187
1, XX, 2017
Ma ke ing and T ade
DOI: 10.15240/ ul/001/2017-1-013
In oduc ion
Loyal cus ome s a e a key ac o o success ul
e ail ac i i ies. Cus ome loyal y can be
de i ned as a highe p obabili y o making
new and epea ed pu chases, spon aneously
ecommending a pa icula e aile and
sp eading he posi i e wo d-o -mou h.
“Loyal cus ome s a e less likely o swi ch o
a compe i o due o p ice inducemen , and hese
cus ome s make mo e pu chases compa ed o
less loyal cus ome s” (Dhal, 2015; Mohelska
& Sokolo a, 2016) Fo he e aile s, cus ome
loyal y is a key ac o in eaching long- e m
comme cial success and p o i abili y. I is much
cheape o e ain and cu a e exis ing cus ome s
han o acqui e new ones. Simila ly, cus ome
loyal y has been inc easingly s udied in heal h
and social se ices o e he pas ew yea s –
pa icula ly in economies whe e he p i a e
sec o domina es his i eld. Many such s udies
ocus on he opic o demand egula ion wi hin
he heal h and social se ices sec o , which
is s ongly mo i a ed by popula ion aging and
ela ed ac o s occu ing in de eloped coun ies
(Ga u o á e al., 2014; Šol és e al., 2014).
In mode n cus ome -o ien ed ma ke ing,
loyal y p og ams (de i ned as ma ke ing
p og ams ha ewa d membe s wi h pu chase
incen i es) a e pe cei ed as he s a egic
ins umen o c ea ing and main aining
e ec i e con ac wi h cus ome s (Bacik e al.,
2015). “In he e ail scene, loyal y p og ams
in ol e a concen a ed e o by e aile s o
build s o e a i c, inc ease baske size and
inc ease equency which c ea es deepe
ela ionship ies wi h i s cus ome base” (Oma ,
Wel, Musa, & Naz i, 2010). An impo an pa in
c ea ing consume s´ sa is ac ion and inducing
consume loyal y is played by he human ac o
(e.g. Wasan and T ipa hi (2015)). I is o en
claimed ha e ail pe sonnel should be well
ins uc ed and mo i a ed owa ds inc easing
he numbe o loyal cus ome s and c ea ing
ha mony be ween co po a e s a egic aims and
he cus ome s’ demand.
As loyal y p og ams (LPs) a e implemen ed,
managed and e alua ed, e aile s need
app op ia e da a and means o iden i y indi idual
pa icipa ing cus ome s. Such da a allow o
e ec i e e alua ion o consume s’ pu chasing
beha iou and habi s. Speci i cally, o he pu pose
o egis e ing, uniquely iden i ying and p ope ly
managing cus ome s in a LP, a ious iden i i ca ion
da a a e equi ed. In he Czech Republic,
es ablishing and managing any such consume /
LP da abase is bound by legal amewo k,
implemen ed o consume and pe sonal da a
p o ec ion. Besides legisla u e, cus ome s o en
ea ha hei pe sonal da a – once passed o
LP-ope a o – migh be misused (e.g. sold o hi d
pa ies and used in an in usi e manne ).
Gene ally speaking, consume s wo y abou
hei p i acy and a e a aid o losing con ol o e
hei pe sonal da a – o some ex en . Impo an ly,
consume s’ p i acy conce ns a e no uni o m.
Indi idual p e e ences, sociodemog aphic and
li es yle ac o s play a signi i can ole. F om
a good- ai h LP o ganize poin o iew, he
quali y o pe sonal da a and con ac in o ma ion
collec ed h ough he LP has an eno mous
e ec on hei abili y o manage and e alua e
LPs p ope ly. This s udy p o ides a s uc u ed
quan i a i e analysis o cus ome s´ willingness
o sha e a ious ypes o pe sonal and con ac
da a wi h LP o ganize s.
1. Li e a u e Re iew
The use o LP- ela ed pe sonal and con ac da a
o he bene i o ma ke e s and businessmen
is a delica e opic, discussed by academic
esea che s and e ail ma ke e s, as well as by
ins i u ions supe ising legal aspec s o such
ac i i ies (see e.g. Alb ech (2006) o Ma ine
and Cannella (2015)).
LOYALTY PROGRAMS AND PERSONAL
DATA SHARING PREFERENCES IN THE
CZECH REPUBLIC
Radek Tahal, Tomáš Fo mánek, Hana Mohelská
EM_1_2017.indd 187EM_1_2017.indd 187 13.3.2017 16:59:1213.3.2017 16:59:12
188 2017, XX, 1
Ma ke ing a obchod
“Loyal y p og ammes, app op ia ely
managed, a e conside ed o allow s uc u ed
and e ec i e ac ions o manage, selec , ela e,
and con ol cus ome s’ buying beha iou ” (La a
& De Mada iaga, 2007). In o de o mee hei
objec i es, LPs need o be bene i cial o bo h
he e aile and he cus ome s. A new s udy by
Nielsen (a co po a e p o ide o in o ma ion
and insigh s in o wha consume s wa ch and
buy), e ealed ha nea ly 60 pe cen o global
esponden s epo ed ha LPs we e a ailable
o hem h ough local e aile s. F om hose
consume s wi h access o LP-o ganizing
e aile s, 84 pe cen epo being mo e likely o
isi such e aile s (Nielsen, 2013).
La a and De Mada iaga (2007) discuss
he impo ance o i nding a desi able balance
be ween he olume o da a equi ed by he
e aile and cus ome s’ willingness o sha ing
hem. As consume s a e disinclined o acili a e
access o ce ain da a ypes o he companies
(LP o ganize s), his can be a ac o ha
mode a es consume willingness o pa icipa e in
such p og amme. “I onically, e en hough loyal y
p og am membe s c a e a mo e pe sonalized,
ele an expe ience, hey also show conce n
abou sha ing he in o ma ion equi ed o
enable he e aile o deli e on his desi e.” (PR
Newswi e, 2013). Consume incen i es, along
wi h us and p i acy assu ances a e he key
s a egic ools in LP consume engagemen .
Nume ous s udies poin ou ha people a e
o en a aid o hei pe sonal da a becoming an
objec o ading among companies (see e.g.
Wade (2010), Spieke mann, Acquis i, Böhme
and Hui (2015)). “Compounding his p oblem
is he common p ac ice o businesses selling
cus ome in o ma ion o o he businesses,
ma ke ing companies, mailing lis s and so
o h, he eby u he inc easing he amoun o
unwan ed o e s and ad e ising” (Ma ke ing
Weekly News, 2013).
In hei co po a e esea ch, Aimia (2011)
desc ibe he young gene a ion (gene a ion Y
o he Millennials, bo n app oxima ely be ween
1977 and 1994) as being less conce ned
when sha ing pe sonal da a wi h a e aile o
shopping easons. O all named ma ke ing
channels in hei su ey, loyal y and ewa d
p og ams a e pe cei ed as he mos p i acy-
iendly by Millennials: ewe han 20% o
Millennial loyal y p og am membe s a e
conce ned abou sha ing pe sonal in o ma ion
wi h loyal y p og am o ganize s.
As Koponen and Mangia acina (2014)
s a e, a good way o illus a e he inc easing
comme cial alue o pe sonal da a is by
conside ing he ecen g ow h o he online
ad e ising sec o . “Ou cul u e places a high
alue on p i acy. Pu ing con ol o he da a –
wha may be collec ed and e ained and wha
mus be dele ed ( he igh o be o go en) –
in each o ou hands h ough a con ac ual
a angemen wi h he collec o o selle ,
p omo es indi idual choice and con ol”
(Ande son, 2015).
I is e ail ma ke e s’ ask o communica e
pe sonal da a equi emen s o consume s as well
as he ad an ages o LPs. Bo h LP-o ganize s
and consume s may be iewed as being in
a p ocess o de eloping a new unde s anding
o he ules o consume engagemen p og ams
(such as LPs) ha no only ma ch consume
p e e ences bu help consume s accomplish
goals in eal- ime – see Documen News (2013)
o de ailed discussion. Cus ome s ough o be
assu ed ha hei da a will be u ilized e hically,
law ully and o imp o ing e ail o e s. “Imagine
a membe o a cus ome loyal y p og am who
is comple ely in o med and ag ees ha da a
abou his shopping habi s a e used o op imize
business p ocesses and o p o ide ad e ising
ma e ial and special o e s” (Ma zne , 2014).
2. Resea ch Focus
This esea ch is ocused on cus ome s´
beha iou , p i acy and da a-sha ing p e e ences
ela ed o LP pa icipa ion in he Czech
Republic. Acco ding o ou p e ious esea ch,
Czech cus ome s a e less inclined o sha ing
hei pe sonal and con ac da a as compa ed
o cus ome s in selec ed EU coun ies (see
e.g. Tahal (2015)). We aim o help ma ke e s
wi h a pa icula ly impo an ask: o i nd he
app op ia e balance be ween he amoun
o pe sonal da a equi ed om consume s
pa icipa ing in a gi en LP and he ex en o
which consume s a e willing o sha e such da a.
In his pape , we use he p ima y su ey da a o
answe wo main esea ch ques ions:
Resea ch ques ion 1: Wha de e mines
consume willingness o p o ide di e en
ypes o con ac and pe sonal in o ma ion
o o ganize s o LPs wi hin he FMCG – a e
he e any signi i can di e ences ha would be
de e mined by socio-demog aphic and li es yle
ac o s?
EM_1_2017.indd 188EM_1_2017.indd 188 13.3.2017 16:59:1213.3.2017 16:59:12
189
1, XX, 2017
Ma ke ing and T ade
Resea ch ques ion 2: Does he ex en o
indi idual pa icipa ion in LPs a ec consume ’s
willingness o p o ide con ac and pe sonal
in o ma ion?
I should be explici ly no ed ha ou esea ch
ac i i ies (and conclusions made) aim owa ds
ai LP schemes, i.e. LPs whe e cus ome s may
ealis ically eel ha in exchange o he pe sonal
in o ma ion p o ided hey a e ewa ded by
adequa e pu chase incen i es and/o imp o ed
and mo e pe sonalized se ices by he e aile .
Ou esea ch o consume willingness o
p o ide con ac and pe sonal in o ma ion o
LP o ganize s ocuses on he ollowing 6 ypes
o pe sonal da a: (a) Name & Su name, (b)
Email, (c) Add ess/ esidence, (d) Bi hda e,
(e) Phone numbe , ( ) Pe sonal ID numbe .
O e all, his classi i ca ion e l ec s common
ma ke ing p ac ice, as LP-based pe sonal
da a a e used o e ail segmen a ion, analysis,
audi ing, o ecas ing, e c. In his pape , he
i s da a ype (Name & Su name) and he las
da a ype (Pe sonal ID numbe ) ha e a limi ed,
benchma k- ype usage: P o iding one’s name in
o de o pa icipa e in a LP is he e y minimum
equi emen . Besides Czech language-speci i c
gende segmen a ion, name bea s li le usable
in o ma ion and i is o en seen as he leas
sensi i e da a ype. The e o e, Name/Su name
da a may se e as a benchma k in measu ing
he basic p opensi y o sha e non-sensi i e
pe sonal da a. In con as , he Pe sonal ID
numbe is a e y sensi i e ype o in o ma ion
wi h a non-negligible misuse po en ial. Ac ually,
LP o ganize s in he Czech Republic a e no
allowed o legally collec and use his ype da a
unde mos p ac ical ci cums ances (excep ions
may apply). Also, Pe sonal ID bea s li le
usable in o ma ion, once da a ypes (b) o
(e) a e con olled o . The e o e, we use he
in o ma ion on consume willingness o sha e
hei ID numbe s as a second benchma k, o
he o he end o he pe sonal da a sensi i i y
spec um.
3. Da a and Resea ch Me hodology
Ou esea ch is based on p ima y su ey
da a om he Czech Republic. A complex,
anonymized and s a i i ed (quo a) sampling was
pe o med du ing he pe iod om No embe
2015 o Ap il 2016, ga he ing da a o a sample
o 411 esponden s om he popula ion o
FMCG consume s aged 15+. The s a i i ed
quo a sampling was based on h ee ac o s:
loca ion, age and gende o he esponden s.
Hence, ou me hodology ensu es consis ency
and ele ance o he esul s – conclusions may
be d awn wi h espec o he popula ion o 15+
consume s. A combina ion o pe sonal and
online da a collec ion was used o ga he socio-
demog aphic in o ma ion, li es yle p e e ences,
a i udes owa ds di e en ypes o wo k and
ee ime ac i i ies o he esponden s. Bo h
quan i a i e (in e al based) and quali a i e
(Yes/No, Like scale-based) ques ions we e
used in he su ey.
The su ey was o ganized and pe o med by
a esea ch eam a he Uni e si y o Economics,
P ague. This eam is led by uni e si y
employees and eache s who supe ise and
coo dina e he asks pe o med by s uden s
specializing in ma ke ing esea ch. This s udy
is pa o a sys ema ic long- e m p ojec o
specialized ma ke ing analyses (see e.g. Tahal
and S ří eský (2014)). Ou empi ical analysis
(logis ic eg ession and ela ed s a is ical
in e ence) is adjus ed o con ol o he quo a
sampling, hus ensu ing p ope conclusions a e
made owa ds he FMCG consume popula ion
based on he esul s and hei in e p e a ion.
Also, o su ey da a alida ion, he Wald-
Wol owi z “Runs” es was used o es he
H0 o o de o obse a ions being a ibu able
o chance agains he H1 o po en ial da a
collec ion mishandling (Wacke ly e al., 2008).
All he su eyed da a (in e al-based, Yes/
No, Like scale) need o be con enien ly s o ed
o subsequen quan i a i e analysis. Gi en
he na u e o ou ques ionnai es, he ga he ed
da a may be con enien ly eco ded as logical
(bina y) a iables. The ans o ma ion o Yes/
No answe s (e.g. o ques ions ela ed o
willingness o p o ide pe sonal da a) is s aigh -
o wa d. Fo in e al-based quan i a i e opics
such as age o income, we use bina ies o
eco d esponden ’s app op ia e in e al en y.
Fo example, he a iable Age_65_plus equals
1 o all he esponden s aged 65 and olde
and is ze o o he wise. Answe s o Like scale-
based ques ions a e also eco ded as bina y
a iables. Fo he sake o ou analysis, o de ed
mul inomial da a may be app oached in a way
simila o he in e al-based quan i a i e
a iables. Fo example, esponden s a e asked
o posi ion hemsel es owa ds a s a emen
“I like eading books” using a i e deg ee Like
scale (“1” = his s a emen desc ibes me e y
EM_1_2017.indd 189EM_1_2017.indd 189 13.3.2017 16:59:1213.3.2017 16:59:12
190 2017, XX, 1
Ma ke ing a obchod
well, …, “5” = his s a emen does no desc ibe
me a all). Subjec i ely pe cei ed impo ance
o indi idual li es yle is add essed by his
ques ion, a he han some ac ual measu e o
eading ime (o page olume coun ). Also, he
ac ha “1” is a be e a ing han “2” (in e ms
o ag eeing wi h he s a emen e alua ed) bea s
o dinal meaning only, i.e. we canno say ha
he di e ence be ween “2” and “4” is somehow
wice as impo an as he di e ence be ween
“4” and “5”. The e o e, he su eyed answe s
o his s a emen we e used o p oduce wo
bina y a iables: LS_books_yes equals 1 o
hose who espond “1” on he Like scale
and ze o o he wise, LS_books_no equals 1
o esponden s who dissocia e hemsel es
om he s a emen by answe ing “5” (and ze o
o he wise). Al hough bina y a iable may be
p oduced o each o he Like scale answe s,
we i nd i empi ically con enien o combine
s a emen s “2” o “4” – i.e. no a e y s ong
posi ion o he esponden – in o a single e e ence
ca ego y. Ou app oach has h ee ad an ages:
Fi s , we e ie e all cases whe e esponden s
ha e s ong pe sonal posi ions on gi en li es yle
ac i i ies and opics such as eading books,
doing spo s, ea ing o ganic ood, e c. Second,
he combined base answe s “2” o “4” may s ill be
included implici ly in he analysis as a e e ence
ca ego y, necessa y o in e p e a ion o he
es ima ed eg ession models. Thi d, using such
e e ence base de-couples he LS_book_yes
and LS_book_no bina ies ha a e no linea ly
dependen and may be bo h used as explana o y
a iables in he same eg ession model.
Using he abo e desc ibed app oach,
we ha e ans o med he su eyed ma e ial
in o a 402- ow da ase (9 esponden s we e
dis ega ded due o missing da a issues) wi h
106 a iables. Six o he a iables desc ibe
consume willingness o sha e pe sonal da a
wi h LP o ganize s. The emaining one hund ed
a iables o m a pool o po en ial/concei able
eg esso s bea ing socio-demog aphic,
li es yle and o he ele an in o ma ion ha
may be used o model consume s’ pe sonal
da a sha ing p e e ences. Fo ou da ase ,
an exhaus i e (b u e- o ce) sea ch o a uly
op imal pa ame ic model speci i ca ion is
compu a ionally inaccessible, as i would equi e
an es ima ion and e alua ion o 6×2100 models.
Hence, in o de o iden i y a small, in o ma i e
and consis en se o explana o y a iables, we
combine a o wa d-s epwise selec ion me hod
( his is a po en ially subop imal algo i hm ha
p oduces nes ed sequences o models) wi h
he non-pa ame ic andom o es app oach.
Di e ences in ou pu s om he wo me hods
a e analysed in o de o de ec any po en ial
sub-op imali y in ou pu om he s epwise
me hod. This app oach allows o e i cien
and compu a ionally easible e alua ion o
indi idual explana o y a iables wi h espec
o p edic ion accu acy o he model, as only
6×1002 models and 6×5,000 andom o es s a e
p oduced and e alua ed ( as and au oma ed
e alua ion p ocedu es a e a ailable in R
and o he so wa e packages). Al hough ou
me hodology does no gua an ee a uly op imal
model speci i ca ion, i may be ega ded as an
accep able app oxima ion wi h a ela i ely
low po en ial o sub op imali y). Fo de ailed
discussion o model selec ion me hodology,
see James e al. (2013). Hence, in o de o
answe he esea ch ques ions (RQs) 1 and 2,
he abo e desc ibed model selec ion p ocess
was used o gene a e a consis en model
speci i ca ion as ou lined by equa ion (1):
yi = β0 + β1 Femalei +
+ β2 Age_15_24i + β3 Age_65_ plusi +
+ β4 Mo a iai +β5 Ea nings_highi +
+ β6 LS_TV_noi + β7 LS_books_noi +
+ β8 LS_In e ne use_noi +
+ β9 LS_Payca d_yesi +
+ β10 LS_exo ics_yesi +
+ β11 LS_cooking_noi +
+ β12 LP_Memb_1_2i +
+ β13 LP_Memb_3_plusi+ui
(1)
whe e yi is a bina y dependen a iable – six
di e en dependen a iables a e used wi h
he le hand side o he equa ion and he e o e
six di e en equa ions a e es ima ed using he
model (1) – Yes/No answe s we e eco ded
o he ques ion “Would you p o ide he
ollowing ype o pe sonal in o ma ion in o de
o become a membe o a loyal y p og am?”
o each esponden and da a ca ego y (a) o
( ) as de i ned abo e. Please no e ha such
ques ion applies iden ically o consume s who
al eady a e membe s and/o ac i e use s o
a LP as well as o indi iduals no pa icipa ing
in LPs. On he igh hand side o (1), βj a e he
coe i cien s o be es ima ed h ough logis ic
eg ession (see Da idson, MacKinnon (2009,
p. 454-465)). Femalei is a bina y explana o y
a iable dis inguishing be ween emale and
EM_1_2017.indd 190EM_1_2017.indd 190 13.3.2017 16:59:1213.3.2017 16:59:12
191
1, XX, 2017
Ma ke ing and T ade
male esponden s, Age_15_24i, is a bina y
indica ing he 15-24 age g oup while Age_65_
plusi depic s indi iduals aged 65 and olde
(upon a iable impo ance e alua ion as
desc ibed abo e, age anges 25-34, 35-49
and 50-64 a e combined in o a single base
ca ego y). Mo a iai desc ibes he esidence
o esponden s and Bohemia se es as i s
e e ence ca ego y. Responden s wi h high
ea nings (de i ned by a mon hly household
income o e 80 housand CZK) a e disce ned
using Ea nings_highi. Like scale-based
li es yle a iables LS_TV_noi, LS_books_noi
and LS_In e ne use_noi ma k esponden s
who dissocia e hemsel es om wa ching
TV, eading books and using he in e ne .
LS_Payca d_yesi indica es whe he he i- h
esponden uses pay ca ds (c edi and debi )
equen ly and LS_exo ics_yesi de e mines
whe he esponden s a e keen on spending
hei holidays a exo ic des ina ions o by he
sea (as he Czech Republic is a landlocked
coun y). LS_cooking_noi disce ns people who
dissocia e hemsel es om cooking (again,
subjec i ely pe cei ed impo ance is add essed
he e). LP_Memb_1_2i iden i i es indi iduals
(consume s) who ac i ely pa icipa e in 1 o 2
LPs. Simila ly, LP_Memb_3_plusi desc ibes
esponden s who ac i ely ake pa in 3 o mo e
LPs. Finally, ui is he po en ially he e oskedas ic
andom elemen .
The in e p e a ion o mos o he li es yle
a iables included in equa ion (1) is ela i ely
s aigh - o wa d, gi en he in ui ion on Like
scale da a ans o ma ion as p o ided abo e.
Howe e , he las wo a iables (LP_Memb_1_2
and LP_Memb_3_plus) equi e some addi ional
explana ion: In he ques ionnai e, esponden s
we e asked wo ques ions ela ed o he quan i y
o LPs. Fi s , esponden s p o ided he o al
numbe o LPs hey a e membe s o . The i s
ques ion se ed mos ly as a lead-in o he nex
ques ion, whe e indi iduals epo ed hei ac ual
(ac i e) LP pa icipa ion. In his a icle, we ocus
on ac i e LP usage a he han a simple LP
membe ship as we i nd ac i e LP pa icipan s
o be a a mo e in e es ing g oup as a as
LP o ganize s a e conce ned. Al hough he
esponden s we e choosing om i e op ions
(0 LPs ac i ely used, 1-2, 3-4, 5-9, 10+), low
obse ed quan i ies in he las wo ca ego ies
led us o combine he las h ee LP ca ego ies
in o a single ca ego y o 3+ LPs ac i ely
used. Hence, we include LP_Memb_1_2 and
LP_Memb_3_plus a iables in equa ion (1) o
desc ibe he e ec o ac i e LP pa icipa ion, while
consume s using ze o LPs o m a base ( e e ence)
ca ego y. The wo LP- ocused a iables discussed
in his pa ag aph a e indispensable o answe ing
he esea ch ques ion 2. In his con ex , i is wo h
men ioning ha he inclusion o bo h a iables
is based on he a iable selec ion ( eg esso
impo ance e alua ion) p ocess desc ibed abo e
– i.e. we didn’ ha e o “ o ce” he wo a iables
in o equa ion (1) in o de o allow o answe ing
esea ch ques ion 2.
The logis ic unc ion used o es ima ion
o he βj coe i cien s in equa ion (1) may be
exp essed as
P(yi = 1 | xi
T) = G(xi
Tβ) =
= exp (xi
Tβ)/[1 + exp(xi
Tβ)], (2)
whe e P(yi = 1 | xi
T) is he p obabili y o success
(i.e. consume willingness o p o ide he
speci i c ype o his/he pe sonal da a), gi en he
obse ed ow ec o o explana o y a iables
xi
T. The exp ession G(xi
Tβ) is a simpli i ed
no a ion o he logis ic unc ion: exp (xi
Tβ)/
[1 + exp(xi
Tβ)] which gua an ees ha all i ed
alues o he dependen a iable lie wi hin he
0, 1 in e al. In model (2), he di ec ion o he
e ec o change in he explana o y a iable xj
on he p obabili y o “success” in he dependen
a iable is always de e mined by he sign o
he co esponding βj coe i cien . Howe e , he
magni udes o he indi idual βj coe i cien s a e
no su i cien ly in o ma i e, gi en he nonlinea
na u e o he logis ic unc ion. The e ec o
a change in xj on he p obabili y o “success”
o he i- h esponden mus be calcula ed
indi idually: as can be seen om equa ion (2),
a change in condi ional p obabili y o success
is calcula ed om a compound unc ion ha
depends on he ollowing a gumen s: βj, all he
emaining coe i cien s in ec o β and all he
obse ed alues o he explana o y a iables
o he i- h esponden (xi
T). Hence, o he i- h
esponden and a chosen bina y explana o y
a iable, say xk, he pa ial e ec om changing
xk om 0 o 1 (while holding all o he explana o y
a iables unchanged) may be simply calcula ed
as
∆G ( . ) = G (β0+ β1x1,i + ... + βk–1,ixk–1,i + βk )
–G (β0+ β1x1,i + ... + βk–1,ixk–1,i )
(3)
EM_1_2017.indd 191EM_1_2017.indd 191 13.3.2017 16:59:1313.3.2017 16:59:13

192 2017, XX, 1
Ma ke ing a obchod
In exp ession (3), we may no e ha he βk
coe i cien is p esen when G(.) is e alua ed
o xk = 1 and omi ed o xk = 0. Fo each
indi idual consume , he exp ession (3) may be
used o e alua ion o changes in condi ional
success p obabili ies. Howe e , o e ec i e
model in e p e a ion, we need o summa ize he
indi idual in o ma ion ob ained om equa ion
(3) ac oss all indi iduals. This is done h ough
he a e age pa ial e ec (APE) s a is ics:
]
,
G(^
β0 + ^
β1x1,i + ... +
–G(^
β0 + ^
β1x1,i + ...
∑
[
APE(xk ) = n–1
+ ^
βk–1,ixk–1,i + ^
βk )
... + ^
βk–1,ixk–1,i )
n
i–1
(4)
whe e ^
βj a e he sample es ima es o βj
coe i cien s. In equa ion (4), he expec ed pa ial
e ec o changing a gi en bina y eg esso xk
om 0 o 1 (ce e is pa ibus) is ob ained o each
o he su ey esponden s and hen a e aged
ac oss indi iduals. Using his app oach, APE(xk)
alues a e usually epo ed along wi h hei
co esponding s anda d e o s and signi i cance
s a is ics. Using exp ession (4), consis en APEs
may be calcula ed o all bina y eg esso s xj in
model (1). Al hough all eg esso s in model (1)
a e bina y, he speci i ca ion chosen p o ides
enough con ol o di e se obse ed ac o s
ha i allows o a s aigh o wa d in e p e a ion
o indi idual APEs; a si ua ion ha is analogous
o he Igno abili y o ea men assump ion (as
in Woold idge (2010, p. 908)).
4. Empi ical Resul s
Indi idual willingness o disclose pe sonal
in o ma ion o LP o ganize s is app oached
using di e se da a e alua ion me hods in o de
answe RQs 1 and 2 and o quan i y impo an
sociodemog aphic and li es yle aspec s o
his ype o consume beha iou . Fi s , able 1
summa izes he o e all consume eadiness
o sha e pe sonal da a. Rows a e o ganized
in descending o de and we may see ha
he e a e p ominen di e ences in pe sonal
in o ma ion sha ing p e e ences ela ed o
he ype o da a. The willingness o sha e
pe sonal da a anges om 91.3% (Name &
Su name) o 3.7% (Pe sonal ID numbe ). This
ange be ween he wo benchma k da a ypes
(as desc ibed abo e) se s an in e p e a ion
amewo k o he emaining da a ypes. In ac ,
as we es o s a is ically signi i can di e ences
in obse ed means ( ela i e “success” a ios,
i.e. he a e age willingness o sha e a speci i c
ype o da a), we i nd ha all mean pai s a e
s a is ically di e en (6 ca ego ies make o 15
possible ca ego y-pai s), wi h he excep ion
o Bi hda e and Phone numbe – means
o hose wo ca ego ies a e no s a is ically
di e en a any easonable signi i cance le el.
Such e alua ion is based on he Wilcoxon
signed ank es ( o ma ched/co ela ed
pai s). Fo de ailed desc ip ion o he es , see
Wacke ly e al. (2008). The ow o de ing and
he posi i e ou come a ios shown in able 1
p o ide an in e es ing insigh in o indi idual
p i acy p e e ences and he gene al a i ude
o consume s o pe sonal da a disclosu e. Fo
example, we may poin ou he qui e low le el
o p epa edness o sha e phone numbe s, ha
Pe sonal in o ma ion ype
& consume willingness
o sha e i
No. o „Successes“
(ou o 402 esponden s)
Posi i e ou come
a io Va iance
Name & Su name 367 0.913 0.080
Email 260 0.647 0.229
Add ess/ esidence 227 0.565 0.246
Bi hda e 156 0.388 0.238
Phone numbe 149 0.371 0.234
Pe sonal ID numbe 15 0.037 0.036
Sou ce: own
Tab. 1: Obse ed willingness o sha e di e en ypes o pe sonal in o ma ion
EM_1_2017.indd 192EM_1_2017.indd 192 13.3.2017 16:59:1313.3.2017 16:59:13
193
1, XX, 2017
Ma ke ing and T ade
consume s p e e o keep undisclosed mo e
han o he con ac in o ma ion such as Email and
Add ess/ esidence. Mos p obably, his a i ude
is ela ed o he a oidance o ad e isemen
oice calls and ex messages, which may be
pe cei ed as a mo e in usi e when compa ed
o emails o pape -based lea l e s. Dele ing
i ele an e-mails is less bo he ing han dealing
wi h unsolici ed phone calls and ex messages
(e.g. Blackbu n (2015)).
The a iance in o ma ion in able 1 is
p o ided mos ly o eade s’ con enience.
Gi en he binomial na u e o he unde lying
dummy a iable ep esen ing willingness o
unwillingness o sha e pe sonal da a, a iance
equals p(1 – p), whe e p is he Posi i e ou come
a io. Table 1 p o ides easonable o e all
insigh in o consume s’ pe sonal da a sha ing
p e e ences. Ye , in o de o answe RQ1 and
RQ2, we need o ocus on he sociodemog aphic
and li es yle aspec s. A he indi idual le el,
many andom elemen s and ac o s play
a signi i can ole in de i ning he ex en o
pe sonal da a sha ing. Howe e , by means o
logis ic eg ession, we a e o en able o quan i y
s a is ically signi i can di e ences in pe sonal
da a sha ing habi s be ween speci i c g oups
o esponden s (ei he sociodemog aphic o
li es yle-based). Such esul s may be p esen ed
in an in o ma i e, easily accessible and o en
ac ionable o m ha may be used o di e se
ma ke ing and LP-managemen pu poses.
Nex , we u n ou a en ion o he
RQ1. Based on he explana o y a iables
(sociodemog aphic and li es yle ac o s),
model (1) was es ima ed o each o he six
dependen a iables (willingness o sha e
di e en pe sonal da a ypes). All es ima ed
models p o ide easonable es ima ion
accu acy and a e s a is ically signi i can a he
5% signi i cance le el – wi h he excep ion o
he model es ima ed o Pe sonal ID numbe ,
which is only signi i can a α = 10%. In logis ic
eg ession, he indi idual coe i cien es ima es
a e no pa icula ly in o ma i e, excep o hei
signs and s a is ical signi i cance. The e o e, we
skip he eg ession ou pu om equa ion (1)
and ocus on he APE alues as de i ned in (4).
In ac , all signs and s a is ical signi i cances o
he βc,j coe i cien s a e unambiguously e l ec ed
in he co esponding APEc(xj), whe e xj is
he j- h explana o y a iable and subsc ip c
deno es he c- h ype o pe sonal in o ma ion
– a dependen a iable in (1). All logis ic
es ima ion ou pu s omi ed om his a icle a e
a ailable om he au ho s upon eques , along
wi h p ima y da a and he R-code used. In
able 2, all APEc(xj) alues a e epo ed along
wi h hei s anda d e o s (he e oscedas ici y
co ec ed alues) and p- alues. Columns in
able 2 a e o ganized by he o e all consume
willingness o sha e pe sonal da a in he same
descending o de as in able 1 – hus allowing
o a simple compa ison.
Fo in e p e a ion o he sociodemog aphic
and li es yle ac o s ela ed o RQ1 – as shown
in able 2 – we shall s a wi h he a iable
Female as an example: Women a e oughly
10% less likely o sha e hei email add ess
and phone numbe wi h LP o ganize s when
compa ed o he e e ence ca ego y (i.e.
men). Fo he emaining ou pe sonal da a
ypes, gende plays no signi i can ole. Such
in e p e a ion is made in a ce e is pa ibus
con ex – we con ol o all he emaining
a iables explici ly included in model (1).
In e es ingly, ou i ndings somewha con adic
a common s e eo ype ha ega ds women as
mo e likely o conceal hei age. This conclusion
is implied h ough he dependen a iable
Bi hda e, o which esponden ’s gende is
no a signi i can eg esso . Fo illus a ion
and eade s’ con enience, he ce e is pa ibus
e ec s on willingness o sha e pe sonal da a
ela ed o he eg esso Female a e included in
i gu e 1 (along wi h co esponding ba s showing
90% signi i cance in e als). Fo example, he
le mos ba (wi hin he Female g oup) shows
ha women a e 3.5% less likely o sha e hei
name wi h LP o ganize s. A he same ime,
he co esponding 90% signi i cance in e al
includes ze o and he e o e his pa icula
e ec is no s a is ically signi i can a α = 0.1
(signi i cance le el o 10%).
The emaining sociodemog aphic ac o s
in l uencing consume willingness o sha e
pe sonal da a wi h LP o ganize s may be b ie l y
summa ized as ollows: F om able 2 we can
see ha people aged 15 o 24 a e 21% mo e
likely o sha e hei bi hda e compa ed o he
age- e e ence g oup o people aged 25 o
64. Fo o he pe sonal da a ypes, signi i can
in l uence o he a iable Age_15_24 is no
iden i i ed. In con as , esponden s aged 65
and olde di e ge om he age- e e ence g oup
a he signi i can ly. They a e almos 28% less
likely o sha e hei email add esses (no e ha
he non-use o he In e ne is con olled o
EM_1_2017.indd 193EM_1_2017.indd 193 13.3.2017 16:59:1313.3.2017 16:59:13
194 2017, XX, 1
Ma ke ing a obchod
Name &
Su name Email Add ess Bi hda e Phone
numbe
Pe sonal
ID numbe
Female
( s.e. )
[ p- alue ]
-0.0352
( 0.0269 )
[ 0.1904 ]
-0.1032 *
( 0.0407 )
[ 0.0112 ]
-0.0014
( 0.0490 )
[ 0.9779 ]
0.0464
( 0.0483 )
[ 0.3371 ]
-0.1015 *
( 0.0489 )
[ 0.0378 ]
0.0047
( 0.0236 )
[ 0.8428 ]
Age_15_24
-0.0144
( 0.0600 )
[ 0.8108 ]
0.0841
( 0.0675 )
[ 0.2130 ]
0.0393
( 0.0743 )
[ 0.5970 ]
0.2106 *
( 0.0751 )
[ 0.0050 ]
0.0820
( 0.0747 )
[ 0.2724 ]
0.0281
( 0.0361 )
[ 0.4364 ]
Age_65_plus
-0.0582 ˙
( 0.0310 )
[ 0.0606 ]
-0.2792 *
( 0.0628 )
[ 0.0000 ]
-0.1520 *
( 0.0624 )
[ 0.0148 ]
-0.0306
( 0.0627 )
[ 0.6261 ]
0.1186 ˙
( 0.0629 )
[ 0.0593 ]
0.0085
( 0.0283 )
[ 0.7629 ]
Mo a ia
-0.0179
( 0.0299 )
[ 0.5503 ]
-0.0839 ˙
( 0.0446 )
[ 0.0600 ]
-0.0431
( 0.0489 )
[ 0.3774 ]
-0.0865 ˙
( 0.0476 )
[ 0.0693 ]
-0.0094
( 0.0475 )
[ 0.8436 ]
0.0300
( 0.0244 )
[ 0.2181 ]
Ea nings_high
-0.1349 ˙
( 0.0689 )
[ 0.0503 ]
-0.0443
( 0.1065 )
[ 0.6775 ]
-0.2708 *
( 0.1085 )
[ 0.0126 ]
-0.0050
( 0.1006 )
[ 0.9607 ]
-0.1118
( 0.0938 )
[ 0.2333 ]
-0.0383 *
( 0.0093 )
[ 0.0000 ]
LS_TV_no
-0.0863
( 0.0586 )
[ 0.1409 ]
-0.0839
( 0.1038 )
[ 0.4187 ]
0.0423
( 0.1037 )
[ 0.6833 ]
0.0203
( 0.1255 )
[ 0.8716 ]
0.0019
( 0.1118 )
[ 0.9866 ]
-0.0385 *
( 0.0094 )
[ 0.0000 ]
LS_books_no
0.0949 *
( 0.0131 )
[ 0.0000 ]
0.1755 *
( 0.0700 )
[ 0.0121 ]
0.1460 ˙
( 0.0873 )
[ 0.0943 ]
0.3214 *
( 0.0836 )
[ 0.0001 ]
0.1028
( 0.0960 )
[ 0.2844 ]
0.0475
( 0.0406 )
[ 0.2418 ]
LS_In e ne _use_no
-0.0828
( 0.0529 )
[ 0.1174 ]
-0.3343 *
( 0.1320 )
[ 0.0113 ]
-0.0191
( 0.0886 )
[ 0.8290 ]
-0.0393
( 0.0915 )
[ 0.6671 ]
-0.2538 *
( 0.0782 )
[ 0.0012 ]
-0.0155
( 0.0207 )
[ 0.4535 ]
LS_Payca d_yes
-0.0361
( 0.0311 )
[ 0.2464 ]
0.1170 *
( 0.0439 )
[ 0.0076 ]
-0.0158
( 0.0508 )
[ 0.7555 ]
-0.0069
( 0.0502 )
[ 0.8908 ]
0.1083 *
( 0.0514 )
[ 0.0353 ]
-0.0276
( 0.0224 )
[ 0.2193 ]
LS_exo ics_yes
0.0424
( 0.0292 )
[ 0.1466 ]
0.0337
( 0.0525 )
[ 0.5205 ]
0.1086 ˙
( 0.0583 )
[ 0.0628 ]
0.1023 ˙
( 0.0581 )
[ 0.0786 ]
0.0647
( 0.0575 )
[ 0.2608 ]
-0.0206
( 0.0198 )
[ 0.2962 ]
LS_cooking_no
0.0103
( 0.0287 )
[ 0.7187 ]
-0.1707 *
( 0.0805 )
[ 0.0339 ]
-0.0290
( 0.0725 )
[ 0.6886 ]
-0.1575 *
( 0.0768 )
[ 0.0404 ]
0.0776
( 0.0808 )
[ 0.3371 ]
0.0073
( 0.0274 )
[ 0.7890 ]
LP_Memb_1_2
0.1254 *
( 0.0230 )
[ 0.0000 ]
0.1015 *
( 0.0422 )
[ 0.0162 ]
0.1450 *
( 0.0511 )
[ 0.0046 ]
0.0428
( 0.0583 )
[ 0.4629 ]
0.1807 *
( 0.0534 )
[ 0.0007 ]
-0.0294
( 0.0199 )
[ 0.1405 ]
LP_Memb_3_plus
0.1555 *
( 0.0237 )
[ 0.0000 ]
0.2412 *
( 0.0484 )
[ 0.0000 ]
0.3267 *
( 0.0548 )
[ 0.0000 ]
0.1735 *
( 0.0643 )
[ 0.0069 ]
0.3548 *
( 0.0603 )
[ 0.0000 ]
0.0577
( 0.0418 )
[ 0.1680 ]
Sou ce: own
No e: * – coe i cien signi i can a α = 0.05; ˙ – coe i cien signi i can a α = 0.1.
Tab. 2: APEs o selec ed ypes o pe sonal da a
EM_1_2017.indd 194EM_1_2017.indd 194 13.3.2017 16:59:1313.3.2017 16:59:13
195
1, XX, 2017
Ma ke ing and T ade
by a sepa a e li es yle explana o y a iable
LS_In e ne use_no), 15% less likely o di ulge
hei add ess/ esidence and also 6% less likely
o p o ide hei names. On he o he hand,
membe s o he age g oup 65+ a e abou 12%
mo e likely o hand o e hei phone numbe s
when compa ed o he e e ence. The e e ence
g oup exhibi s a a he uni o m beha io in
e ms o pe sonal da a sha ing p e e ences.
Hence, i is no con enien o s udy he age
anges 25-34, 35-49 and 50-64 indi idually: we
combine hem in o a single 25-64 e e ence and
s udy how he younge and olde consume s di e .
Responden domiciled in Mo a ia a e
oughly 8.5% less likely o sha e hei email
and bi hda e compa ed o he e e ence
g oup (Bohemia combined wi h he sepa a ely
su eyed egion o P ague). Consume s wi h
high ea nings a e 27% less likely o p o ide LP
o ganize s wi h hei add ess/ esidence, mos
p obably as a secu i y p ecau ion. Simila ly,
hey a e 13.5% less likely o p o ide hei name.
Educa ion (su eyed as p ima y/seconda y/
uni e si y deg ee) does no ha e a signi i can
e ec on indi idual willingness o sha e di e en
ypes o pe sonal da a, once he o he ac o s
in model (1) a e con olled o . The e o e,
educa ion- ela ed a iables a e omi ed om
equa ion (1) and om able 2.
Nex , we b ie l y summa ize he RQ1- ela ed
li es yle ac o s ha in l uence consume s’
willingness o sha e da a mos p ominen ly.
Gene ally speaking, people who dissocia e
hemsel es om eading books (LS_books_
no equals 1) a e mo e likely o p o ide hei
pe sonal da a o LP o ganize s when compa ed
o he e e ence g oup (ac i e book eade s
combined wi h people wi hou a s ong
posi ion on his opic). The di e ence is mos
p ominen o he Bi hda e dependen a iable
– we obse e a 32% inc ease in p obabili y.
Responden s who dissocia e hemsel es om
using he In e ne (LS_In e ne use_no) a e less
likely o sha e hei pe sonal da a: we obse e
a dec ease o 33% in willingness o sha e email
( o ob ious easons, people who dissocia e
hemsel es om using he In e ne a e less
likely o ac ually ha e an email add ess), a 25 %
dec ease in likelihood o sha ing phone numbe
and e en an 8 % dec ease in willingness o
sha e name (howe e , his esul na owly
misses s a is ical signi i cance a α = 0.1).
In con as , a s a is ically signi i can e ec o
LS_TV_no is obse ed only o he Pe sonal
ID numbe (a dec ease o 3%). Consume s who
epo being ac i e payca d use s (42% o he
esponden s s ongly iden i y hemsel es wi h
using a c edi o debi ca d as measu ed by he
Like scale-based a iable LS_Payca d_yes)
a e 12% mo e likely o sha e email add ess and
11% mo e likely o sha e phone numbe wi h
LP o ganize s, while he e ec o his li es yle
ac o is no s a is ically signi i can o he o he
ou ypes o pe sonal da a. Fo illus a ion, he
a iable LS_Payca d_yes is also included in
i gu e 1.
Fig. 1: Illus a ion o selec ed esul s om able 2
Sou ce: own
EM_1_2017.indd 195EM_1_2017.indd 195 13.3.2017 16:59:1313.3.2017 16:59:13