Fang, F ancis Haoyu; Vlaicu, Raz an
Wo king Pape
Local banking supply and p i a e i m ac i i y: E idence
om b anch closu es
IDB Wo king Pape Se ies, No. IDB-WP-1605
P o ided in Coope a ion wi h:
In e -Ame ican De elopmen Bank (IDB), Washing on, DC
Sugges ed Ci a ion: Fang, F ancis Haoyu; Vlaicu, Raz an (2024) : Local banking supply and p i a e
i m ac i i y: E idence om b anch closu es, IDB Wo king Pape Se ies, No. IDB-WP-1605, In e -
Ame ican De elopmen Bank (IDB), Washing on, DC,
h ps://doi.o g/10.18235/0013076
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/302207
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/3.0/igo/
Local Banking Supply and P i a e Fi m
Ac i i y:
E idence om B anch Closu es
F ancis Haoyu Fang
Raz an Vlaicu
WORKING PAPER No IDB-WP-1605
In e -
A
me ican De elopmen Bank
Depa men o Resea ch and Chie Economis
July 2024
* Uni e si y o Minneso a
** In e -Ame ican De elopmen Bank
Local Banking Supply and P i a e Fi m
Ac i i y:
E idence om B anch Closu es
F ancis Haoyu Fang*
Raz an Vlaicu**
In e -
A
me ican De elopmen Bank
Depa men o Resea ch and Chie Economis
July 2024
Ca aloging-in-Publica ion da a p o ided by he
In e -Ame ican De elopmen Bank
Felipe He e a Lib a y
Fang, F ancis Haoyu.
Local banking supply and p i a e i m ac i i y: e idence om b anch closu es
/ F ancis Haoyu Fang, Raz an Vlaicu.
p. cm. — (IDB Wo king Pape Se ies ; 1605)
Includes bibliog aphical e e ences.
1. B anch banks-B azil. 2. Banks and banking-B azil. 3. Communi y banks-
B azil. 4. Business en e p ises-B azil. 5. Unemploymen -B azil. I. Vlaicu,
Raz an. II. In e -Ame ican De elopmen Bank. Depa men o Resea ch and
Chie Economis . III. Ti le. IV. Se ies.
IDB-WP-1605
h p://www.iadb.o g
Copy igh © 2024 In e -Ame ican De elopmen Bank ("IDB"). This wo k is subjec o a C ea i e
Commons license CC BY 3.0 IGO (h ps://c ea i ecommons.o g/licenses/by/3.0/igo/legalcode). The
e ms and condi ions indica ed in he URL link mus be me and he espec i e ecogni ion mus be
g an ed o he IDB.
Fu he o sec ion 8 o he abo e license, any media ion ela ing o dispu es a ising unde such license
shall be conduc ed in acco dance wi h he WIPO Media ion Rules. Any dispu e ela ed o he use o
he wo ks o he IDB ha canno be se led amicably shall be submi ed o a bi a ion pu suan o he
Uni ed Na ions Commission on In e na ional T ade Law (UNCITRAL) ules. The use o he IDB's name
o any pu pose o he han o a ibu ion, and he use o IDB's logo shall be subjec o a sepa a e
w i en license ag eemen be ween he IDB and he use and is no au ho ized as pa o his license.
No e ha he URL link includes e ms and condi ions ha a e an in eg al pa o his license.
The opinions exp essed in his wo k a e hose o he au ho s and do no necessa ily e lec he iews o
he In e -Ame ican De elopmen Bank, i s Boa d o Di ec o s, o he coun ies hey ep esen .
Abs ac
P i a e i ms es ablish ela ionships wi h banks in local ma ke s o ob ain adequa e
inancing o hei ope a ions h ough c edi and loans. As majo banks educed hei
b anch ne wo ks in ecen yea s, many i ms ha e los access o hei local bank. We
e alua e he impac o a la ge numbe o b anch closu es on i m ope a ions, wages and
employmen , and economic ou pu in B azil om 2011 o 2021. We adop a di e ence-
in-di e ences s a egy wi h s agge ed ea men iming, employing bo h wo-way ixed
e ec s and Callaway-San ’Anna es ima o s. Ou s udy inds ha bank b anch closu es
esul in a educ ion in es ablishmen s wi h ac i e ope a ions om 1.2% ini ially o
8.1% wi hin 4-7 yea s, a 0.5 decline in weekly hou s o o mal employmen , and a
comp ession in he eal wage dis ibu ion. Mic o i ms, ade and se ice i ms, and
ag icul u al i ms a e ound o be he mos ulne able. Ou s udy highligh s he im-
po ance o physical bank b anches ha p o ide inancial access and mee he localized
inancial demand o se e al ypes o i ms.
JEL classi ica ions: G21, R11, J21
Keywo ds: Bank b anch closu es, Employmen , Fi m ac i i y, Economic impac ,
Financial access, B azil
F.H. Fang: Depa men o Applied Economics, Uni e si y o Minneso a, 1994 Bu o d A e, Sain Paul,
MN 55108. Email: ang0[email p o ec ed]. R. Vlaicu: Resea ch Depa men , In e -Ame ican De elopmen
Bank, 1300 New Yo k A e NW, Washing on, DC 20577. Email: [email protected] g. We hank Jason Ke win,
Paul Glewwe, Elizabe h Da is, S ephen Polasky, Joe Ri e , Janna Johnson, And ew Goodman-Bacon,
Richa d Thako , Cyn hia Kinnan, P iya Mukhe jee, Tama a McGa ock, Vinicius Vidigl, Jianxuan Lei,
Rachel (Chundan) Lan, Sa ah Wahby, Giang Thai, Leo Zucke o esea ch assis ance, and pa icipan s a
he PhD Lab, Glewwe Lab and 2023 second yea pape semina in he Depa men o Applied Economics.
1 In oduc ion
The impo ance o local bank access o p i a e i ms canno be o e s a ed, as local banks
p o ide essen ial inancial se ices, including c edi , loans, and paymen s. Local access is
pa icula ly i al o small and medium-sized en e p ises (SMEs) ha may no ha e he
same esou ces as la ge co po a ions o secu e inancing h ough al e na i e channels and
ace highe cos s o swi ching lende s (Beck e al., 2008; Nguyen, 2019). Th ough pe sonal
in e ac ions, local banks ob ain so in o ma ion abou ma ke condi ions and business own-
e s ha allows hem o be e ailo inancial p oduc s o he speci ic needs o local i ms
(Be ge e al., 2008). Mo eo e , he p oximi y o banks o i ms can lead o imp o ed c edi
a ailabili y, as banks a e mo e likely o lend o businesses wi h which hey ha e es ablished
ela ionships (Cole and Gun he , 1998).
The p esence o banks can signi ican ly in luence a i m’s labo demand and p oduc i i y,
he eby a ec ing he labo ma ke equilib ium and local economic ou pu . Banks’ p esence
can ha e a mixed impac on a i m’s inancing cos depending on di e en ac o s like bank
owne ship o ma ke powe (Ryan e al., 2014; Ba h e al., 2013). Fu he mo e, s udies
o en ind c edi a ailabili y has a posi i e impac on labo demand and wages by lowe ing
he i m’s inancial ic ions and e en ually acili a ing job c ea ion and p oduc i i y (Popo
and Rocholl, 2018; Fonseca and Van Doo nik, 2022).
Howe e , he landscape o bank access is changing. Recen Wo ld Bank da a2highligh s
a dec ease in comme cial bank b anches pe capi a om i s peak in 2016 and iden i ies a
global end owa ds digi al banking and away om adi ional b ick-and-mo a . This shi
should ha e implica ions o i ms ha ely hea ily on local banking ela ionships o hei
inancial needs. The decline in physical bank b anches can lead o educed access o inancial
se ices o i ms loca ed in u al o unde se ed u ban a eas, po en ially hampe ing hei
su i al, ope a ions, and g ow h. This leads us o he co e ques ions o ou s udy: Wha
is he impac o closing bank b anches on local i m ope a ions, employmen , and economic
ou pu ? And wha kinds o i ms a e mos ulne able o expe iencing such shi s?
To s udy his ques ion, B azil p o ides a sui able con ex . The numbe o bank b anches
expe ienced an inc easing end a e he B azilian go e nmen implemen ed i s “Banks o
All” p og am (“Banco pa a Todos”) in 2004. Howe e , his upwa d end was dis up ed in
2014, and a downwa d end has pe sis ed since hen. As o he end o 2022, B azil had los
mo e han 5,000 bank b anches, which is one- ou h o i s peak numbe . Many municipali ies
los all o hei bank b anches in his p ocess. This c ea es quasi-expe imen al a ia ion in
2See: Wo ld Bank, Numbe o Comme cial Bank B anches pe 100,000 Adul s, a h ps://da a.
wo ldbank.o g/indica o /FB.CBK.BRCH.P5
1
local bank access ha can be used o s udy he e ec s o banking supply on i m ac i i y.
Figu e 1 p esen s he e olu ion o he bank b anch landscape in B azil spanning he las wo
decades om 2002 o 2022.
Ou empi ical analysis elies on a p ima y sample o p i a e sec o i ms and employees in
B azil spanning he yea s om 2011 o 2021. We exploi es ablishmen -le el c oss-sec ional
da a o e ime in addi ion o municipali y-le el panel da a on economic and employmen
condi ions ac oss B azil. We link bank b anch-le el da a wi h i m-le el labo census da a
and municipali y-le el da a on popula ion and local economic accoun s. Conside ing he
geog aphic a ia ion in bank b anch concen a ion, as well as he he e ogenei y a ising om
di e en closu e pa e ns, we chose o concen a e on municipali ies wi h only one b anch.
This subse cons i u es he la ges g oup among banked municipali ies and exhibi s b oad
geog aphic a ia ion in B azil. To iden i y he e ec s o bank b anch closu es on local
o mal employmen , i is impe a i e o es ablish bo h a ea men g oup and a sui able
compa ison g oup. We adop a s agge ed ea men iming design by e aining in he sample
municipali ies whe e he local b anch closu e was no e e sed du ing he s udy pe iod.
We hen u ilize a di e ence-in-di e ences (DiD) amewo k wi h a ia ion in ea men
iming o iden i y he e ec s o bank b anch closu es on local employmen , exploi ing he
closu es o bank b anches as a sou ce o plausible exogenous a ia ion. We discuss po en ial
endogenei y challenges in his con ex , including issues such as e e se causali y, omi ed
a iables, an icipa ion, and spillo e e ec s. To add ess hese issues, se e al iden i ica ion
assump ions mus be made, including he pa allel ends assump ion and limi ed an icipa ion
assump ion. Ou subsequen es ima es will be alid p o ided hese assump ions hold, pa -
icula ly ha ea men - ela ed omi ed a iables do no p oduce sys ema ic pos - ea men
e ec s in one g oup.
To p o ide an o e iew o he impac o bank b anch closu es and cons uc obus
es ima es, we p o ide h ee es ima ion app oaches in ou analysis: he wo-way ixed e ec
(TWFE) model, he Callaway-San ’Anna (CS) es ima o wi h ne e - ea ed obse a ions as
a con ol g oup, and he CS es ima o wi h no -ye - ea ed obse a ions as a con ol g oup.
Conside ing ha issues o ea men e ec he e ogenei y and s agge ed iming may a ec he
eliabili y o he he wo-way ixed e ec es ima o , we compa ed se e al newly de eloped
s agge ed DiD es ima o s and selec ed CS es ima o s; hose es ima o s a e discussed in ecen
li e a u e and can o e come he iden i ica ion issues associa ed wi h TWFE. Mo eo e , he
CS es ima o has o he bene i s ha apply in ou con ex : I can accommoda e less s able
pa allel ends (o en a conce n in long s udy pe iods) and i p o ides anspa en con ol
g oup op ions (ne e - ea ed e sus no -ye - ea ed.) We p esen CS es ima es using bo h
ypes o con ol g oups o add ess po en ial selec ion conce ns.
2
Figu e 1: Bank B anch T ends in B azil, 2002-2022
No es: This g aph p esen s he numbe o bank b anches and municipali ies wi h bank b anches
ha appea ed in B azil’s ESTBAN da ase s each Decembe om 2002 o 2022. The ed line
indica es he end in he numbe o bank b anches, co esponding o he y-axis on he le side.
The blue line indica es he numbe o municipali ies wi h a leas one bank b anch, co esponding
o he y-axis on he igh side. Sou ce: ESTBAN da ase , Banco Cen al do B asil.
We conduc se e al analyses o in es iga e he impac o local bank b anch closu es.
Fi s , we conduc an es ablishmen -le el analysis o see i bank b anch closu es impac i m
ope a ions, including ma ke en y and exi . We ind ha oughly 1 pe cen o es ablishmen s
become inac i e be ween 0 and 3 yea s a e a bank b anch closu e, and he sha e o inac i e
es ablishmen s inc eases om 1.2 pe cen o 8.4 pe cen be ween 4 and 7 yea s a e he
closu e. Howe e , we ind no e ec on he o e all o speci ic ype o i m en e ing o exi ing.
We hen e alua e he impac on local o mal employmen , using a municipali y-le el panel
cons uc ed om wo ke census da a. Ou es ima es sugges ha bank b anch closu es
could ha e a la ge and quicke impac on wages han on employmen . While s a is ically
imp ecise, ou poin es ima es show bank b anch closu es dec ease he a e age wage by 1.5
o 1.7 pe cen in he i s h ee yea s. Fo employmen , ou es ima es sugges bank b anch
closu es do no ha e a signi ican sho - e m e ec bu po en ially a long- e m nega i e e ec .
S udying wage dispa i ies, we ind sho - e m dec eases in he s anda d de ia ion o
wages, bu long- e m dec eases in he di e ence be ween he 90 h and 10 h pe cen ile wage
b acke s. This could sugges ha b anch closu es disp opo iona ely a ec di e en wage
b acke s—in he sho un, po en ially he middle-income b acke , and in he long un, he
3
lowe wage b acke . We conduc a he e ogenei y analysis o cha ac e ize which ypes o i ms
a e mo e ulne able; we ind ha mic o i ms, ade and se ice i ms, and ag icul u e i ms
a e disp opo iona ely impac ed by local bank b anch closu es. Finally, we analyze economic
ou pu accoun s and ind a dec easing end in se ice ou pu due o bank b anch closu es
bu no in indus ial and ag icul u al ou pu .
This pape con ibu es o se e al s ands o li e a u e. Fi s , i adds o he exis ing
li e a u e on he economic e ec s o banking supply. The e ec s o bank de egula ion ha e
been ex ensi ely s udied ac oss a ious pe iods in di e en coun ies. Among he ou comes
conside ed a e income inequali y, new inco po a ion g ow h, eal income, ou pu g ow h, i-
nancial inclusion, and household weal h accumula ion in he Uni ed S a es (Beck e al., 2010;
Black and S ahan, 2002; Ji h Jaya a ne e al., 1996; Cele ie and Ma ay, 2019); po e y
in India (Bu gess and Pande, 2005); income le els in Mexico (B uhn and Lo e, 2014); and
b anch occupa ion, income and employmen in Thailand (Ji e al., 2021). In B azil, s udies
discuss shocks ha could impac bank b anch supply and hei e ec s, including bank me g-
e s and acquisi ions a ec ing inancial a iables, loans, and employmen (Joaquim and an
Doo nik, 2019); na ional policies o bank b anch expansion o u al a eas (Fonseca and Ma-
ay, 2022); and bank obbe ies a ec ing adop ion o digi al inance ools (A gen ie i Ma iani
e al., 2023). 3
Ou s udy ela es o exis ing wo k in es iga ing he e ec s o c edi supply on employ-
men , consump ion, and p oduc i i y. S igli z and Weiss (1981) and Kocha (1997) discuss
he equilib ium in he loan ma ke , speci ically c edi a ioning, household demand o c edi ,
and i s ole in ag icul u al de elopmen . C edi ma ke shocks ha e been ound o nega i ely
impac employmen g ow h in small businesses in he Uni ed S a es (G eens one e al., 2020)
and esul in dec eased consump ion, ea nings, and employmen in India (B eza and Kin-
nan, 2021). Field expe imen s ha e yielded simila esul s, wi h Ka lan and Zinman (2010)
demons a ing he posi i e ou comes o expanding c edi supply on employmen , income,
ood consump ion, and o e all well-being. Bane jee and Du lo (2014) u ilize a a ge ed
di ec lending p og am o es c edi cons ain s among Indian i ms and sugges se e e
cons ain s and high ma ginal a es o e u n on capi al.
Finally, he inance li e a u e has demons a ed he signi icance o bank b anches in local
c edi supply. Gilje e al. (2013) demons a e how bank b anch ne wo ks help in eg a e
lending ma ke s in he Uni ed S a es, e en in he p esence o he secu i iza ion ma ke .
Hasan e al. (2020) s udy he ole o local bank b anches and conclude ha hey se e as
i eplaceable lende s o small and medium en e p ises. Khwaja and Mian (2008) show ha
liquidi y shocks o banks esul in dec eased loan amoun s o i ms and a educed p obabili y
3See also Meslie e al. (2022) and B agoli e al. (2022).
4
es ima o (Dube e al., 2023).
In ou s udy, since di e en municipali ies ha e bank b anch closu es in di e en yea s,
we adop he CS es ima o as an al e na i e obus es ima o as i can p o ide lexibili y
in choosing con ol g oups and obus con idence in e als ia boo s apping, acco ding o
Bake e al. (2022) and Ro h e al. (2023). The CS es ima o calcula es he ea men e ec
by choosing di e en con ol g oups and assigning hem weigh s di e en om hose o he
TWFE model. Thei es ima ed pa ame e can be exp essed as ollows:
ATTne (g, ;ω) = E[Y −Yg−ω−1|Gg= 1] −E[Y −Yg−ω−1|C= 1] (6)
ATTny(g, ;ω) = E[Y −Yg−ω−1|Gg= 1] −E[Y −Yg−ω−1|D +ω= 1] (7)
In equa ions (7) and (8), AT T ne measu es he g oup- ime ea men e ec wi h he
limi ed ea men an icipa ion assump ion imposed. I compa es he di e ence be ween he
ou come a ime and he ou come a ime g−ω−1 in he ea men g oup o g oup Ggand
he same di e ence in he ne e - ea ed g oup, i.e., hose municipali ies wi h a single b anch
open du ing he s udy pe iod. ATTny calcula es he g oup- ime a e age e ec using he
di e ence wi hin he same g oup Ggminus he di e ence in he no -ye - ea ed samples, i.e.,
samples ha en e he ea men g oup a e pe iod +ω. Choosing di e en con ol g oups
can ha e di e en bene i s in he con ex o ou s udy: he ne e - ea ed g oup can p o ide a
s able compa ison ega ding he ime e ec s, while he no -ye - ea ed g oup helps mi iga e
selec ion conce ns by compa ing g oups ha all e en ually ge ea ed. Fu he mo e, he CS
es ima o agg ega es ATT(g, ) o AT T (T) using he ollowing equa ion:
ATT(e) = X
g∈G
I{g+e≤τ}P{G=g|G+e≤τ}ATT(g, g +e) (8)
whe e es ands o he ime since he ea men and τs ands o he end o he s udy pe iod,
2021. I mul iplies he ea ed sha e o he sample by he g oup a e age ea men e ec o
eco e AT T (e).
The CS es ima o equi es he ollowing key assump ions o iden i ica ion: (i) limi ed
ea men an icipa ion— he ea ed i ms o municipali ies should ha e limi ed an icipa ion
o ω≥0 e ms ahead; and (ii) condi ional pa allel ends based on he “no -ye - ea ed”
o “ne e - ea ed” g oups—had hey no been ea ed, he ea men g oup i ms o munic-
ipali ies would ha e main ained a pa allel end compa ed wi h hei con ol g oup pee s.
The empi ical challenge in es ima ing he impac o a local bank b anch closu e is ha
he s ic exogenei y o he ea men can be ques ioned in wo dimensions: e e se causali y
and omi ed a iable bias. Re e se causali y occu s when he closu e o local bank b anches
11
ollows wo sening economic condi ions; omi ed a iable bias ep esen s he exis ence o ex-
an e unobse ed ac o s ha a ec he ea men g oups. Fo example, his can happen
when banks base hei closu e decisions on p edic ions o u u e b anch pe o mance. I
banks can obse e ex an e local ac o s and use hem o o m p edic ions, while we canno ,
hese ac o s become omi ed a iables, biasing ou es ima es.
We base ou iden i ica ion on he pa allel ends (PT) assump ion. The i s pa o PT
equi es ha he ea men and con ol g oups main ain a pa allel end. I banks close
b anches in esponse o de e io a ing local economic condi ions, he ou comes o municipal-
i ies wi h closed b anches would likely display a downwa d p e-closu e end, di e ging om
he con ol g oup. Fo he pos - ea men pe iod, PT assumes no omi ed a iables change
he cu en end excep o he ea men . PT canno be s ic ly es ed, as we canno
obse e an un ea ed coun e pa o each ea ed i m and municipali y. Howe e , we do
p o ide p e- end analysis o a gue PT does hold and no e ha ou es ima es will emain
unbiased only i PT is sa is ied.
We also conside issues o an icipa ion and spillo e s in ou s udy and a e able o con-
clude he conce ns a e minimal. An icipa ion e e s o he possibili y ha local businesses
may an icipa e he closu e o local bank b anches and adjus hei employmen pa e ns in
ad ance. Local bank b anch closu es a e due o bank s a egic decisions, and i is easonable
o assume mos i ms will no ha e access o inside in o ma ion abou an upcoming b anch
closu e in a speci ic loca ion. So, we assume no an icipa ion and se ω= 0. Spillo e e -
ec s occu when i ms ha los access o he local bank b anch seek c edi in a neighbo ing
municipali y, which can po en ially a ec ha municipali y’s i ms’ pe o mance. Figu e 1
shows ha ea ed and compa ison g oups a e geog aphically sepa a ed; i is hus likely ha
spillo e s will be ele an in his con ex . Howe e , na ionwide bank b anch closu es can
s ill ha e an impac on bo h g oups, leading o po en ial unde es ima es o he impac .
4 Empi ical Resul s
We i s epo esul s ega ding i m ope a ions, using bo h es ablishmen -le el and municipali y-
le el da a. Then we look a employmen a iables. Finally, we conduc a he e ogenei y
analysis by i m ype and economic sec o .
4.1 Fi m Ope a ions
To e alua e he impac o losing access o bank b anches o i m ope a ions and numbe
o i ms, we employ he labo census da a ha co e s he uni e se o o mal i ms. The
12
publicly a ailable e sion o he RAIS da ase con ains an “ac i e es ablishmen ” a iable—
an indica o o i ms/es ablishmen s conduc ing any economic ac i i y du ing he yea —and
he numbe o employees. 4We adop he “ac i e es ablishmen ” as ou p ima y ou come
measu e o i m ope a ions. Addi ionally, since he RAIS da ase i sel comes om he
annual labo census in B azil and equi es all egis e ed i ms o epo , we can ack he
numbe s o i ms egis e ed in he da ase o p opo ions o ce ain ypes o i ms o measu e
whe he i ms a e en e ing o lea ing he ma ke .
Hence o h, we epo ei he he o e all e ec o he e en s udy e ec s o he a iables o
in e es , bu we some imes choose o ocus on one kind o e ec o di e en a iables unde
discussion. In he main ex we display he CS es ima es using he ne e - ea ed g oup o
e en s udies and epo he e en s udy plo s wi h all h ee es ima o s in Appendix B.
The es ima es a e gene ally e y simila , sugges ing selec ion issues a e mino , and he PT
assump ion holds well ac oss di e en con ol g oups.
Figu e 3 shows he e en s udy esul s, namely dynamic eg ession coe icien s o i m
ope a ions ia he bina y a iable “ac i e es ablishmen .” The e ealed pa e n sugges s
ha bank b anch closu es make oughly 1 pe cen o es ablishmen s become inac i e in he
ollowing 3 yea s, wi h he e ec inc easing om 1.1 o 1.4 pe cen be ween he i s and
second yea s a e a closu e. Ou es ima es sugges ha bank closu es could ha e a se e e
impac in he long un: The sha e o inac i e es ablishmen s inc eases om 1.2 pe cen o
8.1 pe cen be ween he 4 h and 7 h yea s a e a closu e.
Ro h (2024) indica es he de aul way o calcula e he p e- ea men e ec wi h he
CS es ima o is o use sho gaps and display iola ions o PT as kinks in p e- ea men
es ima es. We ollow his app oach and calcula e he p e- ea men e ec using long gaps
ins ead, hus making he p e- ea men es ima es compa able wi h he dynamic TWFE
model and isual checks o p e- ea men e ec s possible. We ind no e idence o an exis ing
p e- end.
While some i ms may espond o sh inking c edi supply by scaling back ope a ions,
o he s may close down. A he same ime, lack o local bank access can c ea e a ba ie o
en y o new i ms. In he absence o da a on i m exi and en y, we epo es ima es using
he numbe o i ms wi hin he municipali y by agg ega ing es ablishmen -le el da a o he
municipali y le el. Fu he mo e, we calcula e he p opo ions o SMEs, ag icul u al i ms,
and non-ze o-employee i ms.
We epo an o e all ATT e ec in Table 1. In gene al, in hese da a we do no ind
e idence ha a bank b anch closu e leads o a change in he o al numbe o local i ms o
4O e 60 pe cen o es ablishmen s epo ze o ac i e employees a he end o he yea in ou subsample.
Fo his eason, i m-le el employee numbe s a e no a eliable measu e o i m ac i i ies.
13
Figu e 3: E ec o Exposu e o Bank B anch Closu e on Fi m Ope a ions a he Local Le el
No e: This igu e shows es ima es om equa ion (7) ega ding he bina y “ac i e es ablishmen ” a iable,
using CS es ima o s wi h ne e - ea ed municipali ies as he con ol g oup. The ba shows he 95%
con idence in e al. Yea s since bank b anch closu e equals 0 in he yea he municipali y loses i s sole
bank b anch. We epo p e- ea men ends s a ing om pe iod -3. The panel is balanced be ween
pe iods -3 and 0; o he pe iods a e es ima ed using incomple e samples. The s anda d e o s a e compu ed
by wild boo s ap. We use he “long2” op ion unde “csdid” in S a a acco ding o Ro h’s (2014)
sugges ion, making he p e- ea men es ima es compa able wi h he dynamic TWFE model and isual
checks o p e- ea men e ec s possible. We use es ablishmen -le el da a o conduc his analysis.
Sou ce: RAIS, au ho s’ calcula ions.
a composi ional change in he ypes o i ms. In column (1), he s a ic TWFE es ima es
indica e a sligh ly posi i e o small e ec on he numbe o i ms and sha es o i ms. Howe e ,
by in es iga ing wi h he mo e obus CS es ima o s and analyzing dynamic pa e ns (no
epo ed), we obse e a lack o consis en e idence o signi ican impac .
4.2 Employmen
Th ough i ms, bank b anch closu es could ha e an impac on local employmen . In his
sec ion, we e alua e he impac o bank b anch closu es on o mal sec o employmen .
In conduc ing ou analysis, we ind ha some municipali ies ha e e y ew paid o mal
wo ke s egis e ed. To a oid bounda y e ec s, we e ame he employmen analysis a ound
hose municipali ies whe e paid p i a e sec o employees o aled mo e han one pe cen o
he municipal popula ion du ing ou s udy pe iod. I appea s ha municipali ies wi h low
14
sha es o paid wo ke s a e andomly dis ibu ed and no a ec ed by he ea men . This
exclusion ules ou simila numbe s o obse a ions ac oss g oups: 43 (6.3%) o con ol g oup
municipali ies and 36 (8.3%) o ea men g oup municipali ies.
Table 1: Es ima es o he Impac o Bank B anch Closu e on Fi ms
N (Fi ms) P(SME) P(Ag Fi ms) P(Nonze o Fi ms)
(1) (2) (3) (4)
Panel A. TWFE
δ w e 5.057* -0.000 0.002 0.000
(2.390) (0.000) (0.002) (0.003)
Panel B. CS-Ne e
δcsne 2.411 -0.000 0.001 0.001
(7.927) (0.000) (0.011) (0.007)
Panel C. CS-No ye
δcsny 1.883 0.000 0.001 0.001
(7.971) (0.000) (0.011) (0.008)
Baseline Mean 281.285 0.996 0.176 0.357
Obse a ions 12,199 12,199 12,199 12,199
No e: This able shows o e all ATT es ima es o he ea men e ec s on numbe s and sha es o i ms
om equa ions (6) and (9). The s anda d e o s a e epo ed in pa en heses. The s anda d e o s o s a ic
TWFE es ima es a e clus e ed a bo h he municipali y and yea le els. The s anda d e o s o he CS
es ima es a e compu ed by wild boo s ap. The signi icance le el in his able is ep esen ed by: *** p<0.01,
** p<0.05, * p<0.10. The baseline mean ep esen s he mean o he ou come a iables o all obse a ions in
he sample du ing 2011-2013. We use a municipali y-le el panel cons uc ed om es ablishmen -le el da a
o conduc his analysis.
Sou ce: Au ho s’ calcula ions
Two key dimensions o unde s anding he equilib ium e ec s o bank b anch closu es on
employmen a e quan i y and p ice. We calcula ed he na u al log o numbe s o employed
wo ke s and he na u al log o wo ke eal wages. We ind ha employmen and wages
display signi ican a ia ion— hey eac o a b oad ange ac o s ega ding wo ke s and
hei wo king en i onmen —so adjus ing o co a ia es can imp o e es ima ion p ecision.
We hus include wo ke s’ demog aphic p o iles wi hin municipali ies as co a ia es, along
wi h job and es ablishmen in o ma ion. We include he ollowing a iables: emale sha e
o wo ke s, a e age educa ion le el (11 le els, adap ing he scale om he RAIS da ase ),
a e age weekly hi ed hou s, a e age es ablishmen size in g oups, ime in employmen , and
whi es as a sha e o wo ke s.
Figu e 4 Panel A shows he esul s om his analysis. The poin es ima es a e ela i ely
obus o he inclusion o co a ia es, which na ow he con idence in e als (see Figu e 12 o
compa ison). Howe e , he con idence in e als do no each s a is ical signi icance a he
95% le el. The impac o bank b anch closu e on employmen can be ead as ei he a null o
15
Panel A. Local Le el o Employmen and Real Wage
(a) (b)
Panel B. Wage Dispa i ies
(c) (d)
Figu e 4: E ec o Exposu e o Bank B anch Closu e on Local Le els o Employmen and
Wages
No e: This igu e shows es ima es om equa ion (7) ega ding he Log(Numbe o Employed People),
Log(Real Wages), SD (Real Wages), and P90-10 Real Wages. The ba shows he 95% con idence in e al.
Yea s since bank b anch closu e equals 0 in he yea he municipali y loses i s sole bank b anch. We epo
p e- ea men ends s a ing om pe iod -3. The panel is balanced be ween pe iods -3 and 0; o he pe iods
a e es ima ed using incomple e samples. The s anda d e o s a e compu ed by wild boo s ap. We use he
“long2” op ion unde “csdid” in S a a acco ding o Ro h’s (2014) sugges ion, making he p e- ea men
es ima es compa able wi h dynamic TWFE model and isual checks o p e- ea men e ec s possible. We
use municipali y-le el da a o conduc his analysis. We include he ollowing co a ia es o na ow down
he con idence in e als: emale sha e o wo ke s, a e age educa ion le el, a e age weekly hi ed hou s,
a e age es ablishmen size in g oups, ime in employmen , and whi es as a sha e o wo ke s. Real wages a e
calcula ed by adjus ing nominal wages by B azil’s annual CPI. The uni o eal wages is 2011 eais.
Sou ce: RAIS, au ho s’ calcula ions.
16
nega i e e ec . Fo he numbe o employed wo ke s, he CS es ima o wi h ne e - ea ed as
con ols gene ally shows a po en ial long- e m dec easing end despi e i s poin es ima es
being s a is ically insigni ican . Fo he a e age wage, i shows a long- e m dec ease o 1.5
o 1.7 pe cen and an inconclusi e sho - e m e ec . In his analysis, bo h CS es ima o s
show a simila end, while he TWFE es ima o has a sligh ly di e en pa e n. The TWFE
sugges s a null e ec on he numbe o employed wo ke s and a small posi i e e ec on wages.
We do no ind e idence o iola ion o PT.
To be e unde s and he impac o bank b anch closu es on local employmen dynamics,
we del e in o wage dispa i ies. We employ wo dis inc measu es o assess he e ec s on wage
dispa i ies: he s anda d de ia ion o eal wages, o gauge o e all wage a iabili y among
wo ke s, and he di e ence be ween he 90 h and 10 h pe cen iles o eal wages, o examine
dispa i ies be ween high- and low-paid jobs.
We p esen ou es ima es in Figu e 4 Panel B. Fo he s anda d de ia ion o eal wages,
we no e a sho - e m dec ease o 142 eais one yea a e a closu e, pe sis ing o wo yea s
be o e g adually diminishing o e app oxima ely i e yea s. In con as , he gap be ween he
90 h and 10 h pe cen ile wages ini ially shows no signi ican sho - e m e ec . This hin s
a a po en ial long- e m widening, beginning h ee yea s pos -closu e, wi h an inc ease o
28 eais, and eaching a di e ence o 231 eais six yea s a e a closu e. These obse a ions
sugges ha in he sho e m, bank b anch closu es do no signi ican ly a ec he wage gap
be ween high- and low-paid jobs, bu hey do lead o a no able educ ion in o e all wage
dispa i ies among wo ke s, po en ially h ough he middle-income wage b acke . In he long
e m, howe e , he loss o a bank b anch appea s o con ibu e o inc easing wage dispa i ies
be ween high- and low-paid posi ions, e en as o e all wage dispa i ies among wo ke s become
less p onounced.
Finally, we a e in e es ed in de e mining whe he bank b anch closu es impac i ms’
employmen pa e ns in o he ways. Fo ins ance, we examine whe he i ms main ain he
same numbe o employees bu educe hei wo king hou s, o i pa - ime posi ions a e
being used as subs i u es o ull- ime jobs. Ou analysis u ilizes weekly con ac ed hou s
and he p opo ion o pa - ime wo k as me ics. The esul s a e epo ed in Figu e 5. We
obse e a dec ease in con ac ed hou s 5 o 6 yea s ollowing a b anch closu e. Howe e ,
upon examining he e en -s udy plo o pa - ime wo k me ics, we do no ind a clea
impac due o bank b anch closu es.
Finally, we s udy whe he local bank b anch closu es ha e an impac on economic ou pu .
I bank b anch closu es lead o educed i m ope a ions and employmen le els, i is possible
ha impac ed i ms would p oduce less ou pu . He e we ely on municipali y-le el GDP
da a and alue-added accoun s in se ices (excluding public adminis a ion, educa ion, e c.),
17
(a) (b)
Figu e 5: E ec o Exposu e o Bank B anch Closu e on O he Local Employmen Pa e ns
No e: This igu e shows es ima es om equa ion (7) ega ding weekly hi ed hou s and a bina y a iable o
pa - ime wo k. The ba shows he 95% con idence in e al. Yea s since bank b anch closu e equals 0 in
he yea he municipali y loses i s sole bank b anch. We epo p e- ea men ends s a ing om pe iod
-3. The panel is balanced be ween pe iods -3 and 0; o he pe iods a e es ima ed using incomple e samples.
The s anda d e o s a e compu ed by wild boo s ap. We use he “long2” op ion unde “csdid” in S a a
acco ding o Ro h’s (2014) sugges ion, making he p e- ea men es ima es compa able wi h he dynamic
TWFE model and isual checks o p e- ea men e ec s possible. We use municipali y-le el da a o conduc
his analysis. We include he ollowing co a ia es o na ow down he con idence in e als: emale sha e o
wo ke s, a e age educa ion le el, a e age weekly hi ed hou s, a e age es ablishmen size in g oups, ime in
employmen , and whi es as a sha e o wo ke s.
Sou ce: RAIS, au ho s’ calcula ions.
indus y, and ag icul u e o conduc ou analysis. We adjus p ices o 2011 eais and p esen
he e ec pa e n in Figu e 10. In se ices, we ind a sho - e m dec easing end in he
i s 3 yea s a e closu e, al hough he es ima es a e imp ecise. We es ima e ha bank
b anch closu es dec ease alued added in se ices by 3,366 housand eais by he 3 d yea .
In indus y and ag icul u e, we do no ind signi ican sho - e m pa e ns bu a po en ial
long- e m dec easing end.
4.3 He e ogenei y Analysis
In his sec ion, we e alua e how he impac o bank b anch closu es di e s by i m size and
economic sec o .
Knowing he a e age e ec o closing bank b anches on i m ope a ions, i will be aluable
o lea n wha kind o i ms a e mo e ulne able o hese e ec s. We conside wo dimensions:
i m size and main economic sec o . The e a e di e en ways o ca ego ize i ms by numbe
o employees in B azil (Ca nei o e al., 2020). IBGE classi ies he size o i ms in ce ain
sec o s using he numbe o employees. We use IBGE’s employee coun -based i m size
ca ego iza ion o i ms/es ablishmen s in manu ac u ing and se ices (see Table 4). In he
18
absence o a se led classi ica ion o he emaining sec o s, such as ag icul u e, we adop
C a o e al. (2018)’s classi ica ion scheme. 5
We classi y i ms in o ou main sec o s: indus y and cons uc ion, e e ed o as man-
u ac u ing by Ca nei o e al. (2020); ade and se ices, e e ed o as se ices; ag icul u e;
and all o he sec o s.
Table 2: Es ima es o he E ec o Bank B anch Closu e on Fi m Ope a ions, by Fi m Size
Ou come Va iable Ac i e Es ablishmen s
Mic o Fi ms Small Fi ms Medium Fi ms La ge Fi ms
(1) (2) (3) (4)
Panel A. TWFE
δ w e -0.009 -0.000 -0.001 0.002
(0.005) (0.000) (0.002) (0.002)
Panel B. CS-Ne e
δcsne -0.014*** -0.000 -0.000 0.000
(0.002) (0.000) (0.000) (0.000)
Panel C. CS-No ye
δcsny -0.012*** -0.000 -0.000 -0.000
(0.002) (0.000) (0.000) (0.000)
Baseline Mean 0.785 0.996 0.997 0.996
Obse a ions 3,158,958 208,168 26,725 5,189
No e: This able p esen s he impac o a bank b anch closu e on i m ope a ions by i m size, using equa ions
(5), (7) and (8). The s anda d e o s a e epo ed in pa en heses. The s anda d e o s o he s a ic TWFE
es ima es a e clus e ed a bo h he municipali y and yea le el. The s anda d e o s o he CS es ima es a e
compu ed by wild boo s ap. The signi icance le el in his able is ep esen ed by: *** p<0.01, ** p<0.05, *
p<0.10. The baseline mean ep esen s he mean o he ou come a iables o all obse a ions in he sample
du ing 2011-2013. We use es ablishmen -le el da a o conduc his analysis. Obse a ions deno e he numbe
o i ms pe size ca ego y.
Sou ce: Au ho s’ calcula ions
We i s s udy he e ec o a bank b anch closu e on i m ope a ions ac oss di e en i m
sizes. Table 2 p esen s he esul s. We ind ha mic o i ms bea almos he ull impac
on ope a ions om local bank b anch closu es, while he e ec s a e oughly ze o o i ms
o all o he sizes. This could be due o hei high le el o ac i e ope a ion s a us: small,
medium, and la ge i ms a e all mo e han 99.6% ac i e in he baseline yea s (be ween 2011
5An al e na i e app oach o i ms in ag icul u e and all o he sec o s is o apply IBGE’s ca ego iza ion
o manu ac u ing i ms. The esul s a e quali a i ely simila .
19
o 2013). The esul s con i m ou p e ious hypo hesis ha la ge i ms ha e g ea e inancial
access and a e less likely o be dis up ed by a local bank b anch closu e. Mic o i ms play a
dominan ole (92.9%) in ou sample and a e sensi i e o local b anch closu es, which could
imply ha mic o i ms’ banking demands a e mo e localized. We include an e en -s udy
plo o his e ec in Appendix B (see Figu e 14). The e ec is simila o he impac pa e n
o ou en i e sample (see Figu e 11).
Table 3: Es ima es o he E ec o Bank B anch Closu e on Fi m Ope a ions, by Sec o
Ou come Va iable Ac i e Es ablishmen s
T ade & Se ice Indus y & Cons uc ion Ag icul u e All O he
(1) (2) (3) (4)
Panel A. TWFE
δ w e -0.008 -0.003 -0.013 0.017
(0.005) (0.006) (0.008) (0.052)
Panel B. CS-Ne e
δcsne -0.014*** 0.002 -0.018*** 0.000
(0.002) (0.005) (0.003) (0.053)
Panel C. CS-No ye
δcsny -0.013*** 0.003 -0.016*** 0.000
(0.002) (0.005) (0.003) (0.054)
Baseline Mean 0.768 0.790 0.914 0.858
Obse a ions 2,315,624 415,022 665,012 3,757
No e: This able p esen s he impac o a bank b anch closu e on i m ope a ions by i m sec o , using
equa ions (5), (7) and (8). The sec o s a e: ade and se ices, indus y and cons uc ion, ag icul u e, and
all o he i ms. The s anda d e o s a e epo ed in pa en heses. The s anda d e o s o he s a ic TWFE
es ima es a e clus e ed a bo h he municipali y and yea le el. The s anda d e o s o he CS es ima es a e
compu ed by wild boo s ap. The signi icance le el in his able is ep esen ed by: *** p<0.01, ** p<0.05, *
p<0.10. The baseline mean ep esen s he mean o he ou come a iables o all obse a ions in he sample
du ing 2011-2013. We use es ablishmen -le el da a o conduc his analysis. Obse a ions deno e he numbe
o i ms pe sec o .
Sou ce: Au ho s’ calcula ions
We now u n o he impac o bank b anch closu es on i m ope a ions by sec o s. The
esul s a e p esen ed in Table 3. While he TWFE es ima es sugges null e ec s ac oss
di e en sec o s, he CS es ima es imply ha bank b anch closu es dep ess he ope a ions
o i ms in ade and se ices as well as in ag icul u e. O e all e ec s on ag icul u e i ms
a e somewha mo e se e e han on ade and se ice i ms despi e ag icul u e i ms’ highe
a e o ac i e es ablishmen s a us in he baseline yea . We p esen e en -s udy plo s o he
ade and se ice sec o as well as he ag icul u e sec o in Figu e 6.
20
Table 5: Summa y S a is ics o Sample Va iables Ac oss G oups
Ne e -T ea ed G oup T ea men G oup Di e ence
Mean N Mean N
(1) (2) (3) (4) (5)
Es ablishmen -Le el Da a
Ac i e Es ablishmen 0.939 2,337,785 0.926 1,138,674 0.013 ***
(1.146) (1.115) (0.001)
Es ablishmen Size (G oups) 1.585 2,337,785 1.573 1,138,674 0.012 ***
(0.965) (0.969) (0.001)
Municipali y-Le el Da a
Numbe o Es ablishmen s 307.770 7,436 233.222 4,763 74.548 ***
(176.866) (128.248) (2.963)
P(SME) 0.995 7,436 0.996 4,763 0.000 **
(0.007) (0.008) (0.000)
P(Ag Fi m) 0.175 7,436 0.196 4,763 -0.021 ***
(0.163) (0.198) (0.003)
Log(Real Wages) 6.812 6,963 6.791 4,367 0.021 ***
(0.172) (0.170) (0.003)
SD (Real Wages) 845.087 6,963 795.49 4,367 49.597 ***
(434.860) (410.458) (8.216)
Real Wages P90 1591.892 6,963 1514.931 4,367 76.961 ***
(496.044) (525.198) (9.796)
Real Wages P10 627.451 6,963 628.28 4,367 -0.83
(69.203) (71.615) (1.354)
P(Female) 0.309 6,963 0.302 4,367 0.007 *
(0.096) (0.101) (0.002)
Time Employed (Mon hs) 35.384 6,963 34.891 4,367 0.493 *
(11.027) (10.803) (0.211)
Weekly Hi ed Hou s 42.913 6,963 42.974 4,367 -0.061 ***
(0.957) (0.899) (0.018)
Age 33.789 6,963 33.868 4,367 -0.080 *
(1.828) (2.011) (0.037)
A e age Es Size 3.672 6,963 3.593 4,367 0.080 ***
(0.910) (0.939) (0.018)
P(Pa -Time Wo k) 0.002 6,963 0.002 4,367 0.000
(0.008) (0.009) 0.000
Educa ion Le el 5.922 6,963 5.87 4,367 0.052 ***
(0.601) (0.664) (0.012)
Numbe o Employed People 1017.736 6,963 736.796 4,367 280.94 ***
(886.796) (669.004) (15.632)
Race: Whi e 0.522 6,963 0.533 4,367 -0.011 *
(0.267) (0.264) (0.005)
Popula ion 10237.16 6,963 8624.363 4,367 1612.797 ***
(6884.674) (6277.038) 0.000
Valued Added: Ag 30494.07 7,436 21364.62 4,763 9129.45 ***
(49670.360) (27651.030) (787.917)
Value Added: Ind 21514.61 7,436 13793.65 4,763 7720.96 ***
(59695.790) (35852.060) (959.715)
Value Added: Se ices 35166.9 7,436 25079.69 4,763 10087.21 ***
(35294.550) (29811.660) (617.299)
27
[Con inue o able no es]
No e: This able p esen s summa y s a is ics o all he a iables in ou sample. Fi m size is ca ego ized in o
9 g oups in e ms o numbe o employees. Fi m sec o ca ego ies a e indus y and cons uc ion, ade and
se ices, and all o he sec o s. Real wages a e de la ed using he annual consume p ice index; hei uni s
a e 2011 B azilian eais. Educa ion le el consis s o 11 g oups acco ding o ca ego ies in he RAIS da ase .
The pa ial wo k s a us is only a ailable om 2017. Value added is de la ed using he annual consume p ice
index; i s uni s a e housands o 2011 B azilian eais. The signi icance le el in his able is ep esen ed by:
∗p < 0.1∗ ∗p < 0.5,∗∗∗p < 0.01.
Sou ce: Au ho s’ calcula ions
Table 6: Popula ion and GDP Compa isons Ac oss G oups
Yea Ne e -T ea ed T ea men O he Mun
(1) (2) (3)
Popula ion 2011-2013 10688.15 9037.55 41421.61
(159.66) (177.69) (2002.94)
2014-2017 11075.77 9295.07 43427.09
(144.78) (159.00) (1805.63)
2018-2021 11138.72 9268.61 44695.30
(149.08) (160.02) (1854.50)
Real GDP 2011-2013 117566.50 85692.04 976888.10
(2300.79) (1808.36) (79505.82)
2014-2017 123471.40 91513.17 965289.10
(2159.20) (1650.16) (66289.58)
2018-2021 135178.90 96514.38 972079.40
(2518.69) (1889.33) (61537.83)
RGDP pe capi a 2011-2013 13.99 13.12 14.92
(0.29) (0.35) (0.16)
2014-2017 14.10 13.31 14.82
(0.25) (0.30) (0.12)
2018-2021 15.20 14.02 15.95
(0.27) (0.33) (0.14)
No e: This able p esen s he a e age municipal popula ion, eal GDP, and eal GDP pe capi a
o each sample and ime pe iod. The simple a e age is p esen ed and he s anda d de ia ions
wi hin g oups a e p o ided in pa en heses. The uni o popula ion is numbe o indi iduals, and
he uni o eal GDP and eal GDP pe capi a a e housands o eais. Real GDP is de la ed using
B azil’s GDP de la o om FRED and IMF; he base yea is 2011. Municipal popula ions a e
IBGE es ima es, no coun s. RGDP pe capi a is calcula ed by he au ho s using he es ima ed
popula ion.
Sou ce: IBGE, au ho s’ calcula ions
28
Table 7: In ol ed Banks: T ea men and Ne e -T ea ed G oup
BANK T ea men Ne e -T ea ed To al Public-Owned
BCO BANESTES S.A. 0 18 18 X
BCO BRADESCO S.A. 9 157 166
BCO DO BRASIL S.A. 332 271 603 X
BCO DO EST. DE SE S.A. 1 12 13 X
BCO DO EST. DO PA S.A. 0 1 1 X
BCO DO ESTADO DO RS S.A. 0 57 57 X
BCO DO NORDESTE DO BRASIL S.A. 0 1 1 X
BCO SANTANDER (BRASIL) S.A. 29 51 80
CAIXA ECONOMICA FEDERAL 0 11 11 X
ITA`o UNIBANCO S.A. 62 97 159
To al 433 676 1,109 703
No e: This able p esen s he banks and he numbe o hei b anches in he ea men and con ol
g oups. The igh column indica es whe he a bank is publicly owned (i.e., owned by he s a e
o ede al go e nmen ). We include he “To al” ow and column o indica e he numbe o bank
b anches in each g oup and each bank.
Sou ce: Au ho s’ calcula ions
Table 8: Pa e ns o Bank B anch Closu es by Yea
Yea F eq. Pe cen Cum.
2014 29 2.61 2.61
2015 21 1.89 4.5
2016 27 2.43 6.93
2017 155 13.98 20.91
2018 34 3.07 23.98
2019 104 9.38 33.36
2020 2 0.18 33.54
2021 61 5.5 39.04
No Los 676 60.96 100
To al 1109
No e: This able shows he pa e n o bank b anch closu es in ou sample. Ou sample consis s o
municipali ies wi h one bank b anch om 2011 o 2013. We exclude municipali ies whose b anch
coun doesn’ all in o he abso bing s age om 2014. A b anch is conside ed los when he only
b anch closes and doesn’ eopen ill he end o he s udy pe iod. ”No Los ” municipali ies a e
hose ha e ain one bank b anch om 2011 o 2021. The equency in he iming o b anch loss
is shown o ou sample, including hei sha es and cumula i e pe cen age.
Sou ce: Au ho s’ calcula ions
29
Figu e 8: Dis ibu ion o Municipali ies by Numbe o Bank B anches, 2011-2021
No e: This igu e shows he dis ibu ion o municipali ies by numbe o bank b anches loca ed in each
municipali y in each yea . The colo s o each ca ego y a e ma ked in he igh legend. This igu e only
p esen s he dis ibu ions o banked municipali ies. Du ing he s udy pe iod, B azil had 5,570
municipali ies in o al.
Sou ce: Au ho s’ calcula ions, ESTBAN da ase .
30
Figu e 9: E ec o Exposu e o Bank B anch Closu e on Fi m Ope a ions: Mic o Fi ms
No e: This igu e shows es ima es om equa ion (7) ega ding he “ac i e es ablishmen ” a iable among
mic o i ms. The ba shows he 95% con idence in e al. Yea s since bank b anch closu e equals 0 in he
yea he municipali y loses i s sole bank b anch. We epo p e- ea men ends s a ing om pe iod -3.
The panel is balanced be ween pe iods -3 and 0; o he pe iods a e es ima ed using incomple e samples.
The s anda d e o s a e compu ed by wild boo s ap. We use he “long2” op ion unde “csdid” in S a a
acco ding o Ro h’s (2014) sugges ion, making he p e- ea men es ima es compa able wi h he dynamic
TWFE model and isual checks o p e- ea men e ec s possible. We use municipali y-le el da a o
conduc his analysis.
Sou ce: RAIS, au ho s’ calcula ions.
31
(a)
(b) (c)
Figu e 10: E ec o Exposu e o Bank B anch Closu e on Economic Ou pu
No e: This igu e shows es ima es om equa ion (7) ega ding alue added in se ices (excluding adminis-
a ion, educa ion, public heal h, and social secu i y), indus y, and ag icul u e. The ba shows he 95%
con idence in e al. Yea s since bank b anch closu e equals 0 in he yea he municipali y loses i s sole bank
b anch. We epo p e- ea men ends s a ing om pe iod -3. The panel is balanced be ween pe iods -3
and 0; o he pe iods a e es ima ed using incomple e samples. The s anda d e o s o he dynamic TWFE
es ima es a e clus e ed a bo h he municipali y and yea le el. The s anda d e o s o he CS es ima es
a e compu ed by wild boo s ap. P ices a e de la ed using he e e ence yea ’s a e age consume p ice index.
The uni o measu emen is housands o 2011 B azilian eais.
Sou ce: B azil’s GDP by Municipali y, au ho s’ calcula ions.
32
Appendix B. E en -S udy Plo s wi h Th ee Es ima o s
In his sec ion, we p o ide e en -s udy plo s wi h esul s om all h ee es ima o s o equa ions (5), (7) and
(8). The es ima o s a e dynamic TWFE, CS-ne e - ea ed, CS-no -ye - ea ed, espec i ely.
Figu e 11: E ec o Exposu e o Bank B anch Closu e on Fi m Ope a ions a he Local
Le el, Th ee Es ima o s
No e: This igu e shows es ima es om equa ions (5), (7) and (8) ega ding he “ac i e es ablishmen ”
a iable. The ba shows he 95% con idence in e al. Yea s since bank b anch closu e equals 0 in he yea
he municipali y loses i s sole bank b anch. We epo p e- ea men ends s a ing om pe iod -3. The
panel is balanced be ween pe iods -3 and 0; o he pe iods a e es ima ed using incomple e samples. The
s anda d e o s o he TWFE es ima es a e obus s anda d e o s clus e ed by bo h yea and
municipali y. The s anda d e o s o he CS es ima es a e compu ed by wild boo s ap.
Sou ce: RAIS, au ho s’ calcula ions.
33
Panel A. Local Le el o Employmen and Real Wage
(a) (b)
Panel B. Wage Dispa i ies
(c) (d)
Figu e 12: E ec o Exposu e o Bank B anch Closu e on Local Le els o Employmen and
Wages, Th ee Es ima o s
No e: This igu e shows es ima es om equa ions (5), (7) and (8) ega ding he a iables Log(Numbe o
Employed People), Log(Real Wages), SD (Real Wages), and P90-10 Real Wages. The ba shows he 95%
con idence in e al. Yea s since bank b anch closu e equals 0 in he yea he municipali y loses i s sole
bank b anch. We epo p e- ea men ends s a ing om pe iod -3. The panel is balanced be ween
pe iods -3 and 0; o he pe iods a e es ima ed using incomple e samples. The s anda d e o s o he
TWFE es ima es a e obus s anda d e o s clus e ed by bo h yea and municipali y. The s anda d e o s
o he CS es ima es a e compu ed by wild boo s ap.
Sou ce: RAIS, au ho s’ calcula ions.
34
(a) (b)
Figu e 13: E ec o Exposu e o Bank B anch Closu e on O he Local Employmen Pa e ns,
Th ee Es ima o s
No e: This igu e shows es ima es om equa ions (5), (7) and (8) ega ding weekly hi ed hou s and a bina y
a iable o pa - ime wo k. The ba shows he 95% con idence in e al. Yea s since bank b anch closu e
equals 0 in he yea he municipali y loses i s sole bank b anch. We epo p e- ea men ends s a ing om
pe iod -3. The panel is balanced be ween pe iods -3 and 0; o he pe iods a e es ima ed using incomple e
samples. The s anda d e o s o he TWFE es ima es a e obus s anda d e o s clus e ed by bo h yea
and municipali y. The s anda d e o s o he CS es ima es a e compu ed by wild boo s ap.
Sou ce: RAIS, au ho s’ calcula ions.
35
Figu e 14: E ec o Exposu e o Bank B anch Closu e on Fi m Ope a ions: Mic o Fi ms,
Th ee Es ima o s
No e: This igu e shows es ima es om equa ions (5), (7) and (8) ega ding he “ac i e es ablishmen ”
a iable among mic o i ms. The ba shows he 95% con idence in e al. Yea s since bank b anch closu e
equals 0 in he yea he municipali y loses i s sole bank b anch. We epo p e- ea men ends s a ing
om pe iod -3. The panel is balanced be ween pe iods -3 and 0; o he pe iods a e es ima ed using
incomple e samples. The s anda d e o s o he TWFE es ima es a e obus s anda d e o s clus e ed by
bo h yea and municipali y. The s anda d e o s o he CS es ima es a e compu ed by wild boo s ap.
Sou ce: RAIS, au ho s’ calcula ions.
36
