Beware of the Woozle effect and belief perseverance in the PLS-SEM literature!

Hensele , Jö g e al.
A icle — Published Ve sion
Bewa e o he Woozle e ec and belie pe se e ance in he
PLS-SEM li e a u e!
Elec onic Comme ce Resea ch
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: Hensele , Jö g e al. (2024) : Bewa e o he Woozle e ec and belie pe se e ance
in he PLS-SEM li e a u e!, Elec onic Comme ce Resea ch, ISSN 1572-9362, Sp inge US, New Yo k,
NY, Vol. 24, Iss. 2, pp. 715-744,
h ps://doi.o g/10.1007/s10660-024-09849-y
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/315229
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)
Elec onic Comme ce Resea ch (2024) 24:715–744
h ps://doi.o g/10.1007/s10660-024-09849-y
1 3
Bewa e o  heWoozle e ec andbelie pe se e ance
in hePLS‑SEM li e a u e!
Jö gHensele 1,2 · NickLee3· EllenRoeme 4· IldikóKemény1,5·
TaşkınDi sehan6,7· JohnW.Cadogan8,9,10
Accep ed: 1 Ap il 2024 / Published online: 25 June 2024
© The Au ho (s) 2024
Abs ac
Pa ial leas squa es s uc u al equa ion modeling (PLS-SEM) is a popula me hod o
da a analysis in ma ke ing, in o ma ion sys ems esea ch, and ela ed ields. Un o u-
na ely, he li e a u e on PLS-SEM con ains a numbe o miss a emen s ha asc ibe cha -
ac e is ics o PLS-SEM ha i does no possess. In ou s udy, we conside why hese mis-
s a emen s con inue o be made and ein o ced. We show how inaccu acies, omissions,
epe i ions o pas misconcep ions, and he in oduc ion o addi ional misconcep ions
lead o he so-called Woozle e ec . As an example, we use pe haps he mos se ious mis-
concep ion abou PLS-SEM namely i s alleged sui abili y o es ima ing he pa ame e s
o e lec i e measu emen models. The Woozle e ec is a cumula i e p ocess by which
alsehoods become es ablished as ac , and he sel -co ec ing mechanisms o science a e
suspended. Th ough a li e a u e e iew, we iden i y a numbe o ac o s ha a e likely o
ha e con ibu ed o he Woozle e ec in he PLS-SEM li e a u e. Fo he Woozle e ec
o disappea , esea che s need o acknowledge ha PLS-SEM, in i s cu en inca na ion,
is no sui able o scien i ic wo k wi h e lec i e measu emen models.
Keywo ds Re lec i e measu emen · Composi e model· Measu emen e o ·
Consis en PLS· Scien i ic sel -co ec ion· Belie pe se e ance
1 In oduc ion
Pa ial leas squa es s uc u al equa ion modeling (PLS-SEM) is a me hod ha
has gained inc easing popula i y in ma ke ing, in o ma ion sys ems, and ela ed
ields o business esea ch [c . 39].1 I is widely cha ac e ized as a echnique o
Ex ended au ho in o ma ion a ailable on he las page o he a icle
1 We ecognise ha he e is some deba e on whe he i is co ec o e e o PLS-SEM as a s uc u al
equa ion modeling echnique [88, 90]. We ag ee wi h Rönkkö e al. ha , wha e e i s echnical co ec -
ness, he use o he PLS-SEM e m has been a ne nega i e because i has misled esea che s o inco -
ec ly assume ha PLS-SEM and co a iance-based s uc u al equa ion modeling me hods (e.g. such
as ha implemen ed in LISREL [63], la aan [91] o Mplus [78]) a e in e changeable. Add essing his
716
J.Hensele e al.
1 3
causal-p edic i e esea ch [11]. Resea che s use i as a ool o es ima ing he ela-
ionships be ween la en a iables, i.e., unobse ed a iables ha we e only indi-
ec ly measu ed by means o a se o indica o a iables [45].
Un o una ely, he PLS-SEM li e a u e con ains a numbe o misconcep ions,
asc ibing cha ac e is ics o PLS-SEM ha i does no possess [8, 37, 88–90, 97, 98,
101, 102]. P obably he mos se e e misconcep ion is PLS-SEM’s alleged sui abil-
i y o es ima e he pa ame e s o e lec i e measu emen models. Al hough many
au ho s claim ha PLS-SEM can handle e lec i e measu emen models [c . 34, 45],
his claim has ne e been suppo ed by induc i e o deduc i e easoning, and i has
been known o mo e han ou decades ha i is ac ually alse [17]. To be clea :
PLS-SEM is no sui able o es ima ing e lec i e measu emen models. I es ima es
he pa ame e s o composi e models, no e lec i e measu emen models. [see e.g.
16, 18, 54, 94]. I analys s in e p e he PLS-SEM ou pu as he coe icien s o a
e lec i e measu emen model, hey will end up wi h biased es ima es ha “may be
wide o he ma k” [16, p.51]. As a esul , analys s using PLS-SEM may ind e ec s
ha do no exis [20, 100], may no ind e ec s ha do exis [97, 101], o may e en
ind e ec s o he opposi e sign [51, 98].
While many s udies ha e al eady shown ha he e a e misconcep ions in he
PLS-SEM li e a u e [c . 37, 88, 89, 90], he ques ion emains why do hese miscon-
cep ions exis in he i s place, and why do hese misconcep ions con inue o be dis-
semina ed and pe sis in he li e a u e?
In his pape , we p opose a mechanism ha has al eady se ed o he domains o
social science o explain he occu ence and dissemina ion o misconcep ions: he
so-called Woozle e ec [35]. In he emainde o ou a icle, we make eade s awa e
o he Woozle e ec and i s equen companion, belie pe se e ance. To illus a e,
we zoom in on he misconcep ion ha PLS-SEM is sui able o es ima ing e lec-
i e measu emen models and iden i y con ibu ing ac o s ha acili a e he eme -
gence o he Woozle e ec . One implica ion o his analysis is ha esea che s using
PLS-SEM o e lec i e measu emen models a e ope a ing as i scien i ic-appea ing
bu alse claims we e ue, a cha ac e is ic o pseudoscience [81]. Awa eness o he
mechanisms a hand can help p e en o he Woozle e ec s in he u u e.
2 Backg ound
2.1 The Woozle e ec
The phenomenon e e ed o as he Woozle e ec , also known as e idence by
ci a ion, a ises when (a) a sou ce, o sou ces, make claims ha a e un ounded o
ha , a bes , do no ha e su icien e idence o suppo hem, and (b) he claims
deba e is ou side he speci ic scope o his a icle. Ne e heless, eade s should unde s and ha we con-
inue o use he e m PLS-SEM he e o consis ency wi h he ele an li e a u e only, and ou use he e
should no be aken as an endo semen o he use ulness o he e m.
Foo no e 1 (con inued)
717
1 3
Bewa e o  heWoozle e ec andbelie pe se e ance in hePLS‑SEM…
gain c edibili y me ely because hey a e equen ly ci ed, no because hey a e
ue. The e m was coined by Be e ly D. Hough on in 1979 a an annual mee -
ing o he Ame ican Socie y o C iminology based on a Winnie- he-Pooh s o y
[35]. A.A. Milne, in his belo ed child en’s book Winnie- he-Pooh, desc ibes
how Winnie and Pigle , while belie ing hey a e hun ing inc easing numbe s o
an imagina y c ea u e called a Woozle, a e ac ually walking in ci cles, ollowing
he acks hey hemsel es ha e le behind (as isualized in Fig.1). The la e ,
which hey belie e a e woozle acks, p o ide hem wi h all he e idence hey
need o suppo and s eng hen hei belie in he exis ence o he woozle. Gelles
and S aus [36] use he Woozle s o y o illus a e how poo p ac ice in esea ch
and sel - e e en ial s udies cause e o and bias. An unchallenged ci a ion o he
sou ce hiding compe ing well-a gued and quali ied iews wi h ph ases such as
“E e y one knows
…
”, “I is clea ha
…
”, “I is ob ious ha
…
”, “I is gene -
ally ag eed ha
…
” gi es way o he Woozle e ec ; i we hea some hing o en,
we assume i is ue [32]. The s a emen s a e made mo e ce ain han he o iginal
au ho in ended by omi ing essen ial quali ie s om he o iginal a icle [21]. The
Fig. 1 Pigle and Winnie go in ci cles hun ing a woozle—bu he acks hey ollow a e me ely hei own
(illus a ion by E nes Howa d Shepa d, 1926)
718
J.Hensele e al.
1 3
Woozle e ec can also be caused by he ailu e o ace e e ences o hei o iginal
sou ce in a esea ch s udy [64].
To sum up, he Woozle e ec is a cogni i e bias ha can cause indi iduals o
belie e a s a emen o con ic ion o be accu a e because o epea ed ci a ion a he
han because hey a e p esen ed wi h ac ual e idence o c i ical analysis. Acco ding
o Du on [21], he Woozle e ec is a mani es a ion o con i ma ion bias and is asso-
cia ed wi h belie pe se e ance and g oup hink. Woozling is he p ocess by which
esea ch is mis ep esen ed, c ea ing my hs and misconcep ions [79].
2.2 Belie pe se e ance
In he o iginal ex , Winnie- he-Pooh lea ns om Ch is ophe Robin ha he has ci -
cled a bush se e al imes, and hus ealizes he u h, which is ha he has only been
ollowing his own acks. Un o una ely, in he p ac ice o esea ch, a i ing a he
igh insigh is no gua an eed. Ra he , e en when esea che s challenge p io s a e-
men s and ci a ions, and e en when hey p o ide empi ical e idence o laws in he
esea ch p ocess, belie s may con inue o p e ail o e all doub . This is he s age o
belie pe se e ance.
Belie pe se e ance can be in e p e ed as a o m o a esea che ’s subjec i e bias
ha is po en ially in ol ed in all phases o he esea ch p ocess. Managemen sci-
ence, jus as many o he sciences, has been la gely c i icized o such subjec i e bias
ha “is sys ema ically in oduced in o esea ch indings h ough he pa icula ana-
ly ical pe spec i es, me hodologies, and alue assump ions ha esea che s choose
o impose upon da a and injec in o heo e ical in e p e a ions, and ha such bias
dis o s empi ical eali y, c ea i ely ans o ming i h ough he pe cep ual il e ing
which is an ine adicable pa o he esea ch p ocess” [2, p.260].
This subjec i e bias, ei he conscious o unconscious, can lead o a ne e -ending
and sel - ein o cing cycle as o he esea che s begin o ci e biased p ocesses and
esul s. Ci a ion ci cles eme ge ein o cing and epea ing known o unknown e o s.
Ul ima ely, o example, cons ain s o assump ions a e de ined away and neglec ed
so ha e o and bias becomes an inc emen al pa o he esea ch. Belie is ans-
o med in o pseudo- ac s.
I he Woozle e ec emains unconscious om esea che s’ poin s o iew, hose
esea che s will jus con inue he biased loop wi hou ecognizing hei e o , hus
s eng hening he sel -en o cemen e ec .
In con as , when esea che s become awa e o hei e o , di e en op ions
eme ge. Di e en heo e ical amewo ks seem o be use ul o analyse he beha -
io al al e na i es o esea che s becoming awa e o e o . Real op ions heo y
analyses he alue o di e en op ions when indi iduals ha e o make i e e s-
ible in es men s (such as ime and e o ) unde unce ain y (such unce ain y o
esea ch esul s and publica ions). Real op ions include he op ions o g ow o
con ac , o de e , swi ch o abandon an in es men [109]. T ans e ed o esea ch-
e s’ op ions, PLS-SEM so wa e de elope s’ and li e a u e-based ad oca es’
commi men s o he app oach in ol e pe sonal in es men s and sel -sac i ices
[73] comp ising in insic and ex insic esou ces such as emo ions, in ellec ual

719
1 3
Bewa e o  heWoozle e ec andbelie pe se e ance in hePLS‑SEM…
endea o , ime, and social s a us [69]. Resea che s aced wi h e idence ha
con adic s hei exis ing belie s abou ce ain aspec s o hei me hodological
app oach may: choose o be o e ly dismissi e o he ideas and da a upon which
he e idence s ands [79] and engage in de ensi e ac i i y and u he commi -
men s o he esea ch app oach, o ; hey may educe hei commi men o ha
esea ch pa h, de e commi men o he pa h, o exi and swi ch o a new pa h.
Al e na i ely, Hi schman’s exi , oice, and loyal y amewo k [56] may se e
as a u he guideline o he analysis. When indi iduals obse e a dec ease in
quali y ega ding an o ganiza ion, a poli ical sys ems o any o he g ouping, such
as esea ch g oups, he indi idual’s op ions a e o exi om his g oup, o oice
dissa is ac ion o po en ial change o o s ay loyal and o u he suppo he
g oup.
Based on hese wo amewo ks, we subsequen ly discuss he di e en beha -
io al op ions o esea che ’s eac ions when ecognizing e o acco ding o hei
likelihood o occu ence:
1. s ay loyal and o u he in es o commi o he esea ch pa h;
2. de e exi om he esea ch pa h un il mo e in o ma ion is a ailable, wi h a pos-
sible swi ch o a new pa h in he u u e;
3. imely exi and swi ch o a new pa h.
Usually, scien is s a e e y commi ed o speci ic heo ies, me hods o posi ions,
since o he wise hey may be dis ega ded o igno ed by hei pee s in he scien-
i ic communi y [77]. The e o e, a e y likely i s op ion o esea che s ecog-
nizing an e o is he conscious pu sui o he Woozle cycle. They would s ay
loyal o he esea ch g oup and u he commi hei esea ch owa ds hei p e-
ious pa h– e en i an e o has been ecognized. In his sense [77, p.B-614;
emphasis in he o iginal] s a es: “Thus,
…
he scien is may conside ‘ a ional’
no o gi e up his a o ed heses a he i s signs o ‘nega i e’ e idence, no ma -
e how s ong ha e idence may appea a he ime. Indeed, he may e en pe sis
in his scien i ic belie s o yea s in he ace o conside able opposi ion”. As a
consequence, belie pe se e ance may occu (wha [68] desc ibes as pa hologi-
cal science). As esea che s a e social beings, belie pe se e ance may e en be
s eng hened i a g oup o esea che s ein o ces biased esea ch [110]. Resea ch-
e s’ desi e o ha mony and con o mi y could hen lead o g oup hink [61] sup-
po ing he decision o con inue he Woozle pa h.
A second op ion could be he de e al o he decision o exi om he Woozle
cycle. In his way, a esea che gains mo e ime o lea e he exis ing pa h o pa a-
digm when e o s a e acknowledged. When esea che s acknowledge e o s, i is
unlikely ha hey will immedia ely enounce an ea lie pa h, bu hey will s a
looking o al e na i es. Once such an al e na i e is a ailable, hey may decla e
ea lie knowledge as in alid [67]. In his way, Kuhn [67] sugges s a ansi ion
pe iod om one pa adigm in c isis o a new pa adigm. The sea ch o al e na i es
will ake ime, so ha he exi decision om he Woozle pa h is de e ed un il a
new pa h is ound.
720
J.Hensele e al.
1 3
A hi d possibili y, which is he mos desi able om a scien i ic poin o iew,
would be o esea che s o acknowledge hei e o s (e.g., in heo ies, me hods,
and/o esul s) and o exi he exis ing pa h in o de o sea ch o a new, less e one-
ous pa h o science. Howe e , Mi o [77, B-614] ound in his s udy ha an o e -
whelming esponse ega ding how esea che s ac ually conduc hei esea ch was
ela ed o esea che s’ own commi men s. He s a es “a scien is has o be ‘commi -
ed’ o (and some imes e en ‘biased’ in a o o ) his a o i e heo y, pe hypo hesis
o posi ion i i is o be gi en a ai hea ing by he scien i ic communi y” [B-614 77
emphasis in he o iginal]. The e o e, his op ion is p obably he leas likely op ion
and p obably he mos di icul one om an indi idual esea che ’s poin o iew.
The la e wo op ions a e likely o be accompanied by “ oice”, i.e., wi h a deba e
on he quali y o he cu en esea ch pa h and simul aneous sea ch o solu ions o
change he si ua ion. As Kuhn [67, 90–91] pu s i : “Con on ed wi h anomaly o
wi h c isis, scien is s ake a di e en a i ude owa ds exis ing pa adigms, and he
na u e o hei esea ch changes acco dingly. The p oli e a ion o compe ing a ic-
ula ions, he willingness o y any hing, he exp ession o explici discon en , he
ecou se o philosophy and o deba e o e undamen als, all hese a e symp oms o
a ansi ion”.
2.3 Mo ing omscience opseudoscience
The Woozle e ec and belie pe se e ance—pa icula ly i hey occu in combina-
ion—a e a om ha mless o scien i ic p og ess. Kuhn [31] associa es he Woo-
zle e ec wi h he mechanisms ha gene a e pseudoscience. Pseudoscience sha es
p ocedu es, no ms, and habi us wi h science, bu does no con ibu e o knowledge
gene a ion, because i is decoupled om he sea ch o u h o u ili y.
Figu e2 illus a es a pa h owa d pseudoscience made up o h ee phases: ini-
ializa ion o he Woozle e ec , Woozle e ec in ull swing, and belie pe se e ance.
Rema kably, his pa h can eme ge wi hou a esea che consciously doing any hing
bad. The poin o depa u e is ypically a no mal s ep in he esea ch p ocess, namely
ha esea che s epo indings accompanied by a quali ica ion. Fo ins ance, his
could be s a emen in he o m “Finding F holds unde Condi ion C.”
A Woozle e ec can be ini ia ed when subsequen esea che s e e o he ind-
ings, bu wi hou he quali ica ion. Fo ins ance, hey would simply s a e ha “Find-
ing F holds.” Making claims wi hou e idence, commi ing logical e o s, d awing
p ema u e conclusions, omi ing incon enien de ails, o e gene alizing indings, o
e en delibe a ely lying a e o he ways o ini ia e a Woozle e ec . Some imes, he
cause can be poo esea ch p ac ice, bu no always.
The Woozle e ec ge s in o ull swing once he subsequen esea che s a e ci ed
in la e publica ions, and he unsubs an ia ed claims o indings wi hou quali ica ion
gain he s a us o gene alizable “ u h” by i ue o being ci ed in he scien i ic eco d.
An impo an pa o he scien i ic en e p ise and a way o coun e ac he Woo-
zle e ec is he co ec ion o e o . Resea che s do his by aising objec ions and
“se ing he eco d s aigh ” [see 106]. This may mean edoing empi ical s udies
and co ec ing p e ious indings. Howe e , his scien i ic sel -co ec ion can be
721
1 3
Bewa e o  heWoozle e ec andbelie pe se e ance in hePLS‑SEM…
se iously hampe ed i scien is s engage in belie pe se e ance. This means ha hey
con inue o p omulga e he w ong indings despi e he e idence o he con a y. In
i s mos se e e o m, hey e en do his agains hei be e judgmen .
While he igh ack in Fig.2 ul ima ely leads o pseudo-science, esea che s who
s a ed o ollow his ack do no ha e o s ay on his ack. In e e y encoun e wi h
he Woozle e ec , hey ha e he op ion o change hei posi ion and hus swi ch acks.
3 An objec ion: PLS‑SEM does no es ima e e lec i e measu emen
models
A guably, he mos se e e misconcep ion o PLS-SEM is i s alleged sui abili y o
es ima ing e lec i e measu emen models. This misconcep ion seems like a pe ec
ins ance o an es ablished Woozle, in combina ion wi h belie pe se e ance. This
Objec ion
O iginal
inding wi h a
quali ica ion
A esea che epo s indings,
and he o she may p o ide
quali ica ions o he indings.
A second esea che e e s o
he i s esea che , bu wi hou
he quali ica ion.
O he esea che s will hen ci e
bo h epo s, and he indings
wi hou he quali ica ion gain he
s a us o gene alizable “ u h”.
Some esea che s p o ide
concep ual, logical, and/o
empi ical coun e -e idence
agains he indings wi hou he
quali ica ion.
Some esea che s keep on
p omulga ing he indings wi hou
he quali ica ion despi e he
coun e -e idence.
Science
Fig. 2 Mo ing om science o pseudoscience: when belie pe se e ance joins he Woozle e ec
722
J.Hensele e al.
1 3
pa icula Woozle is a he easily shown o be alse, o ins ance by means o algeb a
[c . 16, 20, 101], scena io analyses [c . 51, 97, 98] o Mon e Ca lo simula ions [c .
19, 100, 108]. Howe e , his woozle is so deeply ing ained in he belie s o many
PLS-SEM esea che s and hose who ely on hei esea ch esul s ha a emp ing o
co ec he alsehood seems like some hing o a Sisyphean ask. The woozle is such
an en enched ac oid ha e en when i has been killed by he e idence i e uses o
die: i becomes in essence a ‘zombie Woozle’.2 Owing o he se e i y o his Woozle
e ec , we look a i in mo e de ail.
3.1 Wha PLS‑SEM does
PLS-SEM es ima es composi e models [see he p oo by 18], i.e., i c ea es cons uc
sco es as composi e a iables [71, 107]. Based on a Mon e Ca lo simula ion, also
Hai e al. [44, p.618] conclude ha PLS-SEM is ac ually a “consis en [es ima o ]
o composi e-based models”.
Composi e models ep esen he si ua ion in which a cons uc is made up o i s
obse ed a iables [14, 52]. Fo a cons uc and h ee obse ed a iables, a compos-
i e would ake he o m o Fig.3. The model equa ion would be as ollows [c . 45]:
This composi e model con ains ou a iables: he h ee obse ed a iables
x1
o
x3
and he composi e
𝜉
. Model pa ame e s include he h ee weigh s
w1
o
w3
(nex o
he a iances and co a iances o he obse ed a iables).
Howe e , he e is a la ge body o li e a u e ha akes a di e en s ance by decla -
ing ha PLS-SEM es ima es e lec i e measu emen models [c . 15, 26, 34, 70, 82,
93, 104, 117]. Hai e al. [45, p.16] s a e his wi h c ys al cla i y: “Resea che s can
include e lec i ely and o ma i ely speci ied measu emen models, which PLS-
SEM es ima es wi hou any limi a ions.”
(1)
𝜉=w1
⋅
x1+w2
⋅
x2+w3
⋅
x3
Fig. 3 A composi e model wi h
h ee obse ed a iables
ξ
x2
x1
x3
w1
w2
w3
2 K ugman [66] coined he e m “zombie idea—an idea ha should ha e been killed by e idence, bu
e uses o die.”
729
1 3
Bewa e o  heWoozle e ec andbelie pe se e ance in hePLS‑SEM…
Table 2 (con inued)
S a emen Type o easoning Scien i ic Anecdo al Flawed
“[PLS] Mode A measu emen is associa ed wi h e lec i e cons uc s and Mode B is associa ed wi h o ma i e
cons uc s (Hensele e al., 2009; Tenenhaus e al., 2005).” [4, p. 365]
Ci ing au ho i y
×
“Ad oca es o PLS claim ha i [
…
] can model bo h e lec i e and o ma i e cons uc s wi hin he same
esea ch model.” [82, p. 467]
Unspeci ic e e al
×
“Then, we demons a e PLS using a la ge da ase and a complex s a is ical model wi h o ma i e and e lec i e
indica o s [
…
].” [72, p. 125]
None
×
“PLS-SEM can accommoda e bo h o ma i e and e lec i e measu emen model ypes.” (117, p. xxiii) None
×
“The mos signi ican ad an age o PLS-SEM is ha bo h o ma i e and e lec i e measu emen models can be
speci ied wi h i [
…
].” [15, p. 8]
None
×
“Resea che s can include e lec i ely and o ma i ely speci ied measu emen models, which PLS-SEM es i-
ma es wi hou any limi a ions.” [41, p. 16]
None
×
“PLS pa h models, depending on hei measu emen model speci ica ions, can be classi ied as ei he only e lec-
i e, only o ma i e, o a combina ion o bo h.” [93, p. 1048]
None
×
“While e o e ms a e no usually deno ed in PLS-SEM s uc u al model ep esen a ions, he me hod explici ly
accoun s o measu emen e o inhe en in he i ems (indica o s) by including he assessmen o i em e o in
he measu emen o he co esponding la en composi e (Sa s ed e al., 2016).” [70, p. 97]
Ci ing au ho i y
×

730
J.Hensele e al.
1 3
Table 3 PLS-SEM is no sui able o es ima ing e lec i e measu emen models: he e idence base
S a emen Type o easoning Scien i ic Anecdo al Flawed
“[O]ne can p o e ha es ima es o he pa ame e s, ob ained by s anda d me hods using p oxies o la en
a iables, induce in gene al an es ima e o
Σ
which is no close o i s ue alue e en when S is [
…
]. [
…
]
PLS is no Fishe -consis en .” [17, p. 80]
P oo o a gumen s
×
“PLS makes no adjus men s ega ding measu emen e o o a enua ion.” [65, p. 354] None
×
“The ac o s uc u e is biased [
…
].” [71, p. 107] P oo o a gumen s
×
“An ob ious disad an age o [
…
PLS
…
] is ha [i does] no p o ide a clea es o he pa h hypo hesis, he
e ec o ‘measu emen e o ’ being o in oduce an appa en depa u e om he known causal model.”
[71, p. 266]
Empi ical e idence, p oo
o a gumen s
×
“[PLS] assumes a o ma i e measu emen model.” [62, p. 206] None
×
“[T]he indica o weigh ings by PLS [
…
esul ] in biased es ima es and do no eally con ol o measu e-
men e o .” [87, p. 2]
P oo o a gumen s
×
“Because he la en a iable p oxies in PLS pa h modeling a e weigh ed composi es o e o - ain ed
obse ed a iables, hese p oxies a e hemsel es ain ed wi h e o . As a esul , PLS pa h modeling
pa ame e es ima es a e known o be biased (Chin 1998; Hsu e al. 2006), excep unde a hypo he ical
limi ing condi ion which Wold (1982) named ‘consis ency a la ge.’ This is an un ealis ic si ua ion ha
only occu s when bo h he sample size and he numbe o obse ed a iables a e in ini e.”
P oo o a gumen s, ci ing
e idence
×
“PLS is like eg ession in igno ing measu emen e o in de e mining i s pa h es ima es,” “[S]aying ha
PLS has ‘consis ency a la ge’ is no ac ually a unique o use ul selling poin .” [33, p. i]
P oo o a gumen s
×
“PLS pa h modeling should celeb a e i s s a us as a pu ely composi e-based me hod.” [37, pp. 997–998] P oo o a gumen s
×
“PLS does no es ima e pa h models wi h la en a iables, bu wi h composi es [
…
] PLS does no calcula e
ac o loadings, bu composi e loadings (McDonald, 1996, p. 248).” [88, pp. 433, 437]
Ci ing e idence
×
“[I] he common ac o model i s he da a, esea che s should choose he common ac o model and apply
[
…
] app op ia e es ima ion echniques [o he han PLS].” [54, p. 186]
P oo o a gumen s
×
“[A]s a means o es ima ing he pa ame e s o la en a iable models, PLS has [a] sho coming: he ela-
ionships be ween linea compounds can ne e duplica e he ela ionships be ween he la en a iables.
The simple and undamen al eason is ha no linea combina ion o he indica o s o a block can e e ep-
lica e he co esponding la en a iable, excep when some measu emen e o s a e ze o ( o an analysis o
he gene al case see K ijnen e al., 1998).” [19, p. 11]
Empi ical e idence, p oo
o a gumen s, ci ing
e idence
×
731
1 3
Bewa e o  heWoozle e ec andbelie pe se e ance in hePLS‑SEM…
Table 3 (con inued)
S a emen Type o easoning Scien i ic Anecdo al Flawed
One iew on PLS “is ha PLS is a dis ibu ion- ee way o consis en ly es ima e co a iance s uc u es o
second-o de ac o models. In ligh o he indings p esen ed in his pape , his iew is less enable o
p oduc i e.” [20, p. 312]
Empi ical e idence, p oo
o a gumen s, ci ing
e idence
×
“PLS ollows a composi e model app oach in he es ima ion o he cons uc measu es [...] PLS will p o-
duce ‘biased’ es ima es i he common ac o model holds.” [94, p. 4003]
Empi ical e idence
×
“[PLS] will p oduce biased es ima es i he common ac o model holds [do s] (Sa s ed e al. 2016) [
…
] To
es ima e he model pa ame e s PLS d aws on composi es o med om he indica o s.” [43, p. 619]
Ci ing e idence
×
“PLS has been he subjec o c i ical scien i ic examina ions highligh ing se e al sho comings such as
inconsis en pa ame e es ima es o e lec i e and causal- o ma i e measu emen models” [98, p. 1313]
Empi ical e idence, ci ing
e idence
×
“The mos wo ying is ha PLS con ains no explici hypo he ical causal link be ween he unobse able
concep ual a iables (A and F) and he composi es PLS p oduces.” [8, p. 1707]
P oo o a gumen s
×
“PLS es ima es [a] composi e model, whe e he composi es a e used as app oxima ions o he la en a i-
ables. [
…
] I is clea ha he composi es can only be impe ec app oxima ions (‘p oxies’) o he la en
a iables in he common ac o and he o ma i e model.” [22, p. 4]
None
×
“[PLS] can be used o consis en ly es ima e composi e models.” [55, p. 1743] Empi ical e idence
×
“PLS is no a la en - a iable me hod a all, bu an indica o weigh ing sys em ha c ea es composi e sco es
o subsequen eg ession analysis (E e mann and Rönkkö, 2021; Goodhue e al., 2012; Rönkkö and
Yli alo, 2010).” [89, p. 1599]
Ci ing e idence
×
“[R]egula PLS-SEM ollows a composi e model p ocedu e o es ima ing s uc u al measu emen s [
…
] [c]
ommon ac o models measu ed wi h e lec i e indica o s should be analyzed wi h consis en PLS-SEM.”
[119, p. 13]
None
×
“In he li e a u e i is well documen ed ha PLS-SEM is unsui able o es ima ing models comp ising la en
a iables.” [100, p. 2]
Empi ical e idence, ci ing
e idence
×
“[PLS’s] use o s uc u al equa ion models wi h la en a iables esembles applying he nai e es ima o .”
[101, p. 340]
Empi ical e idence, ci ing
e idence
×
“[B]ias unde pa h analysis wi h composi e sco es is subs an i e [
…
]. Consequen ly, me hods ha ake in o
accoun andom measu emen e o [
…
] should be p e e ed o e pa h analysis ia composi es when i
comes o es ing ela ionships be ween heo e ical cons uc s.” [97, p. 4212]
Empi ical e idence, ci ing
e idence
×
732
J.Hensele e al.
1 3
lawed. As ypes o scien i ic easoning, we ound cases o deduc i e easoning such
as p oo s o chains o a gumen s, cases o induc i e easoning using empi ical e i-
dence, and ci a ions o scien i ic wo k ha used deduc i e o induc i e easoning.
Anecdo al easoning comp ises s a emen s wi hou any explana ions o a gumen s,
hea say such as e e al o unspeci ied ‘ad oca es’, o ci a ions o wo k o his kind.
Flawed easoning includes s a emen s based on w ong conclusions o in alid ind-
ings as well as ci a ions o lawed wo k.
By and la ge, h ee di e en claims can be dis inguished: (1) PLS-SEM equi es
many eliable indica o s o be sui able o es ima ing e lec i e measu emen mod-
els  ( o he acco ding e idence base, see Table 1). (2) PLS-SEM is sui able o
es ima ing e lec i e measu emen models ( o he acco ding e idence base, see
Table2). (3) PLS-SEM is no sui able o es ima ing e lec i e measu emen models
( o he acco ding e idence base, see Table3).
Table2 clea ly shows ha he e is no scien i ic suppo o he second claim, ha
PLS-SEM is sui able o es ima ing e lec i e measu emen models. In con as o
li e a u e ega ding he i s and hi d claims (see Tables1 and 3, espec i ely), he
li e a u e claiming ha PLS-SEM is sui able o es ima ing e lec i e measu emen
models has ne e p o ided any o m o scien i ic easoning, a si ua ion ha uns
coun e o demands o me hodological esea che s o p o ide s ong e idence o
suppo claims hey make ega ding esea ch me hods [111].
3.3 Fac o s acili a ing heeme gence o  heWoozle e ec
The ini ializa ion o he Woozle e ec o PLS-SEM’s alleged sui abili y o es i-
ma ing e lec i e measu emen models canno be a ibu ed o a single publica ion.
Ra he , he e we e se e al ci cums ances ha con ibu ed o he eme gence o he
Woozle e ec . Reason [84] de eloped a model o explain he b eakdown o a com-
plex socio- echnical sys em such as me hodological esea ch, he so-called swiss
cheese model. Based on he obse a ion ha acciden s o en esul om a a ie y o
delayed-ac ion human ailu es commi ed long be o e a s a e o eme gency is ec-
ognized, his model iden i ies as he cen al cause he ad e se con luence o many
causal ac o s, each o which is necessa y bu singly insu icien o cause a sys em
ailu e.
The scien i ic en e p ise can be unde s ood as a sys em ha ies o p oduce u h
[7]. A numbe o scien i ic no ms, p inciples, and s anda ds ha e been es ablished
o help he scien i ic en e p ise sa egua d i s ole as a p oduce o u h and weigh
ou alsehoods, acili a ing wha is gene ally unde s ood as good esea ch p ac ices.
Fo ins ance, Me on [75] o mula ed he no m o disin e es edness, among o h-
e s, which asks scien i ic ins i u ions o ac o he bene i o a common scien i ic
en e p ise ins ead o pe sonal gain. As Fig. 5 shows, scien i ic no ms, p inciples,
and s anda ds can be compa ed o slices o swiss cheese con aining holes. While in
mos cases, alsehoods a e in e cep ed by a leas one slice, he e can be ins ances in
which a alsehood passes all sa egua ding mechanisms o science.
In his spi i , we iden i y a numbe o ac o s ha ha e acili a ed he eme -
gence o he Woozle e ec in he PLS-SEM li e a u e: P io i izing dissemina ion,
733
1 3
Bewa e o  heWoozle e ec andbelie pe se e ance in hePLS‑SEM…
academic dependency, model mis ep esen a ion, lawed easoning, lack o e e enc-
ing, and belie pe se e ance.
P io i izing dissemina ion On many occasions, s a emen s made in he PLS-SEM
li e a u e sugges ha maximising he dissemina ion o he me hod was he p io -
i y, a he han wha may ha e been be e o he disin e es ed pu sui o he u h.
Fo example, ins ead o s a ing ou igh he u h ha PLS-SEM is no consis en
unless he e is a la ge numbe o eliable e lec i e indica o s, he in en o o PLS-
SEM, H. Wold, asc ibed consis ency o PLS-SEM wi h a quali ica ion: “PLS es s
con en wi h consis ency [
…
], albei in he quali ied sense o consis ency a la ge”
[116, p.28]. The e m “consis ency a la ge”, which H. Wold in oduced especially
o his occasion, can be ega ded as a clea euphemism. Had H. Wold op ed o
he co ec cha ac e iza ion o PLS-SEM as an inconsis en es ima o o e lec i e
measu emen models, much con usion could ha e been a oided. Howe e , a he
same ime, i is likely ha such a clea quali ica ion would ha e hinde ed he dis-
semina ion o he me hod. “Al hough no all use ul es ima o s a e unbiased, i ually
all economis s ag ee ha consis ency is a minimal equi emen o an es ima o . The
Nobel P ize-winning econome ician Cli e W. J. G ange once ema ked, ‘I you
can’ ge i igh as n goes o in ini y, you shouldn’ be in his business.’ The implica-
ion is ha , i you es ima o o a pa icula popula ion pa ame e is no consis en ,
hen you a e was ing you ime” [118, p.169]. Euphemis ic e minology such as
‘consis ency a la ge’ appea s o ha e no o he ob ious pu pose han o gloss o e
he ue cha ac e is ics o PLS-SEM, in an a emp o ensu e wide dissemina ion o
he me hod. I appea s ha he wish o dissemina ion o one’s ideas ou weighed he
c i ical scien i ic i ue o “a kind o u e hones y”, he p inciple ha i he e a e
“[d]e ails ha could h ow doub on you in e p e a ion”, hen hey “mus be gi en,
i you know hem” [24, p.11].
Academic dependency Se e al o H. Wold’s Ph.D. s uden s, in pa icula B. Hui
and T. Dijks a, ha e wo ked ex ensi ely on he cha ac e is ics o PLS-SEM. The
dependencies and academic ies may ha e led o a mo e a o able assessmen o
PLS-SEM han migh o he wise ha e been expec ed. B. Hui emphasizes he ‘close
pe sonal ela ionship wi h p o esso Wold’ [59, p.iii], and his disse a ion is qui e
unc i ical abou PLS-SEM’s abili y o es ima e e lec i e measu emen models. Fo
ins ance, he e is a a he posi i e desc ip ion o he e m ‘consis ency a la ge’: “An
index es ima ing a la en a iable is consis en a la ge i (i) his index is cons uc ed
as a unc ion o all he a ailable obse ed indica o s gene a ed by his la en a iable;
and (ii) as he numbe o a ailable obse ed indica o s gene a ed by his la en a i-
able inc eases, he index app oaches he unde lying unobse ed case alue o his
gene a ing la en a iable” [59, p.14]. In con as , T. Dijks a, who ob ained his doc-
o a e somewha la e han B. Hui, ook a mo e c i ical iew o PLS-SEM: “‘Con-
sis ency a la ge’ is a ph ase due o H. Wold and i means ha he PLS es ima o s
will no be consis en [
…
] (indeed, hey may be wide o he ma k [
…
])” [16, p.51].
S ill, Dijks a [16, p.42] lea es an ob iously un enable assump ion o his p omo-
o uncommen ed: “H. Wold assumes u he mo e ha he measu emen e o s a e
unco ela ed wi h each o he and wi h all la en a iables.” In ac , unless pe ec ly
eliable indica o s a e a ailable, PLS-SEM canno p oduce unco ela ed indica o
esiduals. Wi hou he dependency inhe en in a mas e -app en ice ela ionship, i
734
J.Hensele e al.
1 3
would ha e been easie o exp ess e en mo e undamen al c i ique. Ob iously, we
a e awa e ha he explosion o g ow h in he use o PLS-SEM in business esea ch
is also due o mo e ecen au ho s who ha e no ob ious academic dependencies o
ies wi h H. Wold. Howe e , i is equally clea ha he majo i y o wo k suppo i e
o PLS-SEM in he ea ly de elopmen o he me hod was conduc ed by hose wi hin
he academic ne wo k o H. Wold.
Model mis ep esen a ion Visual elemen s play an impo an ole in o ming a i-
udes and belie s [76]. This mechanism has played a c i ical ole in he dissemina ion
o he misconcep ion ha PLS-SEM is sui able o es ima ing e lec i e measu e-
men models. Many igu es in a la ge po ion o li e a u e on PLS-SEM [1, 12, 34,
38, 41, 46, 82, 92, 105, 113, 115, 116, c .] as well as he g aphical use in e ace o
many so wa e implemen a ions (in pa icula , PLS-G aph, Sma PLS, SPAD-PLS,
and VisualPLS) make eade s and analys s belie e ha PLS-SEM es ima es a e lec-
i e measu emen model, al hough PLS-SEM in ac es ima es a composi e model
[51]. Figu e6 shows a ypical isualiza ion o a ‘ e lec i e’ measu emen model in
PLS-SEM and con as s i wi h he e lec i e measu emen model ha PLS-SEM
p e ends o es ima e and wi h a composi e model exp essed in e ms o he Hense-
le –Ogasawa a speci ica ion [120].7 I becomes clea ha he g aphical ep esen a-
ion o a measu emen model is incomple e and ambiguous when only one cons uc
and i s indica o s a e plo ed. In pa icula , he g aphical ep esen a ion o ‘ e lec-
i e’ measu emen models in PLS-SEM so wa e obscu es he ac ha PLS-SEM
es ima es composi e models, no e lec i e measu emen models Thus, he claim ha
PLS-SEM and co a iance-based SEM es ima e he“same model” [Sa s ed , Adle ,
Falsehood
In e es in
dissemina ion
Academic
dependency
Model
mis ep esen a ion
Flawed
easoning
Lack o
e e encing
Belie
pe se e ance
Fig. 5 A swiss cheese model o sys em b eakdown in he PLS-SEM li e a u e
7 The Hensele –Ogasawa a (H–O) speci ica ion [53, 99] exp esses pa –whole ela ionships in e ms o
loadings, no weigh s. The model shown a he igh in Fig.6 is equi alen o he model shown in Fig.3.

735
1 3
Bewa e o  heWoozle e ec andbelie pe se e ance in hePLS‑SEM…
e  al. 2024] is ob iously un ue. The incomple e and ambiguous g aphical ep e-
sen a ion o ‘ e lec i e’ measu emen modelspa es he way o he Woozle e ec .
Full anspa ency o he ac ual model speci ica ion in PLS-SEM could ha e wo ked
agains he Woozle e ec .
Flawed easoning One o he ea lies publica ions s a ing ha PLS-SEM is sui -
able o es ima ing he pa ame e s o e lec i e measu emen models is [26], which
is based on a wo king pape published he yea be o e. Fo nell and Books ein [26]
obse e ha e lec i e measu emen models equi e a pa ame e iza ion in e ms o
loadings and e o a iances, and ha he co a iances be ween measu emen e o s
should be ze o. A i s sigh , Fo nell and Books ein [26] seem o p o ide ho ough
a gumen a ion. They obse e ha PLS-SEM exp esses he ela ionships be ween
la en and obse ed a iables in he o m o Eq.2, and he e o e conclude ha “[a]s
is e iden om [ he] equa ions [
…
], he unobse ed cons uc s can be iewed ei he
as unde lying ac o s o as indices p oduced by he obse able a iables. Tha is, he
obse ed indica o s can be ea ed as e lec i e o o ma i e” [26, p.441]. Howe e ,
he conclusion was lawed, because as explained in Sec .3.2 his is only he case i
he numbe o indica o s end o in ini y.
In la e PLS-SEM li e a u e, a pa icula pa e n o lawed easoning can be
obse ed, namely in e ing he sui abili y o PLS-SEM o a pa icula ype o
esea ch p oblem om he ac ha PLS-SEM is al eady used o ha pu pose. Fo
example, Hensele e al. [50] in e PLS-SEM’s sui abili y om he ac ha 30 s ud-
ies in in e na ional ma ke ing used PLS-SEM o es ima e he pa ame e s o e lec-
i e measu emen models. Simila ly, om he ac ha ce ain e alua ion c i e ia a e
used in empi ical esea ch, Hai e al. [47] conclude ha he use o hese e alua ion
c i e ia is a bes p ac ice. An a gumen by Sa s ed e al. [93] alls in o he same
pa e n: They conclude ha since use s a ely apply a co ec ion o a enua ion o
PLS-SEM esul s, use s do no need such a co ec ion. Such a s a emen is ob iously
lawed. He e we see PLS ad oca es commi ing a kind o na u alis ic allacy, d aw-
ing an ough om an is.
All o hese in e ences implici ly assume ha he ci ed o coun ed use s o he
me hod we e able o make an in o med choice and selec ed an app op ia e esea ch
me hod o he p oblem a hand– i.e. ha hey had made he co ec choice hem-
sel es. Howe e , i analys s based hei choice on misin o ma ion, such as he
Fig. 6 The g aphical isualiza ion o measu emen models in he PLS-SEM li e a u e and PLS-SEM
so wa e con enien ly lea es ague which model is ac ually es ima ed [51]
736
J.Hensele e al.
1 3
alleged sui abili y o PLS-SEM o es ima ing he pa ame e s o e lec i e measu e-
men models, o hemsel es simply based on p ecedence, o poo easoning, hen he
choice beha io says no hing abou he sui abili y o a me hod. 30 people who did
some hing w ong is no e idence ha doing ha hing is igh .
Lack o e e ences As Table1 showed, he e is a la ge amoun o publica ions
[c . 13, 15, 25, 27, 34, 45, 46, 50, 86, 93, 103, 104, 107, 117] ha asc ibe PLS-
SEM he sui abili y o es ima ing e lec i e measu emen models, bu do no ci e
any e idence. Had any o hese au ho s had mo e doub s abou PLS-SEM’s sui -
abili y o es ima ing he pa ame e s o e lec i e measu emen models and ied
o ind e idence o i in he me hodological li e a u e, hey could ha e no iced
ha he e is no e idence. Ce ainly, he e is he o ic—in ha many au ho s s a e
explici ly ha PLS-SEM can model e lec i e measu es—bu a no poin is any
o ha he o ic backed up by e idence, ei he empi ical example o ma hema ical
p oo .
Belie pe se e ance Belie pe se e ance can be hough o as he las hole in he
swiss cheese model ha allows a alsehood o emain in he scien i ic eco ds. In
he PLS-SEM li e a u e, belie pe se e ance looms la ge. Fo ins ance, how can
esea che s who a e on eco d saying ha PLS-SEM es ima es a e only consis -
en -a -la ge [46], ha PLS-SEM “will p oduce biased es ima es i he common
ac o model holds” [44, p.618], ha PLS-SEM equi es a co ec ion o a enu-
a ion o ob ain consis en esul s [43], and ha he use o PLS-SEM o e lec i e
measu emen models wi hou any co ec ion o a enua ion esul s in an ino di-
na e amoun o Type I e o s [108], also hen s a e ha “[ ]esea che s can include
e lec i ely and o ma i ely speci ied measu emen models, which PLS-SEM es i-
ma es wi hou any limi a ions” [45, p.16]? The e a e many possible explana ions
o such inconsis ency among published s a emen s by he same au ho s. How-
e e , we sugges ha belie pe se e ance may be a e y likely explana ion o his
con adic ion.
I is o en said ha he inal ba ie o he p oli e a ion o alsehood in he sci-
en i ic eco d is ha o he na u ally sel -co ec ing na u e o he scien i ic me hod.
Tha is, i inco ec s a emen s a e published and used, coun e -a gumen s and new
in o ma ion can be used wi h he in en ion o “se ing he eco d s aigh ” [c . 106].
This may mean ha empi ical s udies mus be edone and p io indings mus be
co ec ed, o e en ha inco ec pape s a e e ac ed. All hese ac i i ies a e no mal
p ocedu es o scien i ic sel -co ec ion, and a e essen ial wi hin he esea ch me h-
odology li e a u e, whe e i is known ha “i is ela i ely easy o make a me hod
appea be e han i ac ually is [
…
and] ha o e op imis ic s a emen s ega ding
a me hod’s pe o mance may be pa ly a ibu ed o he nonneu al a i ude o he
au ho s, who a e na u ally in e es ed o p esen hei me hod in a posi i e ligh ” [80,
p. 2]. Bea ing hese obse a ions in mind, s a emen s such as “e o s o ‘se he
eco d s aigh ’ ha e no place in se ious science” [95, p.269] appea wo ying. I
seems ha belie pe se e ance does no only p e en esea che s om ha ing an
unbiased look a he subjec ma e a hand, bu i also educes he accep ance o
gene al scien i ic no ms and p inciples.
737
1 3
Bewa e o  heWoozle e ec andbelie pe se e ance in hePLS‑SEM…
4 Conclusions andimplica ions
The e is a my h in he PLS-SEM li e a u e ha PLS-SEM is sui able o es i-
ma ing e lec i e measu emen models. Howe e , analys s using PLS-SEM may
d aw w ong conclusions i hei models include e lec i e measu emen . One
could easily imagine si ua ions in which i ms ha e in es ed millions o dolla s
o esea che s ha e in es ed yea s o wo k in o de eloping a p omising in e -
en ion, and pu ely because PLS-SEM makes inco ec p edic ions, hey would
e oneously conclude ha he in e en ion is ine ec i e. As hings s and, PLS-
SEM is unsui able o s uc u al equa ion models con aining la en a iables, and
esea che s applying PLS-SEM o his pu pose ace he isk o conduc ing pseu-
doscience: I looks like scien i ic p inciples a e being ollowed, when in ac hey
a e no . The co e issue he e is ha e en he mos ecen PLS-SEM li e a u e [c .
45, 93] ei e a es and ein o ces he alse na a i e ha PLS-SEM is use ul as
a ool o scien is s seeking he u h, when i can lead he esea che o d aw
en i ely alse conclusions simply because i does no p oduce consis en es ima es
o e lec i e measu emen models.
Ou pape p esen ed he Woozle e ec as a wo ying phenomenon in he scien-
i ic li e a u e ha p o ides a iable explana ion o why au ho s epea and ein-
o ce alse na a i es. Using he PLS-SEM li e a u e as a case s udy, we showed
how a euphemism plan ed he seed o he Woozle e ec ; an omission o a quali i-
ca ion con ibu ed o he ini ializa ion o he Woozle e ec ; and inco ec g aphi-
cal ep esen a ions, lawed easoning, and lack o e e ences b ough he Woozle
e ec in o ull swing. Belie pe se e ance was iden i ied as a mechanism ha hin-
de s scien i ic sel -co ec ion, he las eso in he scien i ic sea ch o u h.
I canno be uled ou ha he PLS-SEM li e a u e con ains mo e Woozle
e ec s han he one we used as an illus a ion. Fo ins ance, la ge pa s o he
PLS-SEM li e a u e also ein o ces he alse claim ha PLS-SEM has ala m bells
and whis les ha wa n he esea che s when measu emen is p oblema ic [c . 42],
when i is clea ha p oblema ic issues go unno iced [leading o alse conclusions,
see e.g. 51, 98] ha could ha e easily been de ec ed by mo e app op ia e me h-
ods. The e o e, as a as e lec i e measu emen is in ol ed, we un o una ely
canno ejec he no ion o mula ed by Wes land [112, p.38] ha PLS-SEM “is
an ideal ool o unsc upulous o lazy esea che s in e es ed in bogus heo ies
wi h andom da a.”
So wha could PLS-SEM p oponen s do o esol e he con adic ion in he
PLS-SEM li e a u e? The ob ious solu ion is o e ain om making he inco ec
claim ha PLS-SEM is sui able o es ima ing e lec i e measu emen models and
o explain ha i simply es ima es composi e models [c . 8, 55]. I o wha e e
eason analys s wan o use he PLS-SEM algo i hm o es ima e e lec i e meas-
u emen models, hey should employ a co ec ion o a enua ion as o ins ance
done in consis en PLS [PLSc, see 19, 20, 83]. P e e ably, analys s should make
use o co a iance-based SEM (CB-SEM) by de aul . While CB-SEM does no
seem o ha e a subs an ial ad an age in e ms o pa ame e accu acy in case o
well-speci ied models [100], i allows he analys o cons ain o ix pa ame e s,
738
J.Hensele e al.
1 3
and i o e s a la ge a ie y o model assessmen ools. Hensele and Schube h
[49] conjec u e ha PLSc migh be ad an ageous in some special cases o model
misspeci ica ion, such as unmodelled co a iances be ween measu emen e o s
wi hin a block o obse ed a iables. Howe e , mo e me hodological esea ch is
needed o p ecisely iden i y which (i any) cases whe e PLSc excels o e CB-
SEM and ice e sa. In any case, analys s elying on PLS-SEM should make su e
o adhe e o guidelines ha a e ee om he Woozle e ec [e.g., 5, 53].
Acknowledgemen s The au ho s hank he edi o -in-chie o Elec onic Comme ce Resea ch, J. Ch is o-
phe Wes land, o his suppo o ou pape idea. They also hank Flo ian Schube h and Ola Fissche o
help ul commen s on an ea lie e sion o his manusc ip and Alexand a Elbakyan o he e o s o make
science accessible.
Funding On behal o all au ho s, he co esponding au ho s a es ha he e is no Con lic o in e es .
This wo k was suppo ed by na ional unds h ough FCT (Fundação pa a a Ciência e a Tecnologia) unde
he p ojec UIDB/04152/2020– Cen o de In es igação em Ges ão de In o mação (MagIC). Open Access
unding was enabled h ough an a angemen o VSNU.
Open Access This a icle is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License,
which pe mi s use, sha ing, adap a ion, dis ibu ion and ep oduc ion in any medium o o ma , as long
as you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e
Commons licence, and indica e i changes we e made. The images o o he hi d pa y ma e ial in his
a icle a e included in he a icle’s C ea i e Commons licence, unless indica ed o he wise in a c edi line
o he ma e ial. I ma e ial is no included in he a icle’s C ea i e Commons licence and you in ended
use is no pe mi ed by s a u o y egula ion o exceeds he pe mi ed use, you will need o ob ain pe mis-
sion di ec ly om he copy igh holde . To iew a copy o his licence, isi h p://c ea i ecommons.o g/
licenses/by/4.0/.
Re e ences
1. A ne , D. B., La e ie, D. A., & Meie s, A. (2003). De eloping pa simonious e aile equi y
indexes using pa ial leas squa es analysis: A me hod and applica ions. Jou nal o Re ailing, 79(3),
161–170. h ps:// doi. o g/ 10. 1016/ S0022- 4359(03) 00036-8
2. As ley, W. G. (1984). Subjec i i y, sophis y and symbolism in managemen science. Jou nal o
Managemen S udies, 21(3), 259–272. h ps:// doi. o g/ 10. 1111/j. 1467- 6486. 1984. b004 10.x
3. Ba clay, D., Higgins, C., & Thompson, R. (1995). The pa ial leas squa es (PLS) app oach o
causal modeling: Pe sonal compu e adop ion and use as an illus a ion. Technology S udies, 2(2),
285–309.
4. Becke , J. M., Klein, K., & We zels, M. (2012). Hie a chical la en a iable models in PLS-SEM:
Guidelines o using e lec i e- o ma i e ype models. Long Range Planning, 45(5–6), 359–394.
h ps:// doi. o g/ 10. 1016/j. l p. 2012. 10. 001
5. Beni ez, J., Hensele , J., Cas illo, A., e al. (2020). How o pe o m and epo an impac ul analysis
using pa ial leas squa es: Guidelines o con i ma o y and explana o y IS esea ch. In o ma ion
and Managemen , 57(2), 103168. h ps:// doi. o g/ 10. 1016/j. im. 2019. 05. 003
6. Bo sboom, D. (2005). Measu ing he mind. Camb idge Uni e si y P ess.
7. B own, R. H., & Malone, E. L. (2004). Reason, poli ics, and he poli ics o u h: How science is
bo h au onomous and dependen . Sociological Theo y, 22(1), 106–122. h ps:// doi. o g/ 10. 1111/j.
1467- 9558. 2004. 00206.x
8. Cadogan, J. W., & Lee, N. (2023). A mi acle o measu emen o acciden al cons uc i ism? How
PLS sub e s he ealis sea ch o u h. Eu opean Jou nal o Ma ke ing, 57(6), 1703–1724.
h ps:// doi. o g/ 10. 1108/ EJM- 08- 2020- 0637
9. Chin, W. W. (2001). PLS-g aph use ’s guide e sion 3.0. So Modeling Inc.