Hempel, Co inna; Roosen, Ju a
A icle — Published Ve sion
G owing impo ance o p ice: In es iga ing ood alues
be o e and du ing high in la ion in Ge many
Ag icul u al Economics
P o ided in Coope a ion wi h:
John Wiley & Sons
Sugges ed Ci a ion: Hempel, Co inna; Roosen, Ju a (2024) : G owing impo ance o p ice:
In es iga ing ood alues be o e and du ing high in la ion in Ge many, Ag icul u al Economics, ISSN
1574-0862, Wiley Pe iodicals, Inc., Hoboken, NJ, Vol. 55, Iss. 6, pp. 1026-1039,
h ps://doi.o g/10.1111/agec.12865
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/313711
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/
Recei ed: 22 Augus 2023 Re ised: 24 June 2024 Accep ed: 13 Augus 2024
DOI: 10.1111/agec.12865
ORIGINAL ARTICLE
G owing impo ance o p ice: In es iga ing ood alues
be o e and du ing high in la ion in Ge many
Co inna Hempel1Ju a Roosen2,3
1Facul y o Li e Science,
Albs ad -Sigma ingen Uni e si y o
Applied Sciences, Sigma ingen, Ge many
2TUM School o Managemen , Chai o
Ma ke ing and Consume Resea ch,
Technical Uni e si y Munich, Munich,
Ge many
3HEF Wo ld Ag icul u al Sys ems Cen e ,
Technical Uni e si y Munich, F eising,
Ge many
Co espondence
Co inna Hempel, Facul y o Li e Science,
Albs ad -Sigma ingen Uni e si y o
Applied Sciences, Sigma ingen, Ge many.
Email: [email p o ec ed]
Abs ac
Conside ing he consump ion-induced in ensi ica ion o global challenges and
he con inuously changing consume needs, i is impo an o unde s and he
d i e s o consume ood choices unde ex e nal p essu es. We applied bes –
wo s scaling o elici he ela i e impo ance o 11 ood alues and conduc ed
la en class clus e analyses based on indi idual sco es, allowing us o gain
insigh s in o dis inc i e consume segmen s. Da a we e collec ed h ough online
su eys o 1000 consume s in Ba a ia, sou he n Ge many, in No embe 2020
and No embe 2022. As expec ed, he ela i e impo ance o ood alue p ice
has s ongly inc eased du ing his pe iod. Simila ly, he p ice-sensi i e segmen
has become la ge in 2022 han in 2020, while he socie al impac -cen e ed seg-
men has become much smalle in 2022. These indings call o a ge -speci ic
measu es o coun e ac his end o inc easing p ice ocus ha challenges
sus ainable die a y ansi ions.
KEYWORDS
bes –wo s scaling, ood alues, human alues, la en class analysis, p ice in la ion
JEL CLASSIFICATION
Q13, Q18
1 INTRODUCTION
A shi owa d a mo e sus ainable die p esen s a majo
oppo uni y o educe g eenhouse gas emissions om
he ood sys em. Wi h he ood sys em as a whole being
esponsible o 21%–37% o he o al an h opogenic g een-
house gases (Rosenzweig e al., 2020), people’s ood in ake
and p e e ences a ec hei heal h as well as he en i on-
men and clima e (Cla k e al., 2019). Recen ly, he ad e se
en i onmen al e ec s o consump ion beha io ha e
been widely epo ed and discussed in he public media.
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). Ag icul u al Economics published by Wiley Pe iodicals LLC on behal o In e na ional Associa ion o Ag icul u al Economis s.
In e na ional clima e mo emen s, such as he you h-led
mo emen “F idays o Fu u e,” ha e ecei ed conside -
able a en ion and aised awa eness o he u gency and
se e i y o clima e change- ela ed global challenges ( on
Zabe n & Tulloch, 2020). The educ ion in mea con-
sump ion in high-income coun ies appea s o be c ucial
o eaching in e na ional clima e a ge s. Ne e heless,
how o change consume beha io e ec i ely emains
an open ques ion (Pa lasca & Qaim, 2022). While he
educ ion o mea consump ion migh be an inexpensi e
op ion o engage in en i onmen ally iendly and ai ood
1026 wileyonlinelib a y.com/jou nal/agec Ag icul u al Economics. 2024;55:1026–1039.
HEMPEL and ROOSEN 1027
consump ion, o he sus ainable ood choices a e pe cei ed
as being cos lie (e.g., o ganic p oduc ion, ai ade, and
animal wel a e). Hence, signi ican inc eases in ood
p ices may es ain consume s om econside ing hei
ood choices o mo e sus ainable al e na i es (Hempel,
2024). Following he COVID-19 pandemic and Russia’s
a ack on Uk aine in Feb ua y 2022, ood p ice in la ion in
Ge many accele a ed, peaking a a mon hly in la ion a e
o 21.2% in Ma ch 2023.
Owing o he opicali y o e en s, no s udy has in es-
iga ed wha d i es consume s’ ood choices du ing his
pe iod o in la ion. Hence, his s udy aims o highligh how
accele a ing ood p ices cause changes in consume ood
alue hie a chies in Ge many. The e o e, he bes –wo s
scaling (BWS) o ood alues adds o he exis ing esea ch
in wo ways. Fi s , i examines hei ela i e impo ance
be o e (2020) and du ing a pe iod o high p ice in la ion
(2022). Second, i s udies he ood alues in a new geo-
g aphical con ex , as ood alues so a ha e mainly been
s udied among US Ame ican consume s. Mo eo e , we use
he ood alues o segmen a ion analyses and e alua ed
he di e ences be ween he segmen s o he 2020 and 2022
samples. To he bes o ou knowledge, no s udy has con-
duc ed a la en class clus e analysis o iden i y consume
segmen s, as p esen ed by Lusk and B iggeman (2009).
Ou s udy also con ibu es o he heo y by applying he
Schwa z Human Values Scale, in e nal locus o con ol
(ILOC), and he sa is ac ion wi h li e scale o cha ac e ize
he esul ing consume segmen s. While Lusk and B igge-
man (2009) d aw on he means-end chain heo y (in which
he means e e o mo e abs ac human alues) o explain
he ole o ood alues as in e media y alues, so a no
s udy has been conduc ed using human alues o explain
ood alue hie a chies. And ye he ole o alues in unde -
s anding consume p e e ences and beha io s has long
been acknowledged (S e n e al., 1999). Human alues a e
a key componen o he sel and pe sonali y; as such, hey
a e a c i ical d i e o beha io . The highe he le el o
e-exp ession o alues, he mo e s able hey a e (Kahle
& Kennedy, 1988). The 10 alues ha comp ise Schwa z’s
alues heo y a e sel -di ec ion, s imula ion, hedonism,
achie emen , powe , secu i y, con o mi y, adi ion, bene -
olence, and uni e salism. These alues can be s uc u ed
in bipola dimensions: “openness o change s. conse -
a ion” and “sel -enhancemen s. sel - anscendence.”
Openness o change comp ises sel -di ec ion and s imula-
ion, whe eas conse a ion comp ises secu i y, con o mi y,
and adi ion. Sel -enhancemen in ol es bene olence and
uni e salism, whe eas sel - anscendence in ol es powe
and achie emen . Hedonism can be a ibu ed o openness
o change and sel -enhancemen . Acco ding o Schwa z
(2012), a a mo e basic le el, he 10 alues o m a con-
inuum o ela ed mo i a ions. ILOC and li e sa is ac ion
ha e addi ionally been chosen o desc ibe he segmen s
because hese psychog aphic ac o s appea o play a el-
e an ole in consume s’ ood choices (Hempel & Roosen,
2022). ILOC implies ha people belie e e en s in hei li es
a e caused by hei own ac ions and no by ac o s beyond
hei con ol. I has been ound o co ela e wi h en i on-
men ally esponsible beha io and beha io al in en ion in
mul iple s udies (Weime e al., 2017). As such, i is di e -
en om pe cei ed sel -e icacy (i.e., he pe cei ed capabil-
i y o conduc he necessa y ac ions o achie e an ou come
in a pa icula si ua ion) and pe cei ed beha io al con ol,
one elemen o he well-known Theo y o Planned Beha -
io (Ajzen, 2002). ILOC appea s o be mo e s able ac oss
a ying beha io s and si ua ions (Cle eland e al., 2012;
Hanss & Do an, 2019). Howe e , hus a , ILOC has mainly
been s udied wi h ega d o heal hy ea ing beha io s; in
hose cases, a domain-speci ic heal h LOC scale has been
used (e.g., Lee e al., 2018). The sa is ac ion wi h li e scale,
de eloped by Diene e al. (1985) measu es li e sa is ac ion,
de ined as an indi idual’s judgmen o hei well-being and
quali y o li e. Hong and Giannakopoulos (1994) epo ed
indings om p e ious esea ch ha li e sa is ac ion is con-
sis en ly ela ed o g ea e ILOC, which is also suppo ed
by hei esul s. Rega ding ood, Hempel and Roosen
(2022) showed ha li e sa is ac ion was associa ed wi h
consume s’ inc eased p e e ences o local and o ganically
p oduced ood du ing he COVID-19 pandemic.
P e ious s udies ha e in es iga ed changes in ood
beha io due o economic shocks such as he G ea
Recession (e.g., Koh e al., 2013; Kuchle , 2011; Smi h e al.,
2014). Fo example, hese indings indica e ha consume s
exchange ood quali y o he quan i y o esh ege ables
(Kuchle , 2011). Howe e , li le change has been epo ed
ega ding consume s’ cooking p ac ices and away- om-
home consump ion du ing imes o ecession (Smi h e al.,
2014). Mo e ecen ly, hese s udies we e complemen ed
by esea ch conduc ed in a US Ame ican con ex du ing
he COVID-19 pandemic (e.g., Ce oni e al., 2022; Ellison
e al., 2021; Lusk & McFadden, 2021). Typically, o e all
ood spending dec eases du ing ecessions, bu ood
expendi u e makes up a g ea e sha e o consume s’ o al
budge s han in he p eceding yea s. Ne e heless, du ing
he COVID-19 pandemic, many people expe ienced a shi
in ime a ailabili y, leading o inc eased home p oduc ion
and mo e esou ces spen on inding lowe -cos op ions
(Lusk & McFadden, 2021). Ellison e al. (2021) e ealed
educ ions in expendi u es on ood away om home and
inc eased spending on ood e-comme ce. Fu he mo e,
he au ho s in es iga ed ood alues and showed ha hey
we e conside ably s able du ing he pandemic; none he-
less, he e was a sligh educ ion in he impo ance o p ice.
The mos impo an ood alue was as e. Ce oni e al.
(2022) obse ed a subs an ial dec ease in he impo ance
1028 HEMPEL and ROOSEN
o ood sa e y among US Ame ican consume s, whe eas
he alues o as e, nu i ion, appea ance, con enience,
and o igin inc eased sligh ly. The au ho s concluded ha
e en a subs an ial shock, such as he COVID-19 pan-
demic, did no gene a e conside able shi s in ood alue
hie a chies.
The ood alue scale p oposed by Lusk and B iggeman
(2009) is a common measu e o s udying ood-choice
d i e s. Whe eas human alues a e widely e e ed o as
concep s o belie s abou desi able end-s a es o exis ence,
ood alues should be in e p e ed as in e media y alues
ha ela e speci ically o people’s ood choices and, as
such, can also be conside ed d i e s o mo i es he eo .
The au ho s p oposed a se o 11 alues ha speci ically
ela e o consume s’ ood choices o in es iga e hei el-
a i e impo ance. Con a y o o he mul i-i em scales in
his con ex , he au ho s aimed o be abs ac a he han
speci ic o de elop a se o ood alues gene ally applicable
when explaining consume s’ choices be ween a wide ange
o ood p oduc s. Acco ding o Lusk and B iggeman (2009),
hey can be classi ied in o sel -cen e ed alues (i.e., na -
u alness, as e, p ice, con enience, appea ance, nu i ion,
and sa e y) and socie y-cen e ed alues (i.e., adi ion, o i-
gin, ai ness, and en i onmen al impac ). We implemen ed
he ood alue scale using he BWS app oach. Based on
he indi idual bes –wo s sco es, we pe o med a la en
class analysis o clus e consume s. O e he pas decades,
BWS has been a o ed o e mo e adi ional me hods o
measu emen (e.g., a ing scales), as i may assis ma ke
esea che s in educing biases (e.g., scala nonequi a-
lence) and is easy o conduc (Ma ley & Lou ie e, 2005).
Mo e ecen ly, Capu o and Lusk (2020) conduc ed a BWS
expe imen o in es iga e consume p e e ences ega ding
ood policies. They es ima ed mixed logi models and cal-
cula ed he p e e ence sha es o each o he selec ed poli-
cies. We chose o conduc a la en class clus e ing app oach
because o i s known bene i s o e s anda d clus e ing
echniques, o example, a k-means app oach, such as he
ac ha la en class clus e ing is a model-based app oach
wi h mo e o mal c i e ia o he choice o he numbe o
clus e s and mo e igo ous s a is ical es ing (Magidson &
Ve mun , 2002). Mo e speci ically, we ollowed he p oce-
du es sugges ed by Lusk and B iggeman (2009,Appendix)
andLooseandLockshin(2013), who used he bes –wo s
sco es o he la en class app oach. The indings o his
s udy will help policy make s o design a ge -speci ic
communica ion measu es, as hey yield in o ma ion on
consume s’ dis inguishing cha ac e is ics necessa y o
de eloping and implemen ing measu es aimed a heal h-
ie , mo e clima e- and en i onmen - iendly ood choices
du ing economically challenging imes.
In he ollowing chap e , we desc ibe he me hodolog-
ical app oach, s a ing wi h he da a collec ion, ollowed
by he implemen a ion o he BWS and la en class clus e
analysis, and p esen he a iables used o he p o il-
ing o segmen s (i.e., sociodemog aphic and beha io al
a iables, ILOC, li e sa is ac ion, and Schwa z Human
Values). Subsequen ly, we p esen he esul s. Fi s , we
show he ela i e impo ance o ood alue o bo h yea s.
Second, we p esen he main di e ences be ween he iden-
i ied consume segmen s by e e ing o ood alues and
desc ibe he segmen s in mo e de ail using he p o iling
a iables. Finally, we de i e implica ions and e lec on he
limi a ions o ou s udy.
2METHODOLOGY
2.1 Da a collec ion
Da a o his s udy we e collec ed ia he online access
panel o a ma ke esea ch company om No embe 9
o 19, 2020, and No embe 18 o 28, 2022, in Ba a ia— he
la ges ede al s a e in Ge many in e ms o a ea and
he second la ges in e ms o inhabi an s. The ma ke
esea ch ins i u e was esponsible o p og amming he
ques ionnai e, collec ing da a, and incen i izing pa ici-
pan s. Quo as we e speci ied o gende , age, educa ion,
employmen s a us, and household size o ob ain a
sample as close o he gene al popula ion as possible.
Only esponden s who we e a leas pa ly esponsible
o g oce y shopping and we e olde han 18 yea s we e
allowed o pa icipa e. Pa icipan s ook app oxima ely
20 min o comple e he su ey. In bo h yea s, he su ey
began wi h ques ions on pa icipan s’ sociodemog aphic
in o ma ion, con inued wi h he BWS o ood alues, and
ended wi h mul i-i em scales measu ing psychog aphic
a iables, such as ILOC and li e sa is ac ion. The BWS
o he ood alues and la en class clus e analysis a e
desc ibed in he subsequen sec ion, ollowed by an in o-
duc ion o he a iables used o p o iling he esul ing
clus e s.
2.2 Food alues and bes –wo s scaling
Ou me hodological app oach is based on a se o 11 alues
ela ed o consume s’ ood choices p oposed by Lusk and
B iggeman (2009). We chose ood alues o he segmen-
a ion o consume s in ou s udy, closely ollowing he
app oach o Lusk and B iggeman (2009), who we e guided
by he idea o “iden i ying people’s belie s ega ding he
p e e abili y o compe ing ou comes esul ing om ood
pu chase and consump ion” (Lusk & B iggeman, 2009,p.
185). As men ioned abo e, hese can be classi ied in o sel -
cen e ed alues (i.e., na u alness, as e, p ice, con enience,
HEMPEL and ROOSEN 1029
appea ance, nu i ion, and sa e y) and socie y-cen e ed
alues (i.e., adi ion, o igin, ai ness, and en i onmen al
impac ). Se e al s udies ha e adop ed he BWS app oach
o de e mine he ela i e impo ance o ood alues (c .
Bazzani e al., 2018; Ce oni e al., 2022;Ve aine al.,2021).
Sho desc ip ions o ood alues can be ound in Table 2
in he esul s sec ion. These desc ip ions we e p o ided o
pa icipan s du ing he su ey. They we e ansla ed in o
Ge man o his s udy. We applied BWS ype 1 o elici
consume impo ance ankings o ood alues. BWS ype
1 s udies aim o scale he cha ac e is ics o i ems o he
cons uc o in e es , such as liking, ag eemen , o impo -
ance (Loose & Lockshin, 2013). Acco ding o Adamsen
e al. (2013),BWSisanop ion oo e come eliabili y
issues common in simple a ing scales such as Like
scales, which ha e been widely applied in ma ke esea ch
s udies o examine consume s’ (dis)ag eemen wi h mul-
iple i ems. In he con ex o ood alue, socially desi able
esponses a e expec ed, as his is a opic on which s ong
public opinion exis s (i.e., ood alues ela ed o sus ain-
abili y migh be o e - epo ed, while sel -cen e ed alues,
such as p ice, as e, and con enience, migh be unde e-
po ed). Con a y o a ing scales, pa icipan s we e o ced
o choose he bes (mos impo an ) and wo s (leas
impo an ) op ions. By choosing ins ead o a ing, di e -
en in e p e a ions o esponse op ions can be a oided,
which e en ually biases he esul s. Mo eo e , i is possible
o ci cum en people’s endency o s a e ha all issues
a e impo an (Lusk & B iggeman, 2009). To a oid he
eliance on scaling o nume ous i ems wi h a se o Like
s a emen s o he a ing o i ems online o ca ego y scales
in consume esea ch, Jaege e al. (2008) also choose o
implemen a BWS app oach o elici p e e ences o ood
p oduc s.
Fo he expe imen al design, we applied a balanced
incomple e block design (BIBD) o he 11 ood alues, yield-
ing 11 choice se s, each including six ood alues. A BIBD
was chosen because i is he mos widely used design o
coun -based analyses, and a heo e ical p oo o he use
o o he expe imen al designs has no ye been epo ed
(Loose & Lockshin, 2013). An example o such a bes –wo s
ask is shown in Figu e 1.
BWS in ol es a se ies o ade-o s be ween a ibu es
o i ems h ough which he esponden s’ p e e ences a e
e ealed. Thus, he BWS is a conjoin analysis echnique.
We conduc ed a coun -based analysis o gene a e he bes –
wo s sco es subsequen ly used o segmen a ion analysis.
The e o e, we calcula ed indi idual bes –wo s sco es o
each ood alue by aking he di e ence be ween he num-
be o imes ha ood alue was selec ed as he mos
impo an and he numbe o imes i was chosen as he
leas impo an ac oss he 11 choice se s he esponden was
p omp ed o ill in. The highes possible sco e o each ood
alue was +6, and he lowes possible sco e was −6, as all
ood alues appea ed six imes in he BWS ask.
2.3 La en class clus e analysis
We used he bes –wo s sco es o he ood alues in a la en
class clus e analysis, ollowing he app oach o Lusk and
B iggeman (2009). Simila ly, Jaege e al. (2008) used he
bes –wo s sco es o u he analyses, as hey e i ied ha
he sco es we e p opo ional o he pa ame e s o a mul i-
nomial logis ic eg ession. The au ho s used gene alized
linea mixed models o analyze he BWS da a. Howe e ,
we assumed ha ood alue p e e ences a e he e oge-
neous among consume s, as was also ound by Lusk and
B iggeman (2009). Hence, we decided o conduc a la en
class clus e analysis based on he indi idual coun sco es
o all 11 ood alues o accoun o his he e ogenei y. Ou
aim was o iden i y he classes (segmen s) ha e ealed
homogeneous p e e ences wi hin classes and he e oge-
neous p e e ences ac oss classes (c . Bi e al., 2019). As
poin ed ou by Rokeach (1973), indi iduals and g oups
ha e di e en alue “p io i ies” o “hie a chies,” which
a e also e e ed o as alue sys ems. Segmen s de ined by
alue sys ems, a he han a single alue, a e mo e eliable
and ha e g ea e in e p e abili y. Indi iduals belonging o
a pa icula segmen sha e he same alue sys em, which
is ep esen ed by a se o unobse able u ili ies assigned
o alue desc ip ions. The ela i e impo ance weigh s o
each alue desc ip ion p o ided he esea che wi h an
objec i e assessmen o he alue p io i ies wi hin each
segmen .
The main eason o using he la en class app oach as
opposed o s anda d clus e analysis echniques (such as
k-means) is ha i is a model-based clus e ing app oach.
Th ough a s a is ical model, he choice o he clus e c i-
e ion is less a bi a y. Simul aneously, i conside ed ha
he e is unce ain y conce ning he p obabili y o class
membe ship o each objec . The basic la en class clus e
model akes he ollowing o m:
𝑓(𝑦𝑖|𝜃)=
𝐾
∑
𝑘=1
𝜋𝑘𝑓𝑘(𝑦𝑖|𝜃𝑘)
In his equa ion, he ec o 𝑦𝑖deno es an objec ’s sco es
on a se o obse ed a iables, 𝐾is he numbe o clus e s,
and 𝜋𝑘 ep esen s he p io p obabili y o belonging o a
la en class o clus e 𝑘. Al e na i e labels o he 𝑦’s a e
indica o s, dependen a iables, ou come a iables, ou -
pu s, endogenous a iables, o i ems. The dis ibu ion o 𝑦𝑖,
gi en he model pa ame e s 𝜃(i.e., 𝑓(𝑦𝑖|𝜃)), is assumed o
be a mix u e o densi ies, which a e speci ic o he classes
𝑓𝑘(𝑦𝑖|𝜃𝑘)(Ve mun & Magidson, 2002).
1030 HEMPEL and ROOSEN
FIGURE 1 Example o one o he bes –wo s asks p esen ed o he su ey pa icipan s. The su ey was conduc ed in Ge many; hus, a
Ge man e sion o he ood alue scale was included in he ques ionnai e. The English e sion is displayed he e o ensu e eadabili y o all
jou nal eade s.
Speci ically, he p obabili y densi y unc ion o Kla en
classes can be exp essed as ollows:
𝑓(𝑦𝑖)=
𝐾
∑
𝑘=1
𝑃(𝑘)
𝐽
∏
𝑗=1
∅(𝑐𝑖𝑗 −
𝑐𝑗𝑘
𝜎2
𝑗𝑘 )
He e 𝑐𝑖𝑗 ep esen s an indi idual i’s bes –wo s sco e o a
alue j,
𝑐𝑗𝑘 is he mean le el o impo ance o alue j,and
𝜎2
𝑗𝑘 is he clus e - and alue-speci ic s anda d de ia ion. ∅
deno es he s anda d no mal (Gaussian) densi y unc ion.
We conside he p obabili y (P) o belonging o one o
he ou la en classes (k) in he ollowing way: 𝑃(𝑘)=
𝑒∝𝑘
∑𝐾
𝑚=1 𝑒∝𝑚(Lusk & B iggeman, 2009).
We used he esul ing p obabili ies o class membe ship
o e alua e an indi idual’s p obabili y o being in one o
ou classes— he class wi h he highes p edic ed p oba-
bili y was selec ed as he p edic ed class. Bi e al. (2019)
compa ed app oaches o assigning indi idual espon-
den s o iden i ied classes. One app oach was also simply
based on he highes p edic ed p obabili y, whe eas he
o he app oach conside ed he di e ence be ween he
highes and nex highes p obabili ies o class membe -
ship. Howe e , hey did no ind any s a is ically signi ican
di e ences be ween he me hods. Hence, we decided o
adhe e o he simple app oach desc ibed abo e.
I is ecommended ha he numbe o clus e s (o la en
classes) should be selec ed o achie e he lowes Bayesian
In o ma ion C i e ion (BIC). I he BIC dec eases wi hou
eaching a minimal u ning poin , a BIC plo can be used
o de e mine he app op ia e numbe o clus e s (F aley &
Ra e y, 1998). We in es iga ed he BIC alues o all clus-
e solu ions be ween 2 and 12; he minimum was eached
o 11 clus e s in bo h samples (2020: BIC =48,516.74;
2022: BIC =46,821.05). Ou BIC plo s yielded elbows in
he wo-clus e (2020: BIC =49,164.57; 2022: 47,692.28) and
ou -clus e (2020: BIC =48,984.72; 2022: BIC =47,300.06)
solu ions. Fou clus e s (o la en classes) we e selec ed
because his solu ion esul ed in alues ha allowed o
a plausible in e p e a ion. We conduc ed he la en class
clus e analysis in S a a 16.
2.4 P o iling a iables used o
cha ac e ize segmen s
We used he sociodemog aphic a iables o age, sex, and
income o be e desc ibe he clus e s because hey a e e-
quen ly used o s a i ica ion. Fu he , hese sociodemo-
g aphic a iables e ealed signi ican di e ences be ween
he ou clus e s in 2020. Fo occupa ion, we e ealed
signi ican di e ences in ull- ime employmen and e i e-
men be ween he hedonic and socie al impac -cen e ed
segmen s, which migh also be e lec ed in hei espec i e
incomes. Fo he o he occupa ion ca ego ies, household
size, and educa ion, we could no ind signi ican di e -
ences be ween he ou segmen s in 2020. Fo he 2022
sample, we selec ed he same sociodemog aphic a iables
o p o iling as he 2020 sample. Mo eo e , as a alidi y
check o ou ini ial segmen a ion, we examined con-
sume s’ s a ed o ganic and local ood pu chases o desc ibe
he segmen s. The su ey esponden s we e asked o gi e
hei ag eemen o wo single-i em a iables on a i e-poin
Like scale (“I mainly buy locally p oduced ood” and “I
mainly buy o ganically p oduced ood”). These s a emen s
a e gene ic and do no e e o any pa icula p oduc ca -
ego y. We ollowed he app oach o Lusk and B iggeman
(2009), who also su eyed esponden s’ p e ious o ganic
ood pu chases and e ealed ha people o whom na u al-
ness, ai ness, and he en i onmen we e mo e impo an
we e mo e likely o ha e p e iously bough o ganic ood.
Fu he mo e, we measu ed human alues in 2022 using
a sho e sion o he Po ai s Value Ques ionnai e,
he well-es ablished 21-i em Human Value Scale de el-
HEMPEL and ROOSEN 1031
oped by Schwa z (1992). I comp ised 21 sho po ai s
desc ibing he impo ance o di e en alues o di e en
people. Di e en e sions o he scale we e implemen ed
wi h di e se emale and male esponden s. The scales
we e in oduced wi h “Following, some people will be
desc ibed. Please indica e he ex en o which each pe son
is o is no like you.” Responden s we e asked o indica e
he likeness on a 6-poin scale, om “no a all like me” o
“ e y much like me.” Rega ding he analysis, we ollowed
he app oach o calcula ing cen e ed sco es o each alue,
as sugges ed by Schwa z in he Eu opean Social Su ey
guide.
Addi ionally, we included ILOC and he sa is ac ion
wi h li e scale in he 2022 sample. To measu e ILOC, a
h ee-i em scale was used (Jakoby & Jacob, 1999). This
scale was alida ed wi h wo la ge samples in 1995 and
1996, in which C onbach’s alpha was .71 and .62. The o ig-
inal Ge man scale was used in his s udy’s ques ionnai e.
A 5-poin Like scale was used, ollowing he ecom-
menda ions o Jakoby and Jacob (1999). In ou s udy,
C onbach’s alpha o ILOC was .63. A i e-i em scale o
measu ing li e sa is ac ion was adop ed om Janke and
Glöckne -Ris (2012). Al hough he au ho s ecommended
using a 7-poin Like scale, a 5-poin Like scale was
applied in his s udy o consis ency because all o he
scales we e measu ed on 5-poin Like scales. C onbach’s
alpha was .87 o he sa is ac ion wi h li e scale.
We applied - es s o independen samples, analysis o
a iance, and chi-squa e es s o e eal signi ican di e -
ences be ween clus e s ega ding hese p o iling a iables.
Whene e he equi emen s o analysis o a iance we e
no me , we decided on a nonpa ame ic al e na i e (i.e.,
K uskal–Wallis es s). All analyses we e pe o med using
SPSS 26 (IBM).
3RESULTS AND DISCUSSION
3.1 Sample desc ip ion
The inal sample comp ised 1020 esponden s in 2020 and
1004 esponden s in 2022. An o e iew o he sociode-
mog aphic in o ma ion o bo h samples, as well as a
compa ison wi h he gene al popula ion s a is ics, can be
oundinTable1. We pe o med a s a is ical compa ison o
bo h samples using - es s and chi-squa e es s o check o
di e ences be ween he 2020 and 2022 samples. We did no
ind signi ican di e ences in gende , amily s a us, house-
hold size, educa ion le el, o occupa ion (all P- alues we e
la ge han .05). Howe e , he es s e ealed di e ences in
age and income be ween 2020 and 2022 (P<.05); income
was lowe , while age was highe in 2020 han in 2022.
3.2 Gene al desc ip ion o he ood
alues
The coun -based analysis o he ood alues e ealed ha
as e is on a e age conside ed as he mos impo an ood
alue, bo h in 2020 and 2022. While sa e y and na u alness
ollow by a on places 2 and 3 in he 2020 sample, as e is
closely ollowed by p ice in he 2022 sample. Appea ance,
adi ion, and con enience a e he h ee ood alues ha
we e ega ded as leas impo an in bo h yea s. The mos
impo an as well as he leas impo an ood alue belong
o he g oup o sel -cen e ed alues (i.e., as e, sa e y,
na u alness, p ice, nu i ion, appea ance, con enience),
whe eas socie y-cen e ed alues (i.e., o igin, en i onmen-
al impac , ai ness, adi ion) a e si ua ed in he mid- ange
o all ood alues, showing a he li le impo ance on
a e age (Table 2).
While we expec ed he as e o be a ed as one o he
mos impo an ood alues based on he indings o o he
s udies on ood alues and ood-choice mo i es (c . Baud y
e al., 2017; Malone & Lusk, 2017), we did no expec o see
such a clea i s place in 2020. Con e sely, o he s udies
ha e iden i ied ood sa e y as he mos impo an (Bazzani
e al., 2018; Lusk & B iggeman, 2009), which was also
among he op h ee ood alues in ou s udy. None heless,
a s udy measu ing ood-choice mo i es ins ead o ood
alues in a F ench sample also e ealed as e as he
s onges mo i e (Baud y e al., 2017). O e all, he a he
low impo ance o socie y-cen e ed alues e ealed in
bo h samples con i ms he indings o p e ious s udies (c .
Bazzani e al., 2018; Ce oni e al., 2022;Ve aine al.,2021).
The s ong inc ease in he impo ance o he ood alue
p ice is su p ising. This is no in line wi h he indings
o Ce oni e al. (2022), who concluded ha subs an ial
shocks, such as he COVID-19 pandemic, do no gene a e
subs an ial shi s in he ood alue hie a chies o US
Ame ican consume s. Addi ionally, Ellison e al. (2021)
ound a sligh dec ease in he impo ance o p ice du ing
he pandemic, as well o US Ame ican consume s.
This appa en con adic ion may be explained by he
di e en e ec s o he COVID-19 pandemic on consume s
and hei ood choices. While some did no expe ience
budge cons ain s, o he s compensa ed o hem h ough
inc eased home p oduc ion and mo e esou ces spen on
inding lowe -cos op ions due o mo e ime a ailabili y
(Lusk & McFadden, 2021). Howe e , du ing in la ion in
2022, all consume s will be a ec ed by high ood p ices
in Ge many, po en ially leading o an o e all g ea e
impo ance o he ood alue p ice. Fu he mo e, his
inding migh be explained by he geog aphical con ex , as
he impo ance o ood alues is also a ec ed by cul u al
ac o s (Ba ena e al., 2015).
1032 HEMPEL and ROOSEN
TABLE 1 Sociodemog aphic cha ac e is ics o he samples in 2020 and 2022 and s a is ics o he Ge man popula ion (DE).
Sample 2020 2022
n1020 DE 1004 DE
Sex (%) Female 51 51 53 51
Male 49 49 47 49
Age (median) In yea s 48 46 44 46
Occupa ion (%) Full- ime employed 51 n.a. 53 n.a.
Re i ed 22 n.a. 18 n.a.
Pa - ime employed 12 n.a. 14 n.a.
O he (e.g., homemake ,
in school, unemployed)
15 n.a. 15 n.a.
Educa ion (%) Seconda y school 49 57 46 55
Uni e si y deg ee 28 n.a. 33 n.a,
Gene al quali ica ion o
uni e si y en ance
22 35 21 37
No deg ee o s ill in
school
18
*19
*
Household (%) 126 20 26 20
241 34 41 33
317 18 15 18
413 19 14 19
>4 3 9 5 10
Income in € (%) <1.000 5 5 10 12
1.000 ≤1.500 8 8
1.500 ≤2.000 12 10 20 21
2.000 ≤2.500 14 11
2.500 ≤3.000 (2022:
2.500 ≤3.500)
14 11 19 20
3.000 ≤4.000 (2022:
3.500 ≤4.500)
17 19 15 48
>4.000 (>4.500) 18 34 23
No indica ed 12 1 12 0
No e: In bo h su eys, only people aged 18 yea s and abo e, as well as hose who we e some imes esponsible o g oce y shopping, we e allowed o pa icipa e.
The e o e, he e may be de ia ions om he gene al popula ion.
*This sha e is la ge in he gene al popula ion han in ou sample, as we only included people aged 18 yea s and abo e.
3.3 P esen a ion o he consume
segmen s based on ood alues
The la en class clus e analyses esul ed in ou con-
sume segmen s (see Table 3 o 2020 and Table 4 o
2022). The i s g oup comp ises heal h- and sa e y-
conce ned consume s because nu i ion and sa e y a e
conside ed ela i ely impo an . The second segmen com-
p ises consume s whose membe s place ela i ely high
impo ance on he socie y-cen e ed alues o igin, en i-
onmen al impac , and ai ness and is called he socie al
impac -cen e ed segmen (o sho : socie y-cen e ed seg-
men ). G oup 3 comp ises p ice-sensi i e consume s who
pe cei e p ice as he mos impo an alue. G oup 4 is
he hedonic segmen , whose membe s a ed as e as he
mos impo an , closely ollowed by p ice in 2020. How-
e e , in 2022, p ice is also he mos impo an ood alue
in he hedonic segmen , simila o he p ice-sensi i e seg-
men . Al hough as e and p ice we e he o e all mos
signi ican ood alues in ou s udy, hey we e o only
mino impo ance in he socie al impac -cen e ed seg-
men . While heal h-conce ned and p ice-sensi i e con-
sume s a e mos sel -cen e ed alues as compa a i ely
impo an , he hedonic segmen is addi ionally cha ac-
e ized by he ex emely low impo ance hey place on
socie y-cen e ed alues. The e o e, hey a e con a y o he
socie al impac -cen e ed segmen (Tables 3and 4).
A compa ison o segmen sizes be ween 2020 and 2022
e ealed ha he heal h/sa e y-conce ned segmen was
he la ges in bo h yea s and inc eased in size om 2020
HEMPEL and ROOSEN 1033
TABLE 2 Mean alues and s anda d de ia ions (SD) o indi idual bes –wo s sco es (BWS) o all ood alues.
BWS 2020 BWS 2022
Food alues Desc ip ion Mean SD Mean SD
Tas e Ex en o which ood consump ion is
appealing o he senses
1.78 2.17 2.29 2.16
Sa e y Ex en o which consump ion o ood will
no cause illness
1.02 2.26 .84 2.07
Na u alness Ex en o which ood is p oduced wi hou
mode n echnologies
.95 2.12 .80 2.29
P ice P ice paid o he ood .67 3.01 1.86 2.93
O igin Whe e he ag icul u al commodi ies we e
g own
.54 2.14 .10 1.74
Nu i ion Amoun and ype o a , p o ein,
i amins, e c.
.50 2.36 .25 2.13
En . impac E ec o ood p oduc ion on he
en i onmen
.05 1.95 −.35 2.10
Fai ness Ex en o which all pa ies in ol ed in he
p oduc ion o he ood equally bene i
−.14 1.93 −.59 1.65
Appea ance Ex en o which ood looks appealing −1.04 2.03 −1.09 2.02
T adi ion P ese ing adi ional consump ion
pa e ns
−1.74 2.26 −1.97 2.29
Con enience Ease wi h which ood is cooked and/o
consumed
−2.59 2.48 −2.14 2.43
TABLE 3 Class means and s anda d e o s in b acke s o he ou la en classes (2020).
G oup 1
“heal h-/sa e y-conce ned”
34.22%, n=349
G oup 2
“socie y-cen e ed”
29.61%, n=302
G oup 3
“p ice-sensi i e”
20.49%, n=209
G oup 4
“hedonic”
15.69%, n=160
Tas e 2.695 (.318) .371 (.135) 1.237 (.204) 3.071 (.222)
Sa e y 1.576 (.167) .913 (.195) .570 (.206) .628 (.218)
Na u alness 1.596 (.169) 1.752 (.143) −.104 (.155) −.512 (.180)
P ice −.667 (.168) −1.360 (.140) 4.293 (.232) 2.640 (.277)
O igin .246 (.170) 2.234 (.168) .011 (.161) −1.283 (.189)
Nu i ion 1.488 (.181) −.020 (.239) .165 (.200) −.200 (.228)
En i onmen al
impac
−.043 (.122) 1.501 (.155) −.289 (.151) −1.958 (.210)
Fai ness −.384 (.119) 1.258 (.182) −.639 (.150) −1.556 (.168)
Appea ance −.951 (.151) −1.633 (.133) −1.395(.184) .285(.204)
T adi ion −2.254 (.163) −1.463 (.151) −1.734 (.208) −1.170 (.262)
Con enience −3.304 (.182) −3.553 (.147) −2.115 (.211) .055 (.221)
o 2022. The size o he hedonic segmen dec eased only
sligh ly wi hin he 2 yea s. The socie al impac -cen e ed
segmen was he second la ges g oup, wi h a sha e o
29.6 % in 2020; none heless, he segmen size dec eased
o 5.5 % wi hin hose 2 yea s. Con e sely, he size o
he p ice-sensi i e segmen inc eases signi ican ly, becom-
ing he second-la ges g oup in 2022 (Tables 3and 4,
i s ows). These indings sugges an in e se ela ion-
ship; while p ice is becoming ele an o an inc easing
numbe o people, he impo ance o he socie al impac
o ood consump ion is dec easing. A chi-squa e es
e ealed ha o all segmen s, he di e ences in size
be ween bo h yea s we e signi ican , assuming a 5% e o
p obabili y.
