Fenk, Lo enz
A icle — Published Ve sion
En i onmen al pho og aph use in co po a e sus ainabili y
epo ing: A machine‐suppo ed isual con en analysis
Business S a egy and he En i onmen
P o ided in Coope a ion wi h:
John Wiley & Sons
Sugges ed Ci a ion: Fenk, Lo enz (2024) : En i onmen al pho og aph use in co po a e sus ainabili y
epo ing: A machine‐suppo ed isual con en analysis, Business S a egy and he En i onmen ,
ISSN 1099-0836, Wiley Pe iodicals, Inc., Hoboken, NJ, Vol. 34, Iss. 1, pp. 1097-1112,
h ps://doi.o g/10.1002/bse.4035
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/313796
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 p://c ea i ecommons.o g/licenses/by/4.0/
RESEARCH ARTICLE
En i onmen al pho og aph use in co po a e sus ainabili y
epo ing: A machine-suppo ed isual con en analysis
Lo enz Fenk
TUM School o Managemen , Technical
Uni e si y o Munich, Ge many
Co espondence
Lo enz Fenk, Technical Uni e si y o Munich,
TUM School o Managemen , A ciss asse
21, 80333 Munich, Ge many.
Email: [email p o ec ed]
Funding in o ma ion
F ied ich Naumann Founda ion
Abs ac
Despi e he p e alence o pho og aphs in co po a e sus ainabili y epo ing, hei use
is no ye su icien ly unde s ood. To he bes o ou knowledge, his pape is he i s
la ge-scale s udy in he ield. In oducing a no el machine-suppo ed app oach, we
assess en i onmen al pho og aph u iliza ion based on a sample o 45,228 pho o-
g aphs con ained in 1,463 sepa a ely disclosed sus ainabili y epo s om Eu opean
i ms be ween 2011 and 2020. We ind ha agains he o e all end o dec easing
pho og aph u iliza ion, he sha e o en i onmen ally hemed pho og aphs has ma k-
edly inc eased. Fu he mo e, ope a ing in an en i onmen ally sensi i e indus y is
s ongly associa ed wi h a subs an ially la ge sha e o pho og aphs depic ing en i-
onmen al subjec ma e . Las ly, we obse e ha companies signal hei supe io
en i onmen al pe o mance h ough g ea e u iliza ion o en i onmen al pho o-
g aphs. By in oducing a no el machine-suppo ed app oach o analyzing pho o-
g aphs, his s udy makes a me hodological con ibu ion o he ield o sus ainabili y
epo ing. Ou esul s also ha e impo an p ac ical implica ions.
KEYWORDS
co po a e sus ainabili y epo ing, en i onmen al, pho og aph use, pho og aph use, compu e
ision
1|INTRODUCTION
To coun e legi imacy h ea s a ising om en i onmen al pollu ion o
poo wo king condi ions, companies may eso o shaping he pe -
cep ion o hei impac on s akeholde s and socie y a la ge (Cho &
Pa en, 2012). Delibe a e a emp s o in luence he pe cep ion o an
o ganiza ion's legi imacy, e.g., by manipula ing con en and appea -
ance o in o ma ion, a e helped by he in o ma ion ad an age
manage s enjoy o e he epo 's eade s (Me kl-Da ies &
B ennan, 2007,2011).
Pho og aphs, in pa icula , a e well sui ed o communica e
sus ainabili y- ela ed opics, as complex and less quan i iable mes-
sages in he indis inc u u e a e o en mo e e ec i ely con eyed
h ough pho og aphs han numbe s o wo ds (Ande son, 1980;
Ga cía-Sánchez & A aújo-Be nado, 2020; Rämö, 2011). Mo eo e ,
isuals enjoy a mo e p ominen place in cogni i e memo y han
Abb e ia ions: ENV PUT, en i onmen al pho og aph u iliza ion; ENV sha e, en i onmen al pho og aph sha e; ESG, En i onmen al, Social and Go e nance; EU, Eu opean Union; PDF, po able
documen iles; PNG, po able ne wo k g aphics; PUT, pho og aph u iliza ion; VIF, a iance in la ion ac o .
In he ealm o co po a e sus ainabili y epo ing, he ole o pho og aphs emains unde explo ed. This pape in oduces a machine-suppo ed app oach based on compu e ision o examine he
use o pho og aphs, and in pa icula en i onmen al pho og aphs. Analyzing 45,228 pho og aphs con ained in 1,463 sus ainabili y epo s om Eu opean companies, ou indings challenge he
p e ailing no ion o sus ainabili y epo s becoming e e mo e densely illed wi h pho og aphs. In addi ion, we obse e a no able inc ease in en i onmen al pho og aph use. Pa icula ly,
companies ope a ing in en i onmen ally sensi i e indus ies exhibi a signi ican ly highe p opo ion o such isuals. This s udy no only acili a es a no el app oach o s udying pho og aph use
bu also holds p ac ical signi icance, shedding ligh on he e ol ing ield o co po a e sus ainabili y epo ing.
Recei ed: 23 Janua y 2024 Re ised: 16 Oc obe 2024 Accep ed: 17 Oc obe 2024
DOI: 10.1002/bse.4035
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s 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.
© 2024 The Au ho (s). Business S a egy and he En i onmen published by ERP En i onmen and John Wiley & Sons L d.
Bus S a En . 2025;34:1097–1112. wileyonlinelib a y.com/jou nal/bse 1097
w i en ex (T e sky, 1974) and can engende aming e ec s o
hough pa e ns (T e sky & Kahneman, 1986). Scien i ic s udies ha e
cau ioned ha pho og aphs a e used beyond deco a ing ex ual in o -
ma ion; howe e , esea ch has no kep pace wi h elabo a e p ac ices
o isual communica ion employed by he p i a e sec o (B ei ba h
e al., 2010; Da ison, 2015; Da ison & Wa en, 2009). P e ious a i-
cles ind ha pho og aphs a e used o symbolically pu sue legi imacy
by c ea ing idealized isions, which do no ypically e lec companies'
business conduc (Boi al, 2013; B ei ba h e al., 2010; H asky, 2012).
Mo e ecen s udies con i m ha manage s use hei disc e ion o
p ojec he image o a esponsible and us wo hy co po a e ci izen,
while nega i e aspec s a e all bu absen in he isual ealm (Ga cía-
Sánchez & A aújo-Be nado, 2020; Nicolò e al., 2022). Howe e , mos
exis ing s udies ely on small samples, which o en come om a single
indus y, single coun y, single epo ing pe iod, o case s udy
(In e nizzi e al., 2021; Rämö, 2011).
Unde s anding pho og aph use is also o g ea p ac ical ele ance,
as e en seasoned in es o s a e swayed by aes he ics in co po a e
epo ing (Townsend & Shu, 2010). Compa ing di e en ypes o disclo-
su e, pho og aphs a e mos ex ensi ely employed in co po a e sus ain-
abili y epo ing, whe e hey cons i u e an in eg al elemen
(Boi al, 2013; B ei ba h e al., 2010; Da ison, 2015;In e nizzi
e al., 2021;Rämö,2011). The medium o sus ainabili y epo s has also
gained impo ance as he salience o co po a e sus ainabili y
has ma kedly inc eased (A idsson & Dumay, 2022). Fu he mo e, he
subs an ial cos o p in ing and dis ibu ing ha d copy, pho og aph- ich
sus ainabili y epo s ha e all bu e apo a ed as co po a e communica-
ion is mos usually sha ed digi ally, which may ha e in luenced how
manage s design sus ainabili y epo s. This p edilec ion o using pho-
og aphs may also be adop ed by small- and medium-sized en e p ises
(de Villie s & Alexande , 2014; Shabana e al., 2017), which a e soon o
be subjec ed o expanded epo ing equi emen s in he Eu opean
Union (Eu opean Pa liamen & Eu opean Council, 2022). Gi en he con-
ce ning use o pho og aphs epo ed in ela ed s udies, we a gue ha i
is impo an o unde s and hei powe o sway eade s' opinions.
Despi e hei impo an ole and well-documen ed po ency, he use o
pho og aphs emains a blind spo in co po a e sus ainabili y epo ing.
Gi en he inhe en limi a ions o exis ing s udies, we con ibu e
gene alizable indings on he use o pho og aphs in Eu opean co po-
a e sus ainabili y epo ing, which ea u e bo h longi udinal co e age
and c oss-sec ional a ia ion. We quan i y he o e all use o pho o-
g aphs and assess he hemes depic ed he ein. Ou machine-
suppo ed analysis o pho og aphic con en ocuses in pa icula on
pho og aphs con eying en i onmen al subjec ma e o accoun o
he ou sized ole o he en i onmen al pilla wi hin co po a e sus ain-
abili y. While he signi icance o en i onmen al conce ns wi hin co -
po a e sus ainabili y has been a cons an , i s ela i e impo ance has
gained new momen um as e iden by he ambi ious egula o y agenda
pu sued by he Eu opean Union (EU), an inc eased gene al public
awa eness, e.g., global clima e demons a ions, and in ensi ied
in ol emen o he inancial communi y, e.g., ‘Clima e Ac ion 100+
g oup’. In addi ion, we a gue ha en i onmen al hemes can also be
mo e eliably iden i ied han mo e abs ac concep s, such as labo
ela ions o co po a e go e nance. Ou esea ch is hus guided by he
ollowing esea ch ques ion:
To wha ex en do companies use en i onmen al pho og aphs in
hei co po a e sus ainabili y epo ?
Mo eo e , we ha e iden i ied h ee suppo i e esea ch
ques ions ha a e in ended o guide ou esea ch and shed ligh on
speci ic aspec s o ou s udy's ocus:
•How has pho og aph use changed o e ime?
•How is he ela ionship be ween pho og aph use and sus ainabili y
pe o mance?
•How does pho og aph use a y ac oss di e en indus ies?
We employ Google's Vision API, a comme cial p e- ained com-
pu e ision se ice (Google LLC, 2023), o de i e sophis ica ed
insigh s in o la ge image da a se s. Ad ances in a i icial in elligence-
powe ed so wa e and compu ing esou ces enable esea che s o
analyze and d aw conclusions based on la ge-scale da a se s. While
machine-suppo ed so wa e has been pu o use in many ields o
academia, i s applica ion in he con ex o co po a e sus ainabili y
epo ing is o da e la gely con ined o ex ual analysis, and ecen
examples include Nieho (2022) and Bhanda i e al. (2022). In pa icu-
la , we use he label de ec ion ea u e o e ie e pho og aphic
con en by asc ibing labels. To classi y whe he pho og aphs depic
en i onmen al subjec ma e , we de eloped a codebook based on
he de ec ed labels asc ibed by he ision models. Ou machine-based
app oach allows o he analysis o much la ge image da a se s han
p e iously easible. To he bes o ou knowledge, his is he i s
la ge-scale s udy on he use o pho og aphs in co po a e sus ainabili y
epo s, which applies a machine-suppo ed isual con en analysis.
We assess o e all pho og aph u iliza ion and ocus ou analysis on he
use o pho og aphs con eying en i onmen al subjec ma e .
Based on a sample o 45,228 pho og aphs con ained in 1,463
sepa a ely disclosed epo s om 2011 o 2020, we canno con i m
he gene al belie ha sus ainabili y epo s a e becoming e e mo e
densely illed wi h pho og aphs. Agains his obse ed end o alling
pho og aph u iliza ion o e he las decade, he pe cen age o en i-
onmen ally hemed pho og aphs has ma kedly inc eased. F om ou
eg ession models, we ind ha , i s ly, ope a ing in an en i onmen-
ally sensi i e indus y (B anco & Rod igues, 2008; Cla kson e al.,
2011) is associa ed wi h a subs an ially highe pe cen age o pho o-
g aphs depic ing en i onmen al subjec ma e . Secondly, we ind ha
companies wi h highe en i onmen al c eden ials signal hei supe io
pe o mance h ough a highe sha e o en i onmen ally hemed
pho og aphs. Bo h indings a e highly signi ican and e y obus wi h
ega d o changes in he me hodology o iden i ying en i onmen al
pho og aphs. In his s udy, we in oduce a no el app oach o asses-
sing pho og aph use in co po a e epo ing. Fu he mo e, we con ib-
u e o he exis ing li e a u e by p o iding empi ical e idence o o e all
pho og aph use and, in pa icula , he u iliza ion o en i onmen al
pho og aphs by Eu opean companies.
The emainde o his a icle is o ganized as ollows: Sec ion 2
consis s o he li e a u e e iew and he de i a ion o ou esea ch
1098 FENK
ques ion. In Sec ion 3, we ou line ou esea ch app oach, which is ol-
lowed by he empi ical esul s in Sec ion 4. Sec ion 5discusses he
indings as well as limi a ions and sugges ions o u u e esea ch.
Las ly, Sec ion 6p o ides he conclusion o his pape .
2|BACKGROUND
2.1 |Cha ac e is ics o sus ainabili y epo ing
O e he las wo decades, sus ainabili y- ela ed opics ha e inc eas-
ingly occupied annual epo s (Da ison & Wa en, 2009). Simul a-
neously, many companies ha e begun o olun a ily disclose a
sus ainabili y epo along wi h hei inancial epo ing. Since 2017,
la ge legal en i ies domiciled in he Eu opean Economic A ea ha e
been legally obliged o annually disclose non- inancial in o ma ion o
he public. The 2014/95/EU di ec i e (Eu opean Pa liamen &
Eu opean Council, 2014) manda es he comp ehensi e discussion o a
wide ange o opics, anging om en i onmen al conside a ions and
labo ela ions o an i-sla e y p ac ices. While he law dic a es ha
he epo includes ce ain opics, companies do no need o abide by
a s ic epo ing amewo k (Fiech e e al., 2022). Mos o he sus-
ainabili y in o ma ion epo ed by companies can s ill be conside ed
olun a ily disclosed in o ma ion (Ch is ensen e al., 2021). In he
absence o a ha monized epo ing s anda d, manage s enjoy la ge
disc e ion o e he design o hei sus ainabili y epo ing.
The companies' mo i a ions o disclosing olun a y sus ainabili y
epo s a y g ea ly (He zig & Schal egge , 2011). These include
showcasing he company in a posi i e ligh , gaining a compa a i e
edge o e compe i o s (F ias-Acei uno e al., 2014), and/o mi iga ing
s akeholde p essu es and in e e ence by go e nmen agencies (G ay
e al., 1995; P ado-Lo enzo & Ga cía-Sánchez, 2010; Re e e, 2009).
Legi imacy heo y is mos commonly used o explain sus ainabili y dis-
closu e by companies. Following Deegan (2002), companies equi e
legi imacy no only om co e s akeholde s bu also seek o epai o
main ain legi imacy in he eyes o socie y, on which hey con inuously
depend o esou ces and app o al (Bea ie & Jones, 2008; Hahn &
Lül s, 2014; Milne & Pa en, 2002). Legi imacy can be unde s ood as a
empo a y social con ac , which is g an ed by socie y o esponsibly
ope a ing companies (Ash o h & Gibbs, 1990; H asky, 2012). Follow-
ing Suchman (1995), o ganiza ional legi imacy can be a ained ei he
as mo al legi imacy, backed up by he i m's business conduc , o
h ough symbolic a emp s o p ojec he image o a ca ing co po a e
ci izen. As hei legi imacy es s on s aying wi hin a se o social
bounda ies (Deegan, 2002), companies may use sus ainabili y disclo-
su e o legi ima ize hei ope a ions o hei s akeholde s (Cho &
Pa en, 2012; Gu h ie & Pa ke , 1989).
On he o he hand, olun a y disclosu e heo y and, by ex ension,
signaling heo y a e also ex ensi ely used o explain olun a y epo ing
o non- inancial in o ma ion o he public (Cla kson e al., 2008;
Mahoney e al., 2013). The e o e, companies wi h supe io sus ainabil-
i y pe o mance choose o olun a ily disclose o he wise unobse able
sus ainabili y in o ma ion o inc ease hei inancial alue (de Villie s &
Ma ques, 2016; Dhaliwal e al., 2011; Hummel & Schlick, 2016). In his
con ex , signaling e e s o he a emp o communica e a i m's supe-
io sus ainabili y achie emen s (Connelly e al., 2011; Rezaee, 2016).
E en so, disclosu e is cos ly; i ms wi h supe io sus ainabili y c eden-
ials bene i om he compa ison di e en ia ing hem om hei pee s
h ough olun a y sus ainabili y disclosu e (Mahoney e al., 2013). To
p esen hese achie emen s h ough co po a e sus ainabili y epo s,
manage s ha e a e sa ile communica ion oolbox a hei disposal.
2.2 |Visual communica ion
Pho og aphs ha e become ubiqui ous in ou isualized and digi ized
wo ld o business. In he con ex o co po a e epo ing, he g ow h
o isual ma e ial o e he las decades has been ecognized by se -
e al s udies (Bea ie & Jones, 2008; Da ison & Ske a , 2006;
Lee, 1994; Usmani e al., 2020). Compa ing di e en enues o co po-
a e epo ing, s udies ha e shown ha pho og aphs a e especially
p e alen in co po a e sus ainabili y epo ing and cons i u e an in e-
g al ea u e o such epo s (Boi al, 2013; In e nizzi e al., 2021;
Rämö, 2011). Many esea che s ha e exp essed conce n ha he use
o iche media gi es companies a e y po en ool, which can be le -
e aged o he o ical o pe suasi e pu poses (Cho e al., 2009). Pho o-
g aphs may cons uc wha co po a e sus ainabili y means o whom
(B ei ba h e al., 2010), di e a en ion om ad e se in o ma ion
(A o a & Lodhia, 2017), o al e he pe cep ion o non- inancial pe o -
mance (Cho e al., 2010; H asky, 2012; Mahoney e al., 2013;
Peeples, 2011). Hopwood (2009) a gues ha sus ainabili y disclosu e
illed wi h isual image y may ob usca e he ue con en and hus
po en ially dec ease co po a e anspa ency. High-quali y pho o-
g aphs, p o essionally sho and a imes co e ing an en i e epo
page, a e in en ionally and delibe a ely used (B ei ba h e al., 2010;
Da ison, 2015; Ga cía-Sánchez & A aújo-Be nado, 2020). G aphic ele-
men s such as pho og aphs na u ally s and ou om he su ounding
ex and a ac he eade 's a en ion h ough a so-called ‘pop-up
e ec ’(T eisman, 1985). Riche media seems o be especially well
sui ed when he e is ambigui y abou he in e p e a ion o he unde -
lying issue (Cho e al., 2009; Da & Lengel, 1986).
Pho og aphs se e se e al pu poses in long ex documen s. They
o ien a e eade s, hus enhancing in o ma ion p ocessing luency
(In e nizzi e al., 2021). In addi ion, eade s a e mo e likely o selec
ex s wi h pho og aphs, which enhances he o e all a ac i eness o
eading a ex documen (Knobloch e al., 2003). On he o he hand,
isuals may be used o guide eade s o mo e a o able sec ions o
he epo (Usmani e al., 2020). The capaci y o pho og aphs o com-
munica e complex messages o a di e se audience makes hem a e y
powe ul ool o dissemina ing in o ma ion (Ande son, 1980;
Coleman, 2010). This is especially ue compa ed o ex ual in o ma-
ion, whe e complex messages may be los o complica ed language
(Rämö, 2011). S udies sugges ha isuals a e no only easie o ecall
in many si ua ions (Pai io, 1969) bu a e also be e e ained o e he
long- e m (Pai io e al., 1968). Indi iduals also end o spend mo e
ime looking a isual ma e ial han ex s (T e sky, 1974) and become
FENK 1099
biased owa d he image i he wo media gi e opposi e cues
(Zillmann e al., 1999).
Fu he mo e, he capaci y o pho og aphs o ‘c ea e eali y’
(P es on e al., 1996) makes hem especially pe suasi e, as hey a e pe -
cei ed o cap u e he eal wo ld (Pesci e al., 2015). Pho og aphs can
be le e aged o mis ep esen eali y (Ca on & Tu co e, 2009), espe-
cially in ai ly con es ed domains such as co po a e sus ainabili y (Hallin
e al., 2021; Meue e al., 2020). Following Bansal and Kis uck (2006),
pho og aphs can he e o e be used o c ea e a con incing accoun o
he subjec ma e i espec i e o ac ual implemen a ion. Pho og aphs
also ha e he capaci y o engende hough pa e ns and engende
aming e ec s o guide he eade s' hinking (T e sky & Kahneman,
1986). The s a egic and delibe a e a emp o i ms o posi i ely in lu-
ence public opinion abou hei image is called imp ession managemen
(Hooghiems a, 2000; Me kl-Da ies & B ennan, 2007).
Following Da ison (2015), co po a e sus ainabili y epo ing has
inhe en cha ac e is ics ha make i ideal o le e aging isual commu-
nica ion. Compa ed o inancial disclosu e, i speaks o a much wide
audience, including non-p o essional and less expe ienced s akeholde s
(Rämö, 2011). Fu he mo e, i ends o be o wa d-looking and mo e
quali a i e in na u e, s emming om he long ime ho izon o some o
he opics co e ed. The dissemina ion o mos ly quali a i e in o ma ion
also esul s in less compa able measu es o pe o mance
(Ande son, 1980;Cho&Pa en,2007;Rämö,2011). In ligh o he sub-
s an ial disc e ion epo ing manage s enjoy o e hei sus ainabili y
disclosu e, pho og aphs a e a e y po en ool used o showcase he
company in he mos a o able ligh . The impo ance o unde s anding
companies' isual epo ing beha io is u he mo e unde sco ed by he
indings o Townsend and Shu (2010), who showed ha e en seasoned
in es o s a e swayed by he aes he ics o co po a e epo ing.
2.3 |Rela ed wo k and esea ch gap
We a gue ha i is e y impo an o shed mo e ligh on he subjec as
esea ch has lagged behind he sophis ica ed p ac ices o he p i a e
sec o (Da ison, 2015). Despi e a numbe o exis ing s udies, he unde -
s anding o pho og aph use in co po a e sus ainabili y epo ing
emains insu icien . As hese s udies ely on he manual assessmen o
pho og aphic con en , he sample size o he unde lying epo ing da a
is inhe en ly limi ed. While he majo i y o s udies p o ide a lo o ana-
ly ical dep h, hei indings can seldom be gene alized beyond he
obse ed phenomenon. Mos a e based on c oss-sec ional da a om a
single indus y o coun y (Boi al, 2013; Cab e a-Na áez & Quinche-
Ma ín, 2021;In e nizzie al.,2021;Rämö,2011;Zenge al.,2022),
mul iple cases s udies (Ali e al., 2021;B ei ba he al.,2010;Co azza
e al., 2020;Pé ez-Cañiza es,2022), o join ly analyze g aphs and pho-
og aphs as isual disclosu e (H asky, 2012; Nicolò e al., 2022). The
emaining con ibu ions, which ea u e bo h longi udinal co e age and
c oss-sec ional a ia ion, include only a ew dozen epo s om a single
coun y (Chong e al., 2019; Ga cía-Sánchez & A aújo-Be nado, 2020).
Mos s udies unde sco e he cen ali y o pho og aphs in co po-
a e sus ainabili y epo s, which a e an in eg al pa o he epo ing
oolbox (Boi al, 2013; B ei ba h e al., 2010; Zeng e al., 2022). While
he ex en o pho og aph use seems o a y ma kedly be ween com-
panies (In e nizzi e al., 2021), many au ho s obse e ha pho o-
g aphs ha e become mo e hea ily used in ecen yea s (Ali
e al., 2021; Anan ha aman e al., 2020; Bea ie e al., 2008; Chong
e al., 2019). Howe e , we belie e ha hese di e ences may no only
be due o sample selec ions bu also o di e en measu es o pho o-
g aph use, e.g., coun ing he numbe o pho og aphs may p oduce
e y di e en esul s han assessing pho og aphs agains he numbe
o epo pages.
Despi e employing di e en measu es, bo h Anan ha aman e al.
(2020) and Nicolò e al. (2022) ind ha companies ope a ing in en i-
onmen ally sensi i e indus ies make mo e use o isual disclosu e.
Mo eo e , some e idence exis s ha pho og aph use nega i ely co -
ela es wi h bo h he numbe o opics disclosed in he epo as well
as CSR pe o mance (Anan ha aman e al., 2020). On a simila no e,
H asky (2012) and Ga cía-Sánchez and A aújo-Be nado (2020) bo h
obse e ha companies ha a e less mo i a ed by sus ainabili y use a
mo e symbolic app oach o pho og aph use, e.g., h ough he use o
non-speci ic pho og aphs. Fu he mo e, he pho og aphs gene ally
con ey idealized isions o sus ainabili y disconnec ed om he ope -
a ional eali y, while nega i e cues a e almos comple ely absen om
he isual le el (Boi al, 2013; B ei ba h e al., 2010; Chong
e al., 2019). Rega ding he subjec ma e depic ed in he pho o-
g aphs, exis ing s udies ind e y di e en pa e ns ha may plausibly
be explained using indus y-speci ic samples. Rämö (2011) and Boi al
(2013) ind la ge pe cen ages o en i onmen ally hemed pho o-
g aphs, while In e nizzi e al. (2021) and H asky (2012) obse e
socially hemed as well as p oduc - and ope a ions- ela ed pho o-
g aphs o be mos ex ensi ely used. Longi udinal s udies by Pé ez-
Cañiza es (2022) and Chong e al. (2019) ind an inc eased ocus on
depic ing people in he pho og aphs. Howe e , as p e iously poin ed
ou , all s udies a e cha ac e ized by a e y limi ed sample size, o en
based on a single indus y, coun y, and/o epo ing pe iod. Thus,
hei indings should be in e p e ed wi h cau ion. Despi e he ichness
o hei analysis, hei esul s may no be gene alizable beyond he
pa icula phenomenon.
Hence, we s ill obse e subs an ial knowledge gaps in he unde -
s anding o pho og aph use in co po a e sus ainabili y epo s. Fi s ly,
he e is a lack o cohe en measu es o assess pho og aphs in epo -
ing documen s and, as a consequence, a basic unde s anding o he
ex en o o e all pho og aph use. Secondly, he ela i e impo ance o
he speci ic subjec ma e depic ed in pho og aphs is only documen-
ed in indus y-speci ic s udies. Thi dly, he e is scan e idence o
how o e all pho og aph use as well as he use o ce ain hemes a -
ies ac oss company cha ac e is ics. Some s udies obse e ha indus-
y a ilia ion and sus ainabili y pe o mance a e posi i ely associa ed
wi h pho og aph use. We a e in e es ed in whe he his ela ionship
also holds o subjec ma e -speci ic pho og aphs.
To add ess hese gaps, we conduc a longi udinal s udy o p o ide
comp ehensi e empi ical e idence o he ole o pho og aphs in co -
po a e sus ainabili y epo ing. Acco dingly, we seek o con ibu e
mo e gene alizable indings on isual epo ing beha io by Eu opean
1100 FENK
companies. To his end, a new me hodological app oach is in oduced
o quan i y o e all pho og aph use and acili a e a machine-suppo ed
con en analysis o la ge pho og aphic da a se s. We hus shed ligh
on he subjec ma e con eyed h ough pho og aphs. Gi en he wide
ange o isual hemes communica ed in he epo s, we ocus ou
analysis on pho og aphs depic ing en i onmen al hemes. While each
s akeholde has a di e en se o p io i ies wi h ega d o co po a e
sus ainabili y, we a gue ha he en i onmen al pilla o co po a e sus-
ainabili y is a guably he mos p essing. In Eu ope, his is mani es ed
by an ambi ious egula o y agenda cen e ed a ound he Eu opean
G een Deal (Eu opean Commission, 2019). I co e s a ange o ini ia-
i es including s ee ing inancial capi al owa ds g een ac i i ies,
i.e., EU Taxonomy (Eu opean Pa liamen & Eu opean Council, 2020),
mo e s ingen sus ainabili y epo ing equi emen s, i.e., Co po a e
Sus ainabili y Repo ing Di ec i e (Eu opean Pa liamen & Eu opean
Council, 2022), and a amewo k o achie ing clima e neu ali y,
i.e., Eu opean Clima e Law (Eu opean Pa liamen & Eu opean
Council, 2021). Fu he mo e, hese p io i ies can also be obse ed
in he heigh ened in e es om he inancial communi y, e.g., he
‘Clima e Ac ion 100+g oup’, g een bond issuance, as well as gene al
public awa eness h ough he p ominence o he epo s by he
In e go e nmen al Panel on Clima e Change o socie al and poli ical
mo emen s.
On a mo e p ac ical no e, we also a gue ha objec s indica ing
en i onmen al subjec ma e can be mo e eliably iden i ied com-
pa ed o mo e abs ac concep s, such as labo ela ions o co po a e
go e nance. To li e up o he ela i e impo ance o he en i onmen-
al pilla wi hin co po a e sus ainabili y, we ocus ou analysis on he
isual communica ion o en i onmen al subjec ma e .
3|METHODOLOGY
In he ollowing sec ion, we in oduce ou no el machine-suppo ed
con en analysis app oach. Fi s ly, we ou line he sample selec ion
p ocess and desc ibe how we access pho og aph da a om co po a e
sus ainabili y epo s. The ea e , he iden i ica ion o en i onmen al
subjec ma e wi hin he pho og aphs is p esen ed. Las ly, we pu
o wa d a measu e o pho og aph u iliza ion and speci y he eg es-
sion models employed.
3.1 |Sample selec ion and da a ex ac ion
Fo his s udy, we collec ed da a om 10 consecu i e epo ing
pe iods be ween 2011 and 2020. We conside all companies om
coun ies in he Eu opean Economic A ea, which ha e con inuously
been pa o he STOXX Eu ope 600 index be ween Janua y 2011
and Decembe 2020. In his way, we can ensu e ha all i ms a e sub-
jec o a e y simila egula o y en i onmen and inancially s able. To
accu a ely assess pho og aph u iliza ion in Eu opean co po a e sus-
ainabili y epo ing, we limi ou sample o s and-alone sus ainabili y
epo s. This allows us o unequi ocally a ibu e e e y pho og aph o
a company's sus ainabili y disclosu e. Fo his s udy, we equi e ha a
sepa a ely published sus ainabili y epo is no addi ionally con ained
as a chap e wi hin he annual epo . Fu he mo e, he epo ing
pe iod has o ma ch he co esponding calend ical yea . The co po a e
sus ainabili y epo s we e downloaded as po able documen iles
(PDFs) om co po a e websi es and he GRI Sus ainabili y Disclosu e
Da abase (Global Repo ing Ini ia i e, 2020) in Sep embe 2020 and
No embe 2021. Ou inal sample con ains 1,463 s and-alone sus ain-
abili y epo s om 221 companies consis ing o 128,875 pages o
sus ainabili y- ela ed disclosu e. We summa ize he sample selec ion
p ocess in Table 1.
We ex ac ed he pho og aphs con ained in he epo s wi h a
Py hon sc ip buil upon he ‘ i z’ unc ionali y module
(Kas man, 2017). The p og am e ie es he image iles as po able
ne wo k g aphics (PNGs). In ou ins ances in which he sc ip was
unable o ex ac he image iles om he epo , we used a web-
based ex ac ion ool (Spike og SAS, 2020). As no all image iles con-
ained in he epo s a e ac ual pho og aphs, we had o clean he da a
se om non-pho og aphic PNG iles (Rämö, 2011), e.g., cha s,
TABLE 1 Sample selec ion p ocess.
S ep Explana ion
Numbe o
companies
We chose he STOXX Eu ope 600 index
which co e s he 600 la ges publicly lis ed
companies in Eu ope.
600
1 We included only companies in ou sample
which ha e con inuously been pa o he
STOXX 600 Eu ope index be ween Janua y
1s , 2011, and Decembe 31s , 2019, he
s a and end da e o he en-yea s udy
pe iod. Companies which ceased o exis o
en e ed he index in mo e ecen yea s we e
no conside ed.
(256)
2 We excluded companies which a e no
inco po a ed in he Eu opean Economic A ea,
i.e. Swiss companies, as hey a e no subjec
o he Eu opean Union's sus ainabili y
epo ing egula ion.
(31)
3 We conside ed only s and-alone
sus ainabili y epo s as opposed o
in eg a ed epo chap e s included in annual
epo s. Thus, we can unequi ocally a ibu e
e e y pho og aph employed o he
company's sus ainabili y epo ing. Mo eo e ,
we equi ed ha he epo ing pe iod has o
ma ch he co esponding calend ical yea .
(92)
Ou inal sample con ains 1,463 s and-alone
sus ainabili y epo s om 221 companies
consis ing o 128,875 pages o sus ainabili y-
ela ed disclosu e. The 1,463 sus ainabili y
epo s con ain 45,228 pho og aphs
221
No e: The sample con ains 1,463 s and-alone sus ainabili y epo s
disclosed by 221 STOXX Eu ope 600 companies sample which ha e
con inuously been pa o he index be ween Janua y 1s , 2011, and
Decembe 31s , 2019.
FENK 1101
da a ables, icons, co po a e logos, d awings, o ca oons. We also
excluded collages because hey ep esen se e al pho og aphs in one
image ile. We emo ed duplica e pho og aphs in each epo , in o de
o ensu e ha ecu ing pho og aphs such as a chap e sepa a o , a e
only included once. The da a cleaning p ocess had wo s eps: i s , we
applied a 50-kiloby e memo y size h eshold on all iles o exclude he
subs an ial numbe o co po a e logos and icons om he da a se .
Second, he emaining p ocedu e was based on a semi-au oma ic
p ocess, whe eby image iles a e lagged based on he subjec ma e
iden i ied by Google's compu e ision models. Ou inal da a se o
pho og aphs consis s o 45,228 pho og aphs con ained in 1,463
sus ainabili y epo s.
3.2 |Classi ica ion o En i onmen al Pho og aphs
Ad ances in a i icial in elligence-powe ed so wa e and compu ing
esou ces enable esea che s o analyze and d aw conclusions om
la ge-scale da a se s in an au oma ed way. Compu e ision acili-
a es sophis ica ed in o ma ion e ie al om g aphic ma e ial wi h-
ou ha ing p io knowledge o he p esen ed inpu s. I allows o
he analysis o much la ge da a se s han p e iously easible and
ensu es high consis ency in he assessmen o pho og aphic con-
en . Apa om open-sou ce amewo ks, he e a e se e al com-
me cial p o ide s o p e- ained compu e ision solu ions, which do
no equi e addi ional aining o he so wa e. In ou s udy, we use
Google's Vision API, which can be accessed h ough i s cloud com-
pu ing in as uc u e (Google LLC, 2023) and has been employed in
many scien i ic p ojec s (Chen & Lin, 2014; Nanne e al., 2020). Ou
app oach, which we ou line below, is in pa simila o he p o ocol
o analyzing g aphic ma e ial on web pages ou lined by A aujo
e al. (2020).
To analyze he pho og aphic da a se , we deploy Google's label
de ec ion ea u e o e ie e he pho og aph con en by assigning
labels. The pho og aphs a e hus agged wi h up o 10 desc ip i e
labels, as well as a co esponding con idence sco e o each label
anging be ween 0 and 1. We pe o med he da a labeling in Ap il
2022. In o al, he da a se o 45,228 pho og aphs we e agged wi h
452,004 labels, o which 3,059 a e unique alues. To classi y
whe he pho og aphs depic en i onmen al subjec ma e , we de el-
oped a codebook based on he lis o unique labels assigned by he
machine. Two esea che s join ly coded he labels, as o whe he
hey unequi ocally indica e ha he pho og aphs display an en i on-
men al heme. We de ined labels indica ing na u e scenes, animals,
o enewable ene gy p ojec s as en i onmen al hemes. Fu he mo e,
we only included labels ha unambiguously indica e ha he ele-
an objec is depic ed in he pho og aph. Hence, we do no con-
side labels ha indica e objec s ha a e e y likely p esen
alongside he ele an objec , e.g., ‘Shephe d’, who is likely p esen
along a sheep lock. Also, we excluded labels ha ha e a double
meaning, e.g., a ‘Py hon’may bo h be a p og amming language o a
snake. Las ly, we excluded he labels ‘Sky’and ‘Cloud’because hey
a e assigned o oo many pho og aphs ende ing hem oo
unspeci ic. The labels ‘Sky’and ‘Cloud’may s ill be assigned o a
pho og aph, bu we do no include hem in he lis o classi ying
labels. F om he ini ial lis o 3,059 unique labels, we classi ied
472 labels as en i onmen al labels. The en i onmen al labels and
hei co esponding con idence sco es a e hen used o de e mine
whe he a pho og aph is classi ied as an en i onmen al pho og aph.
In o de o quali y as an en i onmen al image, we s ipula e ha a
pho og aph has o be equipped wi h a leas one en i onmen al label
wi h a con idence sco e o o e 0.9 o wo en i onmen al labels wi h
a con idence sco e o o e 0.7. This esul s in 12,462 pho og aphs
labeled as en i onmen al pho og aphs ep esen ing 27.55% o all
pho og aphs in he da a se .
3.3 |Quan i ying pho og aph use
To da e, he e is no b oadly accep ed measu e o assessing pho o-
g aph u iliza ion in co po a e sus ainabili y epo ing. Mos s udies
use he sum o pho og aphs o de e mine he le el o pho og aph
u iliza ion, he sha e o a ce ain image ca ego y (Ali e al., 2021;
Chong e al., 2019; Rämö, 2011) o in a u he i e a ion include he
size o he pho og aphs in o hei assessmen (Ga cía-Sánchez &
A aújo-Be nado, 2020; H asky, 2012; In e nizzi e al., 2021; Nicolò
e al., 2022). Howe e , he e a e only wo g oups o au ho s who
calcula e pho og aph u iliza ion agains he ex en o he epo in
which hey a e con ained (Anan ha aman e al., 2020; B ei ba h
e al., 2010). We a gue ha 20 pho og aphs con ained in a 40-page
epo amoun o a subs an ially di e en use o pho og aphs han
20 pho og aphs as pa o a 100-page epo . Thus, we assess he use
o pho og aphs agains he epo co pus measu ed by he numbe o
pages o he epo documen . The numbe o epo pages is o en
e e ed o as epo ing quan i y (A idsson & Dumay, 2022; Fi ka &
D abble, 2012). Acco dingly, he use o all pho og aphs as well as
speci ically en i onmen al pho og aphs is assessed ela i e o he
epo ing quan i y o he pa icula epo . We call hese measu es
pho og aph u iliza ion (PUT) and en i onmen al pho og aph u iliza ion
(ENV PUT), espec i ely. In addi ion o hese absolu e measu es o
pho og aph use, we calcula e en i onmen al pho og aph use agains
he numbe o all pho og aphs con ained in he pa icula epo ,
which we call en i onmen al pho og aph sha e (ENV sha e). In his
s udy, he di e en measu es o pho og aph use se e as ou depen-
den a iables, which a e de i ed below. O e all pho og aph u iliza-
ion (PUT) o a pa icula epo is measu ed agains epo ing
quan i y:
PUT ¼pho og aph coun
page coun ð1Þ
En i onmen al pho og aph u iliza ion (ENV PUT) is equally
assessed ela i e o epo ing quan i y:
ENV PUT ¼en i onmen al pho og aph coun
page coun ð2Þ
1102 FENK
The en i onmen al pho og aph sha e (ENV sha e) o a pa icula
epo is measu ed as a sha e o all pho og aphs con ained in he
epo :
ENV sha e ¼en i onmen al pho og aph coun
pho og aph coun ð3Þ
3.4 |Reg ession model speci ica ion
The abo e-de i ed measu es o pho og aph use se e as he depen-
den a iables. Thus, we cons uc h ee eg ession models o es i-
ma e he pho og aph used in co po a e sus ainabili y epo s. The
independen and con ol a iables which a e used in all h ee eg es-
sion models a e speci ied below:
I:PUT
¼ß0þß1ENVsensi i eindus yi, þß2ENVpe o mancei,
þß3SOCpe o mancei, þß4GOVpe o mancei, þß5SHcoun yi,
þß6RQi, þß7CO2equ e enuei, þß8ESG epo scopei,
þß9ESGcon o e syi, þß10A gBoa dTenu ei, þß11SIZEi,
þß12OPMi, þεi,
ð4Þ
II :ENV PUT
¼ß0þß1ENVsensi i eindus yi, þß2ENVpe o mancei,
þß3SOCpe o mancei, þß4GOVpe o mancei, þß5SHcoun yi,
þß6RQi, þß7CO2equ e enuei, þß8ESG epo scopei,
þß9ESGcon o e syi, þß10A gBoa dTenu ei, þß11SIZEi,
þß12OPMi, þεi,
ð5Þ
III :ENV Sha e
¼ß0þß1ENVsensi i eindus yi, þß2ENVpe o mancei,
þß3SOCpe o mancei, þß4GOVpe o mancei, þß5SHcoun yi,
þß6RQi, þß7CO2equ e enuei, þß8ESG epo scopei,
þß9ESGcon o e syi, þß10A gBoa dTenu ei, þß11SIZEi,
þß12OPMi, þεi,
ð6Þ
The inancial and non- inancial da a we e ob ained om he Re ini-
i da abase (Kind e al., 2023; Landau e al., 2020; Re ini i Inc., 2023).
As we a gued in Sec ion 2.3, we a e in e es ed in iden i ying he ela-
ionships be ween sus ainabili y pe o mance and pho og aph use as
well as how pho og aph use a ies ac oss indus ies. These a iables
se e as independen a iables in ou eg ession. The a iables ENVpe -
o mance,SOCpe o mance, and GOVpe o mance deno e he h ee pilla
sco es which comp ise he commonly used En i onmen al, Social and
Go e nance (ESG) sco es by Re ini i . The mo e g anula pilla sco es
measu e i m pe o mance in he pa icula sus ainabili y domain.
ENVsensi i eindus y deno es whe he a company ope a es in an en i-
onmen ally sensi i e indus y. Following Cla kson e al. (2011) and
B anco and Rod igues (2008), we de ine he Pulp and Pape , Oil, Gas
and Coal, Chemicals, Me als and Mining, Cons uc ion as well as Build-
ing Ma e ials as en i onmen ally sensi i e indus ies gi en hei high
pollu ion p opensi y. The indus y a ilia ion is in line wi h he Indus y
Classi ica ion Benchma k (FTSE Russell, 2023).
In addi ion, we include se e al con ol a iables in ou eg ession
models. SHcoun y indica es whe he a company is domiciled in a
common law o ci il law coun y. S udies by Fi ka and D abble (2012)
and Kolk and Pe ego (2008) ha e shown ha cul u al di e ences play
an impo an ole in explaining di e ences in how companies
app oach sus ainabili y epo ing. RQ deno es he na u al loga i hm o
he sus ainabili y epo 's page coun as a p oxy o he amoun
o in o ma ion companies choose o disclose in hei epo ing. To
accoun o he di e ence be ween he en i onmen al sensi i i y o
he indus y in gene al and he indi idual g eenhouse gas emissions
o i ms, we include CO2equ e enue measu ing he na u al loga i hm
o CO2-equi alen emissions di ided by ne annual e enue. Fu he -
mo e, ESG epo scope inco po a es he sha e o a company's comme -
cial ac i i ies co e ed by i s non- inancial disclosu e. ESGcon o e sy is
included in ou models o con ol he le el o con o e sy su ounding
a company. We pe o m a na u al loga i hm ans o ma ion o all ESG
sco es o ensu e an app oxima e no mal dis ibu ion o he a iables.
A gBoa dTenu e deno es he a e age enu e o company boa d
membe s. Las ly, we con ol o company size, SIZE, measu ed by he
na u al loga i hm o ne annual e enue, as well as ope a ing p o i
ma gin, OPM, winso ized a he 1% le el. Non-Eu o igu es a e Eu o-
denomina ed using he Eu opean Cen al Bank's a e age exchange
a e o he pa icula epo ing pe iod.
‘I' and ‘ ’deno e i m i and yea . εi, ep esen s he e o e m.
The Hausman es o all models e u ns ha andom e ec s es ima-
o s and ixed e ec s es ima o s a e bo h consis en . Hence, we p o-
ceed wi h mo e e icien andom e ec s es ima o s. Fu he mo e, we
clus e he obus s anda d e o s by economic agen (i) as well as
ime pe iod ( ). The esul s o ou andom e ec s panel eg ession a e
shown below. Fo Model III, s a ed by equa ion (6), we impose an
addi ional es ic ion o conside only epo s wi h a leas i e pho o-
g aphs (n =1,185). We a gue ha , o epo s wi h ew pho og aphs,
he dis ibu ion o he dependen a iable would be hea ily dis o ed
by a high pe cen age o ex eme alues.
4|RESULTS
4.1 |Desc ip i e s a is ics and co ela ions
Ou sample includes 1,463 sepa a e epo s om 221 companies ha
ha e con inuously been pa o he STOXX Eu ope 600 index o la ge
Eu opean companies. Table 2lis s he numbe o obse a ions by
epo ing pe iod. The conside ed i ms a e based in 15 di e en
Eu opean coun ies, o which he Uni ed Kingdom ep esen s he la g-
es segmen wi h 65 sample companies, ollowed by 31 om F ance
and 30 om Ge many (see Table 3). Wi h ega d o he indus y
composi ion, Table 4 epo s he indus y segmen a ion in line wi h he
wo-digi codes o he Indus y Classi ica ion Benchma k (FTSE
Russel, 2023). Table 5shows he desc ip i e s a is ics summa y, includ-
ing minimum, mean, median, maximum, and s anda d de ia ion alues
o all con inuous dependen , independen , and con ol a iables. O all
conside ed epo s in he da a se , 498 (34.04%) we e published by
FENK 1103
companies om sha eholde -o ien ed coun ies domiciled in I eland o
he Uni ed Kingdom, and 965 om companies loca ed in con inen al
Eu ope. Conside ing he indus y sensi i i y classi ica ion, 366 epo s
(25.02%) we e disclosed by companies ope a ing in en i onmen ally
sensi i e indus ies. We calcula e Pea son's co ela ion coe icien s
o all independen and con ol a iables. The s onges co ela ion
be ween he a iables CO2equ e enue and ENVsensi i eindus y s ands
a 0.607. The addi ional es o a iance in la ion ac o (VIF) yields no
ac o exceeding 2.0. The a iable SIZE exhibi s he highes VIF alue o
1.68, 1.76, and 1.90 in Models I, II, and III, espec i ely. Hence, we see
no conce n wi h ega d o mul icollinea i y.
O e all, we ind ha pho og aph u iliza ion measu ed by he sum
o all pho og aphs ela i e o epo ing quan i y has allen be ween
2012 and 2017 by a ound 13% (see Figu e 1). Since 2017, he o e all
le el o pho og aph u iliza ion has been ela i ely s able a a ound
one pho og aph o e e y h ee epo pages. Agains his o e all
end o alling pho og aph u iliza ion, he sha e o pho og aphs con-
eying en i onmen al hemes has s eadily inc eased om 23.12% in
2013 o 31.13% in 2020 (see Figu e 2).
4.2 |Reg ession Resul s
Table 6p esen s he esul s o andom e ec s eg essions as speci ied
in Sec ion 3.4. The F- es o all models is highly signi ican .
Model I es ima es he o e all use o pho og aphs agains he
leng h o he epo . We ind ha less ex ensi e sus ainabili y epo s
ea u e subs an ially mo e pho og aphs ela i e o epo leng h,
i.e., an inc ease in epo ing quan i y by 10% is associa ed wi h a
dec ease o 0.021 ewe pho og aphs pe epo page (p < .0001).
Con olling o he di e ence in epo leng h, we also ind ha
epo s o con inen al Eu opean companies ea u e 0.225 ewe pho-
og aphs pe epo page han in epo s o hei UK o I ish pee s
(p < .0001). Models II and III conside only he use o en i onmen al
pho og aphs which a e measu ed agains he leng h o he epo
(ENV PUT) o as a sha e o all pho og aphs con ained in a pa icula
epo (ENV sha e). The highly signi ican nega i e associa ions
be ween epo ing quan i y and PUT, as well as he coun y o
domicile and PUT obse ed om Model I, also hold o en i onmen al
pho og aph u iliza ion. Howe e , hese wo ela ionships a e no sig-
ni ican o he ENV sha e measu e. Wi h ega d o he eg essions on
he use o en i onmen al pho og aphs, we ind ha he wo main
ela ionships shown below hold equally in bo h Models II and III.
We epo ha companies ope a ing in en i onmen ally sensi i e
indus ies use subs an ially mo e en i onmen al pho og aphs on an
absolu e le el, as well as a sha e o pho og aphs depic ing en i on-
men al subjec ma e . Bo h co ela ions a e highly signi ican a he
p < .0001 le el. Holding all else equal, he sha e o en i onmen al
TABLE 3 Numbe o companies by coun y o domicile.
Coun y Numbe o obse a ions
Aus ia 3
Belgium 7
Denma k 7
Finland 10
F ance 31
Ge many 30
I eland 3
I aly 13
Luxembou g 2
Ne he lands 11
No way 5
Po ugal 3
Spain 12
Sweden 19
Uni ed Kingdom 65
No e: The sample con ains 221 sample companies.
TABLE 4 Numbe o companies by indus y sec o .
ICB Indus y
Code Indus y name
Numbe o
obse a ions
10 Technology 9
15 Telecommunica ions 14
20 Heal h Ca e 13
30 Financials 41
35 Real Es a e 7
40 Consume
Disc e iona y
36
45 Consume S aples 22
50 Indus ials 38
55 Basic Ma e ials 15
60 Ene gy 11
65 U ili ies 15
No e: The sample con ains 221 companies in o al. ICB Indus ies Codes
deno e he wo-digi indus y codes o equi ies by FTSE Russell (2023).
TABLE 2 Numbe o epo s by epo ing pe iod.
Yea Numbe o obse a ions
2011 158
2012 162
2013 156
2014 162
2015 152
2016 147
2017 140
2018 130
2019 133
2020 123
No e: The sample con ains 1,463 epo obse a ions.
1104 FENK
gambling companies. Accoun ing, Audi ing & Accoun abili y Jou nal,
35(4), 1009–1034. h ps://doi.o g/10.1108/AAAJ-11-2019-4239
Eu opean Commission. (2019). Communica ion 2019/640 (The Eu opean
G een Deal). O icial Jou nal o he Eu opean Union.
Eu opean Pa liamen & Eu opean Council. (2014). Di ec i e 2014/95/EU
(Non-Financial Disclosu e). O icial Jou nal o he Eu opean Union.
Eu opean Pa liamen & Eu opean Council. (2020). Regula ion 2020/852
(EU Taxonomy). O icial Jou nal o he Eu opean Union.
Eu opean Pa liamen & Eu opean Council. (2021). Regula ion 2021/1119
(Eu opean Clima e Law). O icial Jou nal o he Eu opean Union.
Eu opean Pa liamen & Eu opean Council. (2022). Di ec i e 2022/2464
(Co po a e Sus ainabili y Repo ing). O icial Jou nal o he Eu opean
Union.
Fiech e , P., Hi z, J.-M., & Lehmann, N. (2022). Real E ec s o a Wide-
sp ead CSR Repo ing Manda e: E idence om he Eu opean Union's
CSR Di ec i e. Jou nal o Accoun ing Resea ch,60(4), 1499–1549.
h ps://doi.o g/10.1111/1475-679X.12424
Fi ka, M., & D abble, M. (2012). Focus and S anda diza ion o Sus ainabil-
i y Repo ing - A Compa a i e S udy o he Uni ed Kingdom and
Finland. Business S a egy and he En i onmen ,21(7), 455–474.
h ps://doi.o g/10.1002/bse.1730
F ias-Acei uno, J., Rod íguez-A iza, L., & Ga cía-Sánchez, I.-M. (2014).
Explana o y Fac o s o In eg a ed Sus ainabili y and Financial Repo -
ing. Business S a egy and he En i onmen ,23(1), 56–72. h ps://doi.
o g/10.1002/bse.1765
FTSE Russell. (2023). Indus y Classi ica ion Benchma k (Equi ies). Re ie ed
om h ps:// esea ch. se ussell.com/p oduc s/downloads/ICB_Rules_
new.pd
Ga cía-Sánchez, I.-M., & A aújo-Be nado, C. (2020). Wha colou is he
co po a e social esponsibili y epo ? S uc u al isual he o ic,
imp ession managemen s a egies, and s akeholde engagemen . Co -
po a e Social Responsibili y and En i onmen al Managemen ,27(2),
1117–1142. h ps://doi.o g/10.1002/cs .1869
Global Repo ing Ini ia i e. (2020). Sus ainabili y Dislosu e Da abase [Com-
pu e so wa e]. Re ie ed om h ps://da abase.global epo ing.o g/
sea ch/
Google LLC. (2023). Py hon google-cloud- ision [Compu e so wa e].
Re ie ed om h ps://pypi.o g/p ojec /google-cloud- ision/
G ay, R., Kouhy, R., & La e s, S. (1995). Co po a e social and en i onmen-
al epo ing. Accoun ing, Audi ing & Accoun abili y Jou nal,8(2), 47–77.
h ps://doi.o g/10.1108/09513579510146996
Guge y, M. (2009). Signaling i ue: olun a y accoun abili y p og ams
among nonp o i o ganiza ions. Policy Sciences,42(3), 243–273.
h ps://doi.o g/10.1007/s11077-009-9085-3
Gu h ie, J., & Pa ke , L. (1989). Co po a e Social Repo ing: A Rebu al o
Legi imacy Theo y. Accoun ing and Business Resea ch,19(76), 343–
352. h ps://doi.o g/10.1080/00014788.1989.9728863
Hahn, R., & Lül s, R. (2014). Legi imizing Nega i e Aspec s in GRI-O ien ed
Sus ainabili y Repo ing: A Quali a i e Analysis o Co po a e Disclo-
su e S a egies. Jou nal o Business E hics,123(3), 401–420. h ps://
doi.o g/10.1007/s10551-013-1801-4
Hallin, A., Ka bom-Gus a sson, T., & Dobe s, P. (2021). T ansi ion
owa ds and o sus ainabili y—Unde s anding sus ainabili y as pe o -
ma i e. Business S a egy and he En i onmen ,30(4), 1948–1957.
h ps://doi.o g/10.1002/bse.2726
He zig, C., & Schal egge , S. (2011). Co po a e Sus ainabili y Repo ing. In
J. Godemann & M. Michelsen (Eds.), Sus ainabili y Communica ion
(pp. 151–169). Sp inge , Cham. h ps://doi.o g/10.1007/978-94-007-
1697-1_14
Higgins, C., Tang, S., & S ubbs, W. (2020). On managing hypoc isy: The
anspa ency o sus ainabili y epo s. Jou nal o Business Resea ch,
114, 395–407. h ps://doi.o g/10.1016/j.jbus es.2019.08.041
Hooghiems a, R. (2000). Co po a e Communica ion and Imp ession
Managemen –New Pe spec i es Why Companies Engage in
Co po a e Social Repo ing. Jou nal o Business E hics,27,55–68.
h ps://doi.o g/10.1023/A:1006400707757
Hopwood, A. (2009). Accoun ing and he en i onmen . Accoun ing, O gani-
za ions and Socie y,34(3–4), 433–439. h ps://doi.o g/10.1016/j.aos.
2009.03.002
H asky, S. (2012). Visual disclosu e s a egies adop ed by mo e and less
sus ainabili y-d i en companies. Accoun ing Fo um,36(3), 154–165.
h ps://doi.o g/10.1016/j.acc o .2012.02.001
Hummel, K., & Schlick, C. (2016). The ela ionship be ween sus ainabili y
pe o mance and sus ainabili y disclosu e –Reconciling olun a y dis-
closu e heo y and legi imacy heo y. Jou nal o Accoun ing and Public
Policy,35(5), 455–476. h ps://doi.o g/10.1016/j.jaccpubpol.2016.
06.001
In e nizzi, A., Bellucci, M., Acu i, D., & Mane i, G. (2021). Fo m and sub-
s ance: Visual con en in CSR epo s and in es o s' pe cep ions. Psy-
chology & Ma ke ing,39(5), 974–989. h ps://doi.o g/10.1002/ma .
21635
Kas man, E. (2017). Py hon i z (Ve sion 0.0.1.de 2) [Compu e so wa e].
Kind, F., Zeppen eld, J., & Lueg, R. (2023). The impac o chie execu i e
o ice na cissism on en i onmen al, social, and go e nance epo ing.
Business S a egy and he En i onmen ,32(7), 4448–4466. h ps://doi.
o g/10.1002/bse.3375
Knobloch, S., Has all, M., Zillmann, D., & Callison, C. (2003). Image y
E ec s on he Selec i e Reading o In e ne Newsmagazines. Commu-
nica ion Resea ch,30(1), 3–29. h ps://doi.o g/10.1177/00936502
02239023
Kolk, A., & Pe ego, P. (2008). De e minan s o he adop ion o sus ainabil-
i y assu ance s a emen s: an in e na ional in es iga ion. Business S a -
egy and he En i onmen ,19(3), 182–198. h ps://doi.o g/10.1002/
bse.643
K ippendo , K. (2018). Con en analysis: An in oduc ion o i s me hodology
(4 h ed.). SAGE.
Landau, A., Rochell, J., Klein, C., & Zwe gel, B. (2020). In eg a ed epo ing
o en i onmen al, social, and go e nance and inancial da a: Does he
ma ke alue in eg a ed epo s? Business S a egy and he En i on-
men ,29(4), 1750–1763. h ps://doi.o g/10.1002/bse.2467
Lee, T. (1994). The changing o m o he co po a e annual epo . The
Accoun ing His o ians Jou nal,21(1), 215–232. h ps://doi.o g/10.
2308/0148-4184.21.1.215
Lyon, T., & Mon gome y, W. (2015). The Means and End o G eenwash.
O ganiza ion & En i onmen ,28(2), 223–249. h ps://doi.o g/10.1177/
1086026615575332
Mahoney, L., Tho ne, L., Cecil, L., & LaGo e, W. (2013). A esea ch no e on
s andalone co po a e social esponsibili y epo s: Signaling o g een-
washing? C i ical Pe spec i es on Accoun ing,24(4–5), 350–359.
h ps://doi.o g/10.1016/j.cpa.2012.09.008
Meng, X., Zeng, S., & Tam, C.-M. (2013). F om Volun a ism o Regula ion:
A S udy on Owne ship, Economic Pe o mance and Co po a e En i-
onmen al In o ma ion Disclosu e in China. Jou nal o Business E hics,
116(1), 217–232. h ps://doi.o g/10.1007/s10551-012-1462-8
Me kl-Da ies, D., & B ennan, N. (2007). Disc e iona y Disclosu e S a e-
gies in Co po a e Na a i es: Inc emen al In o ma ion o Imp ession
Managemen ? Jou nal o Accoun ing Li e a u e,26, 116–196.
Me kl-Da ies, D., & B ennan, N. (2011). A Concep ual F amewo k o
Imp ession Managemen . Accoun ing and Business Resea ch,41(5),
415–437. h ps://doi.o g/10.1080/00014788.2011.574222
Meue , J., Koelbel, J., & Ho mann, V. (2020). On he Na u e o Co po a e
Sus ainabili y. O ganiza ion & En i onmen ,33(3), 319–341. h ps://
doi.o g/10.1177/1086026619850180
Milne, M., & Pa en, D. (2002). Secu ing o ganiza ional legi imacy. Accoun -
ing, Audi ing & Accoun abili y Jou nal,15(3), 372–405. h ps://doi.o g/
10.1108/09513570210435889
Nanne, A., An heunis, M., an de Lee, C., Pos ma, E., Wubben, S., & an
Noo , G. (2020). The Use o Compu e Vision o Analyze B and-
FENK 1111
Rela ed Use Gene a ed Image Con en . Jou nal o In e ac i e Ma ke -
ing,50, 156–167. h ps://doi.o g/10.1016/j.in ma .2019.09.003
Nicolò, G., Riccia delli, A., Raimo, N., & Vi olla, F. (2022). Visual disclosu e
h ough in eg a ed epo ing. Managemen Decision,60(4), 976–994.
h ps://doi.o g/10.1108/MD-01-2021-0034
Nieho , S. (2022). Aligning digi alisa ion and sus ainable de elopmen ?
E idence om he analysis o wo ld iews in sus ainabili y epo s.
Business S a egy and he En i onmen ,31(5), 2546–2567. h ps://doi.
o g/10.1002/bse.3043
Pai io, A. (1969). Men al image y in associa i e lea ning and memo y. Psy-
chological Re iew,76(3), 241–263. h ps://doi.o g/10.1037/h0027272
Pai io, A., Roge s, T., & Smy he, P. (1968). Why a e pic u es easie o ecall
han wo ds? Psychonomic Science,11(4), 137–138. h ps://doi.o g/10.
3758/BF03331011
Pa en, D. (2002). The ela ion be ween en i onmen al pe o mance and
en i onmen al disclosu e: a esea ch no e. Accoun ing, O ganiza ions
and Socie y,27(8), 763–773. h ps://doi.o g/10.1016/S0361-3682
(02)00028-4
Peeples, J. (2011). Toxic Sublime: Imaging Con amina ed Landscapes. En i-
onmen al Communica ion,5(4), 373–392. h ps://doi.o g/10.1080/
17524032.2011.616516
Pé ez-Cañiza es, P. (2022). Visualising Co po a e Social Responsibili y.
Ibé ica,43, 155–178. h ps://doi.o g/10.17398/2340-2784.43.155
Pesci, C., Cos a, E., & Sooba oyen, T. (2015). The o ms o epe i ion in
social and en i onmen al epo s: insigh s om Hume's no ion o
‘imp essions’.Accoun ing and Business Resea ch,45(6–7), 765–800.
h ps://doi.o g/10.1080/00014788.2015.1084224
P ado-Lo enzo, J.-M., & Ga cía-Sánchez, I.-M. (2010). The Role o he
Boa d o Di ec o s in Dissemina ing Rele an In o ma ion on G een-
house Gases. Jou nal o Business E hics,97(3), 391–424. h ps://doi.
o g/10.1007/s10551-010-0515-0
P es on, A. M., W igh , C., & Young, J. J. (1996). IMag[in]ing annual
epo s. Accoun ing, O ganiza ions and Socie y,21(1), 113–137.
h ps://doi.o g/10.1016/0361-3682(95)00032-5
Rämö, H. (2011). Visualizing he Ph one ic O ganiza ion: The Case o Pho-
og aphs in CSR Repo s. Jou nal o Business E hics,104(3), 371–387.
h ps://doi.o g/10.1007/s10551-011-0916-8
London S ock Exchange G oup PLC. (2023). Eikon Da abase [Compu e
so wa e].
Re e e, C. (2009). De e minan s o Co po a e Social Responsibili y Dis-
closu e Ra ings by Spanish Lis ed Fi ms. Jou nal o Business E hics,
88(2), 351–366. h ps://doi.o g/10.1007/s10551-008-9968-9
Rezaee, Z. (2016). Business sus ainabili y esea ch: A heo e ical and in e-
g a ed pe spec i e. Jou nal o Accoun ing Li e a u e,36(1), 48–64.
h ps://doi.o g/10.1016/j.accli .2016.05.003
Shabana, K., Buchhol z, A., & Ca oll, A. (2017). The Ins i u ionaliza ion o
Co po a e Social Responsibili y Repo ing. Business & Socie y,56(8),
1107–1135. h ps://doi.o g/10.1177/0007650316628177
Spike og SAS. (2020). PDF Ex ac o [Compu e so wa e]. Re ie ed om
h ps://www.ex ac pd .com/
Suchman, M. (1995). Managing Legi imacy: S a egic and Ins i u ional
App oaches. The Academy o Managemen Re iew,20(3), 571. h ps://
doi.o g/10.2307/258788
Townsend, C., & Shu, S. (2010). When and how aes he ics in luences
inancial decisions. Jou nal o Consume Psychology,20(4), 452–458.
h ps://doi.o g/10.1016/j.jcps.2010.06.013
T eisman, A. (1985). P ea en i e p ocessing in ision. Compu e Vision,
G aphics, and Image P ocessing,31(2), 156–177. h ps://doi.o g/10.
1016/S0734-189X(85)80004-9
T e sky, A., & Kahneman, D. (1986). Ra ional choice and he aming o
decisions. Jou nal o Business,59(4), 251–278. h ps://doi.o g/10.
1086/296365
T e sky, B. (1974). Eye ixa ions in p edic ion o ecogni ion and ecall.
Memo y & Cogni ion,2(2), 275–278. h ps://doi.o g/10.3758/
BF03208995
Une man, J. (2000). Me hodological issues - Re lec ions on quan i ica ion
in co po a e social epo ing con en analysis. Accoun ing, Audi ing &
Accoun abili y Jou nal,13(5), 667–681. h ps://doi.o g/10.1108/
09513570010353756
Usmani, M., Da ison, J., & Napie , C. (2020). The p oduc ion o s and-
alone sus ainabili y epo s: isual imp ession managemen , legi imacy
and “ unc ional s upidi y”.Accoun ing Fo um,44(4), 315–343. h ps://
doi.o g/10.1080/01559982.2020.1782566
Weda i, L., Jubb, C., & Mo adi-Mo lagh, A. (2021). Co po a e clima e-
ela ed olun a y disclosu es: Does po en ial g eenwash exis among
Aus alian high emi e s epo s? Business S a egy and he En i on-
men ,30(8), 3721–3739. h ps://doi.o g/10.1002/bse.2836
Zeng, X., Momin, M., & Nu unnabi, M. (2022). Pho o disclosu e in human
igh s issues by o une companies: an imp ession managemen pe -
spec i e. Sus ainabili y Accoun ing, Managemen and Policy Jou nal,
13(3), 568–599. h ps://doi.o g/10.1108/SAMPJ-06-2019-0243
Zillmann, D., Gibson, R., & Sa gen , S. (1999). E ec s o Pho og aphs in
News-Magazine Repo s on issue Pe cep ion. Media Psychology,1(3),
207–228. h ps://doi.o g/10.1207/s1532785xmep0103_2
How o ci e his a icle: Fenk, L. (2025). En i onmen al
pho og aph use in co po a e sus ainabili y epo ing: A
machine-suppo ed isual con en analysis. Business S a egy
and he En i onmen ,34(1), 1097–1112. h ps://doi.o g/10.
1002/bse.4035
1112 FENK
