Enhancing credit risk assessments of SMEs with non-financial information

Wahls øm, Ranik Raaen; Becke , Linn-K is in; Fo nes, T ude Nons ad
A icle
Enhancing c edi isk assessmen s o SMEs wi h non-
inancial in o ma ion
Cogen Economics & Finance
P o ided in Coope a ion wi h:
Taylo & F ancis G oup
Sugges ed Ci a ion: Wahls øm, Ranik Raaen; Becke , Linn-K is in; Fo nes, T ude Nons ad (2024) :
Enhancing c edi isk assessmen s o SMEs wi h non- inancial in o ma ion, Cogen Economics &
Finance, ISSN 2332-2039, Taylo & F ancis, Abingdon, Vol. 12, Iss. 1, pp. 1-33,
h ps://doi.o g/10.1080/23322039.2024.2418910
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/321640
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/4.0/
Cogen Economics & Finance
ISSN: 2332-2039 (Online) Jou nal homepage: www. and online.com/jou nals/oae 20
Enhancing c edi isk assessmen s o SMEs wi h
non-financial in o ma ion
Ranik Raaen Wahls øm, Linn-K is in Becke & T ude Nons ad Fo nes
To ci e his a icle: Ranik Raaen Wahls øm, Linn-K is in Becke & T ude Nons ad Fo nes (2024)
Enhancing c edi isk assessmen s o SMEs wi h non-financial in o ma ion, Cogen Economics
& Finance, 12:1, 2418910, DOI: 10.1080/23322039.2024.2418910
To link o his a icle: h ps://doi.o g/10.1080/23322039.2024.2418910
© 2024 The Au ho (s). Published by In o ma
UK Limi ed, ading as Taylo & F ancis
G oup
View supplemen a y ma e ial
Published online: 05 No 2024.
Submi you a icle o his jou nal
A icle iews: 1486
View ela ed a icles
View C ossma k da a
Ci ing a icles: 2 View ci ing a icles
Full Te ms & Condi ions o access and use can be ound a
h ps://www. and online.com/ac ion/jou nalIn o ma ion?jou nalCode=oae 20
ECONOMETRICS & DATA ANALYTICS | RESEARCH ARTICLE
Enhancing c edi isk assessmen s o SMEs wi h non- inancial
in o ma ion
Ranik Raaen Wahls øm , Linn-K is in Becke and T ude Nons ad Fo nes
NTNU Business School, No wegian Uni e si y o Science and Technology, T ondheim, No way
ABSTRACT
We in es iga e non- inancial a iables o p edic ing bank up cy in small and medium-
sized en e p ises (SMEs). The a iables encompass managemen , boa d and owne ship
s uc u es and a e sou ced om uni e sally accessible in o ma ion, ende ing hem
a ailable o all s akeholde s and allowing o he analysis o all SMEs wi hin a ma ke .
Using a la ge and ecen sample o SMEs, we empi ically examine he a iables ha p e-
dic bank up cy o e ime ho izons o one, wo and h ee yea s. Ou analysis inco po-
a es s a e-o - he-a disc e e haza d models, he leas absolu e sh inkage and selec ion
ope a o (LASSO), ex eme g adien boos ing (XGBoos ), adap i e boos ing (AdaBoos ),
bagging and andom o es . We also es obus ness using balanced da ase s gene a ed
using he syn he ic mino i y o e sampling echnique (SMOTE). We ind ha including
non- inancial a iables enhances bank up cy p edic ions compa ed o using inancial
a iables alone. Mo eo e , ou esul s show ha among ou a iables, he mos signi i-
can non- inancial p edic o s o bank up cy a e he age o chie execu i e o ice s
(CEOs), chai pe sons and boa d membe s, as well as owne ship sha e and place o he
boa d membe s’ esidences.
IMPACT STATEMENT
This esea ch highligh s he c i ical ole o in eg a ing non- inancial in o ma ion wi h a-
di ional inancial a iables o enhance he p edic ion o SME bank up cy. While inancial
a iables emain he mos signi ican p edic o s, he inclusion o non inancial ac o s sig-
ni ican ly imp o es he accu acy o he assessmen o SMEs’ inancial heal h, bene i ing
in es o s, policymake s, and inancial ins i u ions by enabling be e isk managemen
and mo e e ec i e suppo schemes. The indings unde sco e he impo ance o di e se
boa d composi ion and local engagemen in educing bank up cy isk, o e ing aluable
insigh s o imp o ing SME go e nance and s abili y.
ARTICLE HISTORY
Recei ed 17 July 2024
Re ised 2 Oc obe 2024
Accep ed 15 Oc obe 2024
KEYWORDS
Small and medium-sized
en e p ises (SMEs);
bank up cy p edic ion;
co po a e go e nance;
non- inancial p edic o s;
LASSO
JEL CLASSIFICATION
CODES
C25; C53; G17; G20;
G33; M41
SUBJECTS
Economics; Finance;
Business, Managemen and
Accoun ing
1. In oduc ion
Imp o ed assessmen s o he inancial s anding o small and medium-sized en e p ises (SMEs) a e impo -
an o many s akeholde s. Fo example, be e assessmen helps SMEs access o mal capi al. This is c u-
cial because SMEs exhibi a highe le el o in o ma ion asymme y han la ge co po a ions, making i
mo e di icul o hem o access o mal capi al, as hey ace challenges in con eying hei s anding o
lende s and in es o s (Beck & Demi guc-Kun , 2006; Beck e al., 2008; Be ge & Udell, 1998,2002; Block
e al., 2018; Masiak e al., 2019). This is pa icula ly p essing now as hey ace challenges due o he wa
in Uk aine and being a he cen e o he economic shocks o he COVID-19 pandemic (G20 & OECD,
2015; O ganisa ion o Economic Co-ope a ion and De elopmen , 2022). SMEs also bene i om an
imp o ed assessmen o hei inancial s anding when con eying c edibili y o cus ome s and supplie s.
Fu he mo e, banks and in es o s bene i om be e assessmen s o SMEs’ inancial s anding, as hey
imp o e isk managemen and educe lending e o s h ough mo e accu a e e alua ions o po en ial
CONTACT Ranik Raaen Wahls øm [email p o ec ed] NTNU Business School, No wegian Uni e si y o Science and
Technology, T ondheim, No way
Supplemen al da a o his a icle can be accessed online a h ps://doi.o g/10.1080/23322039.2024.2418910.
This a icle has been co ec ed wi h mino changes. These changes do no impac he academic con en o he a icle.
ß2024 The Au ho (s). Published by In o ma UK Limi ed, ading as Taylo & F ancis G oup
This is an Open Access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License (h p://c ea i ecommons.o g/licenses/by/4.0/), which
pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed. The e ms on which his a icle has been
published allow he pos ing o he Accep ed Manusc ip in a eposi o y by he au ho (s) o wi h hei consen .
COGENT ECONOMICS & FINANCE
2024, VOL. 12, NO. 1, 2418910
h ps://doi.o g/10.1080/23322039.2024.2418910
new bo owe s and enhanced moni o ing o exis ing bo owe s. Addi ionally, be e assessmen enables
a mo e p ecise de e mina ion o he isk-weigh ed alue o loan po olios. In summa y, his may
inc ease lending o SMEs, which signi ican ly posi i ely a ec s he p o i abili y and di e si ica ion o loan
po olios (Al man & Saba o, 2007; Die sch & Pe ey, 2004). Mo eo e , assessing SMEs’ inancial s anding
is impo an o egula o s, who pe o m on-si e supe ision o banks by analyzing hei loan po olios.
Simila ly, enhanced assessmen s o SMEs’ inancial s anding acili a e a be e analysis o he inancial
s anding o la ge popula ions o i ms encompassing en i e ma ke s. This is c ucial o in es o s in hese
ma ke s and policymake s when designing suppo schemes o businesses o p edic ing losses in ax
e enues esul ing om bank up cy. In summa y, imp o ed assessmen s o SMEs’ inancial s anding
bene i he economy as a whole, especially because SMEs make up a la ge p opo ion o he economy.
1
Fi m’s s andings a e ypically e alua ed h ough c edi isk modeling, which commonly elies on ei he
his o ical accoun ing-based inancial a iables o secu i ies ma ke in o ma ion (Al man e al., 2010). Ea ly
s udies on c edi isk modeling used only accoun ing-based inancial a iables (eg Al man, 1968; Bea e ,
1966). Howe e , using only accoun ing-based inancial a iables o c edi isk modeling can be p oblem-
a ic o se e al easons (Aga wal & Ta le , 2008; Hillegeis e al., 2004). Fi s , because accoun s ep esen
he pas , hey may no be use ul o p edic ing u u e c edi isk. Second, he ue alue o i ms’asse s
may di e om he book alue because o accoun ing p inciples and conse a ism. This may lead o an
inco ec assessmen o a i m’s s anding. Thi d, accoun s a e subjec o managemen manipula ion.
Finally, accoun s a e p epa ed on a going-conce n basis and ha e limi ed u ili y in p edic ing bank-
up cy. SMEs may be pa icula ly exposed o hese challenges because hey usually ha e less de ailed
and anspa en accoun ing-based inancial da a han la ge co po a ions (Be ge & F ame, 2007).
Mo eo e , as SMEs ha e ewe obliga ions ega ding accoun ing da a disclosu e, inhe en ly smalle
accoun ing igu es and a e mo e ulne able o ex e nal e en s han la ge companies, inancial a ios
may be weak p edic o s o SME bank up cy (Ciampi, 2015; Ciampi e al., 2021).
The limi a ions o using accoun ing-based inancial a iables can be o e come by using ma ke -based a -
iables. The li e a u e has p o en ha such a iables e ec i ely inc ease he powe o bank up cy p edic ion
models (Bea e e al., 2005; Cha a & Ja ow, 2004; Hillegeis e al., 2004; Me on, 1974; Shumway, 2001; Tian
e al., 2015). Howe e , he ma ke da a o SMEs a e una ailable. Thus, i is c ucial o in es iga e non- inancial
a iables o p edic he bank up cy o such companies (Al man e al., 2010; Ciampi e al., 2020,2021).
This s udy in es iga es he non- inancial p edic o s o bank up cy in p i a ely held SMEs. So non- inan-
cial in o ma ion, con ex -dependen quali a i e da a ha a e no easily ans e able, has been ound o
imp o e c edi assessmen s (Co n
ee, 2019; G une e al., 2005). Howe e , such in o ma ion is ypically exclu-
si ely accessible o banks ha acqui e i h ough close ela ionships wi h bo owe s (Co n
ee, 2019; Libe i &
Pe e sen, 2019). Mo eo e , so non- inancial in o ma ion may be unsui able o majo decisions in ol ing
mul iple decision-make s because o he challenges in ans e ing i om he collec o o o he s and o
mino decisions due o he high labo cos s associa ed wi h i s collec ion (Co n
ee, 2019). Howe e , ha d
non- inancial in o ma ion, which e e s o quan i a i e, explici and con ex -independen da a, is usually
documen ed as numbe s o can be easily con e ed in o a nume ical o m, ensu ing ha i can be communi-
ca ed o o he s wi hou any loss o de ail. Rela ed s udies ha use ha d non- inancial in o ma ion o assess
SME c edi isk include Wilson and Al anla (2014), who examined newly inco po a ed companies, including
SMEs and o he i ms, and ound ha he boa d o di ec o s’cha ac e is ics con ibu e o bank up cy p edic-
ion. Howe e , hey do no conside inancial a iables because hey ocus solely on newly es ablished i ms
wi h no p io accoun s a ailable. In con as , we ocus on imp o ing c edi assessmen s o all SMEs in an
economy using all a ailable bank up cy p edic o s de i ed om bo h inancial and non- inancial in o ma-
ion. Ano he s udy ela ed o ou s is ha o Al man e al. (2023a,2023b), who ound ha models based on
inancial a iables show imp o ed p edic i e powe when paymen beha io , managemen - ela ed and
employee- ela ed a iables a e inco po a ed. Mo eo e , Ciampi (2015) demons a ed ha including man-
agemen - ela ed a iables in exis ing models based on inancial a iables enhances he p edic ion o de aul
among small en e p ises. Howe e , bo h o hese s udies use non- inancial a iables ha a e no eadily
a ailable o all s akeholde s and consequen ly analyze subsamples o he economy. Fo ins ance, employee-
ela ed in o ma ion abou SMEs is ypically only a ailable in e nally wi hin SMEs, whe eas paymen beha io
a iables a e usually only accessible o banks ha ha e long-s anding cus ome ela ionships wi h he SMEs.
Mo eo e , e en o banks, paymen beha io a iables a e no always a ailable– o example, when assessing
2 R.R. WAHLSTRØM ET AL.
loan applica ions om new bo owe s. These limi a ions make such a iables unsui able o assessing la ge
popula ions o SMEs, such as when in es o s wan o assess he inancial s anding o SMEs in a la ge popula-
ion o when policymake s a e designing business suppo schemes o o ecas ing losses in ax e enues
esul ing om bank up cies. By con as , we examine non- inancial p edic o s de i ed om uni e sally a ail-
able in o ma ion on all SMEs. These p edic o s enable s akeholde s o assess he inancial s anding o SMEs
ac oss an en i e economy. Thus, we con ibu e o he li e a u e by in es iga ing non- inancial p edic o s o
SME bank up cy ha no only assis banks in educing lending e o s and de e mining isk-weigh ed asse s
bu also suppo o he use s o bank up cy p edic ion models, including egula o s in ol ed in de e mining
banks’capi al equi emen s, policymake s, p i a e and public in es o s and c edi a ing agencies (Al man
e al., 2017; Rajan e al., 2015). Based on his, we de i e he ollowing esea ch ques ion:
Can non- inancial in o ma ion a ailable o all SMEs in an economy imp o e assessmen s o hei collec i e
inancial s anding?
In gene al, by ocusing on p i a ely held SMEs, we add ess an impo an gap in he li e a u e, as mos
exis ing esea ch on c edi isk assessmen in es iga es la ge co po a ions (da Sil a Ma os & Shasha, 2024;
Kuizinien_
e e al., 2022; Ma enda e al., 2022; Zhao e al., 2024). This ocus is c ucial gi en he economic sig-
ni icance o SMEs and hei dis inc inancial amewo ks and challenges compa ed wi h la ge co po a-
ions. O e all, consis en wi h he u u e esea ch a enue p oposed by he exis ing li e a u e on c edi isk
modeling (Ciampi e al., 2021; Habib e al., 2020), we make a aluable con ibu ion by demons a ing how
he in eg a ion o non- inancial a iables in o bank up cy p edic ion models mi iga es he in o ma ion
asymme y su ounding SMEs and enhances he abili y o all s akeholde s o e alua e hei s anding.
We employ a new and unique sample o 818,927 SME inancial s a emen s om 2014 o 2019. Fo
each inancial s a emen , we de i e he common inancial a iables ound in Al man (1968), Al man and
Saba o (2007) and Pa aschi e al. (2023) as benchma ks, as well as 20 non- inancial a iables ep esen -
ing managemen , boa d and owne ship s uc u es. We use hese a iables o p edic SME bank up cy
o e one, wo and h ee yea s. We employ he leas absolu e sh inkage and selec ion ope a o (LASSO)
me hod o explo e he impo ance o he a iables and selec he mos app op ia e a iable se s. This
me hod is commonly used in inance li e a u e (eg Calomi is & Mamaysky, 2019; Chae, 2024; Chinco
e al., 2019; Coad & S hoj, 2020; Hau sch e al., 2015; Ogne a e al., 2020; Tian e al., 2015); and is ound
by Pa aschi e al. (2023) o be supe io o o he me hods o selec ing co po a e bank up cy p edic o s.
This me hod has also been used in ecen li e a u e on SME bank up cy and de aul p edic ion (Al man
e al., 2023; Pa aschi e al., 2023). We e alua e he selec ed a iable se s in- and ou -o -sample when
used in disc e e haza d models wi h logis ic eg ession (LR) (Shumway, 2001).
Ou empi ical esul s show ha inancial a iables a e impo an p edic o s o SME bank up cy.
Fu he mo e, we ind ha bo h he in-sample i and ou -o -sample p edic ion pe o mance imp o e when
non- inancial a iables a e included. Among hese, he age o chie execu i e o ice s (CEOs), chai pe sons
and boa d membe s, along wi h he owne ship sha e o boa d membe s and whe he hey eside in he
same coun y as he SMEs’headqua e s (HQs), a e he mos impo an non- inancial bank up cy p edic o s.
Howe e , we also ind ha bank up cy p edic ion models consis ing solely o non- inancial a iables do
no achie e accep able pe o mance. This unde sco es he impo ance o inancial a iables and aligns
wi h p e ious s udies on p edic ing bank up cy in la ge companies (eg Liang e al., 2020).
Ou indings a e obus ac oss 18 pe mu a ions o h ee inancial a iable se s as benchma ks, wo
yea s as es popula ions (2018 and 2019) and h ee di e en ho izons o p edic ing bank up cy (one,
wo and h ee yea s). Fu he mo e, ou esul s a e obus o using balanced da ase s gene a ed wi h he
syn he ic mino i y o e sampling echnique (SMOTE). Finally, ou indings a e obus o employing he
machine lea ning me hods ex eme g adien boos ing (XGBoos ), adap i e boos ing (AdaBoos ), bagging
and andom o es . These me hods model nonlinea ela ionships ha ha e been shown o subs an ially
imp o e bank up cy p edic ions (Lohmann e al., 2023; Lohmann & Ohlige , 2018).
2. Backg ound
The ini ial discussions on bank up cy p edic ion in he li e a u e ocused on analyzing companies’
accoun ing igu es (Smi h & Winako , 1930), while Bea e (1966) demons a ed how indi idual inancial
COGENT ECONOMICS & FINANCE 3

a ios can p edic company bank up cy in uni a ia e models. Al man (1968) in oduced he i s mul i-
a ia e bank up cy p edic ion model, he Z-sco e model, which con inues o be widely used by p ac i-
ione s and academics (e.g, Bl€
ochlinge & Leippold, 2018; Campello e al., 2018; Chang e al., 2019;
Cha a & Ja ow, 2004; Tian e al., 2015). The model comp ises i e inancial a iables ha a e p ima y
aspec s o a company’s inancial p o ile: liquidi y, p o i abili y, le e age, sol ency (co e age) and ac i i y.
Al man’s(1968) model was ini ially designed o p edic he bank up cy o la ge lis ed companies and
includes a a iable de i ed in pa om he ma ke alue o equi y. Howe e , i was la e e ined o a -
ge p i a e i ms by inco po a ing only inancial a iables de i ed exclusi ely om accoun ing alues
(Al man e al., 1977,2019).
Edmis e (1972) was he i s o highligh ha ea ly bank up cy p edic ion models la gely igno ed
SMEs. Mo i a ed by his, he de eloped a model o p edic small business de aul s. Al man and Saba o
(2007) expand on Edmis e ’s(1972) wo k by de eloping a bank up cy p edic ion model speci ically o
SMEs. Simila o he a iable se in Al man (1968), he se o a iables in Al man and Saba o (2007) con-
sis ed o i e inancial a iables ca ego ized as he main aspec s o a company’s inancial p o ile.
Pa aschi e al. (2023) also conside ed SME bank up cy p edic ion and employ he LASSO me hod o
empi ically iden i y he se o en a iables ou o 155 inancial a iables ha yield he bes p edic ions.
The au ho s u he demons a ed ha hese a iables imp o e c edi isk assessmen s, esul ing in sig-
ni ican ly highe bank p o i s. The en selec ed a iables cap u e SMEs’le e age, liquidi y, sol ency
(co e age), age and p o i abili y. Mo eo e , inancial a iables ha e been applied o de i e p oxies o
ea nings managemen , which ha e been ound o imp o e SME bank up cy p edic ions (S
e e in &
Veganzones, 2021).
In he con ex o non- inancial bank up cy p edic o s, we di e en ia e be ween hose de i ed om
so and ha d non- inancial in o ma ion (Co n
ee, 2019; Libe i & Pe e sen, 2019). So non- inancial in o -
ma ion e e s o con ex -dependen quali a i e in o ma ion ha is no easily ans e able and is ypically
collec ed by banks abou bo owe s h ough close ela ionships. The li e a u e sugges s ha such in o -
ma ion is aluable o assessing he c edi isk o small and opaque bo owe s. Fo example, Be ge and
Udell (2002) a gued ha ela ionship lending, condi ional on so non- inancial in o ma ion, educes
in o ma ion p oblems in small i ms. Fu he mo e, Be ge e al. (2005) ound ha smalle banks lend
mo e o smalle i ms because hey ha e a compa a i e ad an age in collec ing and ac ing on so non-
inancial in o ma ion. Mo eo e , S ein (2002) sugges ed ha consolida ion in he banking indus y leads
o a decline in small business lending because so in o ma ion canno be c edibly ansmi ed wi hin
la ge hie a chies. This is consis en wi h he indings o Rajan e al. (2015) ha he models used o p e-
dic ing he de aul s o secu i ized subp ime mo gages be o e he U.S. subp ime mo gage c isis in 2007
se e ely unde es ima ed he likelihood o de aul o opaque bo owe s. Fu he mo e, using a sample o
Ge man SMEs, G une e al. (2005) ound e idence sugges ing ha combining inancial and so non-
inancial a iables leads o mo e accu a e de aul p edic ions han using inancial o so non- inancial
a iables. Co n
ee (2019) eplica ed G une e al. (2005) using c edi iles om a social ela ional bank
specializing in p o iding ex e nal deb unding o genuinely small and opaque i ms ha p io i ize social
o e inancial goals. Co n
ee (2019) ound ha including so non- inancial a iables yields be e p edic-
ions han using only inancial a iables. Howe e , compa ed o G une e al. (2005), he e ealed ha
so non- inancial in o ma ion ends o be mo e aluable han inancial a iables. He a gued ha his
may be because he uses a smalle and mo e opaque sample o bo owe s han ha o G une e al.
(2005). This sugges s ha he la ge and mo e anspa en he bo owe , he lowe he p edic i e alue
o so non- inancial in o ma ion compa ed o inancial a iables.
In con as , ha d non- inancial in o ma ion is quan i a i e, explici and con ex -independen . I is ypic-
ally eco ded as o can be easily educed o numbe s, making i easy o con ey o o he s wi hou losing
in o ma ion. Va iables de i ed om such in o ma ion include paymen beha io a iables, ha is, indica-
o s o la e paymen s o c edi o s, which p e ious s udies ound o inc ease bank up cy p edic ion accu -
acy compa ed o using only inancial a iables (eg Al man e al., 2023; Back, 2005; Lai inen, 1999; Wilson
e al., 2000). Fu he mo e, he a iables de i ed om ha d non- inancial in o ma ion include manage-
men - ela ed a iables, which a e indica o s based on he cha ac e is ics o he managemen and boa d
o di ec o s. Fo example, Wilson and Al anla (2014) epo ed ha he boa d o di ec o s’cha ac e is ics
con ibu e o p edic ing he bank up cy o newly inco po a ed companies wi h limi ed publicly a ailable
4 R.R. WAHLSTRØM ET AL.
da a, including SMEs and o he i ms. Mo eo e , Ciampi (2015) ound ha de aul p edic ions among
small en e p ises imp o e when managemen - ela ed a iables a e included in exis ing models based on
inancial a iables. Al man e al. (2023a,2023b) de eloped an SME de aul p edic o by conside ing
inancial a iables in combina ion wi h paymen beha io , managemen - ela ed and employee- ela ed
a iables and ound ha he p edic i e powe o models based on inancial a iables imp o es when
in oducing hese addi ional a iables. O he ha d non- inancial a iables used o assess SME c edi isk
a e de i ed om he local banking ma ke (A cu i & Le a o, 2020), published legal judgmen s (Yin
e al., 2020) and bag-o -wo ds models applied o con en sc aped om co po a e websi es (C osa o
e al., 2023). Addi ional ha d non- inancial ac o s ha may help con ey he c edi isk include ecei ing
go e nmen g an s. S hoj e al. (2021a,2021b) showed ha ecei ing such g an s can ha e a ce i ica-
ion e ec , inc easing he likelihood o ob aining a long- e m bank loan by being ‘ce i ied’by he go -
e nmen . Mo eo e , Lohmann and Ohlige (2020) a gued ha including quali a i e in o ma ion om
i ms’annual epo s, such as s uc u al and linguis ic cha ac e is ics, enhances he disc imina o y powe
o bank up cy p edic ion models based on inancial a iables.
In eg a ing non- inancial p edic o s can also in oduce e hical challenges ha mus be ca e ully con-
side ed as hey can pe pe ua e o exace ba e disc imina ion. Fo ins ance, Fus e e al. (2022) highligh ed
ha machine lea ning echniques exhibi imp o ed accu acy in p edic ing mo gage de aul s when con-
side ing bo owe s’e hnici y, leading o ad e se consequences o speci ic bo owe g oups. The legal
implica ions a e also signi ican . Reliance on non- inancial p edic o s could expose s akeholde s o legal
challenges i hey lead o disc imina o y p ac ices in lending o in es men decisions. The e o e, we
a gue ha he non- inancial a iables used in ou s udy should be employed o assess he collec i e
inancial s anding o all i ms in a la ge popula ion and should be used wi h cau ion when applied o
decision suppo a he indi idual le el. We u he mi iga e he bias in he non- inancial a iables by
no conside ing hem in isola ion bu wi h o he inancial and non- inancial a iables.
3. Va iables
We in es iga e he 20 non- inancial bank up cy p edic o s p esen ed in Table 1, bo h indi idually and in
conjunc ion wi h he benchma k se s o accoun ing-based inancial p edic o s in Al man (1968), Al man
and Saba o (2007) and Pa aschi e al. (2023), as p esen ed in Table 2.
The 20 non- inancial a iables lis ed in Table 1 a e di ided in o h ee ca ego ies. Fi s , hey include
managemen s uc u e a iables indica ing he cha ac e is ics o SMEs’CEOs. Second, hey cons i u e
boa d s uc u e a iables indica ing boa d cha ac e is ics ha can signi ican ly impac i m pe o mance
by suppo ing, con olling and e alua ing managemen (K ause e al., 2016; Wi he s & Fi za, 2017).
Finally, Table 1 includes owne ship s uc u e a iables.
Ou non- inancial a iables indica e he ages o he CEOs, chai pe sons and boa d membe s. These
a e included because p e ious li e a u e a gues ha pe sons o highe and lowe ages in hese oles can
esul in lowe bank up cy p obabili ies (Pla & Pla , 2012). On he one hand, olde age may mean
g ea e expe ience, which can help a oid bank up cy. Howe e , as olde age may b ing abou conse a-
ism, younge CEOs and boa d membe s may help a oid bank up cy by being mo e willing o y new
ideas and adap o changing business en i onmen s. Pla and Pla (2012) ound ha i ms ha wen
bank up had, on a e age, younge CEOs and boa d membe s han hose ha did no go bank up .
Fu he mo e, An ulo -Fan ulin e al. (2021) in es iga ed he p edic abili y o municipal bank up cy and
ound ha he age o council membe s is one o he mos impo an p edic o s.
Fu he mo e, we hypo hesize ha SME go e nance imp o es i i s managemen and boa d a e oo ed
in he same local communi y. Thus, we include in Table 1 a iables ha indica e whe he CEOs, chai pe -
sons and boa d membe s eside in he same coun y as he SME HQs. This is consis en wi h Wilson and
Al anla (2014), which ound ha bank up cy was associa ed wi h ewe boa d membe s li ing in he
same coun y as he company’s egis e ed add ess.
Table 1 also includes a iables ha indica e he gende s o CEOs, chai pe sons and boa d membe s.
We include his, as p e ious li e a u e sugges s ha women end o be mo e isk-a e se han men (eg
Bo ghans e al., 2009; Cha ness & Gneezy, 2012; Dwye e al., 2002; Jianakoplos & Be nasek, 1998). The
li e a u e on bank up cy p edic ion con i ms his assump ion by sugges ing ha mo e women among
COGENT ECONOMICS & FINANCE 5
he managemen and boa d membe s yield lowe p obabili ies o bank up cy. Fo example, Cho e al.
(2021) ound ha among Chinese i ms du ing 2005–2016, he likelihood o bank up cy was nega i ely
associa ed wi h he p opo ion o emale execu i es. C. J. Ga c
ıa and He e o (2021) ound ha he likeli-
hood o bank up cy among EU i ms du ing 2002–2019 is nega i ely associa ed wi h g ea e boa d gen-
de di e si y. Wilson and Al anla (2014) ound ha bank up companies ha e ewe emale boa d
membe s. Mo eo e , An ulo -Fan ulin e al. (2021) show ha he gende o council membe s is among
he mos impo an a iables o p edic ing de aul in municipali ies.
Mo eo e , we include in Table 1 a iables indica ing whe he he CEO si s on he boa d, whe he he
SME has wo CEOs and whe he a single indi idual se es as he SME’s CEO and chai pe son, e e ed o
as CEO duali y (K ause e al., 2014). The exis ing li e a u e highligh s he po en ial ad an ages and disad-
an ages o CEO duali y. On he one hand, i may be un o una e because i educes he independence
be ween he boa d and managemen , educing he boa d’s abili y o con ol he managemen (Jensen,
Table 1. Non- inancial a iables.
Va iable name Desc ip ion
Managemen s uc u e CEO age The na u al loga i hm o he age o he CEO
CEO woman Dummy; one i he CEO is a woman
CEO duali y Dummy o CEO duali y; one i a single indi idual se es as bo h he CEO and
chai pe son o he boa d
CEO on boa d Dummy; one i he CEO si s on he boa d
CEO coun y Dummy; one i he CEO esides in he same coun y as he SME’s
headqua e s (HQs)
Two CEOs Dummy; one i he company has wo CEOs
Boa d s uc u e Chai pe son age The na u al loga i hm o he age o he chai pe son o he boa d
Chai pe son woman Dummy; one i he chai pe son o he boa d is a woman
Chai pe son coun y Dummy; one i he chai pe son o he boa d esides in he same coun y as
he SME’s headqua e s (HQs)
Boa d size The na u al loga i hm o he numbe o boa d membe s
Boa d age a g The na u al loga i hm o he a e age age among all boa d membe s
Boa d age s d S anda d de ia ion o age among boa d membe s
Boa d women P opo ion o boa d membe s who a e women
Boa d coun y P opo ion o boa d membe s who esides in he same coun y as he SME’s
headqua e s (HQs)
Boa d non-owne s P opo ion o boa d membe s who a e no sha eholde s
Owne ship s uc u e Owne ship concen a ion 1 A e age holdings o sha eholde s
Owne ship concen a ion 2 S anda d de ia ion o sha eholde s’holdings
Owne ship CEO Owne ship sha e wi h he CEO
Owne ship chai pe son Owne ship sha e wi h he chai pe son o he boa d
Owne ship boa d Owne ship sha e wi h he boa d membe s
No e: The 20 non- inancial a iables o in e es di ided in o ca ego ies ep esen ing he managemen , boa d and owne ship s uc u es.
Table 2. Benchma k inancial a iable se s.
Va iable name Ca ego y
Al man (1968) EBIT/ o al asse s Co e age
Re ained ea nings/ o al asse s P o i abili y
Sales/ o al asse s Ac i i y
To al equi y/ o al liabili ies Le e age
Wo king capi al/ o al asse s Liquidi y
Al man and Saba o (2007) Cu en liabili ies/ o al equi y Le e age
EBITDA/in e es expense Ac i i y
EBITDA/ o al asse s P o i abili y
Re ained ea nings/ o al asse s Co e age
Sho - e m liquidi y/ o al asse s Liquidi y
Pa aschi e al. (2023) (Cu en liabili ies - sho - e m liquidi y)/ o al asse s Le e age
Accoun s payable/ o al asse s Liquidi y
Dummy; one i paid-in equi y is less han o al equi y Sol ency
Dummy; one i o al liabili y exceeds o al asse s Le e age
In e es expenses/ o al asse s Sol ency
In en o y/cu en asse s Liquidi y
Log(age in yea s) Age
Ne income/ o al asse s P o i abili y
Public axes payable/ o al asse s Liquidi y
Sho - e m liquidi y/cu en asse s Liquidi y
No es: The benchma k a iable se s o accoun ing-based inancial a iables a e hose o Al man (1968), Al man and Saba o (2007) and
Pa aschi e al. (2023). The a iables a e so ed alphabe ically.
6 R.R. WAHLSTRØM ET AL.
1993). Howe e , CEO duali y may also be bene icial i i leads o a mo e lexible leade ship, which
imp o es company e iciency (Combs e al., 2007; Dowell e al., 2011). The empi ical e idence on how
CEO duali y a ec s he p obabili y o bank up cy is inconclusi e. Fo ins ance, Ciampi (2015) ound ha
CEO duali y educes he isk o de aul . Du u e al. (2016) ound ha CEO duali y nega i ely a ec s i m
pe o mance when independen boa d membe s accoun o a small p opo ion o he boa d. Howe e ,
his nega i e e ec is mi iga ed as he p opo ion o independen boa d membe s inc eases. E en ually,
he au ho s ound ha he impac becomes posi i e when he p opo ion o independen boa d mem-
be s inc eases e en mo e. By con as , Pla and Pla (2012) and Manzaneque e al. (2016) ound no sig-
ni ican e ec o CEO duali y on he likelihood o company bank up cy.
We also include non- inancial a iables ha measu e boa d size and he p opo ion o boa d mem-
be s who a e no sha eholde s. Ha ing mo e boa d membe s esul s in g ea e di e si y and access o
in o ma ion, he eby inc easing he boa d’s e iciency and independence by boos ing i s abili y o con-
ol managemen and di ec he company in he igh di ec ion (Dal on e al., 1999; Manzaneque e al.,
2016; Pea ce & Zah a, 1992). Howe e , boa d e iciency can dec ease wi h mo e membe s i his esul s
in a poo e low o in o ma ion (Gues , 2009). Pla and Pla (2012) and Manzaneque e al. (2016) ound
a nega i e ela ionship be ween he size o a company’s boa d and he likelihood o bank up cy. By con-
as , Ciampi (2015) epo ed ha boa d size did no signi ican ly help p edic de aul s.
Fu he mo e, we include wo non- inancial a iables ha measu e owne ship concen a ion.
2
Theo e ically, high owne ship concen a ion has p os and cons (Ciampi, 2015). On he one hand, la ge
sha eholde s can bene i i m pe o mance because hey ypically ha e mo e expe ise ele an o he
i m han smalle sha eholde s. They also ha e g ea e incen i es o moni o managemen and boa ds
e ec i ely. Howe e , hey may also p omo e i m ine iciency i hey exe cise hei con ol igh s o p i-
a e bene i s. Ciampi (2015) ound ha a highe owne ship concen a ion, whe e a single sha eholde
has he majo i y o sha es, educes he p obabili y o an SME de aul . Tang e al. (2020) and Liang e al.
(2020) e ealed ha he owne ship s ake o he majo i y o s akeholde s helps p edic co po a e bank-
up cy. In con as , Manzaneque e al. (2016) sugges ed ha owne ship concen a ion does no signi i-
can ly impac inancial dis ess.
Finally, Table 1 p esen s h ee non- inancial a iables indica ing he p opo ion o owne ship by CEOs,
chai pe sons and boa d membe s. We include hese a iables because we hypo hesize ha he p obabil-
i y o bank up cy is nega i ely associa ed wi h he owne ship sha e o SMEs’key pe sonnel; highe own-
e ship should p o ide g ea e incen i es o a oid bank up cy. This co esponds wi h Lilien eld-Toal and
Ruenzi (2014), who ound ha CEO owne ship inc eases he pe o mance o lis ed i ms.
4. Da a
Ou da a consis o all unconsolida ed annual inancial s a emen s o p i a ely held No wegian limi ed-
liabili y SMEs o he accoun ing yea s 2014–2019. Al hough ela ed s udies also ocus on da a om only
one economy and ime pe iod,
3
we ecognize he limi a ions o he gene alizabili y o ou analyses, spe-
ci ically on No wegian SMEs. Howe e , we conside No way o e 2014–2019 o be a easible es en i -
onmen o se e al easons. Du ing ou sample yea s, No way expe ienced s able economic condi ions
and a lack o signi ican business cycle luc ua ions, alling be ween he a e ma h o he Eu opean deb
c isis and he onse o he inancial shock caused by he co ona i us c isis. Mo eo e , No way is in e-
g a ed wi h he Eu opean in e nal ma ke as pa o he Eu opean Economic A ea (EEA) and is consid-
e ed a high-income Eu opean coun y.
4
Addi ionally, 99% o all No wegian i ms a e SMEs, compa able
o o he Eu opean coun ies, employing 56% o he wo k o ce.
5
The annual inancial s a emen s o ou da a we e p o ided by he No wegian go e nmen agency
B ønnøysund Regis e Cen e (BRC).
6
I is manda o y o all No wegian limi ed liabili y companies o
epo hei annual inancial s a emen s o he au ho i ies, and hese a e subsequen ly s o ed wi h he
BRC. Addi ionally, he BRC has p o ided us wi h he indus y classi ica ion o he i ms when epo ing
hei annual inancial s a emen s by he No wegian S anda d Indus ial Classi ica ion (SIC2007).
7
Mo eo e , he BRC has supplied us wi h he da es o bank up cy ilings o all i ms in ou da a ha
ha e iled o bank up cy. Fu he mo e, in o ma ion on i ms’CEOs, chai pe sons, boa d membe s, and
owne s used o de i e ou non- inancial a iables was p o ided by Enin AS.
8
COGENT ECONOMICS & FINANCE 7
gi en he la ge da a sample size and he memo y-in ensi e and compu a ionally complex na u e o
hese me hods, hey a e imp ac ical wi hou signi ican compu a ional esou ces ha a e no a ailable
o us. Mo eo e , we do no de i e coe icien es ima es and hei s a is ical signi icance om he LASSO
me hod bu me ely use i o selec a iables ha a e la e used in LR models o which we de i e coe i-
cien es ima es and hei s a is ical signi icance. Howe e , we co ec he s anda d e o s used o de i e
z-sco es by clus e ing hem a he i m le el.
5.3. Machine lea ning me hods
We also es he obus ness o ou esul s by p edic ing bank up cy using machine lea ning me hods
ins ead o LASSO and LR, o he wise applying he same es se ing (see Sec ion 5.1). Speci ically, ollow-
ing da Sil a Ma os and Shasha (2024), we employ ou machine lea ning me hods wi hin h ee classes
o ensemble me hods: boos ing (XGBoos and AdaBoos ), bagging and andom o es . Indeed, hese
h ee classes o ensemble me hods ha e been ound o ou pe o m o he me hods o bank up cy p e-
dic ion (Ba boza e al., 2017). XGBoos (Chen & Gues in, 2016), which was also used o es he obus -
ness in Al man e al. (2023a) and AdaBoos (F eund & Schapi e, 1997) build ensembles o decision ees
sequen ially, whe e each ee co ec s he e o s o he p e ious ones. Fu he mo e, bagging, o boo -
s ap agg ega ing (B eiman, 1996), ains mul iple decision ee classi ie s on subse s o he aining da a
selec ed andomly wi h eplacemen and hen agg ega es hei p edic ions o imp o e s abili y and
accu acy. Finally, andom o es (B eiman, 2001) ains mul iple independen decision ees wi h an-
domly selec ed subse s o a iables and ou pu s he mode o he classes o classi ica ion. Fo each
machine lea ning me hod, we une he hype pa ame e s using a g id sea ch wi h a c oss- alida ion
scheme.
12
To a oid da a leakage, uning is pe o med on he aining da a. Fo u he desc ip ion o he
machine lea ning me hods, we e e o hei desc ip ions in he bank up cy p edic ion s udies by
Ba boza e al. (2017) and Rado ano ic and Haas (2023).
5.4. E alua ion me ics
We e alua e ou bank up cy p edic ion models using se e al e alua ion me ics. Fi s , we e alua e hem
using a e age p ecision (AP) and he a ea unde he ecei e ope a ing cha ac e is ic cu e (AUC), which
a e widely used o bina y classi ica ion p oblems, including bank up cy p edic ion (eg Al man e al.,
2023a; Pa aschi e al., 2023; Tian e al., 2015). The AUC is he a ea unde he plo o he alse posi i e
a e agains he ue posi i e a e ac oss all obse a ions when a ying he disc imina ion h eshold
ac oss he wo limi alues o 0 and 1 (Hosme e al., 2013). AP is calcula ed simila ly, bu based on he
plo o p ecision agains ecall and is ecommended o e AUC o e alua ing model pe o mance wi h
imbalanced da ase s (Sai o & Rehmsmeie , 2015). Highe AUC and AP alues indica e a be e -pe o ming
model. As sugges ed by Pa aschi e al. (2023), we ollow Hosme e al. (2013) by conside ing AUC 2
½0:7, 0:8Þaccep able, AUC 2½0:8, 0:9Þexcellen and AUC 0:9 ou s anding.
Fu he mo e, hey we e e alua ed based on he accu acy a io (AR), a pe o mance me ic de i ed
om he cumula i e accu acy p o ile (CAP) cu e (Engelmann e al., 2003; Mai e al., 2019). I is calcu-
la ed as he a io o he a ea be ween he CAP cu e o he model and he andom model o he a ea
be ween he CAP cu e o he pe ec model and he andom model. Mo eo e , we use he
Kolmogo o -Smi no (KS) s a is ic (Hodges, 1958), a non-pa ame ic es o compa e he dis ibu ions o
wo samples by he maximum di e ence be ween hei cumula i e dis ibu ion unc ions. In he con ex
o bank up cy p edic ion model e alua ion, he KS s a is ic can be used o compa e he p edic ed p oba-
bili ies o bank up and non-bank up classes. Highe AR and KS s a is ics indica e be e model pe -
o mance, e lec ing a g ea e abili y o dis inguish be ween bank up and non-bank up cases.
Addi ionally, we e alua e hem based on Hinge loss (C amme & Singe , 2001), measu ing he dis-
ance be ween he p edic ed and ac ual classi ica ions encoded as −1 and 1, as well as he logis ic loss
(Log loss), which is he alue o he log-likelihood unc ion shown in Equa ion (2), di ided by he num-
be o obse a ions. Mo eo e , ollowing Tian e al. (2015) and Pa aschi e al. (2023), we use he Akaike
(1974) In o ma ion C i e ion (AIC), e alua ing he goodness o i o he model simila ly o he log loss
bu penalized o he numbe o pa ame e s o p e en o e i ing, balancing model complexi y and i .
14 R.R. WAHLSTRØM ET AL.

We also use he Bayesian In o ma ion C i e ion (BIC), which is simila o he AIC bu in oduces a s on-
ge penal y o he numbe o pa ame e s. Addi ionally, we assess he models using B ie (1950) sco e,
which quan i ies he mean-squa ed di e ence be ween he ac ual and p edic ed bank up cy p obabil-
i ies. Lowe Hinge loss, Log loss, AIC, BIC and B ie sco es indica e be e models.
We also epo he B ie Skill Sco e (BSS), which compa es he B ie sco e o a p edic i e model wi h
ha o a e e ence model (Rouls on, 2007). As a e e ence model, we use a model ha always p edic s
he a e age o y2 0,1gN, which is he bank up cy equency, in he aining da a. Fu he mo e, we ol-
low bank up cy p edic ion li e a u e (eg Campbell e al., 2008; Pa aschi e al., 2023; Tian e al., 2015)by
e alua ing he in-sample i o LR models using McFadden’s(1974) pseudo-R squa ed R2¼1−
lðb,b0Þ
lðb0Þ2
½0, 1whe e he denomina o is he log-likelihood o a model con aining only he in e cep coe icien
b0:Highe BSS and R2 alues indica e be e models.
Finally, ollowing he bank up cy p edic ion li e a u e (eg Cha a & Ja ow, 2004; Pa aschi e al., 2023;
Shumway, 2001; Tian e al., 2015), we e alua e hem based on decile ankings. When applying his
me hod, he obse a ions a e di ided in o deciles based on hei p edic ed p obabili y o bank up cy
p o ided by he model, and he p opo ion o ac ual bank up cies wi hin each decile is hen epo ed.
This me hod allows a clea assessmen o he model’s disc imina o y powe by showing how well bank-
up and non-bank up i ms can be dis inguished ac oss di e en isk le els.
6. Resul s
6.1. Main esul s
Tables 6,7and 8p esen he es ima ion esul s o he LR models when p edic ing bank up cy o e ho i-
zons o one, wo and h ee yea s, espec i ely, using a iable se s selec ed by he LASSO me hod om
he popula ion o non- inancial a iables in Table 1 and he se o inancial a iables in Al man (1968),
Al man and Saba o (2007) and Pa aschi e al. (2023) (see Table 2). Addi ionally, we epo he esul s
using all he inancial s a emen s om 2018 and 2019, espec i ely, as ou -o -sample es samples. Fi s ,
we use he accoun ing yea 2018 as ou es sample and le he LASSO me hod use all he inancial
s a emen s om he ou p e ious yea s o selec he bes a iable se among he popula ion o 25 a i-
ables, consis ing o he i e inancial a iables in Al man (1968) (see Table 2) and he 20 non- inancial
a iables in Table 1. The selec ed a iables and he es ima ion esul s when used in LR models a e p e-
sen ed in he i s columns o Tables 6,7and 8. Nex , we epea he p ocedu e using he 2019 accoun -
ing yea as ou es sample and epo he esul s in he second column. We hen epea he p ocedu e
wi h he a iables in Al man and Saba o (2007) ins ead o hose in Al man (1968) and show he esul s
in he hi d and ou h columns. Finally, he p ocedu e is epea ed wi h a popula ion o 30 a iables,
including hose in Pa aschi e al. (2023) and he 20 non- inancial a iables in Table 1. The esul s a e
p esen ed in he las wo columns o Tables 6,7and 8. In each able, Panel A p esen s he a iables
selec ed by he LASSO me hod and hei LR coe icien es ima es and z-sco es in pa en heses. Panels B
and C epo he in-sample i and ou -o -sample p edic ion pe o mance, espec i ely, when using he
a iable se s selec ed by he LASSO me hod, as shown in Panel A, in he op ows, and when using
exclusi ely inancial a iables in he bo om ows.
6.1.1. The impo ance o inancial a iables
The ables show ha inancial a iables a e undoub edly impo an because he LASSO me hod selec s
mos inancial a iables ac oss all a iable se s, p edic ion ho izons and accoun ing yea s as ou es
sample. Fu he mo e, he selec ed inancial a iables a e consis en ac oss all se ings. Speci ically, he
same inancial a iables a e selec ed in all se ings, excep o ‘in e es expenses / o al asse s’which is
selec ed when using he inancial a iable se o Pa aschi e al. (2023) only when 2019 is he es pe iod
and when p edic ing bank up cy o e a ho izon o h ee yea s (see Panel A o Table 8).
The impo ance o he inancial a iables is u he con i med by he associa ed LASSO pa h plo s.
Figu es IA.1, IA.2 and IA.3 in he In e ne Appendix p esen he LASSO pa h plo s o he a iables
selec ed among he non- inancial a iables and hose in Al man (1968), Al man and Saba o (2007) and
Pa aschi e al. (2023), espec i ely, when p edic ing bank up cy o e a ho izon o one yea . Panels A
COGENT ECONOMICS & FINANCE 15
Table 6. Es ima ion esul s when p edic ing bank up cy o e a ho izon o one yea .
Panel A. Va iables selec ed using he LASSO me hod and hei LR coe icien es ima es when p edic ing bank up cy
o e a ho izon o one yea .
Al man (1968) Al man and Saba o (2007) Pa aschi e al. (2023)
2018 2019 2018 2019 2018 2019
CEO age −0.65(−5.18) −0.57(−4.31) −0.84(−6.26) −0.78(−6.03) −0.35(−2.73) −0.35(−2.65)
Chai pe son age −0.78(−6.30) −0.48(−2.74) −0.84(−4.84) −0.54(−3.19) −0.40(−3.13) −0.23(−1.32)
Boa d age a g −0.56(−2.81) −0.21(−1.07) −0.74(−3.85) −0.40(−1.97)
Boa d coun y −0.82(−14.97) −0.92(−17.46) −0.87(−15.83) −0.97(−18.33) −0.78(−14.16) −0.88(−16.77)
Boa d non-owne s −0.29(−2.88) −0.40(−3.92) −0.30(−2.93)
Owne ship boa d 0.72(13.50) 0.50(4.96) 0.95(18.04) 0.66(6.56) 0.60(11.31) 0.41(4.02)
EBIT / o al asse s −1.72(−21.24) −1.78(−22.87)
Re ained ea nings / o al
asse s
−0.38(−9.75) −0.33(−8.97) −0.70(−22.88) −0.66(−22.71)
Sales / o al asse s 0.14(15.07) 0.15(16.59)
Wo king capi al / o al
asse s
−0.77(−13.80) −0.78(−14.57)
EBITDA / o al asse s −1.99(−23.86) −2.07(−25.82)
Sho - e m liquidi y / o al
asse s
−1.77(−15.06) −1.75(−15.53)
(Cu en liabili ies - sho -
e m liquidi y) / o al
asse s
0.12(1.76) 0.20(3.06)
Accoun s payable / o al
asse s
1.27(12.77) 1.18(12.31)
Dummy; one i paid-in
equi y is less han o al
equi y
−0.69(−9.44) −0.78(−10.84)
Dummy; one i o al liabili y
exceeds o al asse s
0.65(9.03) 0.59(8.42)
In en o y / cu en asse s 0.43(5.66) 0.44(6.02)
Log(age in yea s) −0.20(−8.93) −0.19(−8.57)
Ne income / o al asse s −1.08(−13.71) −1.10(−14.71)
Public axes payable / o al
asse s
4.12(22.33) 4.03(22.84)
Sho - e m liquidi y /
cu en asse s
−1.49(−13.69) −1.37(−13.25)
In e cep −0.04(−0.10) 0.87(2.20) 2.23(5.62) 3.34(8.75) −2.28(−5.30) −1.06(−2.48)
Panel B: In-sample i when p edic ing bank up cy o e a ho izon o one yea .
Al man (1968) Al man and Saba o (2007) Pa aschi e al. (2023)
2018 2019 2018 2019 2018 2019
Financial and
non- inancial
a iables
R20.137 0.151 0.129 0.141 0.197 0.209
AUC 0.834 0.846 0.827 0.838 0.876 0.885
AP 0.036 0.042 0.031 0.035 0.057 0.062
AR 0.664 0.690 0.650 0.672 0.749 0.767
KS s a is ic 0.536 0.558 0.511 0.527 0.612 0.631
Hinge loss 1.004 1.004 1.004 1.004 1.004 1.004
Log loss 0.023 0.023 0.023 0.023 0.021 0.021
AIC 24,345 25,113 24,573 25,421 22,684 23,415
BIC 24,446 25,237 24,673 25,533 22,841 23,595
B ie sco e 0.004 0.004 0.004 0.004 0.004 0.004
BSS 0.012 0.014 0.011 0.012 0.028 0.030
Decile 1 0.578 0.592 0.548 0.571 0.672 0.690
Decile 2 0.156 0.160 0.154 0.150 0.137 0.136
Decile 3 0.074 0.081 0.097 0.090 0.065 0.055
Decile 4 0.048 0.045 0.052 0.052 0.035 0.038
Decile 5 0.042 0.035 0.046 0.043 0.028 0.025
Decile 6-10 0.103 0.088 0.104 0.095 0.064 0.055
Exclusi ely inancial
a iables
R20.118 0.127 0.099 0.104 0.185 0.193
AUC 0.816 0.826 0.794 0.800 0.868 0.874
AP 0.031 0.034 0.024 0.025 0.050 0.053
AR 0.628 0.649 0.585 0.597 0.733 0.745
KS s a is ic 0.514 0.530 0.470 0.486 0.603 0.607
Hinge loss 1.004 1.004 1.004 1.004 1.004 1.004
Log loss 0.024 0.024 0.024 0.024 0.022 0.022
AIC 24,892 25,815 25,411 26,503 22,999 23,866
(con inued)
16 R.R. WAHLSTRØM ET AL.
and B in each igu e p esen he plo s using 2018 and 2019 as he ou -o -sample es samples. In sum-
ma y, Figu es IA.1, IA.2 and IA.3 ha e six panels p esen ing he LASSO pa h plo s o he six modeling
pe mu a ions shown in he six columns o Table 6. Simila ly, he six panels o Figu es IA.5, IA.6 and IA.7
in he In e ne Appendix p esen he LASSO pa h plo s o he six modeling pe mu a ions shown in he
Panel C. Ou -o -sample p edic ion pe o mance when p edic ing bank up cy o e a ho izon o one yea .
Al man (1968) Al man and Saba o (2007) Pa aschi e al. (2023)
2018 2019 2018 2019 2018 2019
Financial and
non- inancial
a iables
AUC 0.855 0.830 0.842 0.804 0.889 0.860
AP 0.043 0.025 0.035 0.022 0.064 0.041
AR 0.707 0.657 0.680 0.607 0.775 0.718
KS s a is ic 0.587 0.519 0.536 0.466 0.632 0.600
Hinge loss 1.004 1.004 1.004 1.004 1.004 1.003
Log loss 0.024 0.017 0.024 0.018 0.022 0.017
AIC 6866 5166 6976 5274 6398 4922
BIC 6955 5275 7065 5373 6537 5080
B ie sco e 0.004 0.003 0.004 0.003 0.004 0.003
BSS 0.015 −0.005 0.011 −0.001 0.032 0.011
Decile 1 0.586 0.563 0.575 0.505 0.683 0.617
Decile 2 0.192 0.150 0.152 0.148 0.143 0.150
Decile 3 0.070 0.089 0.081 0.103 0.063 0.099
Decile 4 0.049 0.070 0.062 0.056 0.046 0.023
Decile 5 0.029 0.021 0.041 0.056 0.017 0.031
Decile 6-10 0.075 0.106 0.089 0.131 0.048 0.080
Exclusi ely inancial
a iables
AUC 0.833 0.835 0.786 0.787 0.876 0.860
AP 0.035 0.022 0.026 0.016 0.055 0.037
AR 0.664 0.668 0.570 0.572 0.749 0.718
KS s a is ic 0.538 0.541 0.486 0.435 0.601 0.583
Hinge loss 1.004 1.004 1.004 1.004 1.004 1.004
Log loss 0.025 0.018 0.026 0.018 0.023 0.017
AIC 7078 5234 7443 5408 6543 4953
BIC 7137 5293 7502 5468 6651 5062
B ie sco e 0.004 0.003 0.004 0.003 0.004 0.003
BSS 0.009 −0.011 0.005 −0.005 0.027 0.008
Decile 1 0.560 0.577 0.573 0.500 0.667 0.608
Decile 2 0.167 0.157 0.102 0.129 0.125 0.169
Decile 3 0.097 0.082 0.049 0.080 0.084 0.082
Decile 4 0.051 0.052 0.065 0.092 0.041 0.040
Decile 5 0.032 0.038 0.060 0.061 0.024 0.023
Decile 6-10 0.094 0.094 0.151 0.138 0.059 0.077
No es: Es ima ion esul s o LR models ha p edic bank up cy o e a one-yea ho izon, using a iable se s selec ed by he LASSO me hod
om a popula ion o non- inancial a iables in Table 1 and a se o inancial a iables. The columns display he esul s o pe mu a ions using
all inancial s a emen s om 2018 and 2019 as ou -o -sample es samples and he a iables in Al man (1968), Al man and Saba o (2007)
and Pa aschi e al. (2023), p esen ed in Table 2, as he se o inancial a iables. Panel A p esen s he a iables selec ed using he LASSO
me hod, along wi h hei coe icien es ima es and z-sco es in pa en heses, de i ed om s anda d e o s clus e ed a he i m le el. Panels B
and C epo he in-sample i and ou -o -sample p edic ion pe o mance, espec i ely, using R2, AUC, AP, AR, KS s a is ic, Hinge loss, Log
loss, AIC, BIC, B ie sco e, BSS and decile ankings. The i s ows in Panels B and C epo me ic alues when using he a iable se s
selec ed by he LASSO me hod, shown in Panel A, om he popula ion o bo h non- inancial and inancial a iables. The bo om ows show
he me ic alues when he inancial a iables a e used exclusi ely pe benchma k se , as shown in 2. To ain he models, a ou -yea olling
window app oach was ollowed in which he models we e ained on all inancial s a emen s om he ou accoun ing yea s p eceding he
es popula ions. These da a a e p esen ed in 3.
Table 6 Con inued.
Al man (1968) Al man and Saba o (2007) Pa aschi e al. (2023)
2018 2019 2018 2019 2018 2019
BIC 24,959 25,882 25,478 26,570 23,122 23,990
B ie sco e 0.004 0.004 0.004 0.004 0.004 0.004
BSS 0.009 0.011 0.007 0.008 0.024 0.026
Decile 1 0.546 0.562 0.549 0.560 0.647 0.660
Decile 2 0.157 0.160 0.116 0.117 0.150 0.142
Decile 3 0.095 0.089 0.065 0.064 0.068 0.071
Decile 4 0.046 0.048 0.068 0.062 0.041 0.038
Decile 5 0.035 0.033 0.055 0.050 0.024 0.025
Decile 6-10 0.121 0.109 0.148 0.147 0.070 0.063
COGENT ECONOMICS & FINANCE 17
Table 7. Es ima ion esul s when p edic ing bank up cy o e a ho izon o wo yea s.
Panel A. Va iables selec ed using he LASSO me hod and hei LR coe icien es ima es when p edic ing bank up cy
o e a ho izon o wo yea s.
Al man (1968) Al man and Saba o (2007) Pa aschi e al. (2023)
2018 2019 2018 2019 2018 2019
CEO age −0.58(−6.96) −0.62(−7.79) −0.78(−9.42) −0.88(−10.93) −0.36(−4.30)
Chai pe son age −0.65(−5.95) −0.77(−7.39) −0.73(−6.92) −0.69(−6.62) −0.35(−3.17) −0.34(−3.10)
Boa d age a g −0.70(−5.66) −0.69(−5.84) −0.83(−6.97) −0.77(−6.51) −0.60(−5.10) −0.38(−3.01)
Boa d coun y −0.92(−27.23) −0.71(−22.61) −0.97(−28.66) −0.92(−28.12) −0.88(−25.92) −0.83(−25.41)
Boa d non-owne s −0.38(−6.04) −0.51(−15.89)
Owne ship boa d 0.62(19.15) 0.84(25.83) 0.42(6.62) 0.56(17.03)
EBIT / o al asse s −1.24(−24.63) −1.17(−24.15)
Re ained ea nings / o al asse s −0.37(−14.53) −0.28(−11.81) −0.68(−34.59) −0.65(−35.15)
Sales / o al asse s 0.15(26.36) 0.17(29.56)
Wo king capi al / o al asse s −0.73(−20.57) −0.76(−22.36)
EBITDA / o al asse s −1.42(−27.13) −1.36(−27.02)
Sho - e m liquidi y / o al
asse s
−1.54(−22.57) −1.46(−22.51)
(Cu en liabili ies - sho - e m
liquidi y) / o al asse s
0.24(5.39) 0.20(4.57)
Accoun s payable / o al asse s 1.44(22.63) 1.46(23.44)
Dummy; one i paid-in equi y
is less han o al equi y
−0.57(−13.77) −0.60(−15.00)
Dummy; one i o al liabili y
exceeds o al asse s
0.35(8.03) 0.36(8.41)
In en o y / cu en asse s 0.53(11.24) 0.45(9.73)
Log(age in yea s) −0.34(−26.40) −0.32(−24.94)
Ne income / o al asse s −0.63(−11.94) −0.61(−12.13)
Public axes payable / o al
asse s
4.33(36.71) 4.17(36.60)
Sho - e m liquidi y / cu en
asse s
−1.37(−20.74) −1.36(−21.44)
In e cep 3.00(12.10) 3.74(15.76) 5.16(21.56) 5.50(23.36) −0.19(−0.72) 0.84(3.24)
Panel B: In-sample i when p edic ing bank up cy o e a ho izon o wo yea s.
Al man (1968) Al man and Saba o (2007) Pa aschi e al. (2023)
2018 2019 2018 2019 2018 2019
Financial and
non- inancial
a iables
R20.126 0.119 0.117 0.112 0.195 0.189
AUC 0.814 0.810 0.802 0.797 0.867 0.862
AP 0.063 0.059 0.055 0.054 0.098 0.095
AR 0.622 0.612 0.597 0.587 0.725 0.717
KS s a is ic 0.479 0.475 0.459 0.451 0.583 0.571
Hinge loss 1.010 1.010 1.010 1.010 1.010 1.010
Log loss 0.053 0.054 0.054 0.055 0.049 0.050
AIC 56,248 59,217 56,875 59,691 51,846 54,557
BIC 56,359 59,318 56,976 59,803 52,003 54,726
B ie sco e 0.011 0.011 0.011 0.011 0.010 0.011
BSS 0.020 0.018 0.019 0.019 0.045 0.043
Decile 1 0.485 0.467 0.457 0.455 0.603 0.591
Decile 2 0.178 0.182 0.182 0.174 0.171 0.174
Decile 3 0.106 0.118 0.115 0.113 0.083 0.089
Decile 4 0.077 0.074 0.075 0.081 0.054 0.053
Decile 5 0.051 0.053 0.060 0.056 0.032 0.034
Decile 6-10 0.102 0.106 0.112 0.120 0.057 0.061
Exclusi ely
inancial
a iables
R20.097 0.096 0.076 0.073 0.179 0.174
AUC 0.789 0.790 0.756 0.754 0.858 0.855
AP 0.051 0.051 0.040 0.039 0.085 0.083
AR 0.572 0.573 0.507 0.503 0.709 0.702
KS s a is ic 0.458 0.461 0.387 0.380 0.564 0.558
Hinge loss 1.010 1.010 1.011 1.011 1.010 1.010
Log loss 0.055 0.056 0.056 0.057 0.050 0.051
AIC 58,131 60,759 59,500 62,369 52,833 55,530
BIC 58,198 60,827 59,567 62,437 52,956 55,653
B ie sco e 0.011 0.011 0.011 0.011 0.011 0.011
BSS 0.013 0.012 0.010 0.009 0.037 0.035
Decile 1 0.430 0.432 0.427 0.421 0.576 0.568
Decile 2 0.197 0.195 0.145 0.145 0.176 0.179
Decile 3 0.125 0.128 0.109 0.109 0.093 0.096
Decile 4 0.073 0.071 0.082 0.087 0.058 0.056
Decile 5 0.046 0.044 0.062 0.063 0.037 0.038
Decile 6-10 0.129 0.130 0.175 0.175 0.061 0.064
18 R.R. WAHLSTRØM ET AL.
six columns o Table 7. Mo eo e , he six panels o Figu es IA.9, IA.10 and IA.11 show he LASSO pa h
plo s o he six modeling pe mu a ions shown in he six columns o Table 8. As explained in Sec ion
5.2, he LASSO pa h plo s a e gene a ed by epea edly minimizing Equa ion (3) wi h a ying k alues.
Ini ially, kis se high enough ha he e m kjjbjj1domina es, causing all es ima ed coe icien s o be
ze o. This can be obse ed on he a -le side o he plo s. As kg adually dec eases, mo ing o he igh
on he plo s, mo e coe icien s become non-ze o and en e he model. Va iables ha become non-ze o
a highe k alues ( u he o he le in he plo s) ha e s onge p edic i e powe and hus g ea e
impo ance compa ed o a iables ha become non-ze o a lowe k alues ( u he o he igh in he
plo s). We obse e ha he LASSO me hod selec s inancial a iables be o e non- inancial ones, as inan-
cial a iables become non-ze o a highe k alues ( u he o he le in he plo s). Mo eo e , in mos
cases, he inancial a iables a e selec ed a much highe k alues han he non- inancial a iables, u -
he indica ing hei ela i e impo ance.
We also obse e ha he coe icien signs o all inancial a iables in Tables 6,7and 8 ollow he
expec ed di ec ions.
13
The only excep ion being he posi i e sign o ‘sales / o al asse s’ o he a iable
se in Al man (1968). Howe e , his a iable is excluded om he e ised e sions o Al man’s(1968)
Panel C. Ou -o -sample p edic ion pe o mance when p edic ing bank up cy o e a ho izon o wo yea s.
Al man (1968) Al man and Saba o (2007) Pa aschi e al. (2023)
2018 2019 2018 2019 2018 2019
Financial and
non- inancial
a iables
AUC 0.797 0.799 0.776 0.782 0.847 0.852
AP 0.056 0.035 0.045 0.029 0.086 0.056
AR 0.587 0.593 0.545 0.561 0.686 0.700
KS s a is ic 0.455 0.467 0.414 0.430 0.546 0.554
Hinge loss 1.010 1.010 1.010 1.010 1.009 1.009
Log loss 0.055 0.038 0.056 0.038 0.051 0.035
AIC 15,771 11,270 16,075 11,388 14,740 10,484
BIC 15,870 11,359 16,164 11,487 14,878 10,633
B ie sco e 0.011 0.007 0.011 0.007 0.011 0.007
BSS 0.013 −0.011 0.010 −0.007 0.036 0.009
Decile 1 0.450 0.448 0.405 0.415 0.548 0.555
Decile 2 0.187 0.193 0.182 0.192 0.182 0.186
Decile 3 0.102 0.116 0.117 0.112 0.091 0.094
Decile 4 0.078 0.078 0.086 0.079 0.063 0.070
Decile 5 0.060 0.055 0.068 0.065 0.043 0.029
Decile 6-10 0.122 0.110 0.142 0.137 0.073 0.065
Exclusi ely
inancial
a iables
AUC 0.797 0.792 0.751 0.749 0.846 0.850
AP 0.049 0.033 0.038 0.024 0.078 0.054
AR 0.588 0.579 0.497 0.494 0.684 0.696
KS s a is ic 0.478 0.474 0.386 0.368 0.544 0.541
Hinge loss 1.011 1.011 1.010 1.011 1.010 1.010
Log loss 0.056 0.039 0.060 0.040 0.051 0.036
AIC 16,021 11,474 17,368 11,797 14,823 10,579
BIC 16,080 11,533 17,428 11,856 14,931 10,688
B ie sco e 0.011 0.007 0.011 0.007 0.011 0.007
BSS 0.007 −0.015 0.004 −0.008 0.030 0.005
Decile 1 0.439 0.450 0.417 0.427 0.543 0.553
Decile 2 0.194 0.188 0.150 0.125 0.189 0.173
Decile 3 0.138 0.126 0.114 0.105 0.099 0.103
Decile 4 0.073 0.067 0.088 0.089 0.056 0.067
Decile 5 0.037 0.041 0.065 0.072 0.041 0.038
Decile 6-10 0.119 0.127 0.167 0.182 0.073 0.066
No es: Es ima ion esul s o LR models ha p edic bank up cy o e a wo-yea ho izon, using a iable se s selec ed by he LASSO me hod
om a popula ion o non- inancial a iables in Table 1 and a se o inancial a iables. The columns display he esul s o pe mu a ions using
all inancial s a emen s om 2018 and 2019 as ou -o -sample es samples and he a iables in Al man (1968), Al man and Saba o (2007)
and Pa aschi e al. (2023), p esen ed in Table 2, as he se o inancial a iables. Panel A p esen s he a iables selec ed using he LASSO
me hod, along wi h hei coe icien es ima es and z-sco es in pa en heses, de i ed om s anda d e o s clus e ed a he i m le el. Panels B
and C epo he in-sample i and ou -o -sample p edic ion pe o mance, espec i ely, using R2, AUC, AP, AR, KS s a is ic, Hinge loss, Log
loss, AIC, BIC, B ie sco e, BSS and decile ankings. The i s ows in Panels B and C epo me ic alues when using he a iable se s
selec ed by he LASSO me hod, shown in Panel A, om he popula ion o bo h non- inancial and inancial a iables. The bo om ows show
he me ic alues when he inancial a iables a e used exclusi ely pe benchma k se , as shown in 2. To ain he models, a ou -yea olling
window app oach was ollowed in which he models we e ained on all inancial s a emen s om he ou accoun ing yea s p eceding he
es popula ions. These da a a e p esen ed in 3.
COGENT ECONOMICS & FINANCE 19

model, which a ge s p i a e companies ac oss all indus ies, because i is highly indus y-sensi i e
(Al man, 2018; Al man e al., 2019).
Mo eo e , Panels B and C o Tables 6,7and 8show ha among he h ee inancial a iable se s, he
one in Pa aschi e al. (2023) yields he highes in-sample i and ou -o -sample p edic ion pe o mance
ac oss all p edic ion ho izons, yea s as ou es sample and whe he using inancial a iables exclusi ely
o en iching hem wi h non- inancial a iables. This is expec ed because Pa aschi e al. (2023) selec ed
hei a iable se wi h a sample simila in many espec s o ha we use in ou s udy.
6.1.2. Enhancing models wi h non- inancial a iables
Fu he mo e, we obse e ha en iching inancial a iables wi h non- inancial ones imp o es bank up cy
p edic ion models. Speci ically, as shown in Panels B and C o Tables 6,7and 8, using a iable se s con-
aining bo h inancial and non- inancial a iables ( op ows in he panels) esul s in a be e in-sample i
and ou -o -sample p edic ion pe o mance ac oss all e alua ion me ics compa ed wi h using exclusi ely
inancial a iables (bo om ows o he panel).
The imp o ed model pe o mance is due o he six non- inancial a iables selec ed by he LASSO
me hod. Speci ically, he LASSO me hod consis en ly selec s ‘chai pe son age’ac oss all 18 pe mu a ions
o he inancial a iable se s, yea s used as he es sample and p edic ion ho izons. Addi ionally, ‘CEO
age’and ‘boa d age a g’a e selec ed in all excep one and wo, espec i ely, cases. In all cases, we
obse e a nega i e coe icien o hese a iables, indica ing ha an olde age esul s in a lowe p ob-
abili y o bank up cy. This inding is consis en wi h Pla and Pla (2012) inding ha olde CEOs and
boa d membe s a e associa ed wi h less bank up cy. Fu he mo e, he LASSO me hod consis en ly
selec s he a iable ‘boa d coun y’in all cases, which indica es whe he boa d membe s eside in he
same coun y as he SMEs’HQs. The coe icien s o his a iable a e nega i e in all cases, indica ing a
lowe p obabili y o bank up cy i boa d membe s eside in he same coun y as he SME. Finally, in mos
cases, he LASSO me hod selec s he a iables ‘boa d non-owne s’and ‘owne ship boa d’. The es ima ed
coe icien s o hese a iables consis en ly ha e nega i e and posi i e signs, espec i ely. This inding
indica es ha a highe p opo ion o boa d membe s who a e no sha eholde s dec eases he SME’s like-
lihood o bank up cy, and a lowe sha e o owne ship by boa d membe s also dec eases he likelihood
o bank up cy. This may be because boa ds wi h a lowe owne ship sha e ha e mo e ou side s among
boa d membe s, which can s eng hen he boa d by p o iding mo e di e si y han boa ds wi h high
owne ship in he SME.
6.1.3. Only non- inancial a iables
Nex , we es he pe o mance o a iable se s consis ing only o non- inancial a iables. Such a iable
se s a e use ul o assessing, o example, newly es ablished i ms wi hou hei i s inancial s a emen .
We c ea e hese a iable se s by allowing he LASSO me hod o conside only he 20 non- inancial a ia-
bles in Table 1.Table 9, Panel A shows he selec ed a iables and es ima ion esul s when hey a e used
in LR models using 2018 and 2019 as es samples and when p edic ing bank up cy o e ho izons o
one, wo and h ee yea s. Panel B p esen s he in-sample i and ou -o -sample p edic ion pe o mance.
While Panel B o Table 9 indica es a be e p edic ion han andom when exclusi ely using non- inan-
cial a iables; speci ically, AUC >0:5, he p edic ion pe o mance is no conside ed accep able as AUC
<0:7 (see Sec ion 5.4). Fu he mo e, we obse e ha ac oss all e alua ion me ics, he in-sample i and
ou -o -sample p edic ion pe o mance a e highe when using a iable se s consis ing exclusi ely o
inancial a iables o inancial and non- inancial a iables combined (see Panels B and C o Tables 6,7
and 8) han when using only non- inancial a iables (see Panel B o Table 9). This u he p o es he
impo ance o inancial a iables o bank up cy p edic ion, while non- inancial a iables should be used
only in combina ion wi h inancial a iables.
Fu he mo e, he impo ance o he non- inancial a iables in Tables 6,7and 8is suppo ed as all
hese a iables a e p esen in Table 9 wi h he same sign o he es ima ed coe icien alues. The only
excep ion is ‘boa d non-owne s’, which has ano he sign o he es ima ed coe icien in some cases in
Table 9 compa ed wi h Tables 6,7and 8. Howe e , he es ima ed coe icien s a e no s a is ically signi i-
can in hese cases. Mo eo e , we obse e in Figu es IA.4, IA.8 and IA.12 in he In e ne Appendix, which
20 R.R. WAHLSTRØM ET AL.
Table 8. Es ima ion esul s when p edic ing bank up cy o e a ho izon o h ee yea s.
Panel A. Va iables selec ed using he LASSO me hod and hei LR coe icien es ima es when p edic ing bank up cy
o e a ho izon o h ee yea s.
Al man (1968) Al man and Saba o (2007) Pa aschi e al. (2023)
2018 2019 2018 2019 2018 2019
CEO age −0.70(−8.56) −0.72(−8.57) −0.92(−11.19) −0.91(−10.99) −0.45(−5.29) −0.45(−5.30)
Chai pe son age −0.93(−8.82) −0.69(−6.27) −0.86(−8.09) −0.75(−6.92) −0.51(−4.67) −0.42(−3.74)
Boa d age a g −0.34(−2.89) −0.40(−3.20) −0.38(−3.14) −0.54(−4.34) −0.06(−0.47) −0.18(−1.41)
Boa d coun y −0.74(−23.24) −0.80(−23.77) −0.97(−28.85) −0.85(−25.04) −0.88(−26.25) −0.77(−22.86)
Boa d non-owne s −0.47(−14.49) −0.33(−5.19) −0.65(−20.20) −0.28(−4.41) −0.50(−15.33)
Owne ship boa d 0.42(6.57) 0.27(4.19)
EBIT / o al asse s −0.88(−17.33) −0.84(−16.61)
Re ained ea nings / o al asse s −0.31(−12.14) −0.32(−12.54) −0.62(−31.11) −0.59(−30.05)
Sales / o al asse s 0.15(25.77) 0.14(23.26)
Wo king capi al / o al asse s −0.60(−16.67) −0.63(−17.41)
EBITDA / o al asse s −0.95(−17.96) −0.90(−16.85)
Sho - e m liquidi y / o al
asse s
−1.34(−20.57) −1.36(−20.71)
(Cu en liabili ies - sho - e m
liquidi y) / o al asse s
0.29(6.40) 0.29(6.60)
Accoun s payable / o al asse s 1.20(18.33) 1.21(18.46)
Dummy; one i paid-in equi y
is less han o al equi y
−0.53(−13.53) −0.52(−12.89)
Dummy; one i o al liabili y
exceeds o al asse s
0.17(3.93) 0.21(4.59)
In e es expenses / o al asse s 1.99(6.44)
In en o y / cu en asse s 0.48(10.16) 0.35(7.37)
Log(age in yea s) −0.29(−22.49) −0.31(−23.55)
Ne income / o al asse s −0.39(−7.15) −0.27(−4.89)
Public axes payable / o al
asse s
3.34(27.74) 3.08(25.52)
Sho - e m liquidi y / cu en
asse s
−1.13(−18.11) −1.10(−17.72)
In e cep 3.43(13.89) 3.04(12.48) 4.81(19.60) 5.20(22.21) 0.48(1.76) 0.76(2.84)
Panel B. In-sample i when p edic ing bank up cy o e a ho izon o h ee yea s.
Al man (1968) Al man and Saba o (2007) Pa aschi e al. (2023)
2018 2019 2018 2019 2018 2019
Financial and
non- inancial
a iables
R20.088 0.088 0.085 0.081 0.137 0.133
AUC 0.777 0.777 0.767 0.764 0.826 0.825
AP 0.041 0.040 0.040 0.035 0.060 0.055
AR 0.548 0.549 0.528 0.522 0.644 0.643
KS s a is ic 0.410 0.416 0.398 0.391 0.505 0.500
Hinge loss 1.010 1.010 1.010 1.010 1.010 1.010
Log loss 0.056 0.052 0.056 0.053 0.053 0.050
AIC 58,635 57,231 58,773 57,678 55,455 54,415
BIC 58,736 57,343 58,885 57,779 55,634 54,594
B ie sco e 0.011 0.010 0.011 0.010 0.011 0.010
BSS 0.010 0.010 0.012 0.010 0.023 0.021
Decile 1 0.393 0.405 0.397 0.383 0.499 0.488
Decile 2 0.189 0.188 0.175 0.182 0.192 0.195
Decile 3 0.122 0.114 0.121 0.119 0.105 0.109
Decile 4 0.091 0.092 0.086 0.095 0.071 0.075
Decile 5 0.073 0.064 0.070 0.068 0.046 0.049
Decile 6-10 0.132 0.137 0.151 0.153 0.088 0.084
Exclusi ely
inancial
a iables
R
2
0.066 0.064 0.050 0.048 0.122 0.119
AUC 0.758 0.753 0.723 0.719 0.814 0.813
AP 0.035 0.033 0.028 0.026 0.052 0.048
AR 0.509 0.502 0.441 0.434 0.621 0.620
KS s a is ic 0.400 0.394 0.325 0.317 0.486 0.484
Hinge loss 1.011 1.010 1.011 1.010 1.010 1.010
Log loss 0.057 0.054 0.058 0.055 0.054 0.051
AIC 60,037 58,707 61,014 59,739 56,434 55,286
BIC 60,104 58,774 61,081 59,806 56,557 55,409
B ie sco e 0.011 0.010 0.011 0.010 0.011 0.010
BSS 0.006 0.006 0.005 0.004 0.018 0.017
Decile 1 0.348 0.351 0.331 0.328 0.463 0.452
Decile 2 0.205 0.195 0.162 0.154 0.197 0.203
Decile 3 0.139 0.140 0.125 0.122 0.119 0.123
Decile 4 0.092 0.093 0.100 0.106 0.073 0.076
Decile 5 0.062 0.061 0.082 0.086 0.052 0.053
Decile 6-10 0.153 0.160 0.201 0.204 0.096 0.093
COGENT ECONOMICS & FINANCE 21
p esen he LASSO pa h plo s o six modeling pe mu a ions shown in he six columns o Table 9, ha
‘boa d non-owne s’is among he las chosen by he LASSO me hod, ha is, a low k alues.
6.2. Robus ness es s
6.2.1. Robus ness using balanced da ase s
As wi h any bank up cy p edic ion da ase , ou sample is highly imbalanced; he a e o inancial s a e-
men s ca ego ized as bank up is much lowe han hose ca ego ized as non-bank up (Bea e e al.,
2010). We acknowledge ha sampling o c ea e a balanced aining da ase would dis o he capabil-
i ies o a bank up cy p edic ion model because he a io o non-bank up o bank up obse a ions in
he da a used o de elopmen de ia es om he eal-wo ld popula ion (Zmijewski, 1984). Howe e , in
his sec ion, we in es iga e he obus ness o ou main esul s when using balanced da ase s because
his app oach has been shown o imp o e classi ica ion accu acy signi ican ly (J. Ga cia, 2022).
Speci ically, we use SMOTE (Fe nandez e al., 2018) o gene a e syn he ic samples o he mino i y class
by in e pola ing be ween exis ing mino i y class samples. SMOTE selec s a mino i y class sample, iden i-
ies i s k-nea es neighbo s and c ea es new syn he ic samples along he line segmen s joining he
Panel C. Ou -o -sample p edic ion pe o mance when p edic ing bank up cy o e a ho izon o h ee yea s.
Al man (1968) Al man and Saba o (2007) Pa aschi e al. (2023)
2018 2019 2018 2019 2018 2019
Financial and
non- inancial
a iables
AUC 0.752 0.785 0.741 0.770 0.810 0.840
AP 0.025 0.009 0.023 0.008 0.035 0.015
AR 0.499 0.569 0.479 0.538 0.615 0.679
KS s a is ic 0.375 0.436 0.358 0.406 0.487 0.531
Hinge loss 1.010 1.010 1.010 1.009 1.010 1.009
Log loss 0.043 0.017 0.043 0.017 0.040 0.016
AIC 12,270 5,058 12,326 5,064 11,679 4,780
BIC 12,359 5,157 12,425 5,153 11,837 4,939
B ie sco e 0.008 0.002 0.008 0.002 0.008 0.002
BSS −0.007 −0.119 −0.006 −0.093 0.000 −0.117
Decile 1 0.342 0.445 0.348 0.415 0.440 0.570
Decile 2 0.182 0.166 0.174 0.147 0.208 0.155
Decile 3 0.138 0.117 0.127 0.125 0.129 0.091
Decile 4 0.106 0.083 0.102 0.102 0.072 0.072
Decile 5 0.073 0.057 0.068 0.087 0.058 0.045
Decile 6-10 0.158 0.132 0.182 0.125 0.092 0.068
Exclusi ely
inancial
a iables
AUC 0.741 0.768 0.700 0.726 0.809 0.840
AP 0.024 0.007 0.019 0.005 0.036 0.013
AR 0.479 0.536 0.396 0.451 0.613 0.679
KS s a is ic 0.380 0.426 0.305 0.328 0.475 0.539
Hinge loss 1.011 1.010 1.010 1.010 1.010 1.010
Log loss 0.043 0.018 0.046 0.018 0.041 0.017
AIC 12,427 5,290 13,248 5,392 11,718 4,950
BIC 12,486 5,349 13,307 5,451 11,827 5,059
B ie sco e 0.008 0.002 0.008 0.002 0.008 0.002
BSS −0.007 −0.124 −0.005 −0.086 0.003 −0.127
Decile 1 0.348 0.411 0.335 0.351 0.430 0.551
Decile 2 0.179 0.192 0.133 0.162 0.206 0.185
Decile 3 0.148 0.109 0.124 0.098 0.131 0.087
Decile 4 0.083 0.083 0.099 0.109 0.082 0.060
Decile 5 0.068 0.045 0.100 0.068 0.057 0.026
Decile 6-10 0.174 0.158 0.209 0.211 0.093 0.091
No es: Es ima ion esul s o LR models ha p edic bank up cy o e a h ee-yea ho izon, using a iable se s selec ed by he LASSO me hod
om a popula ion o non- inancial a iables in Table 1 and a se o inancial a iables. The columns display he esul s o pe mu a ions using
all inancial s a emen s om 2018 and 2019 as ou -o -sample es samples, and he a iables in Al man (1968), Al man and Saba o (2007)
and Pa aschi e al. (2023), p esen ed in Table 2, as he se o inancial a iables. Panel A p esen s he a iables selec ed using he LASSO
me hod, along wi h hei coe icien es ima es and z-sco es in pa en heses, de i ed om s anda d e o s clus e ed a he i m le el. Panels B
and C epo he in-sample i and ou -o -sample p edic ion pe o mance, espec i ely, using R2, AUC, AP, AR, KS s a is ic, Hinge loss, Log
loss, AIC, BIC, B ie sco e, BSS and decile ankings. The i s ows in Panels B and C epo me ic alues when using he a iable se s
selec ed by he LASSO me hod, shown in Panel A, om he popula ion o bo h non- inancial and inancial a iables. The bo om ows show
he me ic alues when he inancial a iables a e used exclusi ely pe benchma k se , as shown in 2. To ain he models, a ou -yea olling
window app oach was ollowed in which he models we e ained on all inancial s a emen s om he ou accoun ing yea s p eceding he
es popula ions. These da a a e p esen ed in 3.
22 R.R. WAHLSTRØM ET AL.
selec ed sample and i s neighbo s. This app oach helps balance he class dis ibu ion wi hou duplica -
ing he exis ing mino i y class samples. We use his app oach on he da a o each accoun ing yea
h ice, depending on whe he we ca ego ize bank up cy o e one, wo, o h ee-yea ho izons, esul ing
in h ee da ase s wi h 1,631,396, 1,620,862 and 1,623,426 inancial s a emen s, espec i ely, each wi h
50% o obse a ions ca ego ized as bank up pe accoun ing yea .
Tables IA.1, IA.2 and IA.3 in In e ne Appendix IA.4 p esen he model pe o mance when using bal-
anced da ase s in p edic ing bank up cy o e ho izons o one yea , wo yea s and h ee yea s, espec -
i ely. In each able, Panels A and B exhibi he in-sample i and ou -o -sample p edic ion pe o mance,
espec i ely. As be o e, we le he LASSO me hod selec a iables om a popula ion o non- inancial a -
iables in Table 1 and a se o inancial a iables, and we measu e he pe o mance o using he selec ed
models in LR models. The pe o mance o hese models is displayed in he ables’ op ows. The bo om
ows show he me ic alues when using exclusi ely he inancial a iables pe benchma k se . The
ables’columns display he esul s o pe mu a ions, using all inancial s a emen s om 2018 and 2019 as
ou -o -sample es samples and he a iables in Al man (1968), Al man and Saba o (2007) and Pa aschi
e al. (2023) as he se o inancial a iables. We also ollow he same ou -yea olling window app oach.
In summa y, Tables IA.1, IA.2 and IA.3 a e compa able o Tables 6,7and 8, bu o when using balanced
da ase s.
We obse e ha Tables IA.1, IA.2 and IA.3 con i m ou indings. Speci ically, ac oss all me ics and es
se ings, a iable se s ha include inancial and non- inancial a iables ( op ows o panels) yield supe -
io in-sample i and ou -o -sample p edic ion pe o mance compa ed o se s ha use only he inancial
a iables (bo om ows o panels). We also no e ha among he h ee inancial a iable se s, he one in
Pa aschi e al. (2023) consis en ly yields he highes in-sample i and ou -o -sample p edic ion pe o m-
ance. Addi ionally, we obse e ha while he AUC is a he same le el as when no employing a bal-
anced da ase (see Tables 6,7and 8), he AP is highe when using balanced da ase s. This is expec ed
because in he con ex o highly imbalanced da ase s, i is common o obse e ha he AP is signi i-
can ly lowe han he AUC. This disc epancy a ises om he di e en sensi i i ies o hese me ics o
class imbalances. The AUC e alua es he model’s abili y o dis inguish be ween classes by conside ing
bo h ue posi i e and alse posi i e a es, which can esul in decep i ely high alues owing o he la ge
numbe o ue nega i es. In con as , AP ocuses on he p ecision- ecall ade-o , di ec ly e lec ing he
model’s pe o mance on he mino i y class. Consequen ly, he AP is mo e sensi i e o he model’s abili y
o co ec ly iden i y he mino i y class, o en esul ing in lowe alues in imbalanced da ase s.
Fu he mo e, Table IA.4 in he In e ne Appendix displays he model pe o mance when conside ing
exclusi ely non- inancial a iables and using he balanced da ase s. As wi h no using a balanced da ase
(see Table 9), he in-sample i and ou -o -sample p edic ion pe o mance is o e all highe ac oss all
e alua ion me ics when using a iable se s o exclusi ely inancial a iables o combined inancial and
non- inancial a iables (see Tables IA.1, IA.2 and IA.3), compa ed wi h using only non- inancial a iables
(see Table IA.4).
Figu es IA.13, IA.14 and IA.15 in In e ne Appendix IA.4 p esen he associa ed LASSO pa h plo s,
when using he balanced da ase s, o he a iables selec ed among he non- inancial ones and hose in
Al man (1968), Al man and Saba o (2007) and Pa aschi e al. (2023), espec i ely, when p edic ing bank-
up cy o e a ho izon o one yea . In each igu e, Panels A and B p esen he plo s using 2018 and 2019
as ou -o -sample es samples. Fu he mo e, Figu es IA.17, IA.18 and IA.19 p esen he same bu o p e-
dic ing bank up cy o e a ho izon o wo yea s, whe eas Figu es IA.21, IA.22 and IA.23 p esen he same
bu o p edic ing bank up cy o e a ho izon o h ee yea s. Mo eo e , Figu es IA.16, IA.20 and IA.24 p e-
sen he LASSO pa h plo s o he a iables selec ed among only he non- inancial a iables when using
a balanced da ase and p edic ing bank up cy o e ho izons o one yea , wo yea s and h ee yea s,
espec i ely.
As wi h no using balanced da ase s, we obse e om he LASSO pa h plo s ha inancial a iables
a e selec ed be o e non- inancial ones, as inancial a iables become non-ze o a highe k alues ( u he
o he le in he plo s), and ha in mos cases, he inancial a iables a e selec ed a much highe k
alues. Thus, he LASSO pa h plo s also con i m he impo ance o he inancial a iables when he
balanced da ase s we e used. Mo eo e , we obse e ha when using he balanced da ase s, mo e non-
inancial a iables a e used han when using non-balanced da ase s (see Sec ion 6.1.2). Howe e , he
COGENT ECONOMICS & FINANCE 23
Engelmann, B., Hayden, E., & Tasche, D. (2003). Measu ing he disc imina i e powe o a ing sys ems. Risk,16,
82–86.
Eu opean Commission. (2020). An SME s a egy o a sus ainable and digi al Eu ope (Technical Repo ).
Fe nandez, A., Ga cia, S., He e a, F., & Chawla, N. V. (2018). SMOTE o lea ning om imbalanced da a: P og ess and
challenges, ma king he 15-yea anni e sa y. Jou nal o A i icial In elligence Resea ch,61, 863–905. h ps://doi.o g/
10.1613/jai .1.11192
F eund, Y., & Schapi e, R. E. (1997). A decision- heo e ic gene aliza ion o on-line lea ning and an applica ion o
boos ing. Jou nal o Compu e and Sys em Sciences,55(1), 119–139. h ps://doi.o g/10.1006/jcss.1997.1504
F iedman, J. H., Has ie, T., & Tibshi ani, R. (2010). Regula iza ion pa hs o gene alized linea models ia coo dina e
descen . Jou nal o S a is ical So wa e,33(1), 22. h ps://doi.o g/10.18637/jss. 033.i01
Fus e , A., Goldsmi h-Pinkham, P., Ramado ai, T., & Wal he , A. (2022). P edic ably unequal? The e ec s o machine
lea ning on c edi ma ke s. The Jou nal o Finance,77(1), 5–47. h ps://doi.o g/10.1111/jo i.13090
G20 & OECD. (2015). G20/OECD high-le el p inciples on SME inancing (Technical Repo ).
Ga c
ıa, C. J., & He e o, B. (2021). Female di ec o s, capi al s uc u e, and inancial dis ess. Jou nal o Business
Resea ch,136, 592–601. h ps://doi.o g/10.1016/j.jbus es.2021.07.061
Ga cia, J. (2022). Bank up cy p edic ion using syn he ic sampling. Machine Lea ning wi h Applica ions,9, 100343.
h ps://doi.o g/10.1016/j.mlwa.2022.100343
G une , J., No den, L., & Webe , M. (2005). The ole o non- inancial ac o s in in e nal c edi a ings. Jou nal o
Banking & Finance,29(2), 509–531. h ps://doi.o g/10.1016/j.jbank in.2004.05.017
Gues , P. M. (2009). The impac o boa d size on i m pe o mance: E idence om he UK. The Eu opean Jou nal o
Finance,15(4), 385–404. h ps://doi.o g/10.1080/13518470802466121
Habib, A., Cos a, M. D., Huang, H. J., Bhuiyan, M. B. U., & Sun, L. (2020). De e minan s and consequences o inancial dis-
ess: Re iew o he empi ical li e a u e. Accoun ing & Finance,60(S1), 1023–1075. h ps://doi.o g/10.1111/ac i.12400
Hau sch, N., Schaumbu g, J., & Schienle, M. (2015). Financial ne wo k sys emic isk con ibu ions. Re iew o Finance,
19(2), 685–738. h ps://doi.o g/10.1093/ o / u010
Hillegeis , S. A., Kea ing, E. K., C am, D. P., & Lunds ed , K. G. (2004). Assessing he p obabili y o bank up cy. Re iew
o Accoun ing S udies,9(1), 5–34. h ps://doi.o g/10.1023/B:RAST.0000013627.90884.b7
Hodges, J., J . (1958). The signi icance p obabili y o he Smi no wo-sample es . A ki F€
o Ma ema ik,3(5), 469–
486. h ps://doi.o g/10.1007/BF02589501
Hosme , D. W., Lemeshow, S., & S u di an , R. X. (2013). Applied logis ic eg ession. John Wiley & Sons.
Jensen, M. C. (1993). The mode n indus ial e olu ion, exi , and he ailu e o in e nal con ol sys ems. The Jou nal
o Finance,48(3), 831–880. h ps://doi.o g/10.1111/j.1540-6261.1993. b04022.x
Jianakoplos, N. A., & Be nasek, A. (1998). A e women mo e isk a e se? Economic Inqui y,36(4), 620–630. h ps://doi.
o g/10.1111/j.1465-7295.1998. b01740.x
K ause, R., Semadeni, M., & Cannella, A. A. (2014). CEO duali y: A e iew and esea ch agenda. Jou nal o
Managemen ,40(1), 256–286. h ps://doi.o g/10.1177/0149206313503013
K ause, R., Semadeni, M., & Wi he s, M. C. (2016). Tha special someone: When he boa d iews i s chai as a
esou ce. S a egic Managemen Jou nal,37(9), 1990–2002. h ps://doi.o g/10.1002/smj.2444
Kuizinien_
e, D., K ila i
cius, T., Dama
se i
cius, R., & Maskeli
unas, R. (2022). Sys ema ic e iew o inancial dis ess iden i-
ica ion using a i icial in elligence me hods. Applied A i icial In elligence,36(1), 2138124. h ps://doi.o g/10.1080/
08839514.2022.2138124
Lai inen, E. K. (1999). P edic ing a co po a e c edi analys ’s isk es ima e by logis ic and linea models. In e na ional
Re iew o Financial Analysis,8(2), 97–121. h ps://doi.o g/10.1016/S1057-5219(99)00012-5
Lee, J. D., Sun, D. L., Sun, Y., & Taylo , J. E. (2016). Exac pos -selec ion in e ence, wi h applica ion o he lasso. The
Annals o S a is ics,44(3), 907–927. h ps://doi.o g/10.1214/15-AOS1371
Liang, D., Tsai, C. F., Lu, H. Y. R., & Chang, L. S. (2020). Combining co po a e go e nance indica o s wi h s acking
ensembles o inancial dis ess p edic ion. Jou nal o Business Resea ch,120, 137–146. h ps://doi.o g/10.1016/j.
jbus es.2020.07.052
Libe i, J. M., & Pe e sen, M. A. (2019). In o ma ion: Ha d and so . The Re iew o Co po a e Finance S udies,8(1), 1–41.
h ps://doi.o g/10.1093/ c s/c y009
Lilien eld-Toal, U. V., & Ruenzi, S. (2014). CEO owne ship, s ock ma ke pe o mance, and manage ial disc e ion. The
Jou nal o Finance,69(3), 1013–1050. h ps://doi.o g/10.1111/jo i.12139
Lohmann, C., M€
ollenho , S., & Ohlige , T. (2023). Nonlinea ela ionships in bank up cy p edic ion and hei e ec on
he p o i abili y o bank up cy p edic ion models. Jou nal o Business Economics,93(9), 1661–1690. h ps://doi.o g/
10.1007/s11573-022-01130-8
Lohmann, C., & Ohlige , T. (2018). Nonlinea ela ionships in a logis ic model o de aul o a high-de aul ins allmen
po olio. Jou nal o C edi Risk,14(1), 45–68. h ps://doi.o g/10.21314/JCR.2017.232
Lohmann, C., & Ohlige , T. (2020). Bank up cy p edic ion and he disc imina o y powe o annual epo s: Empi ical
e idence om inancially dis essed Ge man companies. Jou nal o Business Economics,90(1), 137–172. h ps://doi.
o g/10.1007/s11573-019-00938-1
Mai, F., Tian, S., Lee, C., & Ma, L. (2019). Deep lea ning models o bank up cy p edic ion using ex ual disclosu es.
Eu opean Jou nal o Ope a ional Resea ch,274(2), 743–758. h ps://doi.o g/10.1016/j.ejo .2018.10.024
30 R.R. WAHLSTRØM ET AL.

Manzaneque, M., P iego, A. M., & Me ino, E. (2016). Co po a e go e nance e ec on inancial dis ess likelihood:
E idence om Spain. Re is a de Con abilidad,19(1), 111–121. h ps://doi.o g/10.1016/j. csa .2015.04.001
Masiak, C., Block, J. H., Mo i z, A., Lang, F., & K aeme -Eis, H. (2019). How do mic o i ms di e in hei inancing pa -
e ns om la ge SMEs? Ven u e Capi al,21(4), 301–325. h ps://doi.o g/10.1080/13691066.2019.1569333
Ma enda, F. R., Sibanda, M., Chikodza, E., & Gumbo, V. (2022). Bank up cy p edic ion o p i a e i ms in de eloping
economies: A scoping e iew and guidance o u u e esea ch. Managemen Re iew Qua e ly,72(4), 927–966.
h ps://doi.o g/10.1007/s11301-021-00216-x
McFadden, D. (1974). Condi ional logi analysis o quali a i e choice beha iou . In P. Za embka (Ed.), F on ie s in
econome ics (Vol. 1, pp. 105–142). Academic P ess.
Meinshausen, N., & B€
uhlmann, P. (2010). S abili y selec ion. Jou nal o he Royal S a is ical Socie y Se ies B: S a is ical
Me hodology,72(4), 417–473. h ps://doi.o g/10.1111/j.1467-9868.2010.00740.x
Me on, R. C. (1974). On he p icing o co po a e deb : The isk s uc u e o in e es a es. The Jou nal o Finance,
29(2), 449–470. h ps://doi.o g/10.2307/2978814
Ogne a, M., Pio oski, J. D., & Zakolyukina, A. A. (2020). Accoun ing undamen als and sys ema ic isk: Co po a e ail-
u e o e he business cycle. The Accoun ing Re iew,95(5), 321–350. h ps://doi.o g/10.2308/acc -52638
O ganisa ion o Economic Co-ope a ion and De elopmen . (2022). Financing SMEs and en ep eneu s 2022: An OECD
sco eboa d.
Pa aschi , F., Schmid, M., & Wahls øm, R. R. (2023). Bank up cy p edic ion o p i a ely held SMEs using ea u e selec ion
me hods (Wo king Pape ). Zeppelin Uni e si y, No wegian Uni e si y o Science and Technology, Swiss Finance
Ins i u e (SFI), Uni e si y o S . Gallen, and Uni e si y o S . Gallen. h ps://doi.o g/10.2139/ss n.3911490
Pea ce, J. A., & Zah a, S. A. (1992). Boa d composi ion om a s a egic con ingency pe spec i e. Jou nal o
Managemen S udies,29(4), 411–438. h ps://doi.o g/10.1111/j.1467-6486.1992. b00672.x
Pla , H., & Pla , M. (2012). Co po a e boa d a ibu es and bank up cy. Jou nal o Business Resea ch,65(8), 1139–
1143. h ps://doi.o g/10.1016/j.jbus es.2011.08.003
Rado ano ic, J., & Haas, C. (2023). The e alua ion o bank up cy p edic ion models based on socio-economic cos s.
Expe Sys ems wi h Applica ions,227, 120275. h ps://doi.o g/10.1016/j.eswa.2023.120275
Rajan, U., Se u, A., & Vig, V. (2015). The ailu e o models ha p edic ailu e: Dis ance, incen i es, and de aul s.
Jou nal o Financial Economics,115(2), 237–260. h ps://doi.o g/10.1016/j.j ineco.2014.09.012
Rouls on, M. S. (2007). Pe o mance a ge s and he B ie sco e. Me eo ological Applica ions,14(2), 185–194. h ps://
doi.o g/10.1002/me .21
Sai o, T., & Rehmsmeie , M. (2015). The p ecision- ecall plo is mo e in o ma i e han he ROC plo when e alua ing
bina y classi ie s on imbalanced da ase s. PLoS One,10(3), e0118432. h ps://doi.o g/10.1371/jou nal.pone.0118432
S
e e in, E., & Veganzones, D. (2021). Can ea nings managemen in o ma ion imp o e bank up cy p edic ion models?
Annals o Ope a ions Resea ch,306(1-2), 247–272. h ps://doi.o g/10.1007/s10479-021-04183-0
Shumway, T. (2001). Fo ecas ing bank up cy mo e accu a ely: A simple haza d model. The Jou nal o Business,74(1),
101–124. h ps://doi.o g/10.1086/209665
Smi h, R. F., & Winako , A. H. (1930). A es analysis o unsuccess ul indus ial companies (Bu eau o Business Resea ch,
Bulle in. 31). Uni e si y o Illinois P ess.
S hoj, S., Ko a
c, D., Shapi o, J. N., & File , R. K. (2023). The impac o delay: E idence om o mal ou -o -cou es uc-
u ing. Jou nal o Co po a e Finance,78, 102319. h ps://doi.o g/10.1016/j.jco p in.2022.102319
S hoj, S., Lapinski, M., & Walde, J. (2021a). Impac e alua ion o business de elopmen g an s on SME pe o mance.
Small Business Economics,57(3), 1285–1301. h ps://doi.o g/10.1007/s11187-020-00348-6
S hoj, S., 
Sk inja i
c, B., & Radas, S. (2021b). Bidding agains he odds? The impac e alua ion o g an s o young mic o and
small i ms du ing he ecession. Small Business Economics,56(1), 83–103. h ps://doi.o g/10.1007/s11187-019-00200-6
S ein, J. C. (2002). In o ma ion p oduc ion and capi al alloca ion: Decen alized e sus hie a chical i ms. The Jou nal
o Finance,57(5), 1891–1921. h ps://doi.o g/10.1111/0022-1082.00483
Tang, X., Li, S., Tan, M., & Shi, W. (2020). Inco po a ing ex ual and managemen ac o s in o inancial dis ess p edic ion: A
compa a i e s udy o machine lea ning me hods. Jou nal o Fo ecas ing,39(5), 769–787. h ps://doi.o g/10.1002/ o .2661
Taylo , J., & Tibshi ani, R. (2018). Pos -selec ion in e ence o -penalized likelihood models. The Canadian Jou nal o
S a is ics ¼Re ue Canadienne de S a is ique,46(1), 41–61. h ps://doi.o g/10.1002/cjs.11313
Tian, S., Yu, Y., & Guo, H. (2015). Va iable selec ion and co po a e bank up cy o ecas s. Jou nal o Banking & Finance,
52,89–100. h ps://doi.o g/10.1016/j.jbank in.2014.12.003
Tibshi ani, R. (1996). Reg ession sh inkage and selec ion ia he lasso. Jou nal o he Royal S a is ical Socie y Se ies B:
S a is ical Me hodology,58(1), 267–288. h ps://doi.o g/10.1111/j.2517-6161.1996. b02080.x
T oyanskaya, O., Can o , M., She lock, G., B own, P., Has ie, T., Tibshi ani, R., Bo s ein, D., & Al man, R. B. (2001).
Missing alue es ima ion me hods o DNA mic oa ays. Bioin o ma ics,17(6), 520–525. h ps://doi.o g/10.1093/bio-
in o ma ics/17.6.520
an de Gee , S., B€
uhlmann, P., Ri o , Y., & Dezeu e, R. (2014). On asymp o ically op imal con idence egions and es s
o high-dimensional models. The Annals o S a is ics,42(3), 1166–1202. h ps://doi.o g/10.1214/14-AOS1221
Wilson, N., & Al anla , A. (2014). Company ailu e p edic ion wi h limi ed in o ma ion: Newly inco po a ed companies.
Jou nal o he Ope a ional Resea ch Socie y,65(2), 252–264. h ps://doi.o g/10.1057/jo s.2013.31
COGENT ECONOMICS & FINANCE 31
Wilson, N., Summe s, B., & Hope, R. (2000). Using paymen beha iou da a o c edi isk modelling. In e na ional
Jou nal o he Economics o Business,7(3), 333–346. h ps://doi.o g/10.1080/13571510050197230
Wi he s, M. C., & Fi za, M. A. (2017). Do boa d chai s ma e ? The in luence o boa d chai s on i m pe o mance.
S a egic Managemen Jou nal,38(6), 1343–1355. h ps://doi.o g/10.1002/smj.2587
Yin, C., Jiang, C., Jain, H. K., & Wang, Z. (2020). E alua ing he c edi isk o SMEs using legal judgmen s. Decision
Suppo Sys ems,136, 113364. h ps://doi.o g/10.1016/j.dss.2020.113364
Zhang, C. H., & Zhang, S. S. (2014). Con idence in e als o low dimensional pa ame e s in high dimensional linea models.
Jou nal o he Royal S a is ical Socie y Se ies B: S a is ical Me hodology,76(1), 217–242. h ps://doi.o g/10.1111/ ssb.12026
Zhao, J., Ouenniche, J., & De Smed , J. (2024). Su ey, classi ica ion and c i ical analysis o he li e a u e on co po a e
bank up cy and inancial dis ess p edic ion. Machine Lea ning wi h Applica ions,15, 100527. h ps://doi.o g/10.
1016/j.mlwa.2024.100527
Zmijewski, M. E. (1984). Me hodological issues ela ed o he es ima ion o inancial dis ess p edic ion models.
Jou nal o Accoun ing Resea ch,22,59–82. h ps://doi.o g/10.2307/2490859
Appendix A. The sample pe indus y
Table A1. The sample pe indus y.
Financial
s a emen s
Numbe o
companies
Bank up cy equency
Indus y le el 1 Indus y le el 2
One-yea
ho izon
Two-yea
ho izon
Th ee-yea
ho izon
Ag icul u e, o es y
and ishing
C op and animal p oduc ion, hun ing 4765 1084 0.36% 0.78% 0.61%
Fo es y and logging 2461 570 0.33% 0.61% 0.37%
Fishing and aquacul u e 12,989 3155 0.14% 0.32% 0.43%
Mining and qua ying Mining o coal and ligni e 14 3 0.00% 0.00% 0.00%
Ex ac ion o oil and na u al gas 633 147 0.16% 0.16% 0.16%
Mining o me al o es 75 15 0.00% 0.00% 0.00%
O he mining and qua ying 2709 538 0.11% 0.33% 0.41%
Mining suppo se ice ac i i ies 2769 617 0.14% 0.43% 0.54%
M
anu ac u ing
Food p oduc s 9188 2007 0.52% 1.40% 1.09%
Be e ages 1103 285 0.54% 1.45% 1.18%
Tex iles 1353 283 0.00% 0.59% 0.30%
Wea ing appa el 809 197 0.37% 2.35% 1.73%
Lea he and lea he p oduc s 98 22 0.00% 0.00% 0.00%
Wood and wood p oduc s 5177 1079 0.31% 0.89% 0.83%
Pape and pape p oduc s 340 67 0.29% 0.59% 0.29%
P in ing and ep oduc ion 3156 669 0.29% 0.92% 0.92%
Re ined pe oleum p oduc s 55 15 0.00% 1.82% 0.00%
Chemicals, chemical p oduc s 925 197 0.32% 0.97% 0.97%
Pha maceu icals 201 51 0.50% 0.50% 0.00%
Rubbe and plas ic p oduc s 1860 374 0.27% 0.75% 0.59%
O he non-me al mine al p oduc s 2818 573 0.00% 0.75% 0.53%
Basic me als 548 114 0.18% 1.28% 0.55%
Fab ica ed me al p od. 8944 1885 0.48% 1.20% 0.91%
Elec onic and op ical p oduc s 1372 293 0.29% 0.44% 0.36%
Elec ical equipmen 1641 353 0.43% 0.98% 1.04%
Machine y and equipmen 4308 892 0.42% 0.79% 0.56%
Mo o ehicles e c. 758 143 0.40% 1.06% 0.92%
O he anspo equipmen 2043 456 0.64% 1.52% 1.37%
Fu ni u e 2143 467 0.84% 1.49% 1.12%
O he manu ac u ing 2500 558 0.16% 0.76% 0.68%
Repai , ins alla ion o machine y 6870 1573 0.28% 0.89% 0.74%
Cons uc ion Cons uc ion o buildings 96,658 23,772 0.46% 1.26% 1.05%
Ci il enginee ing 3956 878 0.46% 1.11% 1.21%
Specialised cons uc ion ac i i ies 76,835 17,755 0.50% 1.45% 1.20%
Domes ic ade, ca
epai shop
Mo o ehicles, ade and epai 30,129 6786 0.42% 1.17% 1.05%
Wholesale ade 70,096 15,636 0.37% 0.99% 0.89%
Re ail ade 87,311 20,031 0.73% 1.80% 1.45%
T anspo a ion and
s o age
Land anspo , pipeline anspo 20,653 4685 0.49% 1.51% 1.35%
Wa e anspo 10,675 2447 0.10% 0.43% 0.36%
Ai anspo 544 113 0.18% 0.37% 0.37%
Suppo ac . o anspo a ion 10,401 2167 0.12% 0.50% 0.46%
Pos al and cou ie ac i i ies 1121 321 1.52% 4.10% 2.59%
Accommoda ion,
ood se ice
Accommoda ion 11,848 2534 0.28% 0.86% 0.74%
Food and be e age se ice ac . 26,109 6949 1.12% 2.95% 2.28%
In o ma ion and
communica ion
Publishing ac i i ies 8021 1701 0.20% 0.49% 0.45%
Mo ion pic u e, TV, music p od. 4532 1190 0.22% 0.46% 0.44%
P og amming, b oadcas ing ac . 295 78 0.00% 0.68% 0.34%
(con inued)
32 R.R. WAHLSTRØM ET AL.
Table A1. Con inued.
Financial
s a emen s
Numbe o
companies
Bank up cy equency
Indus y le el 1 Indus y le el 2
One-yea
ho izon
Two-yea
ho izon
Th ee-yea
ho izon
Telecommunica ions 2405 575 0.17% 0.54% 0.71%
Compu e p og amming, consul ancy 26,886 7062 0.22% 0.53% 0.49%
In o ma ion se ice ac i i ies 2927 738 0.10% 0.48% 0.38%
P o ess., scien i ic,
ech. ac .
Legal and accoun ing ac i i ies 20,103 4350 0.05% 0.17% 0.20%
Head o ices, managemen consul . 37,651 9148 0.18% 0.29% 0.27%
A chi ec u e, enginee ing ac . 43,187 10,557 0.20% 0.44% 0.42%
Scien i ic esea ch and de elopmen 3030 765 0.10% 0.40% 0.40%
Ad e ising and ma ke esea ch 6291 1571 0.17% 0.87% 0.92%
O he p o ., scien i ic, echn. ac . 12,268 3187 0.37% 0.82% 0.71%
Ve e ina y ac i i ies 2216 494 0.09% 0.05% 0.05%
Adminis a i e,
suppo se ice
Ren al and leasing ac i i ies 11,403 2667 0.30% 0.73% 0.66%
Employmen ac i i ies 6701 1718 0.70% 1.63% 1.49%
T a el agency, ou ope a o s 5355 1378 0.35% 0.88% 0.90%
Secu i y, in es iga ion ac i i ies 1045 253 0.48% 1.91% 1.72%
Buildings, landscape se ice ac . 9082 2221 0.53% 1.21% 0.87%
Business suppo ac i i ies 8924 2066 0.18% 0.52% 0.44%
Educa ion Educa ion 10,236 2548 0.20% 0.51% 0.52%
Human heal h,
social wo k
Human heal h ac i i ies 26,538 6115 0.08% 0.14% 0.15%
Residen ial ca e ac i i ies 704 152 0.00% 0.43% 0.43%
Social wo k wi hou accommoda ion 9066 1951 0.08% 0.19% 0.17%
A s, en e ainmen
and ec ea ion
A s and en e ainmen ac i i ies 5023 1309 0.24% 0.48% 0.40%
Lib a ies, museums, o he cul u e 436 85 0.00% 0.00% 0.00%
Gambling and be ing ac i i ies 778 165 0.13% 0.64% 0.77%
Spo s, amusemen , ec ea ion 9288 2160 0.23% 0.54% 0.55%
O he se ice ac i i ies Membe ship o ganisa ions 516 120 0.00% 0.00% 0.19%
Repai , pe sonal, household goods 1037 256 0.87% 1.83% 1.35%
O he pe sonal se ice ac i i ies 12,022 2782 0.23% 0.62% 0.57%
To al 818,927 192,118 0.39% 1.04% 0.88%
No es: Ou sample o SMEs’ inancial s a emen s, he numbe o unique companies and bank up cy equency o e di e en ime ho izons
pe indus y a he one- and wo-digi le els, in acco dance wi h he No wegian S anda d Indus ial Classi ica ion (SIC2007) based on he
UN’s ISIC Re . 4 and EU’s NACE Re . 2.
COGENT ECONOMICS & FINANCE 33