Addictive consumption, imperfect substitutes and self control: A model and an application to slot machines

Deiana, Claudio; D agone, Da ide; Giua, Ludo ica
Wo king Pape
Addic i e consump ion, impe ec subs i u es and sel
con ol: A model and an applica ion o slo machines
Quade ni - Wo king Pape DSE, No. 1197
P o ided in Coope a ion wi h:
Uni e si y o Bologna, Depa men o Economics
Sugges ed Ci a ion: Deiana, Claudio; D agone, Da ide; Giua, Ludo ica (2024) : Addic i e
consump ion, impe ec subs i u es and sel con ol: A model and an applica ion o slo machines,
Quade ni - Wo king Pape DSE, No. 1197, Alma Ma e S udio um - Uni e si à di Bologna,
Dipa imen o di Scienze Economiche (DSE), Bologna,
h ps://doi.o g/10.6092/unibo/amsac a/8037
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/306818
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/4.0/
ISSN 2282-6483
Addic i e Consump ion,
Impe ec Subs i u es and Sel Con ol:
A Model and an Applica ion o Slo Machines
Claudio Deiana
Da ide D agone
Ludo ica Giua
Quade ni - Wo king Pape DSE N°1197
Addic i e Consump ion, Impe ec Subs i u es and Sel Con ol:
A Model and an Applica ion o Slo Machines
*
Claudio Deiana

Uni e si `a di Caglia i, CRENoS and IZA
Da ide D agone

Uni e si `a di Bologna
Ludo ica Giua
§
Uni e si `a di Caglia i and CRENoS
No embe 20, 2024
Abs ac
We p opose a model o addic i e consump ion o s udy he demand o impe ec subs i u es
in ol ing subs ances like alcohol, nico ine and opioids, as well as beha io al addic ions like
gambling and digi al addic ion. We s udy a 2017 I alian policy aimed a educing gambling
by limi ing he numbe o a ailable slo machines. Despi e he educ ion in slo machines,
he policy p oduced an unin ended 25% inc ease in ne expendi u e, pa icula ly among
low-weal h and low-educa ed indi iduals who also engage in o he addic i e beha io s.
This esul can be a ionalized as he consequence o changes in sel -con ol cos s due o
social con agion e ec s.
Keywo ds: Addic ion; Gambling; Ho izon al di e en ia ion; Sel -con ol; Slo machines,
Temp a ion.
JEL codes: I18, L43, L83.
*
We hank Paolo Buonanno, Elias Ca oni, Rocco d’Es e, Ma co F ancesconi, Gio anni Mas obuoni, And ea
Mo o, Ma co Nieddu, Ma hias Pa ey, Giacomo Pasini, Giulio Zanella and pa icipan s a he SIEP Con e ence
(Ve ona, 2023), he HERB wo kshop (Bologna, 2023), he EuHEA Con e ence (Vienna, 2024), he BHEPPE
semina (Bologna, 2023) o commen s and sugges ions. Any e o s a e he aul o he au ho s only. The
au ho s decla e no con lic o in e es .

Uni e si `a di Caglia i, CRENoS and IZA, Depa men o Economics and Business, ia S. Ignazio da Laconi
17, 09123 Caglia i (I aly); email: [email p o ec ed].

Uni e si `a di Bologna, Depa men o Economics, Piazza Sca a illi 2, 40126, Bologna, I aly; e-mail: da-
ide.d [email protected] .
§
Co esponding au ho . Uni e si `a di Caglia i and CRENoS, Depa men o Economics and Business, ia S.
Ignazio da Laconi 17, 09123 Caglia i (I aly); email: ludo[email p o ec ed].
1
Non- echnical summa y
Addic i e beha io s, including subs ance use diso de s and beha io al addic ions, pose signi -
ican challenges o policymake s due o hei ad e se e ec s on physical and men al heal h,
in e pe sonal ela ionships, economic s abili y, and quali y o li e. Go e nmen s wo ldwide ha e
implemen ed s a egies such as axa ion, p ohibi ion, educa ional p og ams, and es ic ions
on access o mi iga e hese ha ms. These in e en ions o en ely on he economic p inciple
ha inc easing he cos o consump ion educes demand. Howe e , he ou comes can de ia e
om expec a ions when such policies inad e en ly al e sel -con ol and emp a ion cos s.
This pape examines he consump ion o addic i e goods ha a e impe ec subs i u es,
combining heo e ical modeling and empi ical analysis. I in eg a es concep s om ho izon al
di e en ia ion, a ional addic ion, and emp a ion cos s o p edic how addic ion le els, a ail-
abili y o al e na i es, and sel -con ol cons ain s in luence beha io . The heo e ical ame-
wo k add esses a b oad ange o addic i e goods, including subs ances like nico ine, alcohol,
and opioids, as well as beha io al addic ions such as gambling and gaming diso de s. Al hough
beha io al addic ions a e less damaging o physical heal h han subs ance use diso de s, hey
can signi ican ly impai unc ioning ac oss pe sonal and p o essional domains.
The empi ical applica ion ocuses on gambling addic ion, speci ically assessing he e ec s
o a 2017 I alian policy aimed a educing gambling oppo uni ies by limi ing he a ailabili y
o slo machines. Using municipali y-le el adminis a i e da a and household expendi u e
su eys, he s udy e alua es he policy’s impac h ough a di e ence-in-di e ences app oach.
The indings show ha he numbe o slo machines declined by 21% and gambling enues by
11%. Howe e , con a y o he policy’s in en , slo machine expendi u es inc eased by 25%,
d i en la gely by low-income and less-educa ed indi iduals. These unin ended consequences
indica e ha he policy may ha e exace ba ed ha m o he e y g oups i sough o p o ec .
Unde s anda d economic assump ions, educed access should discou age gambling by in-
c easing anspo a ion and oppo uni y cos s. Howe e , he indings sugges ha educed
access also led o inc eased wai ing imes, which ampli ied c a ings and emp a ion cos s, es-
pecially among mo e addic ed indi iduals. This dynamic likely inc eased gambling in ensi y,
unde sco ing he impo ance o psychological ac o s in d i ing addic i e beha io s
2
1 In oduc ion
Subs ance use diso de s and beha io al addic ions aise policy conce ns due o hei impac on
indi idual physical and men al heal h, pe sonal ela ionships, economic and inancial s abili y,
and o e all quali y o li e. To add ess hese conce ns, policymake s ha e adop ed a ange o
s a egies, including p ohibi ion and axa ion o addic i e subs ances, educa ion p og ams, age
limi s, and ime cu ews (see, e.g., E ans e al.,1999;Adda and Co naglia,2006;Ca pen e
and Dobkin,2009;Ange e al.,2011;Cawley and Ruhm,2012;Hansen,2015;Ahamme e al.,
2022;Moo e and Mo is,2024). The a ionale unde lying hese o ms o egula ion is ha
consump ion, e en when addic i e, is expec ed o dec ease when i s cos inc eases (Becke and
Mu phy,1988). Howe e , when he policy in e en ion also a ec s sel -con ol and emp a ion
cos s, his p edic ion may no hold.
In his pape , we p opose a heo e ical and empi ical in es iga ion in o he consump ion o
goods ha a e impe ec subs i u es, addic i e, and induce sel -con ol cos s. Building on he
Ho elling (1929)’s model o ho izon al di e en ia ion, he Becke and Mu phy (1988)’s heo y o
a ional addic ion, and he Gul and Pesendo e (2001)’s model o emp a ion and sel -con ol,
we de i e p edic ions abou he ex ensi e and in ensi e ma gins o addic i e consump ion as
unc ions o he le el o addic ion, cha ac e is ics and a ailabili y o di e en addic i e op ions,
emp a ion cos s, and access cos s. The heo e ical model p o ides a concep ual amewo k
o unde s and consump ion ela ed o addic i e subs ances such as, e.g., nico ine (Pesko and
Wa man,2022), alcohol (Calco ,2019) and opioids (Ag awal e al.,2023), and o beha io al
addic ions such as gambling and gaming diso de s (WHO,2022).1
As an empi ical applica ion, we ocus on gambling addic ion. Speci ically, we es ima e he
e ec s o a 2017 policy in I aly ha manda ed a la ge educ ion in he numbe o a ailable slo
machines. Using no el I alian adminis a i e da a on a ious ypes o gambling ac i i ies a
he municipali y le el and a ep esen a i e su ey da a on household expendi u e, we es ima e
he e ec s o he policy wi h a s anda d di e ence-in-di e ences app oach.
We ind ha in ea ed municipali ies he numbe o slo machines dec eases by 21%, and
he numbe o enues hos ing hem dec eases by 11%. Howe e , he policy leads o unin ended
consequences, wi h ne slo machine expendi u e ising by 25% (30.20 EUR). This inc ease is
d i en by highe spending among playe s wi h low weal h and low educa ion who also engage
1Despi e being non-subs ance- ela ed, and hus gene ally less ha m ul o heal h, beha io al addic ions can
also yield ad e se consequences, impai ing an indi idual’s abili y o unc ion ac oss a ious li e domains (Ala i
e al.,2012;WHO,2018,2024). The ICD11 – he WHO (2022)’s la es In e na ional Classi ica ion o Diseases–
desc ibes gambling and gaming as diso de s due o addic i e beha io s ha de elop as a esul o epe i i e,
ewa ding ac i i ies. O he diso de s, such as shopping diso de , exe cise addic ion (Be czik e al.,2012), and
digi al addic ions ela ed o in e ne use, including sma phone addic ion (Allco e al.,2022), a e no included
in he ICD-11. See Pe y (2015) o a comp ehensi e o e iew.
3

in o he addic i e beha io s. Since slo machine playe s a e p edominan ly om disad an aged
backg ounds (Mas oba is a e al.,2019;Resce e al.,2019), ou esul s imply ha he policy
has ha med he e y popula ion i aimed o p o ec .
The heo e ical model allows o gain insigh s on he possible mechanisms d i ing his ind-
ing. In he absence o sel -con ol and emp a ion cos s, he educ ion in slo machines should
inc ease he anspo a ion cos s and educe willingness o play because o , e.g., inc eased
conges ion and longe wai ing imes. This would unambiguously educe he ex ensi e and
in ensi e ma gin o slo machines play. The obse ed e idence sugges s ha , on he con a y,
emp a ion cos s do play a ole because wai ing imes can also inc ease c a ings o playing,
and he mo e so in mo e addic ed indi iduals. When hese c a ings and he associa ed emp-
a ion cos s a e la ge enough, he in ensi e ma gin o playing and, hus o al expendi u e on
slo machines, can e en ually inc ease.
Theo e ically, ou pape b idges h ee dis inc s eams o li e a u e: addic ion, ho izon al
p oduc di e en ia ion and sel -con ol. The li e a u e on addic ion is ypically ocused on
indi idual choices ela ed o an addic i e good ha ea u es habi - o ma ion and sel -con ol
p oblems. The o me is based on he assump ion ha cu en p e e ences depend on pas
consump ion expe iences (Pollak,1970), an idea ha is ypically o malized as accumula ion
o an addic ion s ock which c ea es ole ance and ein o cemen (Becke and Mu phy,1988).
The la e has been s udied as he consequence o eg e (O phanides and Ze os,1995), ime-
inconsis en beha io (S o z,1955;Laibson,1997;G ube and K¨oszegi,2001;Piccoli and
Tiezzi,2021), cos ly emp a ion (Gul and Pesendo e ,2001,2007), p ojec ion bias (Loewen-
s ein e al.,2003), dual-sel models (Thale and She in,1981;Loewens ein and O’Donoghue,
2004;Ozdeno en e al.,2012), en i onmen al cues (Be nheim and Rangel,2004), o a combi-
na ions o hese ac o s (Allco e al.,2022). Mos exis ing con ibu ions, howe e , ocus on a
single addic i e good.2In his pape we ex end he analysis o allow o impe ec subs i u es
and ho izon al p oduc di e en ia ion `a la Ho elling (1929). This allows o ake in o accoun
ha people a e he e ogeneous, and hey can sa is y hei needs h ough goods ha a e impe -
ec subs i u es, including he op ion o abs aining i he addic i e al e na i es a e no good
enough.
Empi ically, we con ibu e o unde s anding he e ec s o a policy aimed a egula ing ad-
dic i e beha io by es ic ing he a ailabili y o consump ion oppo uni ies. Speci ically, we
add o he economics li e a u e on gambling beha io . By p o iding he i s causal analysis o
a policy aimed a educing gambling access, ou esul s complemen hose o Badji e al. (2023)
and Bake e al. (2024), who analyze he e ec s o he opposi e ype o egula ion—g ea e
2Fo excep ions see, e.g., Dockne and Feich inge (1993); Bask and Melke sson (2004); Cawley and D agone
(2024).
4
gambling accessibili y. Addi ionally, we con ibu e o empi ical s udies on gambling addic-
ion and he subs i u abili y be ween di e en ypes o gambling (Kea ney,2005;Gu yan and
Kea ney,2010). Finally, by emphasizing he possible oles o emp a ion and social con agion
in gambling choices, we con ibu e o unde s anding beha io al d i e s o gambling (S e zka
and Win e ,2023), such as he “lucky s o e” e ec (Gu yan and Kea ney,2008), he impac
o unme liquidi y needs and inancial cons ain s on gambling beha io (He skowi z,2021),
and he ole o media exposu e (De Paola and Scoppa,2014).
The pape is o ganized as ollows. In he nex Sec ion we p esen a model o addic i e
consump ion wi h impe ec subs i u es and sel -con ol cos s. In Sec ion 3we desc ibe he
empi ical applica ion o he model o he case o he gambling indus y in I aly. Sec ion 4de-
sc ibes he empi ical s a egy and Sec ion 5 epo s he empi ical esul s. Sec ion 6concludes.
2 Becke and Mu phy mee Ho elling, Gul and Pesendo e
2.1 A model o addic i e consump ion wi h impe ec subs i u es and sel -
con ol cos s
Conside a scena io whe e addic i e consump ion can occu a wo enues , deno ed as A
and Band loca ed a he ex emes o a [0,1] segmen . Indi iduals a e dis ibu ed along he
[0,1] segmen and pay a ma ginal anspo a ion cos τ ≥0 o each enue ∈ {A, B}and
consume. Once in a enue, indi iduals alloca e hei budge mbe ween suni s o addic i e
consump ion and a composi e nume ai e z. The cos o each addic i e uni consumed a enue
is p .
In each pe iod , indi iduals i s choose whe he o each a enue, hen hey choose he
op imal amoun o addic i e consump ion. The o me choice in o ms abou he ex ensi e
ma gin o addic i e consump ion, while he la e one desc ibes he in ensi e ma gin. Addic i e
consump ion con ibu es o building an indi idual le el o addic ion a≥0. Addic ion e ol es
o e ime depending on pas and cu en consump ion, acco ding o
a( + 1) = γ(s( ) + a( )) ,(1)
whe e pa ame e γ∈[0,1) and a(0) = a0. A e playing, indi iduals e u n o hei ini ial
loca ion (“home”), he addic ion le el changes and a new pe iod begins.
Consump ion o he composi e good yields u ili y Z(z), wi h Zz>0,Zzz ≤0. I an
indi idual a ends a enue , consuming sa enue (deno ed as s ) yields, o a gi en le el
o addic ion a, he ollowing u ili y:
U(s ;a) (2)
5
The u ili y unc ion (2) is s ic ly inc easing in addic i e consump ion (Us>0). Consis en
wi h Becke and Mu phy (1988) and Becke e al. (1991), we assume ha gi en le els o playing
a e less sa is ying when pas consump ion has been g ea e (Ua<0) and ha he mo e a pe son
has played in he pas , he mo e hey like playing oday (Usa >0). These wo assump ions
desc ibe ole ance and ein o cemen , which a e ypical p ope ies o addic ions.3
U ili y om addic i e consump ion depends on he cha ac e is ics o he speci ic enue.
These cha ac e is ics can li e ally desc ibe ea u es o he enue whe e addic i e consump ion
occu s, such as he possible p esence o ameni ies, he quali y o he se ice, he p esence o
o he consume s, bu hey can also be in e p e ed mo e b oadly as ea u es o he wo a ailable
consump ion goods. Wi h he la e in e p e a ion sAand sBdesc ibe impe ec subs i u es
o addic i e consump ion o which consume s ha e di e en as es, condi ional on consuming.
This is he case o , e.g., aping as an impe ec subs i u e o smoking, bee o wine, o he oin
o o he opioids.
Addic i e consump ion ypically ea u es emp a ion and sel -con ol cos s. Speci ically,
we assume ha addic ed indi iduals a e emp ed o spend all a ailable budge mon addic i e
consump ion, dis ega ding he composi e good z. In he spi i o Gul and Pesendo e (2001,
2004), indi iduals can pa ially o e ule such emp a ion, bu a a cos . Speci ically, an indi-
idual consuming s pays a emp a ion cos ha depends on he dis ance be ween he emp ing
choice o spending all budge on he addic i e good and he ac ual choice. Fo conc e eness,
we conside he ollowing cos unc ion (Allco e al.,2022;Cawley and D agone,2024):
C(s ;a, σ )≡a·σ ·(m−s )≥0 (3)
Exp ession (3) depends on a o accoun o he possibili y ha he emp a ion cos is highe
when addic ion is highe , and nil in case o no addic ion, Pa ame e σ ≥0 desc ibes he
ma ginal emp a ion cos a enue , and i can depend on ac o s such as p e ious exe ion o
sel -con ol (Mu a en and Baumeis e ,2000;Loewens ein and O’Donoghue,2004), cogni i e
load (Shi and Fedo ikhin,1999), as well as en i onmen al cues (Loewens ein,1996,2000;
Be nheim and Rangel,2004), wai ing imes (House e al.,2018,2021), social con agion and
pee -p essu e (Lundbo g,2006;Cla k and Loh´eac,2007).
Taking in o accoun he emp a ion cos s, he indi idual objec i e unc ion is:
U(s , z;a, σ )≡ U (s ;a) + Z(z)− C (s ;a, σ )
=U(s ;a) + Z(z) + aσ (s −m) (4)
3The u ili y unc ion is also assumed o be s ic ly conca e. No e ha pa ial de i a i es o he u ili y
unc ion a e deno ed wi h subsc ip s, as in, e.g., Usa ≡∂2U(·)
∂s∂a .
6
I can easily be e i ied ha he objec i e unc ion (4) is inc easing in consump ion (Us>0),
ea u es ein o cemen and ole ance (Usa >0, Ua<0), and ha i dec eases when he
emp a ion pa ame e is highe (Uσ≤0).
Indi iduals ha do no a end ei he enue spend all budge on he composi e good and
ob ain he ese a ion u ili y U(0) ≡U(0, m;a, σ0), which nega i ely depends on addic ion a
and he emp a ion pa ame e σ0.
2.2 Sol ing he model
The p oblem is sol ed by backwa d induc ion. In he second s age o each pe iod , and
condi ional on being a enue ∈ {A, B}, an indi idual wi h addic ion s ock aop imally
chooses addic i e consump ion sand he amoun o consump ion o he composi e good z ha
sol e:4
max
s ,z U(s , z;a, σ ) (5)
s. . m=p s +z(6)
The op imal amoun o addic i e consump ion (s∗
, z∗
) exhaus s he a ailable budge and,
assuming an in e io solu ion, i sa is ies he amilia condi ion whe e he ma ginal a e o
subs i u ion be ween sand zequals he ela i e p ice:
(s∗
, z∗
) : Us(s∗
, z∗
)
Uz(s∗
, z∗
)=p (7)
The pa icipa ion choice in he i s s age depends on he indi idual loca ion ialong he uni
line, and on he indi idual addic ion le el a. The o me a ec s he anspo a ion cos o
each he enues, he la e a ec s he u ili y om consump ion and he ma ginal incen i es o
consume. An indi idual chooses o each a speci ic enue, o he ou side op ion o no addic i e
consump ion, by compa ing he ese a ion alue U(0) and he maximized u ili y o consuming
a ei he enue, i.e.:
V(i, a, A)≡U(s∗
A, z∗
A;a, σA)−τA·i, (8)
V(i, a, B)≡U(s∗
B, z∗
B;a, σB)−τB·(1 −i) (9)
4Unlike a ional addic ion models, which assume indi iduals accoun o how cu en choices impac u u e
addic ion, he e we assume ha indi iduals a e myopic. Reade s in e es ed in o wa d-looking addic ion models
wi hou ho izon al p oduc di e en ia ion can e e o Becke and Mu phy (1988), Chaloupka (1991), and
D agone and Raggi (2021) o ime-consis en beha io , and o Piccoli and Tiezzi (2021) and Allco e al.
(2022) o models wi h ime-inconsis en agen s.
7
his inding is sugges i e, i does no allow o causal conclusions abou he e ec i eness o he
2017 bill. This analysis will be conduc ed in Sec ions 4and 5.
3.2 Da a sou ces
We use wo main da ase s: adminis a i e da a on gambling a municipali y le el and su ey
da a on expendi u e a he household le el. Fo addi ional insigh s on he empi ical esul s,
we use he ADM geolocalized da a on he uni e se o obacco shops in 2016 and he I alian
Census 2011 da a on socio-demog aphic cha ac e is ics.
Gambling da a
The gambling da a a e sou ced om ADM. They span om 2015 o 2019, hey co e 7,877
I alian municipali ies, and hey con ain he uni e se o legal gambling ac i i ies in I aly. ADM
epo s municipali y-le el yea ly da a on he coun s o slo machines (bo h Newslo s and Vide-
olo e ies), he coun s o licensed enues whe e slo machines a e ins alled, and he o al
expendi u e ne o winnings on a ious o ms o gambling.
Table 1p esen s he main summa y s a is ics on he 2015 pe capi a ne expendi u e a he
municipali y le el, ca ego ized by gambling ype: slo machines, spo s be s, lo e ies, bingo,
sc a ch ca ds, and online be ing. Slo machines accoun o he highes pe capi a expendi u e,
a e aging EUR 121.35. This igu e is la gely d i en by expendi u e on Newslo s, which alone
accoun s o EUR 101.83. Expendi u e on o he ypes o gambling is subs an ially lowe . Fo
ins ance, lo e ies, which cons i u e he second-highes ca ego y, collec an expendi u e ha
is app oxima ely hal o ha spen on slo machines. Online be ing collec s only EUR 0.88
on a e age in 2015. I is no ewo hy ha , al hough his igu e nea ly doubles o EUR 1.42 by
Mean S d. de . Min Max
Slo Machines 121.35 345.64 -1.00 28,416.85
— Newslo s 101.83 218.18 -1.00 17,992.04
— VLT 19.51 140.43 0.00 10,424.82
Lo e ies 60.77 213.05 -4,359.54 17,614.33
Sc a ch ca ds 7.01 41.86 -1,022.57 3,404.31
Spo be s 4.09 30.49 -22.65 2,575.94
Bingo 1.24 15.89 -70.48 625.04
Online 0.88 6.20 -1.26 491.62
Table 1: Pe capi a ne expendi u e in 2015, by ype o gambling. Municipali y-le el pe
capi a expendi u e, ne o winnings, in he yea 2015. Sample: 7,877 municipali ies.
14

(a) Newslo s pe capi a (b) Newslo s ne expendi u e pe capi a
Figu e 4: Newslo s: pe capi a numbe and ne expendi u e in 2015. Geog aphical dis ibu-
ion o he pe capi a numbe o Newslo s (panel a) and he pe capi a ne expendi u e on Newslo s
(panel b) by municipali y in 2015. Black lines indica e egional bo de s (20 NUTS-2 a eas).
2019, i s o e all impo ance in he I alian ma ke du ing his pe iod emains ela i ely small.
The geog aphical dis ibu ion o pe capi a Newslo s is posi i ely co ela ed wi h pe capi a
ne expendi u e, as shown in Figu e 4 o 2015.12 In line wi h he 35% educ ion manda ed
by he 2017 bill, be ween 2016 and 2018 he numbe o Newslo s dec eases by abou one hi d
in each egion. This is accompanied by a dec ease in he numbe o enues (see Figu e B.1 in
Appendix B).13 The numbe o Videolo e ies and hei co esponding enues, which a e no
subjec o he policy, emains subs an ially cons an o e ime o sligh ly inc eased.
Household expendi u e da a
To examine he he e ogeneous e ec s o addic ion le els and socio-economic s a us on gambling
expendi u e, we use he I alian Household Expendi u e Su ey (HES). Published by he I alian
Na ional Ins i u e o S a is ics (ISTAT), he su ey is ep esen a i e o he I alian popula ion
and co e s a b oad ange o household expenses, including gambling and alcohol consump-
ion, as well as demog aphic cha ac e is ics o household membe s and sel - epo ed economic
condi ions. We analyze da a om he 2014 o 2019 su ey wa es, wi h an o e all sample o
o e 100,000 households. The da a includes o al expendi u e, and speci ic spending de ails on
12 The pai wise co ela ion is equal o 0.25 and is s a is ically signi ican a any con en ional le el.
13 Piedmon and Valle d’Aos a a e an excep ion as a consequence o mo e s ingen local policies. As shown
in he obus ness checks sec ion, his does no d i e he empi ical esul s.
15
gambling and on alcoholic be e ages, such as bee , cide , wine, liquo s, and spi i s. I co e s
egional in o ma ion, household composi ion (numbe o membe s), and de ailed cha ac e is-
ics o he household head, including gende , age, na ionali y, educa ion le el, employmen
s a us, occupa ion, economic sec o , and ype o employmen con ac .
On a e age, o al household mon hly expendi u e is a ound EUR 2,500. Households who
epo no expendi u e on gambling amoun o 87% o he sample. Figu e B.2 shows he
dis ibu ion o expendi u e on gambling o he emaining 13%. Fo his subsample, a e age
epo ed expendi u e on gambling is EUR 23.77.
4 Empi ical s a egy
4.1 Concep ual amewo k
We empi ically assess he causal e ec s o he 2017 bill using a s anda d di e ence-in-di e ences
app oach. The p edic ions om he heo e ical model p esen ed in Sec ion 2p o ide a na u al
amewo k o in e p e ing he empi ical es ima es. In ac , slo machine play can be en e -
aining (Conlisk,1993;Bu ge e al.,2020), al hough i may lead o nega i e consequences,
such as ad e se e ec s on men al heal h, including dep ession, anxie y, and s ess (Muggle on
e al.,2021;Wa dle and McManus,2021;Badji e al.,2023), as well as loss o money and
inancial ha dship. Fu he mo e, slo machines a e designed o c ea e addic ion (Sch¨ull,2012),
pa icula ly due o isual and senso y ea u es ha encou age ein o cemen h ough epea
play (Ha igan e al.,2010;James e al.,2016).
Deno ing he le el o gambling addic ion as a, and he amoun o slo machine play– he
numbe o spins– as s, he equi emen s o he u ili y unc ion used in he Sec ion 2a e sa is ied
(Us>0, Ua<0, and Usa >0). Mo eo e , playing slo machines equi es eaching he physical
enue ( ) whe e hey a e loca ed, a a anspo a ion cos (τ ).14 Once a he enue, indi iduals
choose how much o play (s ). The ne expec ed cos o playing is p≡P−E(w)>0, whe e
Pis he cos pe play (by law, i is equal ac oss enues) and E(w) is he expec ed mone a y
win.15
Deno e enue Aas a ea ed uni , and enue Bas he con ol uni . Empi ically, we aim a
14 Slo machines a e impe ec subs i u es because hey a e loca ed in di e en enues. In his sense,
he ho izon al di e en ia ion componen o he model is li e ally unde s ood in e ms o loca ion. Howe e ,
anspo a ion cos s can also be in e p e ed as access o s igma cos s o being a playe . We will conside his
al e na i e in e p e a ion when discussing he empi ical esul s in Sec ion 5.
15 Taking li e ally he no ion o playing, ou app oach is complemen a y o he app oach based on isk-
lo ing p e e ences o e mone a y ou comes, which explains why people choose unce ain p ospec s e en when
he ne expec ed gain is nega i e. See F iedman and Sa age (1948); Ha ley and Fa ell (2002); Le i (2004)
o an al e na i e app oach ocused on he ole o unce ain y in gambling choices. Fo an app oach aimed a
desc ibing op imal s opping ime o playing, based on cumula i e ea nings, see Lien and Zheng (2015).
16
es ima ing he ollowing di e ence-in-di e ences (DiD) es ima o :
β=E′
A−EA−E′
B−EB= ∆EA−∆EB(11)
whe e E′
deno es ne expendi u e a enue a e he implemen a ion o he policy, and ∆EA
and ∆EBa e he di e ences in ne expendi u e a enue Aand B, espec i ely.
Using he e minology o he heo e ical model, we conjec u e ha he educ ion in he
numbe o a ailable slo machines p oduces di e en e ec s. Fi s , a educ ion in he numbe
o a ailable slo machines is likely o inc ease he anspo a ion cos τA o each he ea ed
enue A. Second, he policy may educe he u ili y de i ed om gambling a A, due o
ac o s such as conges ion o longe wai ing imes, which may make slo machine play a A
less enjoyable. Bo h mechanisms imply ha he expec ed e ec o he bill is o educe he
ex ensi e and in ensi e ma gins o slo machine play a A(hence ∆EA<0) and o inc ease
he numbe o playe s a B(hence ∆EB>0). This is consis en wi h Badji e al. (2023), who
show ha people esiding in close p oximi y o gambling enues a e mo e likely o gamble and
less likely o be happy.
Thi d, i is possible ha he bill in luences he ma ginal emp a ion cos σAo playing a
A, due o, e.g., longe wai ing imes ha inc ease he desi e o consume (Loewens ein,1987;
House e al.,2018,2021) o he cos o exe ing sel -con ol (Mu a en and Baumeis e ,2000;
Vohs and Fabe ,2007;Hagge e al.,2010;Baumeis e e al.,2018), o possible social con agion
e ec s induced by highe concen a ion o playe s in he same enue (Rocklo and Dye ,2007;
Rocklo e al.,2011,2012,2017;Hop ga ne e al.,2021). Since he model p edic s ha
g ea e emp a ion leads o highe consump ion (bu also less consume s) a A, i s e ec on
ne expendi u e can oppose hose p oduced by inc eased anspo a ion cos s and educed
u ili y.16
The sign o he empi ical es ima e o βcan sugges which o he mechanisms abo e domi-
na es. In he absence o changes in he emp a ion pa ame e σA, an inc ease in anspo a ion
cos s and a educ ion in he u ili y om playing is p edic ed o unambiguously dec ease ex-
pendi u e a A, while expendi u e a Bshould inc ease (see P oposi ions 1and 2). In such a
case, he di e ence-in-di e ence es ima e βis p edic ed o be nega i e. I , howe e , he policy
also p oduces an inc ease in he emp a ion pa ame e σA, o a educ ion in he s igma cos
associa ed o being a playe , ne expendi u e a enue Amay inc ease (hence ∆EA>0). I
16 In p inciple, longe wai ing imes and inc eased c owding can educe he u ili y o playing, bu hey may
also aise emp a ion cos s. Fu he mo e, a highe concen a ion o playe s in ewe enues could educe he
pe cei ed s igma associa ed wi h being a playe , which migh be in e p e ed as a educ ion in “ anspo a ion
cos s” (i τAis unde s ood in a non-li e al sense), leading o mo e playe s a enue A. We do no ake an a-p io i
s ance on which o hese po en ial channels domina es.
17
such e ec is la ge enough, βis posi i e (see Equa ion 11).
4.2 T ea ed uni s
The 2017 bill equi ed Newslo s o be educed in a eas whe e hey we e mo e abundan and
leas p o i able. Hence, we classi y municipali ies as ea ed uni s ( he empi ical coun e pa s o
enue Ain he model) i wo c i e ia a e sa is ied: (i) he numbe o slo machines pe capi a a
he municipali y le el is abo e he na ional a e age as o Decembe 2016, and (ii) o al e enue
pe de ice a municipal le el is below he 15 h pe cen ile o he egional dis ibu ion.17
Panel (a) o Figu e 5illus a es he ea men assignmen , using he Emilia-Romagna egion
as an example. Municipali ies a e i s ca ego ized based on whe he he pe capi a numbe
o Newslo s in he municipali y exceeds o alls below he na ional a e age. This co esponds
o c i e ion (i) and is g aphically ep esen ed by he ed and pink cu es in panel (a). Wi hin
each g oup, we iden i y municipali ies whe e he a e age e enue pe de ice in 2016 is below
he 15 h pe cen ile This co esponds o c i e ion (ii) and by he municipali ies o he le o
he e ical line in panel (a) o Figu e 5. The e o e, he ea ed municipali ies a e hose ha
ha e a highe - han-a e age numbe o Newslo s pe capi a ( ed cu e) and lowe - han-a e age
(a) Municipali ies by e enue pe de ice, Emilia Romagna
0 5.000e-06 .00001 .000015 .00002 .000025
Densi y
0 50000 100000 150000
Re enue pe de ice a municipal le el
Newslo s pe capi a < na ional a e age Newslo s pe capi a > na ional a e age
(b) T ea ed municipali ies, I aly
Figu e 5: Assignmen o ea ed and con ol uni s. Panel (a) shows he dis ibu ions o munici-
pali ies based on he e enue pe de ice in he Emilia-Romagna egion in 2016. The pink and ed cu es
co espond o he municipali ies wi h a numbe o Newslo s pe capi a below and abo e he na ional
a e age, espec i ely. The e ical line indica es he 15 h pe cen ile o he egional dis ibu ion o he
e enue pe de ice. The dashed ed cu e deno es he ea ed uni s. Panel (b) shows he geog aphical
dis ibu ion o he ea ed municipali ies (221, in ed). Black lines indica e egional bo de s (20 NUTS-2
a eas).
17 The 15% h eshold is mo i a ed by he 15% educ ion in slo machines manda ed in 2017. The empi ical
esul s a e obus o using di e en h esholds (see Appendix C).
18
e enues pe de ice (le o he e ical line). These ea ed uni s a e indica ed by he dashed
ed cu e. This p ocedu e is epea ed o each o he 20 I alian egions. Panel (b) displays he
geog aphical dis ibu ion o he ea ed municipali ies (221, 3% o he sample) ac oss he 107
p o inces and he 20 egions o I aly.
Figu e B.3 illus a es he ime a ia ion in he pe capi a numbe o Newslo s and he pe
capi a ne expendi u e on o al slo machines ac oss ea ed and con ol municipali ies. Panel
(a) shows he compliance o he ea ed municipali ies wi h he 2017 bill. Be o e 2017, ends
a e pa allel ac oss ea ed and con ol uni s; a e wa ds, he gap na ows. Panel (b) displays
he e olu ion o pe capi a ne expendi u e. Expendi u e inc eases o e ime in ea ed uni s,
while i emains subs an ially cons an in con ol municipali ies.
4.3 Empi ical model
We es ima e he ollowing empi ical model:
Ymp =α+βCm×
1
( ≥2017) + γmp +δ p +ϵmp (12)
whe e Ymp is he ou come o in e es o municipali y min p o ince pand yea . Speci ically,
we conside he pe capi a numbe o Newslo s, he pe capi a expendi u e on slo machines,
he pe capi a numbe o enues hos ing Newslo s, he pe capi a numbe o Videolo e ies
(i.e. slo machines no a ec ed by he cu ), and he expendi u e on o he ypes o gambling
(sc a ch ca ds, lo e ies, spo s be s, and online be s).
The coe icien o in e es is β, as i cap u es he causal e ec o educing he numbe o
slo machines (Cm) on he ou come. In he heo e ical model, his co esponds o Equa ion 11.
To accoun o ime- a ying he e ogenei ies a sub- egional le el, he empi ical model includes
municipali y ixed e ec s (γmp) and a se o yea -p o ince dummy a iables (δ p). Reg essions
a e popula ion-weigh ed and s anda d e o s a e clus e ed a municipali y le el.
To es o he possible exis ence o di e en ials be ween ea men and con ol g oups in
he p e-policy pe iod, and o ule ou ha he educ ion in Newslo s is endogenously ela ed o
p e- ea men di e en ials in he ou comes, we conside he ollowing e en -s udy speci ica ion:
Ymp =α+
2019
X
j=2015
βjCm×
1
[ =j] + γmp +δ p +ϵmp (13)
wi h yea 2016 as he baseline. We also use his speci ica ion o examine po en ial he e oge-
neous e ec s based on he dis ance o obacco shops, which se es as a p oxy o he ans-
po a ion cos s discussed in P oposi ion 1.
19

To u he suppo he causal in e p e a ion o ou esul s, we un addi ional es s as
obus ness checks. Fi s , we pe o m sensi i i y analyses on he de ini ions o ea ed and
con ol g oups. Second, we e-es ima e he model conside ing di e en sub-samples. Thi d,
we conduc se e al alsi ica ion exe cises, including exe cises ocusing on una ec ed ou comes
and ea men andomiza ion. The esul s a e epo ed in Appendix C.
Finally, we use household-le el da a om he Household Expendi u e Su ey (HES) o
in es iga e he possible co ela ion be ween gambling and isky heal h beha io s, and o un-
de s and wha pa s o he popula ion a e mos a ec ed by he policy. This analysis is policy-
ele an because he li e a u e on isky heal h beha io s has consis en ly shown ha indi id-
uals who engage in one isky beha io , such as d inking, a e mo e likely o engage in o he s,
such as smoking, subs ance use, and gambling (Cawley and Ruhm,2012). Mo eo e , Resce
e al. (2019) no e ha in I aly, slo machines a e p edominan ly used by indi iduals o lowe
socio-economic s a us, aising conce ns abou income- ela ed inequali ies in gambling, pa -
icula ly among he mos ulne able popula ions. We es ima e he ollowing eg ession wi h
Pseudo-Poisson Maximum Likelihood (PPML):
Wh p = exp (ϑ+ϕTh ×
1
(p≥2017) ×Ah p +ψp+ζ +Xh pη) + νhp (14)
whe e Wh p is he mon hly household expendi u e o household hin egion and pe iod p
(mon h o he yea in 2014m1-2019m12). We conside epo ed expendi u e on gambling and
o al expendi u e.
Since he HES da ase only p o ides geog aphical in o ma ion a he egional le el, wi h-
ou speci ying municipali ies, we mus adap he iden i ica ion s a egy o he egional scale.
Households a e de ined as ea ed (Th = 1) i hey eside in a egion ha is ela i ely mo e
exposed o he educ ion in Newslo s in 2017 — ha is, in a egion whe e he sha e o he
popula ion li ing in ea ed municipali ies is abo e he median o he dis ibu ion. Va iable
Ai p iden i ies households who a e in he op qua ile o he dis ibu ion o expendi u e in
alcohol (bee , cide , wine, liquo s and spi i s). This p oxies o indi iduals wi h high le els
o addic ion o alcohol and, possibly, o addic ion in gene al. Wi h his in e p e a ion, he
coe icien ϕcap u es he causal e ec o he educ ion in Newslo s on mon hly expendi u e
o households wi h highe addic ion le els.
The empi ical model in Equa ion 14 includes pe iod (ψp) and egion ixed e ec s (ζ )
o cap u e he e ogenei ies o e ime and ac oss egions. Xh p desc ibes a se o household
cha ac e is ics: numbe o componen s, household head’s gende , age, na ionali y, educa ion,
employmen s a us, occupa ion, economic sec o and ype o con ac .18
18 The ca ego ies o employmen s a us a e: employed, unemployed, homemake /s uden , e i ed, o o he ;
20
5 Empi ical esul s
5.1 E ec s o he policy: Adminis a i e da a
Table 2displays ou main esul s, es ima ed using Equa ion 12. As shown in column 1, com-
pliance wi h he bill was high, as he numbe o Newslo s pe capi a in ea ed municipali ies
dec eased by 21% a e 2017. Column 2 o Table 2shows ha also he numbe o enues hos -
ing Newslo s declined, by app oxima ely 11%. The numbe o Newslo s pe enue dec eased
by a simila amoun (column 3).
The co esponding e en -s udy analyses a e illus a ed in Figu e 6. Panel (a) shows ha
he bill educed he pe capi a numbe o Newslo s by 0.0005 in 2017 and by o 0.0015 in 2018
and 2019. Gi en he 2015 a e age o 0.0054, his means ha he educ ion in he pe capi a
numbe o Newslo de ices eached 10% by he end o 2017 and 31% by he end o 2019. Panel
(b) and (c) show ha he educ ion in he pe capi a numbe o enues was 4% in 2017 and
17% in 2019, while he numbe o Newslo s pe enue dec eases by 7% in 2017 and by 14% in
2018 and 2019. Toge he , hese esul s sugges ha he policy likely inc eased anspo a ion
cos s o playe s and led o highe c owding in he emaining enues.
Con a y o he goal o he policy, pe capi a expendi u e on slo machines signi ican ly
inc eases by 25% (EUR 30.20) in ea ed municipali ies compa ed o con ol uni s. The esul s
a e shown in column 4 o Table 2. The co esponding e en -s udy analysis, shown in panel (d) o
(1) (2) (3) (4)
Pe capi a
Newslo s Pe capi a enues Newslo s pe enue Pe capi a
expendi u e
T ea ×Pos -0.0012*** -0.0002** -0.4088*** 30.1966***
(0.0003) (0.0001) (0.1052) (9.3459)
Obse a ions 39,385 39,385 39,385 39,385
Municipali ies 7,877 7,877 7,877 7,877
Mean ou come in 2015 0.0054 0.0015 3.6800 121.3451
Elas ici y -21.42 -10.69 -11.11 24.88
Municipali y FE ✓ ✓ ✓ ✓
P o ince ×Yea ✓ ✓ ✓ ✓
Table 2: E ec on Newslo s, enues and expendi u e. The ou comes a e he numbe o Newslo s
pe capi a (column 1), he numbe o enues hos ing Newslo s (column 2), he numbe o Newslo s
pe enue (column 3) and he pe capi a ne expendi u e on slo machines (column 4). Elas ici y is
calcula ed and hen mul iplied by 100. The a iable T ea is de ined as in Sec ion 4;Pos ep esen s he
pe iod a e 2016. Popula ion-weigh ed eg essions a e es ima ed based on Equa ion 12, wi h s anda d
e o s clus e ed a he municipali y le el.
* p<.10 ** p<.05 *** p<.01.
o occupa ion: execu i e/manage , adminis a i e s a , manual wo ke s, business owne s/independen p o es-
sionals, o sel -employed; o he economic sec o : ag icul u e, manu ac u ing, o se ices; and o he ype o
con ac : ull- ime o pa - ime.
21
(a) Pe capi a numbe o Newslo de ices
-.002
-.0015
-.001
-.0005
0
.0005
2015 2016 2017 2018 2019
(b) Pe capi a numbe o Newslo enues
-.0004
-.0003
-.0002
-.0001
0
2015 2016 2017 2018 2019
(c) Numbe o Newslo s pe enue
-.8
-.6
-.4
-.2
0
.2
2015 2016 2017 2018 2019
(d) Pe capi a expendi u e on slo machines
-20
0
20
40
60
2015 2016 2017 2018 2019
Figu e 6: E en -s udy: E ec on Newslo s, expendi u e, and c owding. Coe icien s and
co esponding 90% and 95% con idence in e als. Panel (a) shows he pe capi a numbe o Newslo
de ices, (b) shows pe capi a expendi u e on slo machines, (c) shows he pe capi a numbe o enues
hos ing Newslo s, and (d) shows he numbe o Newslo s pe enue. Reg essions a e popula ion-weigh ed
and include municipali y and yea -p o ince ixed e ec s, wi h 2016 as he baseline yea . S anda d e o s
a e clus e ed a he municipali y le el.
Figu e 6, displays a ne expendi u e inc ease o 10% in he i s yea , o 28% in 2018, and o 32%
in 2019. These indings con adic he in ended objec i e o he bill bu hey can be explained
wi hin ou heo e ical model as a consequence o inc eased emp a ion. In he absence o
emp a ion cos s, educing he numbe o slo machines would be expec ed o dec ease bo h
he in ensi e and ex ensi e ma gins o slo machine play due o highe anspo a ion cos s
and educed enjoymen . Howe e , i he bill inc eases he cos o emp a ion, o example,
h ough longe wai ing imes, c owding, o social con agion e ec s in ea ed municipali ies,
he opposi e e ec may occu (see P oposi ion 1).
I should be no ed ha , since ADM did no elease sepa a e da a o 2016, he pe capi a
ne expendi u e on slo machines shown in column 4 o Table 2combines expendi u e on
bo h Newslo s and Videolo e ies. Howe e , o he o he yea s, sepa a e expendi u e da a a e
22
a ailable, enabling us o e- un he analysis while dis inguishing be ween Newslo s and VLTs,
excep o 2016. This allows us o assess he di ec impac o he policy on Newslo s, as well
as any spillo e e ec s on Videolo e ies, which we e no di ec ly a ec ed by he policy.
The esul s a e shown in Table B.1 in he Appendix. Consis en wi h he p e ious indings,
o al expendi u e on slo machines (Newslo s and VLTs combined) signi ican ly inc eases. As
shown in column 3, he inc ease is p ima ily d i en by inc eased spending on Newslo s (EUR
22.27, co esponding o 70% o he o e all inc ease). No ably, while he policy does no a ec
he numbe o Videolo e ies (column 4), i also leads o an inc ease in expendi u e on his ype
o de ice (EUR 9.72, column 5). This spillo e e ec is an addi ional unin ended consequence
o he bill.
In Appendix Cwe show ha he esul s p esen ed abo e a e obus o addi ional checks
using (i) di e en de ini ions o ea ed and con ol g oups, (ii) di e en sub-samples, and (iii)
alsi ica ion exe cises.
5.2 Close enues, mo e gambling
The p e ious Sec ion has shown ha , al hough he policy e ec i ely educed he numbe o
slo machines and enues, i also p oduced unin ended e ec s, wi h a 25% inc ease in slo
machine expendi u es. These esul s a e no consis en wi h highe anspo a ion cos s o a
dec ease in he enjoyabili y o slo machine play in he ea ed municipali ies. As s a ed in
P oposi ions 1and 2, hese d i e s should ins ead lead o a educ ion in he numbe o playe s
and in he amoun o play and, consequen ly, a educ ion in slo machine expendi u e.
Expendi u e in he ea ed municipali ies could inc ease i he educ ion in he numbe o
slo machines and enues a ec s emp a ion cos s. This may be he case i playe s concen a e
in he emaining enues, po en ially p oducing social con agion e ec s, pee p essu e, and
compe i ion among playe s. Mo e c owded enues may also educe he s igma associa ed
wi h gambling, making slo machine play mo e appealing.19 I hese ac o s inc ease he
emp a ion o play, and possibly he numbe o playe s, hen expendi u e in he ea ed uni s
can inc ease a e he policy. No ably, his e ec would be mo e p onounced in places whe e
he anspo a ion cos s a e lowe and among mo e addic ed playe s (P oposi ion 1).
To in es iga e he ole o anspo a ion cos s, we c ea e a measu e o p oximi y o obacco
shops a he municipal le el—since Newslo s a e o en loca ed in hese enues—using da a
om he 2011 Census ac s and he 2016 lis o obacco shop add esses published by ADM.
We compu e he minimum dis ance be ween he cen oid o each census ac and he nea es
19 The educ ion in he s igma associa ed wi h being a slo machine playe , due o an inc ease in he numbe
o people playing slo machines in he same enue, can be o mally desc ibed as a dec ease in τA. As shown in
P oposi ion 1, his would inc ease he ex ensi e ma gin o slo machine play in he ea ed uni s.
23
Hagge , M. S., C. Wood, C. S i , and N. L. Cha zisa an is (2010). Ego deple ion and he
s eng h model o sel -con ol: A me a-analysis. Psychological Bulle in 136(4), 495.
Hansen, B. (2015). Punishmen and de e ence: E idence om d unk d i ing. Ame ican
Economic Re iew 105(4), 1581–1617.
Ha igan, K. A., K. Collins, M. J. Dixon, and J. Fugelsang (2010). Addic i e gameplay:
Wha casual game designe s can lea n om slo machine esea ch. In P oceedings o he
In e na ional Academic Con e ence on he Fu u e o Game Design and Technology, pp.
127–133.
Ha ley, R. and L. Fa ell (2002). Can expec ed u ili y heo y explain gambling? Ame ican
Economic Re iew 92(3), 613–624.
He skowi z, S. (2021). Gambling, sa ing, and lumpy liquidi y needs. Ame ican Economic
Jou nal: Applied Economics 13(1), 72–104.
Hop ga ne , N., T. San os, M. Aue , M. G i i hs, and D. Helic (2021). Social acili a ion
among gamble s: A la ge-scale s udy using accoun -based da a. In P oceedings o he
In e na ional AAAI Con e ence on Web and Social Media, Volume 15, pp. 185–195.
Ho elling, H. (1929). S abili y in compe i ion. The Economic Jou nal 39(153), 41–57.
House , D., J. Liu, D. H. Reiley, and M. B. U bancic (2021). Checking ou emp a ion: A
na u al expe imen wi h pu chases a he g oce y egis e . Jou nal o Economic Beha io &
O ganiza ion 191, 39–50.
House , D., D. Schunk, J. Win e , and E. Xiao (2018). Temp a ion and commi men in he
labo a o y. Games and Economic Beha io 107, 329–344.
James, R. J., C. O’Malley, and R. J. Tunney (2016). Why a e some games mo e addic i e han
o he s: The e ec s o iming and payo on pe se e ance in a slo machine game. F on ie s
in Psychology 7, 46.
Kea ney, M. S. (2005). S a e lo e ies and consume beha io . Jou nal o Public
Economics 89(11-12), 2269–2299.
Laibson, D. (1997). Golden eggs and hype bolic discoun ing. The Qua e ly Jou nal o
Economics 112(2), 443–478.
Le i , S. D. (2004). Why a e gambling ma ke s o ganised so di e en ly om inancial ma ke s?
The Economic Jou nal 114(495), 223–246.
Lien, J. W. and J. Zheng (2015). Deciding when o qui : Re e ence-dependence o e slo
machine ou comes. Ame ican Economic Re iew 105(5), 366–370.
Loewens ein, G. (1987). An icipa ion and he alua ion o delayed consump ion. The Economic
Jou nal 97(387), 666–684.
Loewens ein, G. (1996). Ou o con ol: Visce al in luences on beha io . O ganiza ional
Beha io and Human Decision P ocesses 65(3), 272–292.
30

Loewens ein, G. (2000). Emo ions in economic heo y and economic beha io . Ame ican
Economic Re iew 90(2), 426–432.
Loewens ein, G. and T. O’Donoghue (2004). Animal spi i s: A ec i e and delibe a i e p o-
cesses in economic beha io . SSRN discussion pape 539843, Social Science Resea ch Ne -
wo k.
Loewens ein, G., T. O’Donoghue, and M. Rabin (2003). P ojec ion bias in p edic ing u u e
u ili y. The Qua e ly Jou nal o Economics, 1209–1248.
Lundbo g, P. (2006). Ha ing he w ong iends? Pee e ec s in adolescen subs ance use.
Jou nal o Heal h Economics 25(2), 214–233.
Mas oba is a, L., A. Minu illo, B. Gene i, A. And eo i, R. Paci ici, and C. Mo ali (2019).
Gioco d’azza do nella popolazione adul a: S udio epidemiologico as e sale di ipo osse -
azionale. In R. Paci ici, L. Mas oba is a, A. Minu illo, and C. Mo ali (Eds.), Gioco
d’azza do in I alia: ice ca, o mazione e in o mazione: Risul a i di un p oge o in eg a o,
Rappo i ISTISAN 19/28, pp. 7–50. Roma: Is i u o Supe io e di Sani `a.
Moo e, T. J. and T. Mo is (2024). Shaping he habi s o een d i e s. Ame ican Economic
Jou nal: Economic Policy 16(3), 367–393.
Muggle on, N., P. Pa pa , P. Newall, D. Leake, J. Ga he good, and N. S ewa (2021). The
associa ion be ween gambling and inancial, social and heal h ou comes in big inancial da a.
Na u e Human Beha iou 5(3), 319–326.
Mu a en, M. and R. F. Baumeis e (2000). Sel - egula ion and deple ion o limi ed esou ces:
Does sel -con ol esemble a muscle? Psychological Bulle in 126(2), 247.
OECD (2018). Heal h a a glance: Eu ope 2018. pp. 26–32.
O phanides, A. and D. Ze os (1995). Ra ional addic ion wi h lea ning and eg e . Jou nal o
Poli ical Economy 103(4), 739–758.
Ozdeno en, E., S. W. Salan , and D. Sil e man (2012). Willpowe and he op imal con ol o
isce al u ges. Jou nal o he Eu opean Economic Associa ion 10(2), 342–368.
Pesko, M. F. and C. Wa man (2022). Re-explo ing he ea ly ela ionship be ween eenage
ciga e e and e-ciga e e use using p ice and ax changes. Heal h Economics 31(1), 137–153.
Pe y, N. (2015). Beha io al addic ions: DSM-5
®
and beyond. Ox o d Uni e si y P ess.
Piccoli, L. and S. Tiezzi (2021). Ra ional addic ion and ime-consis ency: An empi ical es .
Jou nal o Heal h Economics 80, 102546.
Pollak, R. A. (1970). Habi o ma ion and dynamic demand unc ions. Jou nal o Poli ical
Economy 78(4, Pa 1), 745–763.
Resce, G., R. Lag a inese, E. Benede i, and S. Molina o (2019). Income- ela ed inequali y in
gambling: E idence om I aly. Re iew o Economics o he Household 17(4), 1107–1131.
31
Rocklo , M. J. and V. Dye (2007). An expe imen on he social acili a ion o gambling
beha io . Jou nal o Gambling S udies 23, 1–12.
Rocklo , M. J., N. G ee , and L. G. E ans (2012). The e ec o me e p esence on elec onic
gaming machine gambling. Jou nal o Gambling Issues (27).
Rocklo , M. J., N. G ee , and C. Fay (2011). The social con agion o gambling: How enue
size con ibu es o playe losses. Jou nal o Gambling S udies 27, 487–497.
Rocklo , M. J., N. Mosko sky, H. Tho ne, M. B owne, and G. M. B yden (2017). En i on-
men al ac o s in he choice o EGMs: A disc e e choice expe imen . Jou nal o Gambling
S udies 33, 719–734.
San os Sil a, J. and S. Ten ey o (2006). The log o g a i y. The Re iew o Economics and
S a is ics 88(4), 641–658.
Sch¨ull, N. D. (2012). Addic ion by Design: Machine Gambling in Las Vegas. P ince on
Uni e si y P ess.
Shi , B. and A. Fedo ikhin (1999). Hea and mind in con lic : The in e play o a ec and
cogni ion in consume decision making. Jou nal o Consume Resea ch 26(3), 278–292.
S e zka, R. M. and S. Win e (2023). How a ional is gambling? Jou nal o Economic
Su eys 37(4), 1432–1488.
S o z, R. H. (1955). Myopia and inconsis ency in dynamic u ili y maximiza ion. The Re iew
o Economic S udies 23(3), 165–180.
Thale , R. H. and H. M. She in (1981). An economic heo y o sel -con ol. Jou nal o Poli ical
Economy 89(2), 392–406.
Vohs, K. D. and R. J. Fabe (2007). Spen esou ces: Sel - egula o y esou ce a ailabili y
a ec s impulse buying. Jou nal o Consume Resea ch 33(4), 537–547.
Wa dle, H. and S. McManus (2021). Suicidali y and gambling among young adul s in G ea
B i ain: Resul s om a c oss-sec ional online su ey. The Lance Public Heal h 6(1), e39–
e49.
WHO (2018). Inclusion o “gaming diso de ” in ICD-11. h ps://www.who.in /news/i em/
14-09-2018-inclusion-o -gaming-diso de -in-icd-11, las accessed on 2024-09-20.
WHO (2022). In e na ional Classi ica ion o Diseases o Mo ali y and Mo bidi y S a is ics,
Ele en h Re ision (ICD-11). Gene a: Wo ld Heal h O ganiza ion, Gene a.
WHO (2024). Addic i e beha io . h ps://www.who.in /heal h- opics/
addic i e-beha iou # ab= ab_1, las accessed on 2024-09-20.
32
A Appendix: P oo s o he heo e ical model
A.1 Sol ing he model
The model is sol ed by backwa d induc ion. Once a a enue ∈ {A, B}, he indi idual
p oblem o sol e is
max
s ,z U(s , z;a, σ ) (15)
s. . m=p s +z(16)
This is a s anda d consume p oblem. The solu ion (s∗
, z∗
) a ei he enue sa is ies:
Us(s∗
, z∗
)
Uz(s∗
, z∗
)=p (17)
and he budge cons ain . In he ollowing, assume ha 0 < s∗
A< s∗
B o any gi en le el o
addic ion. To simpli y he no a ion, we use U(s∗
) as a sho hand o U(s∗
, z∗
;a, σ ), whene e
i does no c ea e con usion.
A he pa icipa ion s age, he choice depends on he compa ison be ween he indi ec
u ili y le els ha an indi idual a loca ion iwi h addi ion s ock aob ains when consuming a
ei he enue, o abs aining al oghe e . Depending on he indi idual le el o addic ion a, he
h eshold loca ions ha deno e he indi e en playe be ween Aand Ba e
iAB (a) = τB+U(s∗
A)−UB(s∗
B)
τA+τB
(18)
The indi iduals indi e en be ween abs aining and consuming a enue a e
iOA (a) = U(s∗
A)−U(0)
τA
;iOB (a) = 1 −U(s∗
B)−U(0)
τB
(19)
To jus i y he slope o he h ee loci o he indi e en consume s d awn in Figu e 1, no e ha
he h eshold posi ions depend on he addic ion le el as ollows:
∂iAB
∂a =1
τA+τB
(Ua(s∗
A)−Ua(s∗
B)) <0,(20)
∂iOA
∂a =1
τA
(Ua(s∗
A)−Ua(0)) >0,(21)
∂iOB
∂a =1
τB
(Ua(0) −Ua(s∗
B)) <0.(22)
Recall we assumed s∗
A< s∗
Band Usa >0 o all a, hence 0 > Ua(s∗
B)> Ua(s∗
A)> Ua(0),
33
Le aAdeno e he in e sec ion poin o iAB wi h he e ical axis (i can be a ini e alue
o in ini y), and ass
=sss
·γ/(1 −γ) he s eady s a e alue o addic ion a a enue . Gi en
he budge cons ain , he maximum easible consump ion is sm=m/p o ei he , which is
associa ed o he addic ion le el am=sm·γ/(1 −γ).
In equa ions 23 and 24, we conside he case in which he alue o he addic ion s ock is
bounded be ween 0 and am. This occu s when he budge cons ain binds be o e eaching he
(uncons ained) s eady s a e, i.e. sss
> sm o ei he . Fo mally, we assume am> aA>ˆa > 0.
I he in e sec ion poin (ˆı, ˆa) is in he in e io o he i s quad an , hen he popula ion is
di ided in o consume s o sA,o sB,o abs aine s, as shown in Figu e 1. Fo gi en dis ibu ion
unc ion ϕ(i, a) o indi idual posi ions and le els o addic ion in he popula ion a ime , he
sha es o addic i e consume s a ei he enue a e:
NA=ZZA
ϕ(i, a) dadi=ZiOA
0Zˆa
0
ϕ(i, a) dadi+ZiAB
0ZaA
ˆa
ϕ(i, a) dadi(23)
NB=ZZB
ϕ(i, a) dadi=Z1
iOB Zˆa
0
ϕ(i, a) dadi+Z1
iAB ZaA
ˆa
ϕ(i, a) dadi+Z1
0Zam
aA
ϕ(i, a) dadi
(24)
A.2 Compa a i e s a ics
In ensi e ma gin o consump ion. We now assess he e ec o changes in he indi idual
le el o addic ion, in ma ginal anspo a ion cos s, and in he ma ginal emp a ion cos s
epo ed in P oposi ion 1. By applying he implici unc ion heo em, he ollowing holds:
∂s∗
∂a =−Usa (s∗
)
Uss (s∗
)>0 (25)
∂s∗
A
∂τA
=∂s∗
B
∂τA
= 0 (26)
∂s∗
A
∂σA
=−UsσA(s∗
A)
Uss s∗
A>0; ∂s∗
B
∂σA
= 0 (27)
No e ha he e ec o a change in he ma ginal emp a ion cos a enue Ais highe o mo e
addic ed indi iduals i (omi ing he a gumen s)
∂
∂a ∂s∗
A
∂σA=1
U2
ss
(UsσUssa −UsaσUss) = 1
U2
ss
(aUssa − Uss)>0 (28)
whe e he las equali y ollows om (4). In he linea -quad a ic speci ica ion ypically used in
he a ional addic ion li e a u e (see, e.g. Becke and Mu phy,1988), exp ession (28) always
holds because Ussa = 0.
34
To s udy he e ec s o a policy x ha educes he enjoyabili y o addic i e consump ion a
A, we assume ha x educes he u ili y and he ma ginal u ili y o addic i e consump ion a
A, bu no a B, i.e. Ux(s∗
A), Usx (s∗
A)<0, Ux(s∗
B) = 0. Then he in oduc ion o he policy
educes consump ion a A, while consump ion a Bis una ec ed:
∂s∗
A
∂x =−Usx (s∗
A)
Uss s∗
A<0; ∂s∗
B
∂x = 0 (29)
This explains he claim on he change a he in ensi e ma gin o P oposi ion 2.
Suppose σA=σB=σ0=σ. A change in all ma ginal cos s o emp a ion σimplies mo e
play a ei he enue:
∂s∗
A
∂σ =−Usσ (s∗
A)
Uss s∗
A>0; ∂s∗
B
∂σ =−Usσ (s∗
B)
Uss s∗
A>0 (30)
Ex ensi e ma gin o consump ion. Based on he exp essions (23) and (24), we ocus on
changes a he ex ensi e ma gins. Conside ing he sha e o consume s o sA, we a e in e es ed
in he ollowing exp ession:
∂NA
∂y =Zˆa
0ϕ(iOA (a), a)∂iOA
∂y dadi
| {z }
ny
OA
+Zm
ˆaϕ(iAB (a), a)∂iAB
∂y dadi
| {z }
ny
AB
(31)
Te ms ny
OA and ny
AB desc ibe he change in he mass o consume s along he OA and he AB
ma gins, espec i ely, as a consequence o a change in y. Thei sign depends on he sign o
∂iOA
∂y and o ∂iAB
∂y ,hence a mo emen o he le o ei he ex ensi e ma gin o co esponds o a
dec ease in he sha e o consume s a A(see Figu e 2 o an example).
Analogously, o s udy changes in he sha e o consume s o sB, we conside
∂NB
∂y =−Zˆa
ˆaϕ(iAB (a), a)∂iAB
∂y dadi
| {z }
ny
AB
−Zˆa
0ϕ(iOB (a), a)∂iOB
∂y dadi
| {z }
ny
OB
(32)
Conside an inc ease in he ma ginal anspo a ion cos o each A. Since
∂i0A
∂τA
=−iOA
τA
<0; ∂iAB
∂τA
=−iAB
τA+τB
<0; ∂iOB
∂τA
= 0 (33)
we conclude ha NAdec eases, while NBand NOinc ease. Conside ing ha no e ec is
p oduced a he in ensi e ma gin o consump ion, we conclude ha an inc ease in he ma ginal
anspo a ion cos o each A educes expendi u e a Aand inc eases expendi u e a B.
35

An inc ease in he ma ginal emp a ion cos σAp oduces analog esul s on he sha es o
consume s and abs aine s, as
∂i0A
∂σA
=UσA(s∗
A)
τA
<0; ∂iAB
∂σA
=UσA(s∗
A)
τA+τB
<0; ∂iOB
∂σA
= 0 (34)
Howe e , as shown be o e, consump ion a Ainc eases, while i emains una ec ed a B.
Hence, expendi u e a Bunambiguously inc eases. The change in expendi u e a A, ins ead,
inc eases only i he inc ease a he in ensi e ma gin (30) mo e han o se s he dec ease a he
ex ensi e ma gin (34).
Suppose σA=σB=σ0=σ. A change in all ma ginal cos s o emp a ion σ, has a di e en
e ec on he ex ensi e ma gin, as
∂i0A
∂σ =Uσ(s∗
A)−Uσ(0)
τA
>0; ∂iAB
∂σ =Uσ(s∗
A)−Uσ(s∗
B)
τA+τB
<0; ∂iOB
∂σ =Uσ(0) −Uσ(s∗
B)
τB
<0
(35)
Hence NBand EBinc ease, while he sha e o abs aine s NOdec eases.
Changing he enjoyabili y o he good implies
∂i0A
∂x =Ux(s∗
A)
τA
<0; ∂iAB
∂x =Ux(s∗
A)
τA+τB
<0; ∂iOB
∂x = 0 (36)
Hence NAdec eases, while NBand NOinc ease when x educes he enjoyabili y o he good.
This implies ha EAdec eases, while EBinc eases.
36
B Appendix: Addi ional ables and igu es
(1) (2) (3) (4) (5)
To al Newslo s VLT
expendi u e Numbe Expendi u e Numbe Expendi u e
T ea ×Pos 31.8748*** -0.0011*** 22.2686*** 0.0000 9.7216**
(8.8048) (0.0003) (6.5536) (0.0001) (3.8582)
Obse a ions 31,508 31,508 31,508 31,508 31,508
Municipali ies 7,877 7,877 7,877 7,877 7,877
Mean ou come in 2015 121.3451 0.0054 101.8331 0.0004 19.5118
Elas ici y 26.27 -20.45 21.87 3.69 49.82
Municipali y FE ✓✓✓✓✓
P o ince ×Yea FE ✓✓✓✓✓
Table B.1: E ec on expendi u e by ype o slo machine. The ou comes a e o al pe capi a
expendi u e on slo machines (column 1), he numbe o Newslo s pe capi a (column 2), o al pe
capi a expendi u e on Newslo s (column 3), he numbe o Videolo e ies pe capi a (column 4), and
o al pe capi a expendi u e on Videolo e ies (column 5). Elas ici y is calcula ed and hen mul iplied
by 100. The a iable T ea is de ined as in Sec ion 4, and Pos ep esen s he pe iod a e 2016. The
esul s a e om popula ion-weigh ed eg essions based on Equa ion 12, wi h s anda d e o s clus e ed
a he municipali y le el. Due o da a a ailabili y, yea 2016 is excluded.
* p<.10 ** p<.05 *** p<.01.
Exposu e Pop Densi y Old-age
index
Young wi h
deg ee Unemployed NEET House P ices Households
in Ha dship TV licence
(1) (2) (3) (4) (5) (6) (7) (8)
Pos ×T ea -0.9108 -2.9035 1.4838 4.3630 4.9784 -0.3034 7.9286 -2.9296
(7.0163) (9.2809) (7.2502) (11.7332) (11.4910) (6.9603) (10.6709) (7.1924)
Pos ×T ea ×Closes o obacco shops 40.9709** 35.1430*** 35.3691*** 35.4921*** 35.6363*** 36.8134*** 37.4609*** 35.5365***
(17.2764) (11.9893) (12.4428) (13.0867) (12.9459) (13.2401) (12.6242) (13.5418)
Pos ×T ea ×High Exposu e -12.6905 5.8661 -18.3797 -11.4608 -12.7358 -12.7674 -19.5852 1.2742
(18.8739) (14.3348) (14.6918) (17.1686) (16.8888) (15.9090) (16.5958) (16.3616)
Obse a ions 39,205 39,205 39,205 39,205 39,205 39,205 39,205 39,205
Municipali ies 7,841 7,841 7,841 7,841 7,841 7,841 7,841 7,841
Municipali y FE ✓✓✓✓✓✓✓✓
P o ince X Yea FE ✓✓✓✓✓✓✓✓
Table B.2: E ec by dis ance om obacco shops, obus ness checks. The a iable Closes o
obacco shops is a dummy equal o one when he median minimum dis ance be ween each census ac and he
closes obacco shop belong o he bo om e ciles o he dis ance dis ibu ion. The a iable High Exposu e is a
dummy equal o one when he municipali ies belong o he op e cile o he dis ibu ion o : popula ion densi y
(column 1), o e -65 o unde -14 a io (column 2), he sha e o 30-34yo wi h a deg ee (column 3), unemploymen
a e (column 4), NEET a e (column 5), housing p ices (column 6), sha e o households in economic ha dship
(column 7), and he sha e o people paying public TV se ices (column 8). 36 municipali ies a e unma ched
hus d opped. Coe icien s associa ed o Pos ×Closes o obacco shops and Pos ×High Exposu e a e no
shown. Popula ion-weigh ed eg essions. S anda d e o s a e clus e ed a he municipali y le el.
* p<.10 ** p<.05 *** p<.01.
37
(a) Newslo s
-1 -.75 -.5 -.25 0 .25 .5
2016-2018 g ow h
Vene o and T en ino Al o Adige
Umb ia
Toscana
Sicilia
Sa degna
Puglia, Basilica a and Molise
Piemon e and Valle d'Aos a
Ma che
Lomba dia
Ligu ia
Lazio
F iuli Venezia Giulia
Emilia Romagna
Campania
Calab ia
Ab uzzo
(b) Venues hos ing Newslo s
-1 -.75 -.5 -.25 0 .25 .5
2016-2018 g ow h
Vene o and T en ino Al o Adige
Umb ia
Toscana
Sicilia
Sa degna
Puglia, Basilica a and Molise
Piemon e and Valle d'Aos a
Ma che
Lomba dia
Ligu ia
Lazio
F iuli Venezia Giulia
Emilia Romagna
Campania
Calab ia
Ab uzzo
(c) VLT de ices
-1 -.75 -.5 -.25 0 .25 .5
2016-2018 g ow h
Vene o and T en ino Al o Adige
Umb ia
Toscana
Sicilia
Sa degna
Puglia, Basilica a and Molise
Piemon e and Valle d'Aos a
Ma che
Lomba dia
Ligu ia
Lazio
F iuli Venezia Giulia
Emilia Romagna
Campania
Calab ia
Ab uzzo
(d) Venues hos ing VLT de ices
-1 -.75 -.5 -.25 0 .25 .5
2016-2018 g ow h
Vene o and T en ino Al o Adige
Umb ia
Toscana
Sicilia
Sa degna
Puglia, Basilica a and Molise
Piemon e and Valle d'Aos a
Ma che
Lomba dia
Ligu ia
Lazio
F iuli Venezia Giulia
Emilia Romagna
Campania
Calab ia
Ab uzzo
Figu e B.1: Pe cen age change in he numbe o Newslo s, VLTs, and enues om 2016
o 2018. The 20 I alian egions a e agg ega ed in o 15 egional a eas, as epo ed in he ADM annual epo s.
Sou ce: ADM annual epo s.
0 .01 .02 .03 .04
Densi y
0 20 40 60 80
Household expendi u e on gambling
Figu e B.2: Household expendi u e on gambling Sample o households wi h posi i e expendi u e
on gambling (13% o he o al sample) du ing he pe iod 2014–2019. Expendi u e is in EUR/mon h.
38
(a) Pe capi a numbe o Newslo s
0
.0025
.005
.0075
.01
2015 2016 2017 2018 2019
T ea ed Con ols
(b) Pe capi a ne expendi u e on slo s
0
50
100
150
200
2015 2016 2017 2018 2019
T ea ed Con ols
Figu e B.3: T ends, aw a e ages. Municipali y-le el a e ages o he pe capi a coun o Newslo s (panel
a) and pe capi a ne expendi u e on slo machines (panel b), by g oup. Red ci cles e e o ea ed uni s, pink
diamonds o con ols uni s.
39