mno m_a e = 5.96, p= 0.04, - es ) and yielded a epo ing le el ha is no di e en om chance (p= 0.10, Chi-squa e- es ); ha is, in expec a ion, people do no chea in his ea men .10 In he nex s ep, we p esen OLS eg essions esul s (see Table 4). We un ou OLS eg essions including a eg ession wi h he in e ac ion e m A e ha es ×No m in e en ion o he ollowing o m: Repo i=α+γ0Ha es i+γ1Ha es i×No mi+δNo mi+βxi+ei, whe e he dependen a iable Repo ediis Repo ed Numbe (1–10) by indi idual i. The indica o a iable Ha es iis equal o 1 i he indi idual is in an a e -ha es g oup and 0 i he indi idual is in a be o e-ha es g oup. The indica o a iable No miis equal o 1 o he no m- eminde game and 0 o he baseline chea ing game. Sociodemog aphic con ols xlike gende , age, educa ion le el, amily s a us, and economic s a us a e included in eg ession specifica ions (3) and (4). All s anda d e o s a e clus e ed a he illage le el. In gene al, he eg essions suppo he p e ious analysis. The fi s eg ession specifica ion (Column 1) shows ha o e all chea ing is no di e en be o e and a e ha es (i.e., he coefficien on A e Ha es is small and no significan ). Fu he - mo e, we find ha when we look a he whole sample (be o e ha es and a e ha es ), he no m in e en ion lowe s chea ing ( he coefficien on No m in e en ion is nega i e and ma ginally significan ). In he second eg ession specifica ion, we es o he in e ac ion be ween ha es and no m in e en ion (Column 2). We find ha he in e en ion is mo e han wice as e ec i e a e ha es han be o e ha es (while he coefficien on No m in e en ion amoun s o −0.30, he in e - ac ion e ec o A e ha es ×No m amoun s o −0.38). Whe eas he doubled e ec size shows ha no m in e en ion is 10 In addi ion, no e ha in all he ea men s, only high numbe s a e o e epo ed significan ly. We find ha , in expec a ion, eigh s a e o e epo ed in base_a e , nines a e o e epo ed in no m_a e , and ens a e o e epo ed in base_be o e, no m_be o e, and base_a e (p< 0.05, one-sided binomial es s), whe eas no o he numbe is o e epo ed significan ly in any ea men . The finding ha chea ing happens a high numbe s is in line wi h he es o he chea ing li e a u e using andom-d aw games ( o a me a-s udy, see Abele e al., 2019). Fo ins ance, excluding he mos ex eme epo ing o ens om he da a se would make he a e age o e epo ing in he ou ea men s indis inguishable. 574 S. Boonmanun , A. Kajackai e and S. Meie Games and Economic Beha io 124 (2020) 569–578 Fig. 2. Dis ibu ion o epo ed numbe s in baseline chea ing game and no m- eminde game be o e and a e ha es . economically significan ly mo e e ec i e a e ha es , he in e ac ion is no s a is ically significan on con en ional le els. No e also ha we lack sufficien powe o de ec a significan in e ac ion e ec . As s a ed ea lie , we se he sample size based on ou financial limi a ions, which led us o isi 48 illages and a o al o 559 obse a ions. An ex-pos powe calcu- la ion shows ha a e ha es —when we expec ed o see a la ge ea men e ec o no ms—gi en he sample size o 280, he means, and s anda d de ia ions, he powe we ha e amoun s o 52.58%. Because we ha e ela i ely low powe e en in he a e -ha es g oup, in which we expec a la ge no ms’ e ec , we a e clea ly unde powe ed o de ec ing significan in e ac ion e ec s o no ms be ween he be o e- and a e -ha es g oups. O e all, we find ha (1) in he baseline chea ing game, chea ing beha io is no di e en be o e and a e ha es ; (2) a social-no m eminde significan ly educes chea ing a e ha es (when he popula ion is iche ) bu no be o e ha es (when he popula ion is poo e ); and (3) he in e ac ion o ha es and social-no m eminde sugges s ha no ms wo k be e a e ha es han be o e ha es , bu he in e ac ion e ec is no s a is ically significan . Finally, ano he way o analyze he e ec o po e y on chea ing is o examine he co ela ional ela ionship be ween po e y and chea ing beha io . Ins ead o compa ing he beha io be o e and a e ha es , we pooled he be o e- and a e - ha es g oups and di ided ou pa icipan pool by hei e ec i e income h ough a median spli , calling he a me s abo e he median “highe income” and hose below he median “lowe income.” We ound co ela ional income e ec s ha a e e y simila o he causal ha es e ec s on chea ing (see Panel B in Fig. 1). In he baseline chea ing game, lowe - and highe -income a me s chea ed o a simila ex en (Mbase_lowe = 6.67 s. Mbase_highe = 6.58, p= 0.76, - es ). In he no m- eminde game, lowe -income pa icipan s did no eac o he no m eminde (Mbase_lowe = 6.67 s. Mno m_lowe = 6.41, p = 0.44, - es ), whe eas he no m eminde was e ec i e o highe -income indi iduals (Mbase_highe = 6.58 s. Mno m_highe = 5.83, p= 0.03, - es ). No e, howe e , ha hese esul s a e only co ela ional. 4. Discussion Ou s udy analyzed how po e y a ec s he endency o chea in a field se ing. Ou expe imen e ealed wo main esul s wi h a se o implica ions. Fi s , po e y i sel does no change humans’ inclina ion o chea . F om a pu ely economic pe spec i e, i is somewha su p ising ha chea ing is no highe be o e ha es han a e ha es . Some p e ious e idence sugges ed ha financial ci cums ances a e co ela ed wi h une hical beha io (Gäch e and Schulz, 2016). We showed ha income i sel is no signi - ican enough o causally a ec one’s p opensi y o chea . Tha is, people in po e y need cash immedia ely, bu hei need does no make hem neglec he mo al disu ili y associa ed wi h chea ing. This esul is in line wi h hose om an expe i- men independen ly conduc ed a he same ime as ou expe imen : Aksoy and Palma (2019) measu ed chea ing beha io be o e and du ing ha es wi h Gua emalan co ee a me s and ound ha chea ing o one’s own benefi was he same be o e and du ing ha es .11 Tha he baseline esul is eplica ed a wo ends o he wo ld makes us e en mo e confiden 11 We we e unawa e o each o he ’s expe imen s and conduc ed hem du ing he exac same weeks. 575 S. Boonmanun , A. Kajackai e and S. Meie Games and Economic Beha io 124 (2020) 569–578 Table 4 Resul s o OLS eg essions on chea ing beha io . Repo ed Numbe (1–10) Dependen Va iable (1) (2) (3) (4) A e Ha es −0.15 (0.27) 0.03 (0.37) −0.13 (0.28) 0.01 (0.38) No m In e en ion −0.49* (0.27) −0.30 (0.41) −0.48* (0.28) −0.33 (0.43) A e Ha es ×No m -−0.38 (0.54) -−0.29 (0.56) Demog aphic Con ols Female −0.23 (0.28) −0.23 (0.29) Age −0.01 (0.02) −0.01 (0.02) Educa ion (yea s in school) 0.05 (0.06) 0.05 (0.06) Ma ied −0.02 (0.34) −0.03 (0.34) No. o child en −0.03 (0.12) −0.04 (0.12) Household annual income <−0.001 (<0.001) <−0.001 (<0.001) Sa ings (dummy) 0.11 (0.29) 0.12 (0.29) Cons an 6.70*** (0.25) 6.61*** (0.29) 7.01*** (1.20) 6.93*** (1.19) No. o pa icipan s 559 559 559 559 R20.0083 0.0094 0.0156 0.0162 No es: S anda d e o s, clus e ed a he illage le el, a e in pa en heses; * p< 0.1, ** p< 0.05, *** p< 0.01. in his esul . Di e en ly om ou s and Aksoy and Palma’s (2019) expe imen al design, in which he changes in finan- cial si ua ion a e na u ally occu ing (and no made salien ), Sha ma e al. (2014)used an exogenous financial-dep i a ion manipula ion be o e a chea ing game. The financial dep i a ion was ei he subjec i e (answe ing ques ions on financial de- p i a ion be o ehand) o objec i e (losing money in p e-s ages o he chea ing game). They ound ha he eeling o being financially disad an aged in a Wes e n s uden sample and a U.S. MTu k sample led o significan ly highe chea ing le els. When combined, he e idence om ou expe imen and he expe imen by Aksoy and Palma (2019)sugges s ha financial cons ain s hemsel es migh no lead o inc eases in chea ing beha io . Howe e , when he financial disad an age is made mo e salien (as in Sha ma e al., 2014), i migh lead o highe chea ing le els. Second, po e y ende s he mo al eminde ine ec i e. The explana ion, which is in line wi h he esul , is ha sca ci y educes one’s cogni i e esou ces and leads o “ unneling.” In o he wo ds, because o sca ci y, financially es ic ed a me s migh ha e no paid a en ion o he no m eminde ha we p o ided hem because hey we e p eoccupied wi h he financial cons ain s hey we e acing. A po en ial addi ional explana ion o he esul is economic: whe eas ich a me s can a o d o in e nalize he no m and ollow he desi ed beha io , poo a me s ha e ewe financial possibili ies o changing hei beha io acco ding o he no m. In e en ions shown o be highly e ec i e and o wo k when people a e iche and less cons ained migh no wo k when people a e financially cons ained. This makes i ha d o change he no ms in poo egions and neighbo hoods. How- e e , ou esul s sugges ha , fi s , iming o he in e en ions migh ma e . Mos households expe ience some mo e-o -less p edic able cycles o financial cons ain s due o pay cycles o seasonali y in income s eams. In e en ions o change beha - io need o ake hose cycles in o accoun and ime hei implemen a ion in pe iods o ewe financial cons ain s. Second, a social-no m eminde was ine ec i e when people we e poo in ou s udy, bu i is jus one o he po en ial in e en ions o change chea ing beha io . This in e en ion equi es a en ion and cogni i e esou ces in o de o be e ec i e— esou ces ha financially cons ained indi iduals may lack. When choosing a pa icula ype o in e en ion, one needs o emembe ha po e y educes people’s cogni i e capaci y and migh a ec in e en ions as a esul . No e ha we es ed jus one in e en ion and ha ou in e en ion equi es slack—cogni i e esou ces—whe eas some o he in e en ions migh ins ead c ea e slack. Simple in e en ions ha change he choice a chi ec u e (Johnson e al., 2012) and ha equi e ewe cogni i e esou ces, such as punishmen o chea ing (Khalme ski e al., 2017), migh be mo e e ec i e. Also, o he simple in e en- ions ha ha e p o ed o wo k well in o he con ex s could possibly be mo e e ec i e han a social-no m eminde , such as le ing people sign an oa h (as is Maza e al., 2008) o changing he pe cei ed likelihood o being caugh and punished (as in Cas o and Sca ascini, 2015; Hallswo h e al., 2015). 576