Boddin, Dominik; e Kaa , Daniel Ma cel; Roszbach, Kaspe
Wo king Pape
C oss-bo de bank lows, egional household c edi booms
and bank isk- aking
Wo king Pape , No. 10/2024
P o ided in Coope a ion wi h:
No ges Bank, Oslo
Sugges ed Ci a ion: Boddin, Dominik; e Kaa , Daniel Ma cel; Roszbach, Kaspe (2024) : C oss-bo de
bank lows, egional household c edi booms and bank isk- aking, Wo king Pape , No. 10/2024,
ISBN 978-82-8379-319-2, No ges Bank, Oslo,
h ps://hdl.handle.ne /11250/3172793
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/310428
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by-nc-nd/4.0/
Wo king Pape
C oss-Bo de Bank Flows, Regional Household C edi Booms
and Bank Risk-Taking
No ges Bank Resea ch
Au ho s:
Dominik Boddin
Daniel e Kaa
Kaspe Roszbach
Keywo ds
C oss-bo de bank lows,
Households, Bank lending, Risk-
aking, C edi booms
10 | 2024
No ges Bank Wo king Pape
Wo king pape s a No ges Bank, a 1992/1 il 2009/2 kan bes illes på e-
pos : se icesen e @no ges-bank.no
F a 1999 og sene e e publikasjonene ilgjengelige på www.no ges-bank.no
Wo king pape s inneholde o skningsa beide og u edninge som anlig is ikke
ha å sin endelige o m. Hensik en e blan anne a o a e en kan mo a
kommen a e a kollege og and e in e esse e. Synspunk e og konklusjone i
a beidene s å o o a e nes egning.
Wo king pape s om No ges Bank, om 1992/1 o 2009/2 can be o de ed
by e-mail: se icesen e @no ges-bank.no
Wo king pape s om 1999 onwa ds a e a ailable on www.no ges-bank.no
No ges Bank’s wo king pape s p esen esea ch p ojec s and epo s (no
usually in hei inal o m) and a e in ended in e alia o enable he au ho o
bene i om he commen s o colleagues and o he in e es ed pa ies. Views
and conclusions exp essed in wo king pape s a e he esponsibili y o he
au ho s alone.
ISSN 1502-8143 (online)
ISBN 978-82-8379-319-2 (online)
C oss-Bo de Bank Flows, Regional Household C edi Booms and
Bank Risk-Taking∗
Dominik Boddin†Daniel Ma cel e Kaa ‡Kaspe Roszbach§
Decembe 2, 2024
Abs ac
This pape p o ides no el mic o-le el e idence ha c oss-bo de bank lows a e
impo an o households’ access o c edi no only in eme ging ma ke s bu also in
ad anced economies. These o eign bank lows can d i e local c edi c edi booms
ha inc ease bank isk. We s udy how he in lux o c oss-bo de bank unding ha
ollowed he ECB’s implemen a ion o non-con en ional mone a y policy in 2014/15
impac ed lending o households, using supe iso y bank-le el da a alongside household-
le el c edi and consump ion da a om Ge many. Regional banks ha a e highly
exposed o luc ua ions in o eign capi al in lows inc ease consume lending o iskie ,
lowe -income households by 50% mo e han o he banks. When deposi in lows om
non-eu o a ea banks ise, his induces less capi alized banks o expand hei lending
on he ex ensi e ma gin. Imp o ed access o c edi enables lowe -income cus ome s o
exposed banks o inc ease non-du able consume spending. Da a om a la ge g oup
o eu o a ea coun ies con i m ou conclusions.
Keywo ds: C oss-Bo de Bank Flows, Households, Bank Lending, Risk-Taking,
C edi Booms, Funding Shocks
JEL Classi ica ions: F3, G2, G5
∗This pape should no be epo ed as ep esen ing he iews o he Bundesbank, Eu osys em o No ges
Bank. The iews exp essed a e hose o he au ho s and do no necessa ily e lec hose o hese ins i u ions.
We hank a e e ee o he No ges Bank wo king pape se ies, Tobias Be g, Na han Con e se, Vale iya Dinge ,
Linda Goldbe g, La s No den, S e en Ongena, Tim Schmid -Eisenloh , Emil Ve ne and pa icipan s in he
No ges Bank Sp ing Ins i u e o aluable commen s and Tobias Schmid o sha ing his code. This pape
uses da a om he Bundesbank Panel on Household Finances and he Eu osys em Household Finance and
Consump ion Su ey, accessed ia scien i ic use iles. Repo ed esul s and ela ed obse a ions and analyses
may no co espond o he esul s o analyses o he da a p oduce s. The con iden ial Bundesbank da a was
accessed a i s Resea ch Da a and Se ice Cen e ia on-si e use (P ojec No. 2023/0041). Te Kaa g a e ully
acknowledges inancial suppo om he Du ch Resea ch Council NWO, g an numbe VI.Veni.211E.023.
†Deu sche Bundesbank (dominik.bo[email p o ec ed])
‡Uni e si y o G oningen (d.m. e.k[email p o ec ed])
§Co esponding au ho : No ges Bank and Uni e si y o G oningen (kaspe . oszbac[email p o ec ed])
1 In oduc ion
An ex ensi e body o li e a u e has documen ed, using agg ega e, bank-le el, o bank- i m
da a ha o eign capi al in lows inc ease o e all bank lending, wi h c edi shi ing owa ds
iskie i ms and coun ies (e.g., Magud e al.,2014;Baskaya e al.,2017;Te Kaa ,2021).
Capi al in lows esul ing om changing na ional o in e na ional inancial condi ions a ec
bank lending h ough bo h secu i ies and in e bank ma ke s as well as in a-conce n lows in
la ge global banks (Ce o elli and Goldbe g,2012;Temes a y e al.,2018;Co ea e al.,2021).
Such wholesale sou ces o o eign bank unding a e known o be impo an in de eloping and
eme ging economies. Whe he o eign unding in lows a ec households, no ably in mo e
ad anced economies, has ecei ed limi ed a en ion. Recen esea ch on eme ging ma ke
sec o -le el da a by Ga be e al. (2019) documen s ha agg ega e household c edi ises
in esponse o capi al in lows. Rela edly, Sa ie e al. (2020) show ha inancial openness
igge s a ealloca ion o esou ces owa ds i ms wi h high expendi u e elas ici y ac i i ies.
Ye , his li e a u e has no de o ed a en ion o whe he o eign capi al in lows a ec he
composi ion and alloca ion o c edi be ween households.
This pape aims o add ess his gap in he li e a u e by in es iga ing he e ec s o
inc eased o eign capi al in lows on he household sec o in Ge many. Speci ically, we ocus
on he pe iod when he Eu opean Cen al Bank (ECB) implemen ed i s nega i e in e es
a e policy and quan i a i e easing p og ams in 2014/2015. Ne c oss-bo de bank lows in o
he eu o a ea inc eased signi ican ly, ising om -3.5% o GDP in 2014:Q1 o almos +3%
in 2016:Q3, p o iding new unds o eu o a ea banks. In Ge many, he la ges eu o a ea
economy, he inc ease in bank in lows was e en mo e p onounced, as we documen below.
To s udy he e ec s o hese in lows, we use g anula household-le el da a combined wi h
de ailed supe iso y bank balance shee in o ma ion. We ind ha he ise in c oss-bo de
bank in lows induced banks wi h g ea e ini ial dependence on non-co e unding (NCF),
i.e., in e bank bo owing, money ma ke unding and deb secu i ies inancing, o aise hei
consume loan supply o low-income households. In economic e ms lowe income households
1
expe ience a 51 pe cen age poin s highe g ow h a e in uncolla e alized consume c edi
compa ed o highe -income households. Lowe income households who ha e hei main bank
ela ionship wi h a mo e exposed bank, i.e., wi h g ea e dependence on non-co e unding,
expe ience an e en la ge g ow h di e en ial o 83 pe cen age poin s. When a bank is
weakly capi alized, he e ec s a e e en s onge , consis en wi h he li e a u e on he isk-
aking channel o mone a y policy ansmission (e.g., Jim´enez e al.,2014). The g ow h in
consume c edi mainly bene i s households on he ex ensi e ma gin, i.e., households, who
did no ecei e uncolla e alized c edi be o e, see inc eases in consume c edi olumes. We
ind no e idence o inc eased isk- aking in banks’ mo gage lending.
The inc ease in banks’ consume lending o iskie households is consis en wi h heo e -
ical p edic ions. Acha ya and Naq i (2012) show ha an inc ease in bank liquidi y caused,
o ins ance, by capi al in lows, wo sens bank agency p oblems and induces loan o ice s o
inc ease hei lending o iskie loan applican s. Ma inez-Mie a and Repullo (2017) a gue
ha a ise in he supply o sa ings, as occu s h ough capi al in lows, educes in e es a e
ma gins and incen i izes banks o main ain p o i abili y by cu ing back on cos s, pa icula ly
on moni o ing and sc eening. This leads o inc eased lending o iskie bo owe s. Simila ly,
Rajan (2006) highligh s ha lowe in e es a es, po en ially s emming om capi al in lows,
can lead o isk- aking and a sea ch o yield by banks.
We s udy he impac o c oss-bo de lows on banks’ lending o households by le e aging
wo g anula household-le el da a se s. The i s da a se , used o ou benchma k anal-
ysis, is he Ge man Panel on Household Finances (PHF), which con ains de ailed su ey
in o ma ion on households’ c edi , income, weal h, consump ion and backg ound cha ac e -
is ics. In ou main analyses we exploi a peculia ea u e o he Ge man banking sys em:
ce ain banks—sa ings and coope a i e banks—a e es ic ed o ope a ing wi hin speci ic
geog aphical bounda ies o adminis a i e egions. These egions align wi h he egional
in o ma ion a ailable o households. As he PHF also includes ques ions abou households’
p ima y banking ela ionships, we can link households o a speci ic bank when hei main
2
ela ionship is wi h a sa ings o coope a i e bank. Using ich supe iso y da a om he
Bundesbank, we quan i y he ela ionship be ween bank lows and lending o households as
a unc ion o banks’ exposu e o c oss-bo de lows. In he second pa o ou analysis, we
p o ide ex e nal alida ion o ou indings by employing household da a om he ECB’s
Household Finance and Consump ion su ey (HFCS) o he eu o a ea. These da a enable
us o con i m, wi hin a b oade sample, ha bank lows disp opo iona ely a ec banks’
lending o lowe -income households.
We exploi he su ge in eu o a ea bank in lows in 2015-17, which was la gely d i en by he
ECB’s implemen a ion o non-con en ional mone a y policy ools, o es ima e di e ence-in-
di e ences eg essions o a ious measu es o c edi and consump ion o hese households
and banks. Ou main ou come a iable o in e es is he g ow h a e o a household’s
consume o mo gage c edi . We measu e a bank’s exposu e o c oss-bo de bank lows as i s
p e-2015 NCF a io, i.e., in e bank bo owing plus money ma ke and deb secu i ies issued
as a sha e o o al asse s. This ollows Baskaya e al. (2017), who demons a e ha banks wi h
highe NCF a ios exhibi a lending beha io mo e sensi i e o c oss-bo de capi al lows.
In ui i ely, banks ha depend hea ily on in e bank unding and o he ypes o non-co e
unds should be mo e a ec ed by c oss-bo de bank lows, while e ail deposi s a e ypically
qui e s icky and hence la gely un ela ed o such lows. To assess whe he mo e exposed banks
especially inc ease lending iskie households, we analyze he in e ac ion be ween he banks’
non-co e dependence and households’ iskiness, p oxied by ini ial income (Maye ,2023,Bee
e al.,2018,Ame ican Exp ess,2022). We u he iden i y he accompanying eal e ec s by
s udying a ious componen s o a household’s consump ion expendi u es.
Ou analysis p o ides h ee main esul s. Fi s , we show ha mo e exposed banks, i.e.,
hose mo e dependen on in e bank unding, inc ease hei lending o low-income households
in esponse o he bank in low shock. Economically, ou es ima es imply ha a low-income
household, in he 25 h pe cen ile o he income dis ibu ion, compa ed o a high-income
household, in he 75 h pe cen ile, expe iences a 51 pe cen age poin highe g ow h a e
3
o consume c edi a e he bank in low shock. This g ow h a e di e en ial ises o 83
pe cen age poin s o low-income households whose main banking ela ionship is wi h a
mo e exposed bank, i.e., one in he 75 h pe cen ile o he NCFR dis ibu ion. In con as ,
mo gage c edi is la gely una ec ed by he in low o o eign bank unds. These e ec s
emain obus when accoun ing o a ange o ixed e ec s and household cha ac e is ics.
Addi ionally, we obse e a weakly posi i e shi in consume lending owa ds younge and
mig an households. The g ow h in c edi is d i en by he ex ensi e ma gin, i.e., by loans o
households who had no p e iously bo owed om exposed banks be o e he o eign in low
shock. We u he es ablish ha he inc ease in lending o low-income households is mos
p onounced o poo ly capi alized banks, which imp o e hei p o i abili y as a esul o
he c edi expansion. Mic o da a om a g oup o eu o a ea coun ies con i m ou indings
in a b oade sample o households. Second, we show ha lowe -income households whose
p ima y banking ela ionship is wi h a bank wi h g ea e dependence on non-co e unding
inc ease hei consump ion expendi u es: Households in he 25 h pe cen ile o he income
dis ibu ion inc ease hei non-du able expendi u es by 28.7% ela i e o hose in he 75 h
pe cen ile. Thi d, using con iden ial supe iso y bank da a, we p o ide a bluep in o he
p ecise channel h ough which o eign bank in lows each Ge man egional banks and hei
household cus ome s. We es ablish ha Ge man banks wi h a highe NCFR be o e he
o eign bank in low shock expe ience a ela i e ise in non-co e unding olumes a e he
shock. Di ec deposi s by non-a ea banks a egional Ge man banks g ow app oxima ely i e
imes as e han indi ec deposi s, whe e o eign banks make deposi s a la ge Ge man banks
ha pass on ”excess liquidi y” o smalle egional banks. Non-eu o a ea banks hus di ec ly
each smalle Ge man banks ia he in e bank ma ke , bu his mechanism is ein o ced
h ough a ickle-down e ec om la ge Ge man banks.
Toge he , hese esul s p o ide new e idence ha o eign bank in lows a e quan i a i ely
impo an o household lending in ad anced economies like Ge many, no only in eme ging
and de eloping ma ke s. Fo eign bank lows ope a e h ough bo h he in e na ional ne wo k
4
o la ge global banks (Ce o elli and Goldbe g,2012,Co ea e al.,2021), as well as ia
egional banks dependen on non-co e unding.
We add ess se e al po en ial h ea s o ou analysis and iden i ica ion s a egy, demon-
s a ing ha ou indings a e highly obus . Fi s , ou exposu e measu e, he NCF a io,
may no be andomly dis ibu ed ac oss banks and could co ela e wi h bank con ols, po en-
ially biasing ou es ima es. We he e o e include a la ge se o bank con ols and addi ionally
in e ac hem wi h household cha ac e is ics. Ou esul s emain quan i a i ely and quali a-
i ely unchanged. Second, ou indings could depend on he speci ic g oss exposu e measu e
we chose. To check his, we e un ou eg essions using banks’ ne exposu e o non-co e und-
ing lows and ind ou esul s a e una ec ed. Thi d, ou analysis assumes ha households
bo ow p ima ily om hei main ela ionship bank. To dispel any conce n ha ou esul s
could be d i en by an unobse ed shi o inc easingly impo an online banks, we e- un
ou main eg essions on a sub-sample o he mos loyal bank cus ome s. The esul s con i m
he obus ness o ou indings o households wi h igh ly de ined banking ela ionships.
Finally, we conduc a placebo es using da a om a pe iod wi hou any subs an ial
change in c oss-bo de bank lows. We ind no shi in mo e exposed banks’ consume
lending o low-income households du ing ha pe iod. Simila ly, we es ima e ou eg essions
wi h placebo ou comes, such as changes in households’ income o ne wo h, o use he sha e
o angible ixed asse s o e o al asse s as a placebo bank-le el exposu e a iable. In all o
hese eg essions, ou coe icien s o in e es u n s a is ically insigni ican , p o iding indi ec
e idence in suppo o he pa allel end assump ion.
We con ibu e o ou s ands o li e a u e. Fi s , a s and o esea ch shows ha (eme g-
ing economy) banks ha e a highly p ocyclical access o non-co e unding om global capi al
ma ke s (Gio anni e al.,2021) and, when mo e dependen on NCF, aise hei loan sup-
ply in esponse o o eign NCF in lows (Baskaya e al.,2017).1Te Kaa (2021) shows ha
1Sa mien o (2022) s udies he ape an um episode and shows ha Colombian i ms expe ienced a
wo sening o c edi access om banks ecei ing unding om ab oad. Using bank-le el da a, Knee and
Raabe (2019) show ha highe capi al a ec s lending by UK banks.
5
Table 1 Summa y S a is ics o Ge man Households / Banks
Va iable Obse a ions Mean SD 5 h 95 h
∆Mo gages 1,536 -15.08 415.86 -1012.67 999.88
∆Consume loans 1,536 -31.12 396.71 -851.74 829.43
Consump ion(non-du able) 1,536 9.26 0.73 8.19 10.31
Consump ion(du able) 1,468 9.79 1.19 8.19 11.09
Consump ion( ood) 1,536 8.53 0.56 7.62 9.39
Consump ion( es au an ) 1,536 6.46 2.12 0.00 8.34
Ln(Nonco e) 14,615 11.26 2.10 8.01 14.61
Ln(In e bank) 14,615 11.18 2.04 8.00 14.51
ROA 13,524 0.04 2.48 0.00 0.42
ROE 13,524 1.89 16.99 0.00 6.64
Ne weal h 1,536 12.05 1.87 8.22 14.31
Income 1,536 10.85 0.75 9.61 11.95
Ren e 1,536 0.31 0.46 0.00 1.00
Age 1,536 59.71 14.30 32.00 80.00
Fo eign 1,536 0.06 0.24 0.00 1.00
Income Exp. 1,536 0.08 0.27 0.00 1.00
Unemployed 1,536 0.29 0.45 0.00 1.00
Sel -Employed 1,536 0.18 0.38 0.00 1.00
Non-Co e 1,536 13.47 5.84 5.13 23.77
G oss In e bank 1,536 12.54 5.65 4.54 21.65
G oss Domes ic In e bank 1,536 0.02 0.98 -1.41 1.63
G oss EA In e bank 1,536 0.02 1.02 -0.38 1.98
G oss Non-EA In e bank 1,536 -0.02 0.36 -0.08 0.10
Ne In e bank 1,536 4.93 7.72 -8.42 16.86
Size 1,536 14.46 1.17 12.64 16.22
ROA 1,534 0.15 0.08 0.02 0.28
Equi y 1,536 5.67 1.02 4.02 7.55
Liquidi y 1,536 1.40 0.43 0.85 2.32
No e. The able epo s summa y s a is ics o he Ge man bank-household da a se . The i s block o
a iables a e he ou come a iables in he di e en s ages o he analysis a ei he he household o bank
le el. The second one con ains he household-le el con ols and he hi d block con ains he bank con ols,
bo h o which a e ixed in he yea 2014. Household-le el da a come om he PHF and span h ee pe iods:
2010-2011, 2014, and 2017. We p o ide da a de ini ions and sou ces in Table A1.
2.1 using hese da a wi h wo excep ions. Fi s , Finland and F ance p o ide da a o only
wo o he su ey wa es. We he e o e use he log o c edi olumes ins ead o log-di e ences
as he ou come a iables o a oid educing ou sample size. Second, he HFCS con ains
income expec a ions only in he hi d su ey wa e, which p e en s us om inco po a ing
his a iable in o ou di e ence-in-di e ences eg essions.
To accoun o he a ying in ensi y o c oss-bo de bank lows ac oss coun ies in he eu o
a ea, we ma ch he Eu opean household da a wi h agg ega e c oss-bo de bank low da a
12
ob ained om he BIS Loca ional Banking S a is ics (BIS-LBS). This measu e is compu ed
as he FX - and b eak-adjus ed change in a coun y’s banking sec o liabili ies is-a- is banks
in all o he coun ies, ne o he co esponding change in he banking sec o ’s o eign asse s,
as a ac ion o nominal GDP.10
Table 2p esen s he summa y s a is ics o he Eu opean sample. The wo c edi a iables
(in logs) ha e mean alues o 2.3 and 3.3, wi h mo gages showing g ea e s anda d de ia ion
han consume loans. He e ogenei y ac oss households is mo e p onounced in ne weal h han
in income. The a e age age o household heads in he sample is 57 yea s, app oxima ely 10%
o households hold o eign ci izenship, and one i h a e en e s. Finally, he a io o ne bank
in lows o e GDP has an a e age alue o 0.6%, anging om -1.4 o 7.0% be ween he 5 h
and 95 h pe cen ile. Table A3 u he p o ides sepa a e summa y s a is ics o mo e- and
less-exposed coun ies.
Table 2 Summa y S a is ics o Eu opean Households
Va iable Obse a ions Mean SD 5 h 95 h
Ln(ConsLoans) 34,980 2.3 4.0 0.0 10.1
Ln(Mo gages) 34,980 3.3 5.1 0.0 12.2
Ne weal h 34,980 12.1 1.9 8.3 14.6
Income 34,980 10.6 0.9 9.2 12.0
Ren e 34,980 0.2 0.4 0 1
Age 34,980 57.1 15.3 31 81
Fo eign 28,270 0.1 0.3 0 1
Bank lows 34,980 0.6 2.9 -1.4 7.0
No e. The able epo s summa y s a is ics o he sample om he Eu opean Cen al Bank’s Household
Finance and Consump ion Su ey (HFCS). This con ains household-le el da a om 22 Eu opean coun ies.
The da a se spans h ee pe iods: 2009-2011, 2013-2014, and 2016-2018. We exclude da a om coun ies ha
do no conduc he su eys as a panel. Ou inal eg ession sample comp ises da a om Belgium, Cyp us,
Ge many, Finland, F ance, I aly and Spain. The summa y s a is ics a e epo ed o all households ha a e
included in Table 7, column (1). We p o ide da a de ini ions and sou ces in Sec ion 2.2.
10B eak-adjus ed means ha he BIS co ec s c oss-bo de lows o b eaks in he epo ing popula ion
and/o epo ing me hodology.
13
3 C oss-Bo de Bank Flow Dynamics
Figu e 1illus a es he dynamics o eu o a ea ne c oss-bo de capi al lows, measu ed as he
nega i e o he cu en accoun , and disagg ega ed in o ne FDI, ne po olio in es men ,
and ne o he in es men in lows. The la e ca ego y p ima ily consis s o c oss-bo de
in e bank c edi . The igu e shows ha o e all capi al lows we e pe sis en ly nega i e
be ween 2011 and 2019. Howe e , a e he ECB’s implemen a ion o a nega i e in e es
a e policy in 2014:Q2 and i s QE p og am in 2015:Q1, po olio in lows as a pe cen age
o GDP declined and u ned nega i e, while o he in es men in lows, including in e bank
in lows, inc eased subs an ially. These dynamics e lec ha , as o eign in es o s sold eu o-
denomina ed go e nmen bonds o accommoda e he ECB’s asse pu chase p og am (Be gan
e al.,2020), he e enues om hose asse sales p o ided new unds o eu o a ea banks.
Figu e 1 The Eu o A ea Financial Accoun
No e. This igu e shows he eu o a ea inancial accoun , wi h he solid line depic ing o al ne capi al
in lows ( he nega i e o he cu en accoun ), and he ba s ep esen ing po olio in es men , FDI, and o he
in es men in lows, espec i ely, all in ne e ms and as a pe cen age o eu o a ea GDP. The low a iables
a e smoo hed by using ou -qua e mo ing a e ages be o e di iding by GDP. The e ical lines ma k he
implemen a ion o nega i e a es in 2014:Q2 and o he ECB’s QE p og am in 2015:Q1. Sou ces: BIS, ECB
and FRED. See Da a Appendix o de ails.
Figu e 2, Panel A, shows ha b eaking down he inancial accoun using BIS-LBS da a
and ocusing on ne c oss-bo de bank in lows p oduces a simila pa e n o highe in lows
14
o banks loca ed in he eu o a ea. When we spli ne bank in lows in o g oss in lows and
ou lows, Panel B indica es ha a change in g oss in lows was d i ing he inc ease in ne
lows, i.e., banks loca ed ou side o he eu o a ea expanded hei in e bank lending o banks
wi hin he eu o a ea. Panel C demons a es ha coun ies in he co e o he eu o a ea we e
he ecipien s o he g owing in lows in 2015-17. This sugges s ha o eign in es o s mainly
p o ided c oss-bo de unds o banks in he no he n eu o a ea, which we e pe cei ed as sa e
a he ime. In Ge many, bank in lows inc eased signi ican ly, ising om -6% in 2013 o +4%
in 2016 (Panel D). In ou main eg ession speci ica ions, we le e age his sha p inc ease in
Ge man bank in lows in a di e ence-in-di e ences se ing ha exploi s he a ying in ensi y
wi h which hese lows a ec di e en banks. Fo ex e nal alida ion, we use eu o a ea da a
o le e age c oss-coun y a ia ion in bank in lows as documen ed in Panel C.
4 Empi ical Speci ica ion
4.1 Ge man Benchma k Speci ica ion
In ou benchma k speci ica ion, we use Ge man su ey da a o iden i y he e ec o c oss-
bo de bank lows on banks’ lending o households. This is achie ed by es ima ing a
di e ence-in-di e ences model ha exploi s he inc ease in c oss-bo de bank lows in o Ge -
many a e he ECB’s implemen a ion o i s nega i e in e es a e policy and QE p og ams
in 2014/15. Ou eg essions a e speci ied as ollows:
∆Yh,b, =α +αh+β·(Pos ×Xh,2014) + ϵh,b, ,(1)
whe e Y ep esen s he loga i hm o ei he o al mo gage o o al consume loans o house-
hold h bo owing om bank b. The key a iable o in e es is he in e ac ion be ween he
Pos -dummy, which equals one o he su ey wa e ollowing he eco e y o bank lows (wa e
3) and ze o o he wise, and a ious p e-in low household cha ac e is ics. These con ols in-
15
Figu e 2 Bank Flows in he Eu o A ea
Panel A: Ne Bank In lows - En i e Eu o A ea Panel B: G oss Bank Flows - En i e Eu o A ea
-4 -2 0 2 4
2013q1 2014q1 2015q1 2016q1 2017q1 2018q1
-6 -4 -2 0 2
2013q1 2014q1 2015q1 2016q1 2017q1 2018q1
G oss In lows G oss Ou lows
Panel C: Bank In lows - Co e s Pe iphe y Panel D: Bank In lows - Ge many
-6 -4 -2 0 2 4
2013q1 2014q1 2015q1 2016q1 2017q1 2018q1
Co e Pe iphe y
-6 -4 -2 0 2 4
2013q1 2014q1 2015q1 2016q1 2017q1 2018q1
No e. This igu e depic s he dynamics o ne c oss-bo de bank in lows in he eu o a ea (Panel A), i s
b eakdown in o g oss in lows and ou lows (Panel B), ne in lows sepa a ely o coun ies in he pe iphe y
(Cyp us, G eece, I eland, I aly, Po ugal, Spain) s co e (all o he coun ies) o he eu o a ea (Panel C),
and o Ge many only (Panel D). Bank lows a e scaled by nominal GDP and hen smoo hed by compu ing
ou -qua e mo ing a e ages. The e ical lines ma k he implemen a ion o nega i e a es in 2014:Q2 and
o he ECB’s QE p og am in 2015:Q1. Sou ces: F ed and BIS-LBS
16
clude he loga i hm o household income, as we a e pa icula ly in e es ed in whe he bank
in lows induce an inc eased c edi alloca ion owa ds low-income, iskie households. Addi-
ional household cha ac e is ics a e included as con ols, in e ac ed wi h he Pos -dummy,
o cap u e hei po en ial e ec s on lending. Equa ion (1) also con ains household and
ime ixed e ec s, deno ed by αhand α , o con ol o unobse ed household-speci ic, ime-
in a ian cha ac e is ics and agg ega e condi ions ha equally impac all households. The
s anda d e o s he e and in he ollowing speci ica ion a e he e oskedas ici y- obus , bu
clus e ing hem a he coun y le el leads o consis en esul s (no epo ed).
In a second s ep, ou benchma k speci ica ion, we expand he eg ession by inco po a ing
a iple in e ac ion e m in ol ing he in e ac ion be ween he Pos -dummy, he a ious
household cha ac e is ics, ixed a hei p e- ea men le els, and a bank’s ini ial NCFR.
The expanded equa ion akes he ollowing o m:
∆Yh,b, =α +αh+γ·(Pos ×Non-co eb,2014) + σ·(Pos ×Xh,2014) +
ν·(Non-co eb,2014 ×Xh,2014) + ω·(Pos ×Xh,2014 ×Non-co eb,2014) + ϵh,b, .(2)
This will be ou p e e ed speci ica ion because i enables us o explo e whe he c oss-bo de
bank lows induce mo e exposed banks o exhibi a heigh ened isk appe i e in hei lending
p ac ices owa ds households, whe e exposu e is measu ed by banks’ NCFR. This ollows
Baskaya e al. (2017), who show o Tu key ha banks wi h g ea e NCFRs a e mo e a ec ed
by c oss-bo de lows han hose elian on cus ome deposi s. We he e o e hypo hesize
ha he coe icien ωwill be nega i e, i.e., banks which a e expec ed o bene i mo e om
he upswing in c oss-bo de bank lows will inc ease hei lending o lowe -income ( isky)
households ela i e o o he banks.
In ou mos sa u a ed model speci ica ion, we include no only household and ime ixed
e ec s, bu also bankg oup-loca ion-income- ime ixed e ec s. He e, “bankg oup” e e s o
whe he a household’s main ela ionship bank is a sa ings o coope a i e bank, “loca ion”
17
ep esen s one o he 401 adminis a i e Ge man egions, “ ime” co esponds o he su ey
wa e, and “income” deno es he decile o he household income dis ibu ion. These ixed
e ec s align wi h Deg yse e al. (2019), who show ha indus y-loca ion-size- ime ixed
e ec s con ol o loan demand in bank- i m ela ionships in a simila manne o Khwaja-
Mian’s i m- ime ixed e ec s (Khwaja and Mian,2008). Simila ly, ou bankg oup-loca ion-
income- ime ixed e ec s in end o abso b any he e ogenei y ha is speci ic o a clus e o
households in a ce ain egion, wi h a ce ain bank g oup p e e ence, o a speci ic income, a
a pa icula poin in ime. By con olling o he bulk o households’ changes in loan demand,
ou es ima ion will iden i y shi s in c edi supply ollowing c oss-bo de bank in lows.
The cen al assump ion unde lying he di e ence-in-di e ences eg essions is ha , in he
absence o c oss-bo de bank lows, banks wi h a highe non-co e dependence would ha e
exhibi ed he same end in lending beha io as banks wi h a lowe dependence. To alida e
his assump ion, as a i s s ep, Figu e 3shows he ime se ies dynamics o he loga i hm o
consume c edi — he ou come a iable we ind mos a ec ed by c oss-bo de bank lows—
o ou dis inc bank-household combina ions: mo e (less) exposed banks and low (high)
income households. As becomes clea om Figu e 3, p io o he inc ease in bank in lows
s a ing in 2015, mo e exposed banks, i.e., hose wi h a NCFR in he uppe 67% o he
dis ibu ion, and less exposed banks (below he 33 d pe cen ile) ollowed he same ends
in lending o low-income households. A e he inc ease in bank in lows, mo e exposed
banks inc ease hei consume lending o hese households, while less exposed banks did no .
Fo high-income households, we see simila consume c edi dynamics independen o bank
in lows and bank exposu e. In Sec ion 5.3, we will also es ima e a p ope placebo eg ession
on a sample pe iod wi hou a su ge in bank lows. When doing so, ou benchma k esul s
disappea , p o iding u he e idence in suppo o he pa allel end assump ion.
Fo he di e ence-in-di e ences es ima es o be unbiased, he ea men s a us should be
assigned andomly. When his condi ion is no sa is ied, o example because banks’ non-co e
a ios a e co ela ed wi h o he bank co a ia es, p ope ly con olling o hese co a ia es will
18
Figu e 3 Pa allel T ends Be o e he Bank In low Shock
100
200
300
400
500
600
700
123
exposed - low income exposed - high income nonexposed - low income nonexposed - high income
No e. This igu e shows he agg ega e log o consume c edi olumes in ou Ge man inal PHF sample
o ou dis inc bank-household combina ions: (i) low-income households (lowes 50%) and exposed banks
( op 67% o non-co e a ios); (ii) low-income households and less exposed banks (lowes 33%); (iii) high-
income households (uppe 50%) and exposed banks; (i ) high-income households and less exposed banks.
The e ical line depic s he s a o c oss-bo de bank lows in o Ge many. Sou ces: PHF, Bundesbank
Supe iso y Da a.
sa is y he condi ional mean ze o assump ion and ensu e unbiased es ima es (Robe s and
Whi ed,2013). The e o e, we include a b oad se bank co a ia es ixed a hei p e-in low
wa e 2 le el in e ac ed wi h he Pos -dummy and he household cha ac e is ics in ma ix
X. We show la e on ha he inclusion o hese in e ac ions has a negligible impac on ou
baseline es ima es, sugges ing ha non- andom ea men alloca ion does no jeopa dise ou
iden i ica ion.
4.2 Ex e nal Validi y: Eu o A ea Da a
To es ablish ex e nal alidi y o ou esul s, we also use da a o nea ly 18,000 households
om se en eu o a ea coun ies: Belgium, Cyp us, Finland, F ance, Ge many, I aly, and
Spain. As desc ibed in Sec ion 2, he Eu opean da a do no allow o a linkage be ween
households and indi idual banks. This limi a ion p e en s us om condi ioning he link
19
be ween c oss-bo de bank lows and household c edi on banks’ exposu e o such lows,
which weakens iden i ica ion in his pa o he analysis.
Ins ead, hese speci ica ions use c oss-coun y a ia ion in he in ensi y o bank in lows.
Speci ically, we es ima e he ollowing eg ession:
Log(Yh,c, ) = α +αh+ζ·(Pos ×Bank In lowsc,2016/17) + κ·(Pos ×Xh,2014) +
τ·(Pos ×Xh,2014 ×Bank In lowsc,2016/17) + ϵh,c, ,(3)
whe e Log(Y) is he loga i hm o mo gage o consume c edi . In he eu o a ea eg essions,
we de ine he ou come a iables in log-le els ins ead o i s di e ences, as in he benchma k
eg essions. The la e app oach equi es da a om a leas h ee su ey wa es and lead o
he exclusion o 6,000 obse a ions om Finland and F ance. The ma ix X, which con ains
con ol a iables, includes all a iables o he Ge man benchma k eg essions, excep o
income expec a ions, which is missing in wa es one and wo o he HFCS su ey. The
Pos -dummy equals one o su ey wa e h ee and ze o o he wise.
Because we canno lean on his o ical bank-le el exposu es o capi al in lows, a key di -
e ence in he eu o a ea eg essions is ha changes in c edi now depend on a coun y’s
ne c oss-bo de bank in lows as a sha e o GDP du ing 2016-17. Consequen ly, he esul s
om hese eg essions should be ea ed as complemen a y a he han causal e idence.
A he coun y-household le el, we expec ha la ge bank in lows will also be associa ed
wi h a s onge shi in c edi owa ds iskie households. The eg essions include household
and wa e ixed e ec s o con ol o he e ogenei y ac oss households and o e ime. Some
speci ica ions add coun y-wa e ixed e ec s o be e abso b loan demand shi s ollowing
c oss-bo de bank lows. S anda d e o s a e clus e ed a he coun y-wa e-le el.
Impo an ly, he c oss-coun y, c oss-household eg essions help es ablish ha ou bench-
ma k esul s o Ge many a e no solely d i en by he adop ion o nega i e a es o QE.
Ins ead, he esul s highligh he ole o ela i e changes in c oss-bo de bank lows. Bo h
20
mone a y policy ins umen s we e se equally ac oss all eu o a ea coun ies. The c oss-
coun y eg essions enable us o disen angle bank in low e ec s om mone a y policy and
in es iga e o wha ex en only coun ies encoun e ing bank in lows expe ienced changes in
he alloca ion o household-le el c edi , as we expec om he bank-household esul s o
Ge man households.
5 Empi ical Findings: C edi Alloca ion
5.1 Benchma k Resul s o Ge man Households
He e, we p esen ou benchma k esul s co esponding o Equa ions 1and 2. In Table 3,
columns (1)-(2), we p esen he esul s o mo gage and consume loans in speci ica ions
ha in e ac he Pos -dummy solely wi h household income, o now dis ega ding banks’
di e en ial exposu es o c oss-bo de bank lows. A e he bank in low shock, low-income
households expe ience an inc ease in consume c edi , while hei mo gage c edi olumes
emain una ec ed. In columns (3)-(4), we accoun o bank he e ogenei y by in e ac ing
he Pos -dummy no only wi h household income, bu also wi h he main bank’s p e-shock
NCFR. Consis en wi h ou expec a ions, he double in e ac ion be ween bank exposu e
and he Pos -dummy is posi i e and s a is ically signi ican a he 1% le el. Con e sely,
he iple in e ac ion e m has a nega i e and s a is ically signi ican coe icien a he 1%
le el, indica ing ha mo e exposed banks inc ease consume lending disp opo iona ely o
low-income households. In his iple in e ac ion model, he coe icien s on he pos -income
double in e ac ion a e no di ec ly compa able o hose o he double in e ac ion model o
columns (1)-(2). When combining he di ec ea men e ec wi h he income in e ac ion
e m, we ind ha he ma ginal ea men e ec on consume c edi supply becomes nega i e
o annual income le els abo e 66,000 eu os, sligh ly exceeding he sample mean. Once again,
we ind no signi ican e ec o mo gage lending.
In columns (5)-(6), we un a ho se ace be ween he household income iple in e ac ion
21
Swi ching Beha io
As men ioned ea lie , we obse e a household’s main bank only in he wo p e-in low wa es.
Ou empi ical s a egy hus implici ly assumes ha an unobse ed ise in swi ching beha io
om egional o na ional banks, possibly wi h a g ea e lending capaci y, be ween wa es wo
and h ee does no d i e ou indings. Gene ally, Ge man households a e e y loyal o hei
banks; only 117 households, o 7%, change hei main bank be ween he i s and second
wa e. Thus, we also do no expec subs an ial swi ching beha io be ween he second and
hi d wa es. To mi iga e any esidual conce ns abou an unobse able swi ching e ec , we
es i households wi h a g ea e endency o swi ch in ”no mal” imes a e d i ing ou main
indings. In Table 5, we exclude all households ha changed hei main bank be ween wa es
one and wo and e-es ima e ou benchma k eg ession. Column (1) shows his educes he
size o ou measu ed e ec somewha bu main ains he signi icance o ou coe icien .
Al e na i e Con ols Fo C edi Demand
In columns (2) and (3), we conduc wo addi ional sensi i i y es s o ou benchma k indings
and con ol in a mo e g anula way o po en ial shi s in he demand o c edi . In column
(2), we exclude households ha we e unemployed be o e he capi al in lows om he es ima-
ion o Equa ion 2, while in column (3) we exclude sel -employed households. Unemployed
households a e mo e likely o ha e li le o no consume c edi ini i ally and may he e o e
be mo e inclined o expe ience a ise in c edi i hey, o example, gain employmen du ing
he pe iod o inc eased o eign bank in lows. Including such households in he benchma k
eg essions migh consequen ly il ou coe icien es ima e owa ds inding a signi ican e ec
on household c edi . Column (2) o Table 5shows ha his is an unwa an ed conce n as
he coe icien es ima e is mo e o less unchanged when unemployed households a e excluded
om he eg ession. Sel -employed households, con e sely, may expe ience a g ea e ise in
c edi du ing he pos -in low pe iod i he in lows boos ed gene al economic ac i i y and
inc eased c edi demand, e ec s no ully cap u ed by ou ixed e ec s.
28
Table 5 Robus ness & He e ogenei y: Bank Swi ching, C edi Demand, Rela ionship Leng h, C edi Type
(1) (2) (3) (4) (5) (6) (7) (8)
No swi che s No UI No sel -employed Age ≥30 Age ≥40 No s uden loans Fo mal c edi T iple bank in e ac ions
Pos ×Income 102.1 61.39 188.8∗160.0∗88.31 97.10 99.37 921.4
(88.29) (100.4) (101.3) (84.31) (86.97) (85.11) (85.25) (594.3)
Pos ×Non-Co e 172.1∗∗∗ 157.5∗∗ 202.9∗∗∗ 178.7∗∗∗ 150.1∗∗ 150.2∗∗ 152.4∗∗ 203.7∗∗∗
(64.04) (71.87) (68.99) (61.97) (68.23) (64.52) (64.65) (70.01)
Pos ×Income ×Non-Co e -11.56∗-15.25∗∗ -20.06∗∗∗ -18.65∗∗∗ -14.56∗∗ -15.15∗∗ -15.27∗∗ -17.73∗∗
(6.278) (7.244) (6.673) (5.884) (6.069) (5.921) (5.926) (6.986)
Household FE Yes Yes Yes Yes Yes Yes Yes Yes
Time FE Yes Yes Yes Yes Yes Yes Yes Yes
O he Bank In e ac ions No No No No No No No Yes
Obs 1,302 1,264 1,090 1,488 1,380 1,536 1,534 1,534
R20.311 0.306 0.308 0.295 0.308 0.313 0.313 0.328
No e. The dependen a iable is he household-le el change in he loga i hm o consume c edi olumes. These eg essions a e based on he PHF
da a. Bank exposu e a iables o igina e om BISTA and GuV. The main eg esso s a e he iple in e ac ions be ween a Pos -dummy equal o one
o he hi d wa e o he PHF su ey and ze o o he wise, bank-le el NCFRs measu ed in wa e 2, and he ollowing household-le el cha ac e is ics
ixed a he wa e 2 alue: log o income, log o ne weal h, a dummy measu ing whe he a household en s he main esidence, age o he household
head, a dummy o o eign ci izenship, and income expec a ions. In column (1), we d op households ha swi ched hei main bank be ween wa e 1
and 2. Column (2) d ops unemployed, column (3) d ops sel -employed households. In columns (4) and (5), we d op households aged below 30 o 40,
espec i ely. Columns (6) and (7) use a igh e de ini ion o consume c edi , excluding s uden loans and loans om iends. In column (8), we con ol
o he co esponding iple in e ac ions be ween he Pos -dummy, he a o emen ioned household cha ac e is ics, and he ollowing addi ional bank
co a ia es: bank size, capi aliza ion, liquidi y, and e u n on asse s. Mos in e ac ion es ima es a e no displayed o conse e space. Da a de ails can
be ound in Table A1. The eg essions include ime and household ixed e ec s. He e oscedas ici y- obus s anda d e o s a e shown in pa en heses.
∗,∗∗ and ∗∗∗ indica e s a is ical signi icance a he 10%, 5%, and 1% le els, espec i ely.
29
Rela ionship Leng h
This could also il ou eg essions owa ds inding a signi ican e ec on lending o house-
holds. Column (3) con i ms ha when excluding sel -employed households, he coe icien
on he iple in e ac ion e m emains nega i e and signi ican a he 1% le el. As no ed
ea lie , ou bank-le el analysis assumes ha households ob ain hei loans om hei main
ela ionship bank. This assump ion is suppo ed by Pu i e al. (2017), who ind ha Ge -
man households ha e e y s ong ies wi h hei sa ings banks. They show ha o e 80 %
o loan applican s ha e been cus ome s o a leas i e yea s. P e ious bank–deposi o ela-
ionships also inc ease access o uncolla e alized c edi , such as consume loans. To p o ide
u he suppo o his no ion, we es ic ou sample and un sepa a e eg essions o ela-
i ely olde households, i.e., hose who a e likely o ha e longe -s anding bank ela ionships.
Columns (4) and (5) o Table 5con i m ha when we es ic ou sample o households aged
30 yea s o mo e, and 40 yea s o mo e, espec i ely, he coe icien es ima es a e nea ly he
same as in he benchma k eg essions.
Di e en C edi Types
Finally, we accoun o he ac ha some sou ces o c edi a e una ec ed by luc ua ions
in c oss-bo de bank unding. The PHF’s de ini ion o consume loans includes consume
ins allmen loans, bank o e d a s, c edi ca d deb , loans om iends o employe s, and
s uden loans. As he la e wo componen s a e independen o bank loan supply, we ede-
ine consume c edi mo e s ic ly by excluding loans om iends o employe s, and s uden
loans. Because he PHF, un o una ely, combines consume ins allmen loans and employe
loans in o one a iable, we exclude only s uden loans in column (6). In column (7), we hen
emo e households ha epo ha ing ob ained loans om hei employe . Nei he modi i-
ca ion a ec s he es ima ed coe icien , al hough he signi icance le el is sligh ly educed as
he sample size sh inks.
30
Con olling o Non-Random T ea men
A po en ial h ea o ou main eg essions is ha banks’ exposu e o c oss-bo de lows
may no be dis ibu ed andomly bu co ela es wi h o he bank cha ac e is ics. As Table
A2 shows, howe e , his is unlikely o be a majo conce n o ou analysis as bo h mo e
and less exposed banks sha e simila cha ac e is ics. Speci ically, bo h ypes o bank ypes
a e compa able in size, p o i abili y and capi aliza ion. Only liquidi y a ios seem o be
signi ican ly smalle o mo e exposed banks.
Ye , as explained in Sec ion 4.1, con olling o hese bank co a ia es inc eases he like-
lihood ha he condi ional mean ze o assump ion is sa is ied and ha we hence ob ain
unbiased es ima es (Robe s and Whi ed,2013). To his end, we un addi ional eg essions
ha con ol o he iple in e ac ions be ween a ich se o bank co a ia es, ixed a hei
p e-in low wa e 2 alues, he pos -dummy and ou household co a ia es. Column (8) o Ta-
ble 5shows ha hei inclusion changes nei he he size no he signi icance o ou coe icien
o in e es . While we do no epo he coe icien s o he addi ional in e ac ion e ms in
Table 5, mos o hem a e s a is ically insigni ican . We do ind, howe e , ha ollowing he
bank in low shock, be e capi alized banks inc ease consume lending o younge , high ne
wo h households and hose wi h o eign ci izenship.
5.3 Placebo Tes
In Sec ion 4.1 we pe o med an ini ial check o he pa allel ends assump ion. Figu e 3indi-
ca ed ha mo e and less exposed banks exhibi ed lending pa e ns up o 2014 and di e ged,
pa icula ly o lending o lowe income households, when bank lows in o Ge many began
inc easing in 2014. Ideally, ou su ey da a would con ain a long p e- ea men ime se ies
o each household o e i y i he pa allel ends assump ion is sa is ied. Gi en ha he
PHF da a span only h ee wa es, we ins ead add ess his limi a ion by conduc ing placebo
eg essions. The i s one es ima es equa ion 2on a p e-in low sample. Fo his, we e- un
ou benchma k eg ession, es ic ing he da a o he i s (2010-2011) and second (2014)
31
su ey wa es. Wi h only wo sample wa es, we canno compu e he ou come a iable in
log-di e ences, ins ead, we use he loga i hm o consume c edi as he dependen a i-
able. Column (1) o Table 6con i ms ha ha ou main esul s hold when we e- un he
benchma k eg ession wi h his al e na i e ans o ma ion o he c edi a iable. We hen
es ima e his eg ession speci ica ion on he p e-in low sample. Column (2) shows ha , in
his placebo eg ession, he di e ence in lending pa e ns be ween mo e and less exposed
banks disappea s. This inding p o ides addi ional suppo o he pa allel ends assump-
ion, sugges ing ha a ec ed and una ec ed banks ollowed simila lending pa hs be o e he
sudden ise in in e na ional bank in lows.
Nex , we pe o m wo addi ional se s o placebo eg essions o con i m ha a ec ed
and una ec ed banks displayed simila lending ends be o e he bank in low shock. Fi s ,
we un ou benchma k eg ession wi h he log-change in consume c edi as he ou come
a iable bu subs i u e he bank exposu e a iable wi h a placebo— he bank-le el sha e o
angible ixed asse s o e o al asse s. C oss-bo de bank in lows p o ide addi ional liquidi y
o banks dependen on non-co e unding, ega dless o hei asse s uc u e, and in pa icula
independen ly o he sha e o a bank’s angible asse s. We he e o e expec his eg ession o
p oduce insigni ican ea men e ec s. Column (3) o Table 6con i ms ha he consume
c edi supply by ”placebo- ea ed” and ”un ea ed” banks e ol es equally, u he alida ing
he pa allel p e- end assump ion.
Second, ins ead o using a placebo ea men a iable, we eplace he dependen a iable
wi h household-le el ou comes expec ed o be un ela ed o c oss-bo de bank lows. These
include g ow h in income, g ow h o ne wo h, changes in he sha e o s ocks in he asse
po olio, changes in he sha e o housing in o al asse s, and changes in housing enu e
s a us. Columns (4)-(8) show ha he iple in e ac ion coe icien on Pos x Income x
NCFR is s a is ically insigni ican o all hese eg essions. Households wi h ela ionships
wi h mo e o less exposed banks exhibi ed no di e ging dynamics in hese placebo ou comes,
p o iding u he suppo o he pa allel end assump ion.
32
Table 6 Placebo Tes s
(1) (2) (3) (4) (5) (6) (7) (8)
Benchma k Ln(ConsLoans) Placebo Ln(ConsLoans) ∆ Ln(ConsLoans) ∆ Ln(Income) ∆ Ln(Ne Wo h) ∆ S ocks ∆ Housing ∆ Tenu e
Pos ×Income 0.0301 -0.0729 37.17 19.60∗∗ -21.92 -0.306 3.206 -0.0959∗∗∗
(0.386) (0.500) (26.28) (8.014) (15.62) (0.37) (3.1) (0.0366)
Pos ×Tangible -162.3
(586.6)
Pos ×Income ×Tangible 32.72
(47.54)
Pos ×Non-Co e 0.427 0.180 -1.698 -6.521 0.49 -0.0647 -0.0004
(0.283) (0.307) (12.84) (17.18) (0.85) (4.993) (0.0482)
Pos ×Income ×Non-Co e -0.0453∗0.0163 0.5 2.443 -0.0769 0.161 -0.0026
(0.0275) (0.0322) (1.191) (1.534) (0.0731) (0.471) (0.0044)
Household FE Yes Yes Yes Yes Yes Yes Yes Yes
Time FE Yes Yes Yes Yes Yes Yes Yes Yes
O he Bank In e ac ions No No
Obs 2,910 1,958 1,536 1,494 1,468 1,536 1,536 1,536
R20.702 0.694 0.29 0.541 0.462 0.383 0.39 0.5
No e. These eg essions a e based on he PHF da a. Bank exposu e a iables o igina e om BISTA and GuV. The dependen a iable is he
household-le el change in he loga i hm o consume c edi olumes (columns 1, 2, and 3), he log-change in income (column 4), he log-change in
ne wo h (column 5), he change in he sha e o s ocks o e a household’s o al po olio alue (column 6), he change in he sha e o housing weal h
o e he o al po olio alue (column 7) and he change in a household’s housing enu e s a us (column 8). When a household epo s ze o s ock
o housing weal h, we se he po olio sha e equal o ze o. Housing enu e equals 1 when a household i s en s he main esidence and hen owns
i ; ze o when enu e s a us does no change; minus one when a household i s owns and hen en s i s main esidence. The main eg esso s a e he
iple in e ac ions be ween a Pos -dummy equal o one o he hi d wa e o he PHF su ey and ze o o he wise, bank-le el NCFRs (columns 1 and
2 and 4 o 8) o bank-le el angible ixed asse s o e o al asse s (column 3) measu ed in wa e 2, and he ollowing household-le el cha ac e is ics
ixed a he wa e 2 alue: log o income, log o ne weal h, a dummy measu ing whe he a household en s he main esidence, age o he household
head, a dummy measu ing whe he a household has a mig an backg ound, and income expec a ions. Mos in e ac ion es ima es a e no displayed o
conse e space. Da a de ails can be ound in Table A1. The eg essions include ime and household ixed e ec s. He e oscedas ici y- obus s anda d
e o s a e shown in pa en heses. ∗,∗∗ and ∗∗∗ indica e s a is ical signi icance a he 10%, 5%, and 1% le els, espec i ely.
33
5.4 Ex e nal Validi y: Eu o A ea Households
Thus a , we ha e es ablished ha Ge man households bene i ed om inc eased c oss-bo de
bank in lows. In his sec ion, we show ha ou main indings ha e ex e nal alidi y in a la ge
da a se o households om se en eu o a ea coun ies. As explained abo e, hese da a do
no con ain a link be ween households and hei banks. The e o e, we ocus on he e ec o
c oss-bo de bank in lows on c edi olumes, wi hou di e en ia ing be ween mo e and less
exposed banks. Ins ead, we measu e households’ exposu e by means o coun y-le el bank
in lows o e GDP, as displayed in Figu e 2.
In Table 7, we p esen e idence ha o he eu o a ea coun ies besides Ge many also
expe ienced a ise consume c edi o low-income households as c oss-bo de bank lows in o
hese coun ies g ew. In column (1), we es ima e Equa ion 3on he la ges possible da a
se , excluding he o eign ci izenship dummy missing o Spain. Consume c edi o low-
income households inc eases signi ican ly in coun ies ha expe ience g ea e c oss-bo de
bank in lows, as can be seen om he implied -s a is ics o he iple in e ac ion e m: Pos
x coun y-le el bank in lows x household income.
Ou coe icien o in e es emains consis en when we eplica e he Ge man eg ession
con ols as closely as possible by including he o eign ci izenship dummy (column 2) and
coun y- ime ixed e ec s (column 3). In column (4), we isola e he log-income iple in e -
ac ions while omi ing o he household in e ac ions, s ill ob aining a signi ican coe icien
es ima e. Simila ly, he esul s hold in column (5), whe e we exclude he 5,546 Ge man
households ha we e included in columns (1)-(4), hough in his case he iple in e ac ion
coe icien alls sligh ly below con en ional signi icance le els. Finally, in column (6), we use
he log o mo gage c edi olumes as he ou come a iable. Consis en wi h ou Ge man
benchma k esul s, we do no see a shi in mo gage c edi ac oss households.
34
Table 7 Resul s o he Eu opean Household Sample
(1) (2) (3) (4) (5) (6)
Ln(ConsLoans) Ln(ConsLoans) Ln(ConsLoans) Ln(ConsLoans) Ln(ConsLoans) Ln(Mo gages)
Pos ×Income -0.197∗∗ -0.134∗∗ -0.122∗∗ -0.089∗-0.170∗-0.059
(0.08) (0.05) (0.04) (0.04) (0.08) (0.01)
Pos ×Income ×Flows -0.034∗-0.027∗-0.035∗∗ -0.025∗∗∗ -0.026 -0.019
(0.02) (0.01) (0.01) (0.01) (0.02) (0.02)
Household FE Yes Yes Yes Yes Yes Yes
Time FE Yes Yes Yes Yes Yes Yes
Coun y-Time FE No No Yes No No No
Household Con ols ×Pos ×Flows Yes Yes Yes No Yes Yes
Obs 34,980 28,270 34,980 35,034 29,434 34,980
No. o Coun ies 7 6 7 7 6 7
R20.726 0.735 0.727 0.725 0.727 0.873
No e: The eg essions a e based on wa es 2 and 3 o he HFCS su ey. The dependen a iable in columns (1)-(5) is he loga i hm o consume loans.
In column (6), i is he loga i hm o mo gages. The main eg esso is coun y-le el ne bank in lows o e nominal GDP, a e aged du ing 2016-2017,
and in e ac ed wi h household-le el income measu ed in wa e 2 as well as a dummy equal o one a e he signi ican change in bank lows (wa e 3)
and ze o o he wise. All columns, apa om column (4), include ime and household ixed e ec s and he ollowing household con ols, measu ed in
wa e 2, in e ac ed wi h he Pos -dummy and coun y-le el bank lows: ne wo h, age, and a en e dummy. Only column (2) includes addi ionally
a dummy o o eign ci izenship. All hese in e ac ions, as well as all lowe -o de in e ac ions o he iple in e ac ions, a e included in all eg essions
unless hey a e abso bed by ixed e ec s, bu we supp ess hei coe icien s o sa e space. Column (3) addi ionally con ols o coun y- ime ixed
e ec s. S anda d e o s, clus e ed a he coun y- ime le el, a e shown in pa en heses. ∗,∗∗ and ∗∗∗ indica e s a is ical signi icance a he 10%, 5%,
and 1% le els, espec i ely.
35
Taken oge he , he esul s in Table 7p o ide e idence ha o he eu o a ea coun ies
exhibi ed a simila inc ease in consume c edi owa ds low-income households in esponse
o he in low o o eign bank unding. The esul s also demons a e ha he indings o
Ge man households we e no d i en by he ECB’s non-con en ional mone a y policy, as all
eu o a ea coun ies aced he same mone a y policy mix. Only eu o a ea coun ies wi h
g ea e bank in lows expe ienced a shi in consume c edi owa ds low-income households.
6 Mechanisms
In his sec ion, we iden i y he mechanisms unde lying ou esul s. We s a by examining
whe he he agg ega e bank in low shock indeed implies highe bank-le el non-co e unding
olumes o mo e ela i e o less exposed banks. Then we s udy o wha ex en ou esul s
a e d i en by egional banks ob aining in e bank liquidi y om ab oad di ec ly, o whe he
c oss-bo de in e bank liquidi y ickles down o egional Ge man banks h ough la ge banks.
Nex , we s udy he ex ensi e e sus he in ensi e ma gin o lending. Finally, we in es iga e
why banks especially aise hei consume loan supply o low-income, highe - isk households,
wi h a pa icula ocus on he ole o bank agency p oblems.
6.1 Mo e Exposed Banks Expe ience G ea e Funding In lows
Ou main eg ession speci ica ion implici ly assumes ha banks wi h highe ini ial non-co e
unding a ios a e mo e exposed o agg ega e bank in low shocks. In his sub-sec ion, we
e i y his assump ion by examining whe he hese banks indeed expe ience g ea e non-co e
unding in lows om ab oad ollowing he shock.
To in es iga e his, we eg ess he loga i hm o each bank’s o al non-co e unding olume
as well as he in e bank componen indi idually on he in e ac ion e m Pos x NCFR, using
he same sample pe iod as in he household eg essions. We include bank and yea ixed
e ec s and clus e s anda d e o s a he bank le el. As Table 8shows, banks wi h a
36
highe NCFR p io o he c oss-bo de bank in low shock indeed expe ience highe non-
co e unding in lows, ega dless o whe he measu ed as o al non-co e unding (column 1)
o as in e bank liabili ies (column 2). These e ec s a e app oxima ely wice as la ge when
ocusing exclusi ely on egional banks (columns 3-4), consis en wi h he household eg ession
indings. Toge he , hese esul s con i m ha banks classi ied as mo e exposed a e indeed
he ones expe iencing highe in lows o wholesale unds as a consequence o he c oss-bo de
in low shock.
Table 8 Do Non-Co e Volumes Inc ease o Mo e Exposed Banks?
All Banks Regional Banks
(1) (2) (3) (4)
Ln(Nonco e) Ln(In e bank) Ln(Nonco e) Ln(In e bank)
Pos ×Non-Co e 0.003∗∗∗ 0.003∗∗∗ 0.006∗∗∗ 0.005∗∗
(0.001) (0.001) (0.002) (0.002)
Bank FE Yes Yes Yes Yes
Time FE Yes Yes Yes Yes
Obs 14,212 14,212 11,735 11,735
R20.96 0.95 0.98 0.97
No e. The dependen a iable is he log-le el o a bank’s non-co e o in e bank unds, espec i ely. The da a
o igina e om BISTA and GuV and co e he pe iod 2010-17. The main eg esso is he double in e ac ion
be ween a Pos -dummy equal o one o he hi d wa e o he PHF su ey and ze o o he wise, and bank-le el
NCFRs measu ed in wa e 2. In columns (1) and (2), we include all banks in he analysis. Columns (3) and
(4) only include egional banks. Da a de ails can be ound in Table A1. Time and bank ixed e ec s a e
included. He e oscedas ici y- obus s anda d e o s clus e ed a he bank le el a e shown in pa en heses. ∗,
∗∗ and ∗∗∗ indica e s a is ical signi icance a he 10%, 5%, and 1% le el.
6.2 Di ec s Indi ec T ansmission
So a , we ha e shown ha egional banks dependen on non-co e unding inc ease hei
consume lending o low-income households. This could be d i en ei he by di ec access o
o eign wholesale liquidi y o by a ickle-down e ec , whe e la ge banks a ac c oss-bo de
bank in lows and pass on hei liquidi y ”su plus” o smalle banks. To disen angle bo h
e ec s we exploi he g anula i y o he supe iso y da a, which allows us o b eak down
37
We p oceed by spli ing he sample in o households bo owing om mo e exposed and less
exposed banks, i.e., hose wi h highe and lowe non-co e a ios. Columns (1)-(4) in Table
11 show ha households bo owing om less exposed banks do no inc ease consump ion.
Low-income households banking wi h mo e exposed c edi p o ide s do, howe e , show an
inc ease in hei non-du able consump ion, pa icula ly ood and be e ages consumed ou side
he home (columns (5)-(8)), al hough he coe icien s in columns (1)-(4) a e no s a is ically
di e en om hose in columns (5)-(8). Du able consump ion by low-income households, on
he o he hand, is no a ec ed by bank in lows. The e ec on non-du able consump ion is no
only s a is ically, bu also economically signi ican . A e he shock, a low-income household,
i.e., one in he 25 h pe cen ile o he income dis ibu ion, has a 28.7% highe consump ion
o non-du ables ela i e o hei p e-in low consump ion, and ela i e o a highe -income
household, i.e. one in he 75 h pe cen ile o he income dis ibu ion.
O e all, hese indings p o ide aluable insigh s in o he e ec s o in e na ional capi al
lows. While c oss-bo de bank in lows ha e been shown o bea he po en ial o inancial
ins abili y isks h ough sudden inc eases in lending, ou analysis highligh s hei ole in
elaxing c edi cons ain s o poo e households wi h p e iously unme demand o c edi .
The imp o emen in hei access o c edi ansla es exclusi ely in o a g ow h o sho e -
e m consume c edi , which hese households use o aise non-du able consump ion, a mo e
ansi o y o m o expendi u e.
8 Conclusions
We s udy he e ec s o c oss-bo de capi al lows on egional Ge man banks’ isk- aking and
hei c edi supply o households. We employ g anula ma ched bank-household da a and
es ablish ha c oss-bo de bank in lows induce egional banks wi h a g ea e non-co e und-
ing dependency o inc ease hei uncolla e alized lending o iskie , lowe -income households.
Howe e , we do no obse e any inc ease in isk- aking in banks’ mo gage lending.
44
When in es iga ing h ough which channels o eign unding lows a ec lending, we ind
ha he ise in c edi by egional Ge man banks occu ed h ough unding in lows om
p ima ily non-eu o a ea banks and o a lesse ex en h ough in e bank unding om o he
Ge man banks. Consis en wi h he p esence o a isk- aking channel simila o ha in
ea lie esea ch on he ansmission o mone a y policy, we es ablish ha wo se capi alized
banks a e esponsible o he ise in c edi , while be e capi alized banks show no g ow h
in household lending. We u he demons a e ha his c edi expansion occu s h ough
he ex ensi e ma gin. Finally, as access o c edi imp o es, lowe -income households who
a e clien s o mo e exposed banks inc ease hei consump ion expendi u es, especially on
non-essen ial non-du ables. We es ablish he ex e nal alidi y o ou main esul s using
c oss-coun y household da a om almos 18,000 households in he eu o a ea.
While p e ious esea ch has shown ha c oss-bo de capi al in lows aise banks’ lending
o isky i ms, we p o ide new household-le el e idence ha a simila isk- aking e ec exis s
in banks’ household lending. We also documen ha c oss-bo de capi al lows gene a e
la ge luc ua ions in c edi supply h ough smalle egional banks in Ge many, an ad anced
economy and he la ges membe s a e o he eu o a ea.
O he esea ch has ecen ly demons a ed ha pa icula ly c edi booms in he household
sec o can lead o boom-bus cycles and p edic inancial c ises. A ise in c edi may hus
aise inancial s abili y isks. A he same ime, g ea e access o c edi allows lowe -income
households o inc ease consump ion and he e o e educe consump ion inequali y, a leas in
he sho un. In he longe un, howe e , poo e households will ace inc eased deb le els.
O e all, ou analysis highligh s he ade-o s policymake s ace when o eign capi al in-
lows in he in e bank ma ke lead o luc ua ions in he a ailabili y o c edi . A comple e
assessmen o he long- e m e ec s o c oss-bo de capi al in lows on (consump ion) inequal-
i y and a g anula unde s anding o he mechanisms behind hese e ec s equi es u he
esea ch.
45
Re e ences
Acha ya, Vi al and Hassan Naq i, “The seeds o a c isis: A heo y o bank liquidi y
and isk aking o e he business cycle,” Jou nal o Financial Economics, 2012, 106 (2),
349–366.
Al a illa, Ca lo, Luc Lae en, and Jos´e-Luis Peyd ´o, “Mone a y and Mac op uden ial
Policy Complemen a i ies: e idence om Eu opean c edi egis e s,” ECB Wo king Pape
No. 2504, 2020.
Ame ican Exp ess, “A e age C edi Sco es by Age, S a e, and Income,” Technical Repo
Decembe 2022.
Baskaya, Yusu Sone , Julian di Gio anni, S¸ebnem Kalemli- ¨
Ozcan, Jos´e-Luis
Peyd ´o, and Mehme Fa ih Ulu, “Capi al lows and he in e na ional c edi channel,”
Jou nal o In e na ional Economics, 2017, 108, S15–S22.
Bee , Rachael, Felicia Ionescu, and Geng Li, “A e Income and C edi Sco es Highly
Co ela ed?,” FEDS No es 2018-08-13-1, Boa d o Go e no s o he Fede al Rese e Sys-
em (U.S.) Augus 2018.
Bellema e, Ma c F and Casey J Wichman, “Elas ici ies and he in e se hype bolic
sine ans o ma ion,” Ox o d Bulle in o Economics and S a is ics, 2020, 82 (1), 50–61.
Be gan , Ka ha ina, Michael Fido a, and Ma in Schmi z,In e na ional capi al lows
a he secu i y le el: e idence om he ECB’s asse pu chase p og amme, In e na ional
Mone a y Fund, 2020.
Blanc, Julia Le and Tobias Schmid , “Es ima ing household consump ion using weal h
su ey da a,” Mimeo, Deu sche Bundesbank, 2018.
Caballe o, Juli´an A, “Do su ges in in e na ional capi al in lows in luence he likelihood o
banking c ises?,” Economic Jou nal, 2016, 126 (591), 281–316.
46
Campbell, John Y. and Joao F. Cocco, “A Model o Mo gage De aul ,” Jou nal o
Finance, 2015, 70 (4), 1495–1554.
Ce o elli, Nicola and Linda S. Goldbe g, “Banking Globaliza ion and Mone a y T ans-
mission,” Jou nal o Finance, 2012, 67 (5), 1811–1843.
Co ea, Rica do, Teodo a Paligo o a, Ho acio Sap iza, and And ei Zla e, “C oss-
Bo de Bank Flows and Mone a y Policy,” Re iew o Financial S udies, 02 2021, 35 (1),
438–481.
Deg yse, Hans, Oli ie De Jonghe, Sanja Jako lje i´c, Klaas Mulie , and Glenn
Schepens, “Iden i ying c edi supply shocks wi h bank- i m da a: Me hods and applica-
ions,” Jou nal o Financial In e media ion, 2019, 40, 100813.
di Gio anni, Julian, S¸ebnem Kalemli- ¨
Ozcan, Mehme Fa ih Ulu, and Yusu Sone
Baskaya, “In e na ional Spillo e s and Local C edi Cycles,” Re iew o Economic S udies,
10 2021, 89 (2), 733–773.
Dinge , Vale iya and Daniel Ma cel e Te Kaa , “C oss-bo de capi al lows and bank
isk- aking,” Jou nal o Banking & Finance, 2020, 117, 105842.
Epu e, Mi cea, I ina Mihai, Camelia Minoiu, and Jos´e-Luis Peyd ´o, “Global inan-
cial cycle, household c edi , and mac op uden ial policies,” Fo hcoming in: Managemen
Science, 2024.
Ga be , Gab iel, A i Mian, Jacopo Pon icelli, and Ami Su i, “Household deb and
ecession in B azil,” in “Handbook o US Consume Economics,” Else ie , 2019, pp. 97–
119.
Gy¨ongy¨osi, Gy˝oz˝o, S e en Ongena, and Ibolya Schindele, “The impac o mone a y
condi ions on bank lending o households,” SFI Wo king Pape , 2024.
47
Hale, Galina, T¨ume Kapan, and Camelia Minoiu, “Shock T ansmission Th ough
C oss-Bo de Bank Lending: C edi and Real E ec s,” Re iew o Financial S udies, 12
2020, 33 (10), 4839–4882.
Holms om, Beng and Jean Ti ole, “Financial in e media ion, loanable unds, and he
eal sec o ,” Qua e ly Jou nal o Economics, 1997, 112 (3), 663–691.
Iye , Rajkamal and Jos´e-Luis Peyd ´o, “In e bank Con agion a Wo k: E idence om
a Na u al Expe imen ,” Re iew o Financial S udies, 04 2011, 24 (4), 1337–1377.
Jim´enez, Gab iel, S e en Ongena, Jos´e-Luis Peyd ´o, and Jes´us Sau ina, “Haz-
a dous imes o mone a y policy: Wha do wen y- h ee million bank loans say abou he
e ec s o mone a y policy on c edi isk- aking?,” Econome ica, 2014, 82 (2), 463–505.
Jonghe, Oli ie De, Hans Dewach e , Klaas Mulie , S e en Ongena, and Glenn
Schepens, “Some bo owe s a e mo e equal han o he s: Bank unding shocks and c edi
ealloca ion,” Re iew o Finance, 2020, 24 (1), 1–43.
Jo d`a, `
Osca , Mo i z Schula ick, and Alan M Taylo , “When c edi bi es back,”
Jou nal o Money, C edi and Banking, 2013, 45 (s2), 3–28.
, , and , “The g ea mo gaging: housing inance, c ises and business cycles,” Eco-
nomic Policy, 2016, 31 (85), 107–152.
Kaa , Daniel Te, “C oss-bo de deb lows and c edi alloca ion: Fi m-le el e idence om
he eu o a ea,” Jou nal o Money, C edi and Banking, 2021, 53 (7), 1797–1818.
Khwaja, Asim Ijaz and A i Mian, “T acing he impac o bank liquidi y shocks: E i-
dence om an eme ging ma ke ,” Ame ican Economic Re iew, 2008, 98 (4), 1413–1442.
Kinde mann, Fabian, Julia Le Blanc, Monika Piazzesi, and Ma in Schneide ,
“Lea ning abou housing cos : Su ey e idence om he Ge man house p ice boom,”
NBER Wo king Pape 28895, 2021.
48
Knee , Ch is iane and Alexande Raabe, “T acking o eign capi al: he e ec o capi al
in lows on bank lending in he UK,” Bank o England S a Wo king Pape No. 804, 2019.
Magud, Nicolas E, Ca men M Reinha , and Es eban R Vespe oni, “Capi al in-
lows, exchange a e lexibili y and c edi booms,” Re iew o De elopmen Economics,
2014, 18 (3), 415–430.
Ma inez-Mie a, Da id and Ra ael Repullo, “Sea ch o yield,” Econome ica, 2017,
85 (2), 351–378.
Maye , E ik J., “Big Banks, Household C edi Access, and In e gene a ional Economic
Mobili y,” Jou nal o Financial and Quan i a i e Analysis, 2023, p. 1–37.
Mian, A i , Ami Su i, and Emil Ve ne , “Household deb and business cycles wo ld-
wide,” Qua e ly Jou nal o Economics, 2017, 132 (4), 1755–1817.
M¨ulle , Ka s en and Emil Ve ne , “C edi alloca ion and mac oeconomic luc ua ions,”
Fo hcoming in: Re iew o Economic S udies, 2024.
Ongena, S e en, G¨unseli T¨ume -Alkan, and Na alja Von Wes e nhagen, “Do ex-
posu es o sagging eal es a e, subp ime, o condui s ab oad lead o con ac ion and ligh
o quali y in bank lending a home?,” Re iew o Finance, 2018, 22 (4), 1335–1373.
, Jos´e-Luis Peyd ´o, and Neel je an Ho en, “Shocks Ab oad, Pain a Home? Bank-
Fi m-Le el E idence on he In e na ional T ansmission o Financial Shocks,” IMF Eco-
nomic Re iew, 2015, 63 (4), 698–750.
Pu i, Manju, J¨o g Rocholl, and Sascha S e en, “Wha do a million obse a ions ha e
o say abou loan de aul s? Opening he black box o ela ionships,” Jou nal o Financial
In e media ion, 2017, 31, 1–15.
49
, J¨o g Rocholl, and Sascha S e en, “Global e ail lending in he a e ma h o he US
inancial c isis: Dis inguishing be ween supply and demand e ec s,” Jou nal o Financial
Economics, 2011, 100 (3), 556–578.
Rajan, Raghu am G, “Has inance made he wo ld iskie ?,” Eu opean inancial manage-
men , 2006, 12 (4), 499–533.
Robe s, Michael R and Toni M Whi ed, “Endogenei y in empi ical co po a e inance,”
in “Handbook o he Economics o Finance,” Vol. 2, Else ie , 2013, pp. 493–572.
Sa ie, Felipe, Liliana Va ela, and Kei-Mu Yi, “The mic o and mac o dynamics o
capi al lows,” Technical Repo , Na ional Bu eau o Economic Resea ch 2020.
Sa mien o, Miguel, “Sudden yield e e sals and inancial in e media ion in eme ging ma -
ke s,” Jou nal o Financial S abili y, 2022, p. 101050.
Schae e , M. and H. S ahl, “Mon hly Balance Shee S a is ics (BISTA), Da a Repo
2023-08 – Me ada a Ve sion BISTA-Doc- 5-0,” Mimeo, Deu sche Bundesbank, Resea ch
Da a and Se ice Cen e, 2023.
Schnabl, Philipp, “The In e na ional T ansmission o Bank Liquidi y Shocks: E idence
om an Eme ging Ma ke ,” Jou nal o Finance, 2012, 67 (3), 897–932.
Schula ick, Mo i z and Alan M Taylo , “C edi booms gone bus : mone a y policy,
le e age cycles, and inancial c ises, 1870–2008,” Ame ican Economic Re iew, 2012, 102
(2), 1029–1061.
S ahl, H. and N. Schelle , “S a is ics o he banks’ p o i and loss accoun s (GuV), Da a
Repo 2023-06 – Documen a ion e sion 12,” Mimeo, Deu sche Bundesbank, Resea ch
Da a and Se ice Cen e, 2023.
Su i, Ami and Alan M Taylo , “Financial c ises: A su ey,” Handbook o In e na ional
Economics, 2022, 6, 291–340.
50
Temes a y, Judi , S e en Ongena, and Ann L. Owen, “A global lending channel
unplugged? Does U.S. mone a y policy a ec c oss-bo de and a ilia e lending by global
U.S. banks?,” Jou nal o In e na ional Economics, 2018, 112, 50–69.
51
Appendix
‘C oss-Bo de Bank Flows, Regional Household C edi Booms and
Bank Risk-Taking’
by D. Boddin, D. e Kaa and K. Roszbach
Decembe 2, 2024
A Addi ional Tables
Table A1 Va iable De ini ions and Sou ces
Va iable De ini ion Uni Sou ce
∆Consume loans The log-di e ence in households’ ou s anding consume c edi olumes % HFCS o PHF, espec i ely
∆Mo gages The log-di e ence in households’ ou s anding mo gage c edi olumes % HFCS/PHF
Consump ion(non-du able) The loga i hm o households’ non-du able consump ion ln(x) PHF
Consump ion(du able) The loga i hm o households’ du able consump ion, de ined as income less ne sa ing less non-du able consump ion ln(x) PHF
Consump ion( ood) The loga i hm o households’ ood a home consump ion ln(x) PHF
Consump ion( es au an ) The loga i hm o households’ ood ou side home consump ion ln(x) PHF
Ne weal h The loga i hm o a household’s ne weal h (asse s less liabili ies) ln(eu o) HFCS/PHF
Income The loga i hm o a household’s o al g oss income ln(eu o) HFCS/PHF
Ren e =1 i household is a en e in he main esidence 0/1 HFCS/PHF
Fo eign =1 i a household’s coun y o bi h is ou side o Ge many 0/1 HFCS/PHF
Age Age o he household head - HFCS/PHF
Income Exp. =1 i a household expec s i s income o ise mo e han in la ion 0/1 PHF
Sel -Employed =1 i a household gene a es sel -employmen income 0/1 PHF
Unemployed =1 i a household ecei es unemploymen bene i s o any o he egula social ans e s 0/1 PHF
Non-Co e Banks’ sum o in e bank deposi s, as well as money ma ke secu i ies and bonds issued, o e o al asse s % Deu sche Bundesbank
G oss In e bank Banks’ in e bank deposi s o e o al asse s % Deu sche Bundesbank
G oss Domes ic In e bank Banks’ s anda dized domes ic in e bank deposi s o e o al asse s % Deu sche Bundesbank
G oss EA In e bank Banks’ s anda dized wi hin-eu o a ea in e bank deposi s o e o al asse s % Deu sche Bundesbank
G oss Non-EA In e bank Banks’ s anda dized non-eu o a ea in e bank deposi s o e o al asse s % Deu sche Bundesbank
Ne In e bank Banks’ in e bank deposi s ne o in e bank loans o e o al asse s % Deu sche Bundesbank
Ln(Nonco e) Banks’ loga i hm o non-co e unding olumes ln(eu o) Deu sche Bundesbank
Ln(In e bank) Banks’ loga i hm o in e bank unding olumes ln(eu o) Deu sche Bundesbank
Size Bank size, de ined as he log o o al asse s ln(eu o) Deu sche Bundesbank
ROA Banks’ e u n on asse s % Deu sche Bundesbank
ROE Banks’ e u n on equi y % Deu sche Bundesbank
Liquidi y Banks’ sum o cash, cen al bank ese es and easu ies held o e o al asse s % Deu sche Bundesbank
Capi aliza ion Banks’ o al capi al o e o al asse s % Deu sche Bundesbank
O he lows Ne o he in es men in lows o e nominal GDP % In e na ional Financial S a is ics
Po olio lows Ne po olio in es men in lows o e nominal GDP % In e na ional Financial S a is ics
FDI Flows Ne o eign di ec in es men in lows o e nominal GDP % In e na ional Financial S a is ics
Bank lows FX and b eak-adjus ed change in banks’ liabili ies less he equi alen change in asse s is-a- is all o he banks o e GDP % BIS-LBS
A1
