The substitutability between brick-and-mortar stores and e-Commerce: the case of books

Gö z, Geo g; Klo z, Phil-Ad ian; Schä e , Jan Thomas; He old, Daniel
A icle — Published Ve sion
The subs i u abili y be ween b ick-and-mo a s o es and
e-Comme ce: he case o books
Jou nal o Cul u al Economics
Sugges ed Ci a ion: Gö z, Geo g; Klo z, Phil-Ad ian; Schä e , Jan Thomas; He old, Daniel (2025) : The
subs i u abili y be ween b ick-and-mo a s o es and e-Comme ce: he case o books, Jou nal o
Cul u al Economics, ISSN 1573-6997, Sp inge US, New Yo k, Vol. 49, Iss. 4, pp. 811-854,
h ps://doi.o g/10.1007/s10824-025-09544-2
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/333382
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/
Vol.:(0123456789)
Jou nal o Cul u al Economics (2025) 49:811–854
h ps://doi.o g/10.1007/s10824-025-09544-2
ORIGINAL ARTICLE
The subs i u abili y be weenb ick‑and‑mo a s o es
ande‑Comme ce: hecase o books
Geo gGö z1· DanielHe old1 · Phil‑Ad ianKlo z2· JanThomasSchä e 1
Recei ed: 21 Ap il 2023 / Accep ed: 2 May 2025 / Published online: 25 June 2025
© The Au ho (s) 2025, co ec ed publica ion 2025
Abs ac
We analyze compe i ion be ween he online and he o line e ail channel by using
da a on he Ge man book ma ke , which is cha ac e ized by ixed book p ices. The
analysis sheds ligh on he ex en o which consume s pe cei e e-Comme ce and
adi ional b ick-and-mo a s o es as subs i u es. We ind ha , on a e age, when a
books o e closes, sales o p in books dec ease by a ound 744 uni s pe mon h. This
explains abou 37% o he o al loss in sales o p in books in ou sample. These
indings indica e impe ec subs i u abili y be ween he online and he o line e ail
channel. Subs i u abili y be ween he channels emains impe ec when we inco po-
a e in o ma ion on e-book sales. The magni ude o he e ec is gen e-dependen .
Fo ins ance, sales o ic ion i les dec ease mo e s ongly han sales o school books.
The da a used in ou s udy we e p o ided by media con ol GmbH, G K GmbH and
Acxiom Deu schland GmbH. We ecei ed unding om he Ge man Publishe s and
Bookselle s Associa ion (“Bö sen e ein des Deu schen Buchhandels e. V.”) o buy
he da a. Mo eo e , he Ge man Publishe s and Bookselle s Associa ion unded a
esea ch p ojec a he Chai o Geo g Gö z om 2018 o 2020. Ou s udy is a p od-
uc o his p ojec . The p ojec i sel was scien i ic in na u e (i. e., no comme cial
esea ch p ojec ). To conduc his p ojec , he posi ions o he coau ho s Jan Thomas
Schä e and Daniel He old we e unded by he Ge man Publishe s and Bookselle s
Associa ion du ing he a o emen ioned pe iod.
Keywo ds P oduc di e en ia ion· Book ma ke · Re ailing· e-Comme ce
We hank he Ge man Bookselle s’ Associa ion, Joel Wald ogel, Samuel de Haas, Da id Finck,
Maximilian Mau ice Gailand Daniel Lüke o con inuous suppo and ui ul discussions.
* Daniel He old
[email p o ec ed]
1 Chai o Indus ial O ganiza ion, Regula ion andAn i us , Depa men o Economics, Jus us
Liebig Uni e si y Giessen, Liche S asse 62, 35394Giessen, Ge many
2 Düsseldo Ins i u e o Compe i ion Economics (DICE), Hein ich Heine Uni e si y Düsseldo ,
Uni e si ä ss aße 1, 40225Düsseldo , Ge many
812
Jou nal o Cul u al Economics (2025) 49:811–854
JEL Classi ica ion L13· L81· D12· L42
1 In oduc ion
Wi h he ise o e-Comme ce, con en ional b ick-and-mo a e ail sec o s ha e
expe ienced a subs an ial inc ease in compe i i e p essu e ac oss a ious indus ies
(see, e.g., Bu and Spa ks 2003; S ini asan e al. 2002). I emains an open ques ion
o wha ex en and in which indus ies consume s pe cei e e-Comme ce and adi-
ional, physical s o es as subs i u es (see, e.g., Wang and Gold a b 2017; Sinai and
Wald ogel 2004; B ynjol sson e al. 2009). In his a icle, we in es iga e he subs i-
u abili y be ween e-Comme ce and physical e aile s in he Ge man book ma ke .
Physical books o es and e-Comme ce appa en ly ha e di e en “se ice” ea-
u es. e-Comme ce ypically o e s con enien p oduc sea ch op ions among po en-
ially huge in en o ies (see, e.g., Wald ogel 2017). Howe e , o de ing online en ails
wai ing cos s and e-Comme ce does no o e he possibili y o physical inspec ion
(Logino a, 2009; Guo and Lai, 2017). Books a e conside ed expe ience goods,
so ha p epu chase unce ain y po en ially a ec s demand (Reins ein and Snyde ,
2005; Hilge e al., 2011). Agains his backg ound, consume s ha e he possibil-
i y o ecei e ad ice a physical books o es o ia online e iews o a ings (see
also Reime s and Wald ogel 2021; Lizze i 1999; Clemen e al. 2007; Ma el and
McCa e y 1990).
Cowen (2008) and Gilbe (2015) add ess ha while educing he impo ance o
physical s o es, e-Comme ce migh expand o e all demand o books by a ac ing
new cus ome s. Howe e , gi en he di e ences in se ices o e ed by he wo chan-
nels, i is unclea o wha ex en consume s iew he o line and online channels as
subs i u es. I hey a e conside ed ela i ely a subs i u es by su icien ly many con-
sume s, i could be ha a dec ease in he numbe o physical books o es (po en ially
igge ed by ad ancing digi iza ion) dec eases sales when (po en ially cap i e) con-
sume s o he o line channel educe hei demand o books. Analyzing he subs i u-
ion pa e ns be ween he wo e ail channels can hus imp o e he unde s anding o
how digi iza ion a ec s he e ail landscape. The Ge man book ma ke is pa icula ly
well-sui ed o such an analysis because i is cha ac e ized by ixed book p ices such
ha he p ice-dimension plays no ole in he compa ison be ween he wo channels.
To analyze o wha ex en consume s pe cei e he online and o line channels as
subs i u es, we in es iga e how closu es o physical books o es a ec book sales. A
no el panel da a se is used which consis s o mon hly sales da a o physical e ail-
e s and e-Comme ce in Ge many co e ing he pe iod 2011–2017. The da a se also
con ains in o ma ion on he numbe o physical e aile s. The ela ionship be ween
he numbe o physical books o es and book sales may be bidi ec ional, i.e., demand
may inc ease in he numbe o ou le s o he numbe o ou le s may inc ease in
demand. We employ an ins umen al a iable (IV) app oach using a p oxy o popu-
la ion as an ins umen o ob ain consis en es ima es.
Ou analysis sugges s ha , on a e age, he closu e o one physical books o e
esul s in a mon hly decline o 744 book sales pe ede al s a e. Be ween 2011 and
2017, books o e closu es collec i ely con ibu ed o a dec ease in mon hly sales o
813
Jou nal o Cul u al Economics (2025) 49:811–854
a ound 1 million uni s, accoun ing o app oxima ely 37% o he o e all decline in
mon hly book sales.
The d op in sales appea s o be gen e-speci ic. Fo example, sales o ic ion i les
dec ease by a ound 208 uni s on a e age pe mon h and ede al s a e when a book
s o e closes, whe eas we ind no e ec o schoolbooks. Rema kably, o ic ion i les
and child en books we ind ha a dec ease exi o physical books o es implies also
a weak dec ease in sales in e-Comme ce. This inding indica es complemen a y
be ween he wo channels in he sense ha e-Comme ce bene i s om he p esence
o physical books o es. O e all, ou indings p o ide e idence ha e-Comme ce is
no pe ec subs i u e o physical books o es.
We con ibu e o he li e a u e on he impac o digi iza ion on “ adi ional” phys-
ical e aile s. The e a e nume ous a icles analyzing compe i ion be ween he wo
e ail channels, such as B own and Goolsbee (2002), Jin and Ka o (2007), O ek e al.
(2011), Gau i e al. (2021) o Cou u e e al. (2021). The impac o digi iza ion on he
e ail landscape has been s udied om di e en angles, including he opics c owd
a ings (Reime s and Wald ogel, 2021), he ole o niche-p oduc s (B ynjol sson
e al., 2003, 2006; Reime s and Wald ogel, 2017) and inno a ion, di usion and cop-
y igh p o ec ion (Wald ogel and Reime s, 2015; Wald ogel, 2017; Reime s, 2016).
Pa icula ly closely ela ed o his a icle a e he pape s o Goolsbee (2001), P ince
(2007) and Goldmanis e al. (2010). Goolsbee (2001) and P ince (2007) analyze
he ma ke o compu e equipmen and ind e idence o in ensi ying compe i ion
be ween he wo channels h oughou he la e 1990s. Goldmanis e al. (2010) ana-
lyzes compe i ion be ween he online and o line channel in h ee ma ke s in he US:
a el agencies, books o es and new ca deale s. Wi h espec o books o es, hey
documen a shi o ma ke sha es om small books o es o la ge chains in ma ke s
ha a e exposed o a la ge deg ee o online compe i ion, leading o a decline in he
numbe o books o es and employmen . In he con ex o he impac o digi iza ion
on he book ma ke , ou pape is closely ela ed o B ynjol sson and Smi h (2000),
Clay e al. (2002) and Che alie and Goolsbee (2003). These s udies, howe e , ana-
lyze ma ke s whe e e aile s compe e in p ices. Wi h ixed book p ices (i.e., in he
absence o p ice compe i ion be ween e aile s), we ind ha he dec ease in sales
igge ed by he closu e o a books o e a e no ully compensa ed by inc eased sales
o o he books o es o in e-Comme ce.
The a icle is also ela ed o he li e a u e and policy discussion on e ical
es ain s, pa icula ly wi h espec o he e alua ion o po en ial e iciency e ec s
a ising om esale p ice main enance (RPM). This e alua ion has become an impo -
an opic o compe i ion policy a leas since he Leegin-Case, which led o RPM
being shi ed om pe se illegali y o a ule o eason based app oach in he US in
2007. In he EU, RPM is s ill p ac ically conside ed pe se illegal (see Akman and
Sokol 2017) o a mo e de ailed o e iew). Howe e , in a ious Eu opean coun ies
such as F ance, Aus ia and Ge many (and o he coun ies such as Japan),1 ixed
book p ice sys ems a e in place, which occu in he o m o RPM. As he goal o
1 See, o ins ance, Global Fixed Book P ice Repo o he In e na ional Publishe s Associa ion, May 23,
2014, h ps:// bi . ly/ 2Wg1 pz. Acco ding o Poo and an Eijk (2017), in 15 OECD-coun ies ( en EU-
membe s), he e is a ixed book p ice sys em in place.
814
Jou nal o Cul u al Economics (2025) 49:811–854
ixed book p ices is usually o p o ec books as a cul u al o me i good, hey cons i-
u e an excep ion when i comes o he applica ion o compe i ion law. Whe he such
an exemp ion om some undamen al pilla s o compe i ion law is jus i ied, c u-
cially depends on ixed book p ices being an app op ia e ool o achie e he policy
goal o p omo e he demand and supply o books. Agains he backg ound o ou
indings, one way h ough which ixed book p ices could a ec book sales would
be secu ing ma gins and he eby po en ially p omo ing ma ke en y o p e en ing
exi o physical e aile s (Bouckae , 2000; Elzinga and Mills, 2008; Ma el and
McCa e y, 1985). In pa icula , Williams (2024) inds ha , on a e age, book sales
a e highe in coun ies wi h RPM, while p ices a e simila . He explains his indings
by a posi i e impac o ixed book p ices on he ne wo k o books o es.
The a icle is s uc u ed as ollows. Sec ion2 con ains a b ie concep ualiza ion
o ou empi ical analysis based on economic heo y. Ou da a se is desc ibed in
Sec .3. Sec ion4 p esen s ou empi ical analysis. Sec ion5 p o ides an o e iew o
he obus ness checks ha a e p esen ed in he Appendix. Sec ion6 concludes.
2 Theo y
The ollowing sec ion p o ides a b ie heo e ical backg ound o he empi ical anal-
yses p esen ed in he emainde o his a icle. The s a ing poin is he no ion ha
he online and o line e ail channels ha e di e en “se ice” ea u es, as was al eady
explained in he in oduc ion. Consume s a y ega ding hei p e e ences owa d
hese ea u es. This is an example o (ho izon al) p oduc di e en ia ion, which, in
he li e a u e, is usually o malized using Ho elling and Salop models whe e con-
sume s’ p e e ences ega ding he wo sales channels a e e lec ed in di e en ans-
po cos s (Balasub amanian, 1998; Bouckae , 2000; Chu e al., 2012; Guo and Lai,
2017; Leg os and S ahl, 2019).
In ou empi ical se ing, we s udy he e ec o a physical books o e’s ma ke exi .
In a Ho elling o Salop model, ce e is pa ibus such an exi has no e ec on consum-
e s buying online o pa onizing a books o e ha is no closing. Ne u ili y o hose
consume s who pa onized a books o e ha is closing dec eases due o an inc ease
in incon enience cos s om ei he pa onizing ano he books o e o pu chasing
online. This can lead o a dec ease in o e all demand i ha inc ease is so high ha ,
o gi en p ices, consume s ne u ili y om pu chasing books a ano he books o e
o online becomes nega i e.2
I is possible in such a se ing ha exi o physical books o es ce e is pa ibus
leads o a dec ease in ma ke co e age and, hus, demand o books. We in e p e
his pa e n as impe ec subs i u ion be ween he online and he o line channel.
Tha is, some consume s who pa onized a physical books o e and who ha e high
2 Such a o maliza ion equi es a model ha allows o pa ial ma ke co e age, i.e., a model in which
no all consume s always pu chase he p oduc by assump ion. To cons uc such a model, one would
need o depa om he assump ion o symme ic incon enience cos s om buying online, as i is usually
used in he li e a u e (see abo e). Ins ead, one could assume ha hese cos s di e be ween consume s.
An ea lie e sion o his pape con ained such a model. I is a ailable om he au ho s upon eques .

815
Jou nal o Cul u al Economics (2025) 49:811–854
incon enience cos s om buying online may s op buying al oge he because nei he
e-Comme ce no he emaining books o es a e pe cei ed as “close enough” subs i-
u es. This no ion may se e as a heo e ical in e p e a ion o he ollowing empi ical
analyses.
3 Da a anddesc ip i e s a is ics
Fo ou empi ical analysis, we use mon hly da a on he numbe o book e aile s as
well as sales olumes and e enue da a on a ede al s a e le el. The da a comp ise
sales o books in b ick-and-mo a s o es as well as e-comme ce sales, o example,
books shipped by Amazon. We dis inguish be ween nine di e en p oduc g oups o
con ol o gen e-speci ic e ec s. The da abase co e s he pe iod om Janua y 2011
o Decembe 2017. We u ilize a combina ion o da a om mul iple sou ces. In wha
ollows, we p esen some desc ip i e s a is ics and we desc ibe he da a sou ces in
mo e de ail.
3.1 B ick‑and‑mo a books o es
The Ge man Publishe s and Bookselle s Associa ion3 main ains se e al da abases
on he numbe o b ick-and-mo a s o es, such as i s membe ship da abase and he
Add ess Book o he Ge man-Language Book T ade4, which we ha e combined o
ob ain an accu a e pic u e o he ma ke and i s de elopmen o e ime. We checked
he consis ency o ou da ase using da a on he o de olume o books o es p o-
ided by h ee o he la ges Ge man wholesale s.5 Appendix A.1 con ains an in-
dep h desc ip ion o he da a gene a ion p ocess.
Figu e1 shows he de elopmen o he numbe o physical books o es in Ge -
many o e ime, based on da a om he Ge man Publishe s and Bookselle s Associ-
a ion. The da a include in o ma ion on ou le s o chain s o es as well as independen
books o es. Be ween 2011 and 2017, he numbe o physical books o es dec eased
om o e 6,300 o less han 4,900.
Figu e2 p o ides summa y s a is ics o books o es on he ede al s a e le el. On
a e age, he numbe o b ick-and-mo a books o es dec eased by 20 pe cen o e
he obse a ion pe iod. In absolu e numbe s, a ound 1,435 books o es exi ed he
ma ke .
Figu e3 shows he numbe o b ick-and-mo a s o es pe 100,000 esiden s in
beginning o 2011 and end o 2017.6
3 Bö sen e ein des Deu schen Buchhandels, see h ps:// www. boe s en e ein. de.
4 Ad essbuch ü den deu schsp achigen Buchhandel, see h ps:// adb- online. de; las accessed Decembe
19, 2023.
5 Lib i GmbH (h ps:// www. lib i. de), G. Umb ei GmbH & Co. KG (h ps:// www. umb e i . de), and Koch,
Ne & Volckma GmbH (KNV, h p:// www. kn . de).
6 No e ha we use labo o ce as a p oxy o popula ion. This is because popula ion da a a e only pub-
lished on a yea ly equency. See Sec .4 o u he discussion.
816
Jou nal o Cul u al Economics (2025) 49:811–854
Fig. 1 De elopmen o physical books o es o e ime. Sou ce: Ge man Publishe s and Bookselle s Asso-
cia ion
Fig. 2 De elopmen o physical books o es on he ede al s a e le el. Sou ce: Ge man Publishe s and
Bookselle s Associa ion
817
Jou nal o Cul u al Economics (2025) 49:811–854
3.2 Sales da a
In o de o s udy he de elopmen o sales o e he obse a ion pe iod, we use da a
om wo sou ces. Fi s , we ha e access o scanne da a p o ided by media con ol
GmbH.7 Second, we use consume panel da a p o ided by G K GmbH.8 Bo h da a-
bases di e wi h espec o co e age, equency and accu acy. The scanne da a a e
based on quasi eal ime da a and comp ises in o ma ion on sales o p in books in
independen books o es, chain s o es as well as online e aile s (see Appendix A.2
o mo e cha ac e is ics o he da a se ). The da a p o ide claims o co e mo e han
80–90% o he Ge man p in book ma ke . Howe e , he scanne da a do no con ain
in o ma ion on he sale o e-Books. To he bes o ou knowledge, he mos eli-
able in o ma ion on e-Book sales in Ge many a ailable o esea che s a e consume
panel da a, which, o he obse a ion pe iod 2014–2017, is p o ided by G K. The
sales da a p ojec ions o G K a e based on oughly 20,000 consume s being su -
eyed on a egula basis and con ain in o ma ion on he pu chases o p in books,
e-Books and audio books. In con as o he panel da a, he su ey da a a e only
a ailable on a qua e ly le el o he pe iod om Q1.2014 o Q4.2017 (see Appendix
A.3 o mo e cha ac e is ics o he da a se ). Fo he yea s 2011 o 2013, su ey da a
Fig. 3 Numbe o physical books o es pe 100,000 esiden s in beginning o 2011 and end o 2017.
Sou ce: Ge man Publishe s and Bookselle s Associa ion
7 see h ps:// www. media- con ol. de/ buch1. h ml; las accessed Decembe 20, 2023.
8 see h ps:// www. g k. com/ de; las accessed Decembe 20, 2023.
818
Jou nal o Cul u al Economics (2025) 49:811–854
a e only a ailable as yea ly agg ega es and o whole Ge many, a he han on he
ede al s a e le el and a e included o illus a i e pu poses only.9
Figu e 4 shows sales igu es o physical books om he scanne da a se . I
becomes ob ious ha he sales o p in books in Ge many exhibi a dec easing
end. In 2011, 297.2 million p in books we e sold in b ick-and-mo a and online
s o es wi h a co esponding e enue o 3.66 billion Eu os. This implies an a e age
book p ice o 12.31 Eu o. In 2017, 264 million books we e sold wi h a e enue o
3.54 billion Eu os (i.e., 13.41 Eu o pe book). The decline in sales was hus a leas
pa ly o se by a p ice inc ease so ha e enues o b ick-and-mo a and online
s o es emained ai ly cons an .
As can be seen in Fig.5, he numbe o p in books sold pe esiden dec eased
be ween 2011 and 2017.
A he same ime, e-Book sales inc eased om 4.30 million in 2011 o 29.15 mil-
lion in 2017. As can be seen om Fig.6, he g ow h a e o e-book sales s alled
a e a sha p inc ease and co esponding e enues e en declined be ween 2016 and
2017. I is impo an o no e ha e-Book sales da a canno be di ec ly compa ed o
he da a on p in book sales ha ha e been p esen ed in Fig.4. F om he compa ison
o G K’s p in book sales da a (see Appendix A.3) wi h he mo e accu a e scanne
da a, we know ha he consume panel p ojec ions migh be o e es ima ing " ue"
sales by up o 25 pe cen . Howe e , e en wi hou co ec ing o his po en ial bias, i
becomes clea ha e-Book sales did no comple ely compensa e he decline in sales
o p in books in all yea s. Fo example, p in book sales d opped by 6.1 Million
copies be ween 2013 and 2014, while e-Book sales only inc eased by 3.31 Million.
Fig. 4 Annual e enues (in million Eu os) and sales (in million books sold) o p in books in b ick-and-
mo a s o es and e-Comme ce. Cumula i e losses since 2011 in pa en heses. Thu ingia and Saa land a e
excluded due o inconsis encies in he da a. Sou ce: media con ol GmbH
9 Fo e-Book sales see h ps:// de. s a i s a. com/ s a i s ik/ da en/ s udie/ 232191/ um a ge/ absa z- on-e- books-
in- deu s chland; las accessed Decembe 20, 2023. Fo o al sales (including e-Books and audio books)
see h ps:// de. s a i s a. com/ s a i s ik/ da en/ s udie/ 416380/ um a ge/ absa z- on- buech e n- in- deu s chland;
las accessed Decembe 20, 2023.
825
Jou nal o Cul u al Economics (2025) 49:811–854
4.2 Reg ession analyses
This sec ion p esen s he esul s o he analysis o he pe iod 2011–2017 and
ocuses on sales o physical books. Be o e we p esen he eg ession esul s, we i s
explain ou 2SLS es ima ion app oach. The s uc u al equa ion o ou basic model
akes he ollowing o m:
whe e he dependen a iable a e he sales o p in books pe capi a in ede al s a e i
and mon h . Ou ea men a iable

#
s o es
i,
e e s o he i ed alues om he i s -
s age o he absolu e numbe o books o es in ede al s a e i in mon h .
The yea -mon h ixed e ec s a e gi en by
𝜉
. As men ioned abo e, book sales
a e dec easing o e ime. This end, which could, o ins ance, be associa ed wi h
mo e consume s subs i u ing eading wi h mo ie s eaming o online gaming, is
cap u ed by hose dummies. Mo eo e , he yea -mon h ixed e ec s also con ol o
he seasonali y o book sales, e.g., he la ge Ch is mas e ec . The a iable
g endsi,
cap u es he Google T ends Index o he opic “book”,
e eade i,
gi es he Google
T ends Index o he sea ch i em “E-book eade ” and
𝜉i
depic s ime-in a ian ixed
e ec s o each ede al s a e i.19
As explained in Sec .4.1, we use
popi,
and pop
2
i,
as ins umen s o he po en-
ially endogenous a iable
#s o esi,
.
#s o esi,
is co ela ed wi h
popi,
and pop
2
i,
,
which is a p e equisi e o he ele ance condi ion. No e ha , o he ele ance con-
di ion o be sa is ied, pop
2
i,
needs o be implemen ed, as o he wise he co ela ion
is no su icien ly s ong. As explained in Sec .4.1, he o hogonali y assump ion
should be sa is ied as well because i is easonable o assume ha popula ion does
no a ec how many books a pe son eads. Thus, he p ojec ion in he i s -s age
eg ession o ou base model can be o malized as ollows:
The co esponding i s -s age eg ession esul s a e p esen ed in Table11 o Appen-
dix B.2.
4.2.1 Baseline esul s
Table2 p esen s he eg ession esul s o Eq. (1).20 We un he eg ession o o al
sales, i. e. he sum o sales in physical books o es and e-Comme ce (column (1)),
(1)
sales
_pc
i,
=𝛽
1
#s o es
i,
+𝛽
2
g ends
i,
+𝛽
3
e eade
i,
+𝜉
i
+𝜉
+u
i, ,
(2)
#
s o es
i,
=
𝛿1
pop
i,
+
𝛿2
pop
2
i,
+
𝛿3
g ends
i,
+
𝛿4
e eade
i,
+
𝜉i
+
𝜉
+
𝜀i, .
19 No e ha we only in es iga e 14 ou o 16 ede al s a es in Ge many. We excluded wo ede al s a es
(Saa land and Thu ingia) om ou analyses due o po en ial e o s in he da a.
20 We p esen he esul s o a nai e OLS es ima ion o Eq. (1) in Table10 o Appendix B.1. In his
eg ession, he numbe o books o es has a posi i e e ec on he p in book sales, which is signi ican on
he 5%-le el. Howe e , as discussed in he p e ious sec ions (esp. Sec .4.1), his nai e OLS eg ession
is p one o endogenei y. Neglec ing his endogenei y issue could lead o a biased es ima ion, which is he
eason why we apply an IV es ima ion app oach.

826
Jou nal o Cul u al Economics (2025) 49:811–854
and o he wo di e en sales channels physical books o es and e-Comme ce sepa-
a ely (columns (2) and (3)).
The coe icien o
#s o es
measu es how a change in he o al numbe o book-
shops a ec s book sales pe capi a in an a e age mon h o an a e age ede al s a e.
Thus, based on column (1) in Table2, we can conclude ha i one physical book-
s o e closes, ce e is pa ibus, sales pe capi a dec ease by 0.000248 on a e age (a e -
age mon h, a e age ede al s a e). The e ec can be disen angled in o he o line
and online channel. The e ec on sales in physical books o es is 0.000183 and in
e-Comme ce 0.0000559.
Be o e u ning o he in e p e a ion o he indings, some echnical aspec s ha e o
be discussed. The po en ially endogenous a iable in Equa ion (1),
#s o es
, is iden i-
ied by he ins umen as indica ed by he Kleibe gen–Paap k Wald F-s a is ic which
clea ly exceeds he IV c i ical alue om S ock and Yogo (2005). Fu he mo e, he
Null o he Ande son–Rubin Wald F-s a is ic can be ejec ed o he eg ession wi h
agg ega ed (column (1) in Table2) and o line sales (column (2)), indica ing ha
he po en ially endogenous eg esso is ele an in he s uc u al equa ion. Addi ion-
ally, a es o exogenei y o he a iable
#s o es
using he di e ence o wo Sa gan-
Hansen s a is ics (also GMM dis ance o C-s a is ic) can be ejec ed.21
A posi i e coe icien o
#s o es
(column (1) in Table2) means ha when a phys-
ical books o e exi s he ma ke , agg ega e sales dec ease. The same applies o sales
in he o line and he online channel in isola ion (columns (2) and (3) in Table2).
This implies ha , on a e age, ma ke exi o physical books o es is no ully com-
pensa ed o by sales in o he physical books o es. (Full compensa ion would occu
i he e ec s we e no s a is ically di e en om ze o. In ha case, depending on
which e ec is conside ed, he closu e o a books o e would no a ec o al sales,
sales in he o line o sales in he online channel.) Ou esul s hus indica e ha , on
a e age, ma ke exi o physical books o es leads o a signi ican dec ease in sales.
The e ec is di e en o he o line and he online channel. One can see ha
o line sales dec ease mo e s ongly upon ma ke exi o physical books o es han
online sales (columns (2) and (3) in Table2).
The posi i e coe icien o e-Comme ce migh seem con using a i s glance. I
indica es complemen a i y be ween he channels, which is consis en wi h he well-
known p oblem o ee- iding on se ice p o ision as in, o example, Telse (1960):
The ea u es o “se ices” o one sales channel a ec sales in he o he channel.
E en hough he exac mechanism canno be clea ly iden i ied based on ou da a,
he e a e po en ial explana ions o his pa e n. One po en ial explana ion could be
“show ooming” (Zhang e al., 2018). Al hough RPM elimina es p ice di e ences
be ween he o line and online channel, non-p ice ac o s like p oduc i and wai
imes, as highligh ed by Gensle e al. (2017), can s ill d i e such an e ec . Ano he
po en ial explana ion would be wo d-o -mou h (Beck, 2007), whe e ini ial expo-
su e o a i le in he o line channel, which is cha ac e ized by a much highe sales
olume in Ge many (see Sec .A.3 in he Appendix), can lead o inc eased online
21 This endogenei y es s a is ic is nume ically equi alen o a Du bin-Wu Hausman es unde condi-
ional homoskedas ici y (see Hayashi (2000)).
827
Jou nal o Cul u al Economics (2025) 49:811–854
pu chases. In Sec . 4.2.2, he e ec will be disen angled o di e en gen es and
ypes o books o es.
To ge a be e unde s anding o he magni ude o ou indings, i is illumina ing
o exp ess he dec ease in sales in absolu e e ms. Popula ion, which we app oxi-
ma e using labo o ce (see abo e), is, on a e age, 3 million pe ede al s a e (42
million o al). Thus, when one books o e closes, a e age book sales dec ease by
0.000248 ⋅3 million =744
uni s pe mon h in an a e age ede al s a e.
Based on hese igu es, i is possible o answe he ollowing ques ion: How
did he closu es o books o es a ec p in book sales in he pe iod 2011–2017? To
answe his ques ion, i s no e ha mon hly sales wen down by 2,765,000 om
24,767,500 in 2011 o 22,002,500 in 2017 (see Fig.4). In he 14 ede al s a es exam-
ined in his sec ion, 1,382 books o es exi ed he ma ke .22 Wi h an a e age dec ease
in sales upon he closu e o a books o e o 744 uni s pe mon h on a e age, s o e
closu es ha e accoun ed o a d op o 1,028,208 in mon hly sales. This sugges s ha
books o e closu es accoun ed o abou 37% o he sales dec ease.23
When in e p e ing hese indings, one has o keep in mind ha hey occu in he
Ge man ma ke , which is cha ac e ized by a ela i ely high densi y o books o es.
Acco ding o ou da a, in Ge many he e is app oxima ely one physical books o e
pe 16,400 inhabi an s in 2018. Fo ins ance, in he USA he numbe o physical
books o es is much lowe . The numbe o s o es ha sold books peaked a 12,000 in
1992 (see Wu e al. 2018), so ha he e was app oxima ely one physical books o e
pe 21,400 inhabi an s. I emains an open ques ion whe he he e ec o a closu e
on sales becomes mo e p onounced he lowe he numbe o physical s o es because
wi h each closu e i becomes inc easingly mo e di icul o consume s o ind an
Table 2 Main es ima ions wi h sales olumes on p in books pe capi a as dependen a iable
S anda d e o s in pa en heses
Signi icance le els:*P < 0.05, **P < 0.01, ***P < 0.001
(1) (2) (3)
Agg ega ed O line e-Comme ce
# S o es 0.000248*** 0.000183*** 0.0000559*
(0.0000611) (0.0000486) (0.0000260)
Google T ends 0.00376*** 0.00317*** 0.000739*
(0.000811) (0.000637) (0.000354)
Google T ends e-Reade 0.000470 0.000169 0.000225
(0.00113) (0.000815) (0.000622)
Fede al S a e FE Yes Yes Yes
Yea -Mon h FE Yes Yes Yes
Ande son–Rubin Wald F-s a is ic 8.009 7.281 2.095
Kleibe gen–Paap k Wald F-s a is ic 2238.1 2238.1 2238.1
# o obse a ions 1176 1176 1176
22 Including Thu ungia and Saa land, 1,435 we e closed be ween 2011 and 2017, see Sec .3.
23 We hank one o he anonymous e iewe s o sugges ing his in e p e a ion.
828
Jou nal o Cul u al Economics (2025) 49:811–854
app op ia e subs i u e o line (e.g., inc easing physical a eling dis ances). I is le
open o u u e esea ch o in es iga e whe he he e ec o he closu e o physi-
cal books o es is di e en in ma ke s wi h ewe o mo e physical books o es han
Ge many.
No e ha he coe icien o Google T ends o he e m “book” is s a is ically
signi ican . This means ha , ce e is pa ibus, wi h inc easing sea ch olume, sales
o books end o inc ease. In con as , he coe icien o Google T ends e-Reade
is insigni ican in Table2 implying ha he sea ch beha io o he consume s o
e-Reade s does no signi ican ly a ec he physical book sales pe capi a. In o he
wo ds, i does no appea o be he case ha consume s subs i u e e-Books o p in
books in a sys ema ic ashion, as would be indica ed by a nega i e coe icien o
ha index. This can also be seen as an indica ion ha consume s do no subs i u e
p in books by e-Books when books o es close, o a leas no one- o -one. E-Book
sales will be discussed in Sec .4.3 in mo e de ail.
Las ly, as a i s obus ness check24, we compa e he esul s o ou IV es ima ion
p esen ed abo e wi h ou al e na i e speci ica ions: an IV es ima ion using yea ly
da a (a), a nai e OLS es ima ion (b) and wo OLS speci ica ions wi h lags (c) and
(d). The IV es ima ion using yea ly da a (a) has he bene i ha seasonal pa e ns in
sales o he numbe o books o es (e.g., he Ch is mas e ec o membe ships o he
Ge man Bookselle s Associa ion unning ou ) play no ole in he es ima ions. How-
e e , he numbe o obse a ions d ops om 1,176 o 98, which s ongly educes he
powe o he analysis. The nai e OLS es ima ion (b) igno es po en ial endogene-
i y ha a ises when book sales a e eg essed on he numbe o books o es (see he
beginning o his chap e ). The lagged OLS speci ica ions (c) and (d) can be seen
as mo e casual a emp s o con ol o his endogenei y han an IV app oach, since
he book sales in pe iod should no di ec ly a ec he numbe o books o es in
−1
. Speci ica ions (c) and (d) we e un in le els and in log le els, espec i ely. The
esul s o he espec i e es ima o s o
#s o es
a e p esen ed in Table3.
The second o i h ows in Table3 depic he es ima ion esul s o he a o e-
men ioned, al e na i e speci ica ions, whe eas he i s ow depic s he esul s
de i ed om he main IV. In compa ison o he la e , he IV wi h yea ly a ia ion
(a) shows ha he e ec s o a closu e on agg ega ed sales o sales in e-Comme ce
a e no signi ican ly di e en om ze o a he 5 %-le el. This is no su p ising
gi en he dec ease in he numbe obse a ions. In e es ingly, he coe icien o
#s o es
is s ill signi ican ly di e en om ze o a he 10 %-le el. Mo eo e , he
e ec s o a closu e on o line sales emains signi ican a he 0.1 %-le el. Wi h
a nai e OLS (b), he coe icien o
#s o es
is no s a is ically signi ican a he 5
%-le el. The coe icien is signi ican a he 10 %-le el o all channels (Agg e-
ga ed, O line and Online). Howe e , gi en ha he nai e OLS does no accoun
o po en ial endogenei y, he e ec may be biased, anyway. The lagged OLS es i-
ma es (c) and (d) a e s a is ically signi ican a he 5 %-le el o he o line chan-
nel. Speci ica ion (d) also shows a signi ican e ec a he 5 %-le el on agg e-
ga e. The magni ude o some o he e ec s in he OLS speci ica ions is gene ally
highe han in he IV es ima ion. The coe icien o agg ega ed sales (column (1))
24 Fu he obus ness checks a e p esen ed in Sec .5.
829
Jou nal o Cul u al Economics (2025) 49:811–854
implies an a e age dec ease in sales pe books o e ha closes o a ound 1,980
uni s (a g. annual sales pe ede al s a e (19,8 million books, see Table1) imes
he coe icien (0.0001)). The e ec s in e-Comme ce a e a om signi ican and
should no be in e p e ed. The es ima ion equa ions o (c) and (d) a e p esen ed
in Appendix C.4.
O e all, i emains o conclude ha , a ha poin , he esul s o he main IV
es ima ion appea o be obus . As expec ed, he signi icance le els s ongly
dec ease when using yea ly da a, howe e , hey a e in line wi h wha we obse e
when using mon hly da a. Also he e ec s o nai e o lagged OLS es ima ions
poin in a simila di ec ion as he IV esul s. I should be no ed, hough, ha
he OLS es ima ions should be in e p e ed ca e ully as hey a e expec ed o be
a ec ed by endogenei y.
4.2.2 Gen e‑speci ic e ec s
In his sec ion, we will check whe he he e a e gen e-speci ic di e ences in he
e ec s o exi o physical books o es on sales. In doing so, we un he IV eg es-
sions explained abo e o he nine di e en book gen es (see in Sec .3) sepa a ely.
Table4 ep esen s he esul s o ou ea men a iable
#s o es
in he espec i e
second s ages o he 2SLS eg essions wi h agg ega ed sales, o line sales and sales
in e-Comme ce as dependen a iables.
The esul s p esen ed in he i s column o Table4 a e la gely in line wi h he
esul s p esen ed abo e. Fo ins ance, i one books o e closes, 0.0000692 p in
books o he ic ion gen e a e sold less in bo h e ail channels on a e age (mon h
and ede al s a e). This means ha
0.0000692 ⋅3 million =207.6
ic ion books a e
Table 3 Compa ison o IV and OLS es ima ions
S anda d e o s in pa en heses
Signi icance le els:*P < 0.05, **P < 0.01, ***P < 0.001
(1) (2) (3)
Agg ega ed O line e-Comme ce
IV 0.000248*** 0.000183*** 0.0000559*
(0.0000611) (0.0000486) (0.0000260)
Yea ly 0.00298 0.00589*** 0.000565
(0.00167) (0.0000521) (0.000971)
OLS 0.000515 0.000470 0.000106
(0.000249) (0.000224) (0.0000493)
Lagged OLS 0.000557 0.000562* −0.00000515
(0.000272) (0.000222) (0.0000983)
Lagged OLS (log le el) 0.0001* 0.00128** −0.00084
(0.0004682) (0.0005409) (0.0017939)
Fede al S a e FE Yes Yes Yes
Time FE Yes Yes Yes
830
Jou nal o Cul u al Economics (2025) 49:811–854
sold less pe mon h, on a e age. Simila in e p e a ions can be applied o he gen es
non ic ion, humani ies, child en books, na u al sciences, guidebooks and a el. In
pa icula , no e ha he e ec o child en books is e en sligh ly s onge han ha
o ic ion (
0.0000701 ⋅3 million =210.3
).
Howe e , he change in he numbe o physical books o es has no signi ican
impac on o line sales o school books and books belonging o he ca ego y social
sciences. In o he wo ds, i physical books o es close, o al sales as well as sales
in he o line and online ca ego y emain una ec ed. This inding is in line wi h
expec a ions as he demand o school books can be expec ed o be highly inelas ic
and independen o he numbe o books o es. In ha gen e, se ice aspec s such as
expe opinion, ad hoc sales, e c., should be almos i ele an .
Table 4 IV eg essions by
gen es
S anda d e o s in pa en heses
Signi icance le els:*P < 0.05, **P < 0.01, ***P < 0.001
Agg ega ed Sales O line Sales e-Comme ce
Fic ion
# S o es 0.0000692*** 0.0000557*** 0.0000151*
(0.0000191) (0.0000166) (0.00000678)
Non ic ion
# S o es 0.0000150** 0.0000112** 0.00000379
(0.00000459) (0.00000359) (0.00000215)
Humani ies
# S o es 0.0000106*** 0.00000695*** 0.00000209
(0.00000200) (0.00000138) (0.00000125)
Child en books
# S o es 0.0000701*** 0.0000573*** 0.0000113*
(0.0000109) (0.00000914) (0.00000455)
Na u al sciences
# S o es 0.00000655*** 0.00000358*** 0.00000319*
(0.00000143) (0.000000474) (0.00000131)
Guidebooks
# S o es 0.0000400*** 0.0000323*** 0.00000975
(0.00000766) (0.00000413) (0.00000588)
T a el
# S o es 0.0000128*** 0.0000106*** 0.00000233*
(0.00000251) (0.00000211) (0.00000118)
School books
# S o es 0.00000456 0.00000533 0.00000281
(0.0000369) (0.0000332) (0.00000520)
Social sciences
# S o es 0.00000276 0.000000745 0.00000275
(0.00000184) (0.000000853) (0.00000148)

831
Jou nal o Cul u al Economics (2025) 49:811–854
The coe icien s in he hi d column in Table4 (Sales e-Comme ce) equi e some
discussion. As al eady explained, hese coe icien s cap u e a complemen a i y
be ween he online and he o line channel in a sense ha online sales a e posi i ely
a ec ed by he numbe o books o es. This is he case o ic ion, child en books,
na u al sciences and a el. Howe e , no e ha his e ec is only signi ican on he
5%-le el. Mo eo e , o ic ion (45.3 uni s) and child en books (33.9 uni s) he e ec
is a s onge han o na u al sciences (9.57) and a el (6.99).
4.3 Es ima ion wi hE‑books
I is easonable o assume ha a leas some consume s subs i u e e-Books o
p in books when physical books o es close, especially since he e is e idence ha
e-Books expanded book sales (see Gilbe , 2015,p. 167–170, o an o e iew). I
is hus app op ia e o check whe he he e ec s o exi o physical s o es on book
sales p esen ed abo e a e o e es ima ed due o missing in o ma ion on e-Book sales.
Thus, we will now deploy he consume panel da a se ha includes in o ma ion on
e-Book sales (see Appendix A.3 o mo e de ails). The da a a e used o epea he
analyses p esen ed in Sec .4.2. Es ima ion esul s a e epo ed in Table5 as ollows.
We un he eg ession o o al book sales (column (1)) and o he wo book
o ma s p in and e-Books sepa a ely (columns (2) and (3)). As shown in Table5,
he e ec o he numbe o physical books o es on book sales emains s a is ically
signi ican in all h ee eg ession app oaches. The epo ed coe icien in column (1)
implies a dec ease in o al book sales pe capi a, including e-Books, o 0.00257 in a
gi en qua e pe ede al s a e (signi ican on he 1%-le el), when a physical book-
s o e closes. (Fo a mo e de ailed discussion o he o de o magni ude o he e ec ,
see below.) The e ec in he numbe o books o es on book sales is also signi ican ly
posi i e when we un he eg ession o p in and e-Books sepa a ely (columns (2)
and (3)). The eby, he numbe o physical books o es has a la ge e ec on p in
book sales pe capi a (0.00239) han on e-Book sales pe capi a (0.000189). Again,
Table 5 Es ima ion wi h o al book sales (p in books and e-Books) pe capi a as dependen a iable
using consume su ey da a om G K
S anda d e o s in pa en heses
Signi icance le els:*p < 0.05, **p < 0.01, ***p < 0.001
(1) (2) (3)
All books P in e-Books
# S o es 0.00257*** 0.00239*** 0.000189***
(0.0000841) (0.0000849) (0.0000228)
Fede al S a e FE Yes Yes Yes
Yea -Qua e FE Yes Yes Yes
Ande son–Rubin Wald F-s a is ic 479.6 443.6 29.27
Kleibe gen–Paap k Wald F-s a is ic 8475.0 8475.0 8475.0
# o obse a ions 320 320 320
832
Jou nal o Cul u al Economics (2025) 49:811–854
he signi ican esul o e-Books indica es complemen a i y be ween he online and
he o line channel.
These indings also show ha he main esul s epo ed in Sec .4.2 a e quali a-
i ely obus when in o ma ion on e-Book sales a e included in o he analyses, as
e-Book sales appa en ly do no compensa e o he dec ease in sales igge ed by he
exi o physical books o es, as his would equi e a nega i e coe icien o
#s o es
.
Thus, a signi ican ly posi i e (o insigni ican ) coe icien does no challenge he p e-
ious indings.
No e ha he e ec s appea o be s onge when we use consume panel da a.
The coe icien o “All books” epo ed in column 1 implies a dec ease in a e -
age qua e ly sales o
0.00257
⋅
3 million
=
7, 710
, i.e., 2,570 uni s pe mon h. This
e ec is a ound 3.5- imes s onge han he one epo ed in he Sec .4.2.1. A possi-
ble explana ion o his di e ence is ha in o ma ion in he consume panel da a se
a e ex apola ed by G K based on su ey da a, whe eas he scanne da a a e based
on ac ual sales (see Sec .3). Thus, he wo da a se s di e in e ms o absolu e sales
igu es. As explained in Sec .3, we conside he scanne da a o be a mo e accu a e
desc ip ion o ac ual book sales han su ey da a.
5 Robus ness checks
Robus ness checks we e pe o med o he analyses p esen ed in he p e ious
sec ions. All es ima ion esul s a e p esen ed in Appendix C. Ou esul s emain
obus wi h espec o he ollowing modi ica ions.
• To ule ou ha book p ices a ec some o he esul s, he change in he
mon hly a e age book p ices is in eg a ed as a “bad” con ol a iable in o he
second s age o he IV eg ession. A e age book p ices a e compu ed based on
scanne da a o he pe iod 2011–2017. Including p ices as a con ol can ule
ou ha dec easing book sales a e d i en by inc easing book p ices and no by
he dec easing numbe o physical books o es. The con ol a iable is “bad”
because he ela ionship be ween sales and p ices is bidi ec ional (see Appen-
dix C.1).
• The numbe o s uden s is included as a co a ia e. This is done o con ol o he
educa ion le els in he 14 Ge man ede al s a es o ou da a se . The ixed e ec s
included in o he eg essions cap u e na ionwide ime ends and ime-in a ian
e ec s ha a ec indi idual ede al s a es. Howe e , sociodemog aphic e ec s
such as educa ion can be ime- a ian and speci ic o a ede al s a e. Those
e ec s po en ially con ound ou analyses (see Appendix C.2).
• The analyses p esen ed in he Sec .4.2 can be pe o med using a combina ion o
su ey and scanne da a o include e-Book sales. The e o e, we use he a io o
e-Book sales om he su ey da a o calcula e absolu e e-Book sales based on
he scanne da a. This e-Book sales da a hen can be used o epea ou es ima-
ions o he pe iod 2014–2017 (qua e ly le el) wi h da a on p in and digi al
book sales (see Appendix C.3).
833
Jou nal o Cul u al Economics (2025) 49:811–854
• We also eg ess he book sales pe capi a on he lagged numbe o physical book-
s o es in an OLS es ima ion app oach. Lagging ou ea men a iable
#s o es
does a leas pa ially sol e he e e se causali y issue be ween he numbe o
books o es and he book sales pe capi a (see Appendix C.4).
• To accoun o po en ial di e ences in he de elopmen o book sales ac oss Ge -
man ede al s a es, we ex end ou IV es ima ion by inco po a ing s a e-speci ic
ime ends. The esul s con i m ou main indings while yielding sligh ly la ge
poin es ima es (see Appendix C.5).
6 Conclusion
We ind ha , o e all, e-Comme ce does no pose a pe ec subs i u e o physical
e aile s om he iewpoin o he consume s. Ou esul s p edic ha when a physi-
cal books o e closes, on a e age, mon hly book sales dec ease by 744 uni s. Taking
in o ha be ween 2011–2017, 1,382 books o es we e closed ac oss he 14 ede al
s a es comp ising ou da a se , ou esul s indica e ha a ound 37% o he d op in
mon hly p in book sales can be aced by o he closu es o books o es.
Appa en ly, a la ge enough numbe o consume s p e e s o buy books a physi-
cal books o es o closu es o hose s o es o ha e a s a is ically signi ican , nega i e
impac on he o al sales o books. Consume s’ p e e ences owa d o line “se ices”
migh po en ially be a ec ed by expe opinion, a mo e ca e ul selec ion and p esen-
a ion o i les, ad hoc pu chases o simply he a mosphe e o physical books o es.
The magni ude o he e ec di e s be ween gen es. The e ec is pa icula ly
s ong o ic ion and child en books i les, whe e we no only ind a signi ican
d op in o line sales bu also in online sales ollowing he ma ke exi o physical
books o es. This inding indica es complemen a i y be ween he channels o some
gen es. On he o he hand, we ind no such ela ionship be ween he wo channels
when i comes o school books. This obse a ion appea s in ui i e as he demand o
schoolbooks should be de e mined by easons o he han ea u es o se ices o e ed
by online and o line books o es.
I is impo an o no e ha he magni ude o he e ec s o ma ke exi o physi-
cal books o es was de e mined o he Ge man ma ke , which has a ela i ely high
numbe o physical books o es. I emains an open ques ion whe he he e ec o a
closu e on sales becomes mo e p onounced he lowe he numbe o physical s o es
because wi h each closu e i becomes inc easingly mo e di icul o consume s o
ind an app op ia e subs i u e o line (e.g., inc easing physical a eling dis ances).
I is le open o u u e esea ch o in es iga e whe he he e ec o he closu e o
physical books o es is di e en in ma ke s wi h ewe o mo e physical books o es
han Ge many.
Ou inding ha consume s pe cei e online and o line e aile s as impe ec sub-
s i u es p o ides an addi ional ace o he policy e alua ion o ixed book p ices.
The goal o hese e ical es ain s is usually o p o ec books as cul u al o me i
goods. I is desc ibed in he li e a u e ha ixed book p ices can p omo e ma ke
en y (Bouckae (2000), Guo and Lai (2017), Elzinga and Mills (2008,p. 1848))
and p e en exi by, o example, secu ing ma gins, in pa icula in he p esence o
834
Jou nal o Cul u al Economics (2025) 49:811–854
online compe i ion (Ma el and McCa e y (1985, 376), Bouckae (2000), Guo
and Lai (2017), Elzinga and Mills (2008,p. 1848), Leg os and S ahl (2019)). The e
is e idence om he UK ha a e he ab oga ion o he Ne Book Ag eemen , UK’s
and I eland’s ixed book p ice sys em abandoned in he 1990s, he book ma ke con-
solida ed (Da ies e al., 2004; Fishwick e al., 1997; Dea nley and Fea he , 2002). In
combina ion wi h ou inding ha a la ge numbe o physical books o es p omo es
book sales, ixed book p ice sys ems may hus suppo he policy goal o secu ing a
b oad supply o books. E en hough sys ema ic analysis is wa an ed, his e iciency
e ec o ixed book p ices would ha e o be aken in o accoun when e alua ing he
wel a e e ec s o supp essing p ice compe i ion among e aile s.
Appendix: A: Da a
A.1: B ick‑and‑mo a books o es
The numbe s p esen ed in Fig.1 a e de i ed om he membe ship da abase o he
Ge man Publishe s and Bookselle s Associa ion. Roughly 85–90% o all book e ail-
e s in Ge many a e membe s o he Ge man Publishe s and Bookselle s Associa-
ion. The da abase con ains in o ma ion on he name, he add ess, he ounding yea
o a books o e, he da e o he beginning as well as he end o a membe ship in he
Associa ion. Since 1963 a unique iden i ie (“Ve keh snumme ”) is assigned o each
business en i y ha is membe o he Ge man Publishe s and Bookselle s Associa-
ion. This iden i ie is used o communica ion be ween business pa ne s, i.e., o
communica ion be ween books o es, publishe s, and wholesale s. Using his iden i-
ie we a e able o ma ch books o es o o he da abases, like he so-called Add ess
Book o he Ge man-Language Book T ade.25 This add ess book was es ablished in
1839, and is now managed by a subsidia y o he Ge man Publishe s and Booksell-
e s Associa ion.26 The da abase con ains o e 30,000 add esses o publishe s, book-
s o es, music s o es, and publishing ep esen a i es in he Ge man-speaking egion.
The add ess book is upda ed egula ly and changes (en y, exi , change o loca ion o
legal name) a e published online. Membe s o he Ge man Publishe s and Booksell-
e s Associa ion can be lis ed in he add ess book o ee, nonmembe s ha e o pay a
small annual ee.
We apply se e al s eps in o de o ans o m he me ged add ess book/membe -
ship da abase in o a mon hly panel ha con ains he numbe o ac i e b ick-and-
mo a s o es o each Fede al S a e. Fi s , we d op all publishe s, head o ices and
wa ehouses as well as online books o es om he da abase. Second, we ob ain in o -
ma ion on he Fede al S a e in which each b ick-and-mo a s o e is loca ed, using
he Google Maps API. Thi d, we g oup s o es by loca ion. In doing so, we a oid
si ua ions whe e a books o e would ecei e a new iden i ie due o, o example, a
25 Ad essbuch ü den deu schsp achigen Buchhandel, see h ps:// adb- online. de; las accessed Decembe
19, 2023.
26 MVB Ma ke ing- und Ve lagsse ice des Buchhandels GmbH, see h ps:// m b- online. de; las
accessed Decembe 20, 2023.
841
Jou nal o Cul u al Economics (2025) 49:811–854
Table 8 Summa y s a is ics by Fede al S a e
Fede al S a e N Mean SD Min Median Max
Baden-Wue embe g Quan i y all channels 84 3,054,151 817,311 2,197,958 2,826,255 6,033,089
Numbe B&M s o es 84 797.6 63.7 698 793 905
Popula ion (labo o ce) 84 5,832,617 175,385 5,538,385 5,825,910 6,179,156
Google T ends Book 84 67.39 6.267 57 66 88
Google T ends Amazon Kindle 84 29.06 13.22 6 26 82
Ba a ia Quan i y all channels 84 3,523,258 967,749 2,578,736 3,254,845 6,786,929
Numbe B&M s o es 84 855.6 65.84 769 833 967
Popula ion (labo o ce) 84 6,991,562 187,336 6,670,375 7,014,883 7,368,733
Google T ends Book 84 66.75 6.856 58 65 87
Google T ends Amazon Kindle 84 30.68 14.63 6 28.5 87
Be lin Quan i y all channels 84 1,095,268 300,601 779,227 1,012,833 2,094,380
Numbe B&M s o es 84 248.4 11.45 230 249 274
Popula ion (labo o ce) 84 1,807,867 53,918 1,710,468 1,823,527 1,900,722
Google T ends Book 84 69.99 9.021 57 68 100
Google T ends Amazon Kindle 84 27.36 11.35 5 25.5 70
B andenbu g Quan i y all channels 84 629,316 213,298 414,295 533,336 1,213,088
Numbe B&M s o es 84 91.93 6.259 80 94 101
Popula ion (labo o ce) 84 1,331,518 8977 1,311,513 1,332,724 1,345,946
Google T ends Book 84 62.45 7.922 47 61 83
Google T ends Amazon Kindle 84 24.83 12.71 0 23.5 75
B emen Quan i y all channels 84 279,569 97,812 180,552 246,384 671,877
Numbe B&M s o es 84 53.01 3.012 49 53 58
Popula ion (labo o ce) 84 337,925 9175 321,313 340,915 350,670
Google T ends Book 84 64.96 8.074 53 63 91
Google T ends Amazon Kindle 84 27.33 13.31 3 25 75

842
Jou nal o Cul u al Economics (2025) 49:811–854
Table 8 (con inued)
Fede al S a e N Mean SD Min Median Max
Hambu g Quan i y all channels 84 764,794 217,571 569,237 705,560 1,587,100
Numbe B&M s o es 84 122.4 11.04 105 121.5 139
Popula ion (labo o ce) 84 972,280 29,745 919,795 974,548 1,025,176
Google T ends Book 84 65.24 7.824 51 63 96
Google T ends Amazon Kindle 84 27.62 13.25 5 26 85
Hesse Quan i y all channels 84 1,800,033 493,730 1,276,866 1,674,170 3,667,458
Numbe B&M s o es 84 505.5 45.83 439 504 576
Popula ion (labo o ce) 84 3,219,236 78,415 3,073,820 3,216,483 3,368,375
Google T ends Book 84 65.48 6.647 55 64 85
Google T ends Amazon Kindle 84 29.05 13.39 6 26 80
Lowe Saxony Quan i y all channels 84 2,639,531 865,852 1,783,210 2,266,358 4,885,420
Numbe B&M s o es 84 504 43.91 439 502.5 579
Popula ion (labo o ce) 84 4,128,667 89,228 3,974,260 4,136,114 4,289,250
Google T ends Book 84 61.05 6.784 49 60 80
Google T ends Amazon Kindle 84 25.29 11.45 3 24 70
Mecklenbu g-Wes e n Quan i y all channels 84 409,381 111,777 281,548 370,741 697,173
Pome ania Numbe B&M s o es 84 74.02 6.499 66 71 86
Popula ion (labo o ce) 84 838,488 11,969 822,790 832,014 862,432
Google T ends Book 84 59.42 7.409 45 59 84
Google T ends Amazon Kindle 84 24.64 12.44 2 22 63
843
Jou nal o Cul u al Economics (2025) 49:811–854
Table 8 (con inued)
Fede al S a e N Mean SD Min Median Max
No h Rhine-Wes phalia Quan i y all channels 84 5,218,629 1,560,673 3,288,458 4,670,050 10,464,162
Numbe B&M s o es 84 1194 101.4 1038 1186 1362
Popula ion (labo o ce) 84 9,255,150 169,422 8,962,720 9,257,316 9,556,575
Google T ends Book 84 62.6 6.559 53 61 82
Google T ends Amazon Kindle 84 27.06 12.93 5 25 73
Rhineland-Pala ina e Quan i y all channels 84 1,178,232 358,714 770,241 1,045,224 2,169,307
Numbe B&M s o es 84 329.8 30.79 284 333.5 383
Popula ion (labo o ce) 84 2,149,753 38,968 2,080,827 2,155,255 2,213,204
Google T ends Book 84 65.88 6.185 55 64 84
Google T ends Amazon Kindle 84 30.79 14.74 7 27.5 84
Saxony Quan i y all channels 84 1,194,030 367,516 852,609 1,097,054 2,535,721
Numbe B&M s o es 84 271.2 21.27 238 269.5 307
Popula ion (labo o ce) 84 2,120,833 9956 2,091,661 2,120,822 2,141,032
Google T ends Book 84 61.9 7.279 51 60 82
Google T ends Amazon Kindle 84 22.48 10.31 5 21 63
Saxony-Anhal Quan i y all channels 84 573,037 201,924 284,508 510,070 1,149,577
Numbe B&M s o es 84 115.5 11.75 97 116 136
Popula ion (labo o ce) 84 1,173,376 17,382 1,142,313 1,176,068 1,205,128
Google T ends Book 84 58.56 6.354 47 58 75
Google T ends Amazon Kindle 84 24.57 13.5 2 21.5 71
844
Jou nal o Cul u al Economics (2025) 49:811–854
Table 8 (con inued)
Fede al S a e N Mean SD Min Median Max
Schleswig-Hols ein Quan i y all channels 84 780,612 237,037 546,278 720,590 1,579,033
Numbe B&M s o es 84 197.3 12.2 176 200 219
Popula ion (labo o ce) 84 1,484,578 35,646 1,422,040 1,489,927 1,544,259
Google T ends Book 84 64.39 7.709 53 63 85
Google T ends Amazon Kindle 84 27.45 13.27 3 26 76
845
Jou nal o Cul u al Economics (2025) 49:811–854
Table 9 Summa y s a is ics by-p oduc g oup
P oduc g oup N Mean SD Min Median Max
Child en Books Quan i y all channels 1176 329,661 303,005 28,008 211,995 2,139,147
Numbe B&M s o es 1176 382.8 338.5 49 251 1362
Popula ion (labo
o ce)
1176 2,974,561 2,560,595 321,313 1,990,775 9,556,575
Google T ends Book 1176 64 7.834 45 63 100
Google T ends Ama-
zon Kindle
1176 27.01 13.08 0 25 87
Fic ion Quan i y all channels 1176 596,405 584,793 77,323 355,013 4,753,941
Numbe B&M s o es 1176 382.8 338.5 49 251 1362
Popula ion (labo
o ce)
1176 2,974,561 2,560,595 321,313 1,990,775 9,556,575
Google T ends Book 1176 64 7.834 45 63 100
Google T ends Ama-
zon Kindle
1176 27.01 13.08 0 25 87
Guidebooks Quan i y all channels 1176 224,833 210,871 19,005 135,797 1,561,292
Numbe B&M s o es 1176 382.8 338.5 49 251 1362
Popula ion (labo
o ce)
1176 2,974,561 2,560,595 321,313 1,990,775 9,556,575
Google T ends Book 1176 64 7.834 45 63 100
Google T ends Ama-
zon Kindle
1176 27.01 13.08 0 25 87
Humani ies Quan i y all channels 1176 53,430 50,631 2944 34,048 362,306
Numbe B&M s o es 1176 382.8 338.5 49 251 1362
Popula ion (labo
o ce)
1176 2,974,561 2,560,595 321,313 1,990,775 9,556,575
Google T ends Book 1176 64 7.834 45 63 100
Google T ends Ama-
zon Kindle
1176 27.01 13.08 0 25 87
Na u al Sciences Quan i y all channels 1176 30,242 27,125 2045 18,891 171,151
Numbe B&M s o es 1176 382.8 338.5 49 251 1362
Popula ion (labo
o ce)
1176 2,974,561 2,560,595 321,313 1,990,775 9,556,575
Google T ends Book 1176 64 7.834 45 63 100
Google T ends Ama-
zon Kindle
1176 27.01 13.08 0 25 87
Non ic ion Quan i y all channels 1176 133,848 130,071 13,981 85,486 1,036,771
Numbe B&M s o es 1176 382.8 338.5 49 251 1362
Popula ion (labo
o ce)
1176 2,974,561 2,560,595 321,313 1,990,775 9,556,575
Google T ends Book 1176 64 7.834 45 63 100
Google T ends Ama-
zon Kindle
1176 27.01 13.08 0 25 87
846
Jou nal o Cul u al Economics (2025) 49:811–854
B: IV App oach
B.1: Nai e OLS es ima ion o Equa ion (1)
See Table10.
Table 9 (con inued)
P oduc g oup N Mean SD Min Median Max
School Books Quan i y all channels 1176 170,586 272,208 5384 76,379 2,135,074
Numbe B&M s o es 1176 382.8 338.5 49 251 1362
Popula ion (labo
o ce)
1176 2,974,561 2,560,595 321,313 1,990,775 9,556,575
Google T ends Book 1176 64 7.834 45 63 100
Google T ends Ama-
zon Kindle
1176 27.01 13.08 0 25 87
Social Sciences Quan i y all channels 1176 30,410 28,521 1610 18,847 189,124
Numbe B&M s o es 1176 382.8 338.5 49 251 1362
Popula ion (labo
o ce)
1176 2,974,561 2,560,595 321,313 1,990,775 9,556,575
Google T ends Book 1176 64 7.834 45 63 100
Google T ends Ama-
zon Kindle
1176 27.01 13.08 0 25 87
T a el Quan i y all channels 1176 83,379 79,644 8028 53,826 434,636
Numbe B&M s o es 1176 382.8 338.5 49 251 1362
Popula ion (labo
o ce)
1176 2,974,561 2,560,595 321,313 1,990,775 9,556,575
Google T ends Book 1176 64 7.834 45 63 100
Google T ends Ama-
zon Kindle
1176 27.01 13.08 0 25 87
Table 10 Nai e OLS es ima ion
S anda d e o s in pa en heses
Signi icance le els:*p < 0.05, **p < 0.01, ***p < 0.001
(1)
Agg ega ed
# S o es 0.000576*
(0.000259)
Google T ends 0.00341*
(0.00144)
Google T ends e-Reade 0.000000498
(0.00117)
Fede al S a e FE Yes
Yea -Mon h FE Yes
# o obse a ions 1176

847
Jou nal o Cul u al Economics (2025) 49:811–854
B.2: Fi s ‑s age IV eg ession esul o Equa ion (2)
See Table11.
C: Robus ness checks
C.1: Book p ices as“bad” con ol
The change in he mon hly a e age book p ices is in eg a ed as a “bad” con ol
a iable in o he second s age o he IV eg ession. We calcula e he change o he
mon hly a e age book p ice o he pe iod 2011–2017 based on ou scanne da a
and use his pa ame e as a “bad” con ol a iable in ou IV es ima ion app oach
om Equa ion (1). O cou se, he p ice is a “bad” con ol a iable in his app oach
because we now eg ess a quan i y pa ame e on a p ice pa ame e in he second
s age o ou 2SLS es ima ion. The esul s o ou IV es ima ion when con olling o
p ice changes is p esen ed in Table12.
Table12 shows ha he newly included a iable
Book p ice
has no signi ican
e ec on he dependen a iable o al p in book sales. The coe icien o his a i-
able is expec ed o be biased due o e e se causali y because quan i ies a e basically
eg essed on p ices. Ne e heless, om Table12 one can see ha he numbe o
books o es is s ill signi ican and has a posi i e sign. This implies ha a dec easing
numbe o physical books o es also leads o lowe p in book sales pe capi a when
we con ol o book p ice changes, which is in line wi h he indings p esen ed in he
main ex .
Table 11 Fi s -s age eg ession
esul s o baseline IV es ima ion
S anda d e o s in pa en heses
Signi icance le els:*P < 0.05, **P < 0.01, ***P < 0.001
(1)
# S o es
Pop 0.000410***
(0.00000652)
Pop. squa ed −4.71e−11***
(9.16e−13)
Google T ends 0.592**
(0.184)
Google T ends e-Reade 0.691***
(0.159)
Fede al S a e FE Yes
Yea -Mon h FE Yes
# o obse a ions 1176
848
Jou nal o Cul u al Economics (2025) 49:811–854
C.2: Numbe o uni e si y s uden s ascon ol
S a e and ime speci ic sociodemog aphic e ec s migh con ound ou baseline anal-
ysis, in pa icula changes in he educa ion le el. To con ol o hose e ec s, we
use da a on he numbe o uni e si y s uden s by ede al s a es30 as a p oxy o edu-
ca ion. Gi en ha hese da a a e only a ailable annually, we agg ega e sales aken
om scanne da a o ha le el and un a IV eg ession on a yea ly basis. In pa icu-
la , we eg ess p in book sales pe capi a on ou ea men a iable, he numbe o
physical books o es, and he lagged numbe o s uden s (as well as o he co a ia es
men ioned in Equa ion (1)) in he second s age o ou 2SLS es ima ion. The esul s
a e p esen ed in Table13.
One can see ha he lagged numbe o uni e si y s uden s has no signi ican
e ec on p in book sales pe capi a in Ge many. No e ha he e is li le a ia ion
in he numbe o s uden s such ha he ede al s a e ixed e ec and he numbe o
s uden s a e closely ela ed. Howe e , he lagged numbe o physical books o es s ill
has a posi i e and signi ican e ec on p in book sales, which again is in line wi h
he indings p esen ed in he main ex . I appea s su p ising ha he esul emains
s a is ically signi ican e en hough he numbe o obse a ion is only one wel h o
ha in he main ex (yea ly da a ins ead o mon hly da a).
Table 12 IV es ima ion wi h
mon hly a e age book p ice
change as bad con ol a iable
S anda d e o s in pa en heses
Signi icance le els:*P < 0.05, **P < 0.01, ***P < 0.001
(1)
Agg ega ed
# S o es 0.000319***
(0.0000683)
Google T ends 0.00419***
(0.00107)
Google T ends e-Reade 0.000595
(0.00112)
Book P ice −0.00887
(0.00638)
Fede al S a e FE Yes
Yea -Mon h FE Yes
Ande son–Rubin Wald F-s a is ic 10.24
Kleibe gen–Paap Wald F-s a is ic 1194.4
# o obse a ions 1176
30 These da a a e p o ided by he Ge man Fede al S a is ical O ice, see h ps:// www. s a i s isc hebib lio h
ek. de/ mi / ecei e/ DESe ie_ mods_ 00000 113.
849
Jou nal o Cul u al Economics (2025) 49:811–854
C.3: Es ima ions using acombina ion o scanne andsu ey da a
In o de o inco po a e e-Book sales in o he scanne da a se , we ollow hese s eps:
1. Agg ega e he scanne da a based on ede al s a e qua e s o he pe iod
2014Q1
−
2017Q4
.
2. U ilize su ey da a o calcula e he a io o e-Book o p in book sales o each
ede al s a e qua e .
3. Apply he calcula ed a io o es ima e e-Book sales using he scanne da a.
Table 13 IV es ima ion on a
yea ly basis when con olling
o he numbe o uni e si y
s uden s
S anda d e o s in pa en heses
Signi icance le els:*P < 0.05, **P < 0.01, ***P < 0.001
(1)
Agg ega ed
# S o es 0.00642*
(0.00316)
Google T ends −0.0351
(0.0550)
Google T ends e-Reade 0.0828
(0.0470)
S uden s −1
−0.000000970
(0.00000912)
Fede al S a e FE Yes
Yea FE Yes
Ande son–Rubin Wald F-s a is ic 2.853
Kleibe gen–Paap Wald F-s a is ic 22.99
# o obse a ions 84
Table 14 IV eg ession esul s
including e-Book sales
S anda d e o s in pa en heses
Signi icance le els:*P < 0.05, **P < 0.01,***P < 0.001
(1)
Sales e-Com-
me ce (incl.
e-Books)
# S o es 0.000547***
(0.0000502)
Fede al S a e FE Yes
Yea -Qua e FE Yes
Ande son–Rubin Wald F-s a is ic 60.75
Kleibe gen–Paap Wald F-s a is ic 13927.8
# o obse a ions 224
850
Jou nal o Cul u al Economics (2025) 49:811–854
Subsequen ly, he ob ained e-Book sales da a can be used o eplica e ou es ima-
ions in Sec .4.2 wi h comp ehensi e da a on bo h p in and digi al book sales.
Es ima ion esul s using p in and calcula ed e-Book sales o he sales channel
e-Comme ce as dependen a iable a e p esen ed in Table14. One can see ha he
numbe o physical books o es s ill has a signi ican posi i e e ec on he book sales
pe capi a in he e-Comme ce when also including e-Book sales in o he eg ession
o he book scanne da a (c . column (3) in Table2 o Sec .4.2.1).
This inding also shows ha he indings epo ed in Sec .4.2 a e obus when
in o ma ion on e-Book sales a e included in o he analyses, as e-Book sales appa -
en ly do no compensa e o he dec ease in sales igge ed by he exi o physical
books o es, as his would equi e a nega i e coe icien o
#s o es
. Thus, a signi i-
can ly posi i e (o insigni ican ) coe icien does no challenge he p e ious indings.
C.4: OLS Es ima ion wi hlagged numbe o books o es
Addi ional a ian s o Equa ion (2) we e un based on OLS using he yea ly o
mon hly lagged numbe o books o es as explana o y a iables. In gene al, his es i-
ma ion s a egy should a leas pa ially esol e po en ial e e se causali y be ween
he numbe o books o es and book sales pe capi a since book sales in should no
a ec he numbe o books o es in
−1
. The es ima ion equa ion can be o mula ed
as ollows:
The esul s based on yea ly da a a e p esen ed in Table3. He e,
salesi,
is ei he
measu ed in le els (column 4) o in logs (column 5).
The esul s he lagged OLS es ima ion using mon hly da a a e depic ed in
Table15 and imply ha he numbe o books o es ha e a posi i e e ec on he sales
pe capi a in he o line sales channel (column (2)). This means ha a closing book-
s o e in
=−1
signi ican ly lowe s he book sales pe capi a in
=0
(signi ican on
salesi,
=
𝛼
+
𝛽books o esi,
−
1
+
𝜇i
+
𝜇
+
𝜖i, .
Table 15 OLS Es ima ion wi h
lagged numbe o books o es as
ea men a iable
S anda d e o s in pa en heses
Signi icance le els:*P < 0.05, **P < 0.01, ***P < 0.001
(1) (2) (3)
Agg ega ed O line e-Comme ce
#S o es −1
0.000557 0.000562* −0.00000515
(0.000272) (0.000222) (0.0000983)
Google T ends 0.00276 0.00238* 0.000378
(0.00168) (0.000920) (0.00111)
Google T ends e-Reade −0.000279 −0.000431 0.000152
(0.00141) (0.00105) (0.00109)
Fede al S a e FE Yes Yes Yes
Yea -Mon h FE Yes Yes Yes
# o obse a ions 1008 1008 1008