World Academy of Science, Engineering and Technology

International Journal of Computer and Information Engineering

Vol:6, No:12, 2012

The Measurement of Endogenous Higher-Order

Formative Composite Variables in PLS-SEM:
An Empirical Application from CRM System
Development
Samppa Suoniemi, Harri Terho, and Rami Olkkonen

practical guidelines how to create, estimate and validate

Abstract—In recent methodological articles related to structural formative models, in sharp contrast to standardized
equation modeling (SEM), the question of how to measure development procedures that have been developed for
endogenous formative variables has been raised as an urgent,
International Science Index, Computer and Information Engineering Vol:6, No:12, 2012 waset.org/Publication/8748

reflective measures over the years [11], [33]. Publications that

unresolved issue. This research presents an empirical application
have addressed these issues have appeared only fairly recently,
from the CRM system development context to test a recently
developed technique, which makes it possible to measure endogenous with arguably the most notable contributions coming from
formative constructs in structural models. PLS path modeling is used [12]-[14], [23], [29], [36], respectively.
to demonstrate the feasibility of measuring antecedent relationships at The scarcity of empirical models with formative structures
the formative indicator level, not the formative construct level. may also be due to the fact that choice of measurement
Empirical results show that this technique is a promising approach to perspective is still often ignored by researchers [10], despite
measure antecedent relationships of formative constructs in SEM.
increasing evidence in literature about the undesirable
consequences of model misspecification [13], [23]. In recent
Keywords—CRM system development, formative measures, PLS
path modeling, research methodology.
years, though, scholars have begun to challenge the “blind
adherence” to the reflective approach with its strict emphasis
I. INTRODUCTION on exploratory factor analysis and internal consistency [9].
Indeed, formative measurement models have been
S TRUCTURAL equation modeling (SEM) has become the
leading analytical approach in contemporary IS research.
SEM owes its popularity to the fact it allows researchers to
increasingly applied by IS researchers (for a list of examples,
see [33]).
In addition to the issues related to formative measurement
examine multiple causal relationships simultaneously. Thus, model assessment, structural models with formative measures
SEM accommodates research that aims to unravel complex pose a particular type of problem, which has remained largely
phenomena and inter-relationships between concepts in unsolved to date. Reference [9] voiced their concerns about
complete theoretical models. “the conceptual plausibility of formatively-measured
Historically, covariance-based SEM (CB-SEM) [24] has constructs occupying endogenous positions in structural
been the standard analysis method in IS research as well as models”, and stressed the urgency of finding a solution to this
behavioral sciences in general. It is parameter-based and the dilemma. This is a challenge with endogenous formative
underlying assumption of the traditional CB-SEM is that the constructs due to different nomological networks of
indicators used to measure latent variables are reflective in antecedents and consequences [23]. As a response, Cadogan &
nature [8]. Partial Least Squares SEM (PLS-SEM) [39] has Lee [5] demonstrated the inappropriateness of developing
challenged CB-SEM has emerged as a complementary theory about antecedents to endogenous formative constructs
analysis method in recent years. Unlike CB-SEM, PLS at the aggregate level (i.e. path relationships between latent
modeling aims to maximize the explained variance of the variables). Rather, antecedents’ relationships to the dependent
dependent endogenous variables in the structural model [19]. formative construct should be assessed at the formative
In addition to reflective latent variables, PLS is also well- indicator level (i.e. path relationships from latent variable to
equipped to estimate formative latent variables. As this paper indicator), which would be unorthodox in SEM. The purpose
focuses on formative measurement, PLS-SEM was chosen as of this paper is to discuss and empirically test the feasibility of
the most appropriate analysis method [20]. Cadogan & Lee’s [5] conceptual solution in the measurement
Compared to reflective models, the use of formative models of endogenous 2nd order formative constructs.
in empirical studies remains scarce [11]. The lack of
popularity of formative models in IS research and other related II. MEASURING ENDOGENOUS FORMATIVE COMPOSITE
disciplines has probably been influenced by the lack of VARIABLES
Cadogan & Lee [5] presented a conceptual solution to
Samppa Suoniemi is with the University of Turku, School of Economics, measure endogenous 2nd order formative constructs. As their
Department of Marketing, Finland (phone: +358-44-549-1055; e-mail:
sawisu@ utu.fi).
novel approach has been neither discussed nor tested in other

International Scholarly and Scientific Research & Innovation 6(12) 2012 1648 scholar.waset.org/1307-6892/8748
 World Academy of Science, Engineering and Technology
International Journal of Computer and Information Engineering
Vol:6, No:12, 2012

puublished articlles, the follow wing discoursee is largely baased on thee issue of measurement
m error estimattion with forrmative
thheir article unlless stated otherwise. meeasurement models.
m Lateent variabless are intentiionally
They [5] identified two important isssues, which provide p inccluded to enabble the estimaation of measuurement errorr at the
suupport for asssessing the relationships between
b antecedents inddicator level. This is achieeved by assign ning each forrmative
annd formative indicators,
i nott the formativve construct. The
T first inddicator a singgle reflective indicator of its respectivee latent
issue is related to the concepptual distinctioon between forrmative vaariable. Its coonceptual justification is questionable,
q t
though
laatent variable and foormative coomposite variable. v [100]-[11].
Thheoretically, the relationnship betweeen antecedennts and In summary, the appropriaate approach is i to test anteccedent-
foormative latennt variable caan be assessed at the forrmative enndogenous forrmative compoosite variable relationshipss at the
coonstruct level.. However, a formative lattent variable requires
r forrmative indicaator level. As reference [5]] put it, “if thee items
a census of alll possible cauuses, which iss usually emppirically (foormative indiccators) are loogically form
mative, then … item
unnrealistic. Thuus, in most casses the construuct is not a forrmative levvel modelingg is most apppropriate”. With W regard to the
laatent variable but in fact a formative coomposite variaable. A constructs adoppted in this study, this is i the case from f a
foormative com mposite variabble is merely a collection, not a theeoretical view
wpoint as welll as from an empirical
e viewwpoint.
ceensus of form mative indicattors. In the casec of a forrmative Thhe conceptualization of an endogenous 2nd order forrmative
coomponent varriable, anteceddents can be only assessedd based composite variaable with anteecedent relatioonships measu ured at
International Science Index, Computer and Information Engineering Vol:6, No:12, 2012 waset.org/Publication/8748

onn their correelations withh the specificc set of forrmative thee formative inndicator level is
i presented inn Fig. 1.
inndicators proposed to form the formativee composite variable.
v
U
Unfortunately, there is no generalizabillity in such results.
Consequently, the solution is i to assess reelationships between
b
anntecedents andd formative inndicators.
The second issue is relaated to the diifferent nomoological
neetworks of forrmative indicaators’ anteceddents. In otherr words,
foormative inddicators mayy be influennced by coommon
anntecedents inn different magnitudes,
m o they mayy have
or
diifferent anteecedents altoogether. Thuus, examininng the
reelationships from
f anteceddents to a formative
f commposite
vaariable may conceal signifiicant relationsships or displaay non-
exxistent relationnships. As a reesult, empirical findings reggarding
anntecedent relaationships woould be ambiiguous at besst. In a
simmilar vein with
w the first issue, the solution is to assess
reelationships between
b anteccedents and formative
f inddicators,
noot the formativve composite variable.
Cadogan & Lee [5] argueed further thaat any variation in a Fig. 1 Endogennous 2nd order formative
f compposite variable with
w
antecedent relaationships at thee formative indiicator level (adoopted
foormative construct must occcur either duee to variationn in one
froom [5])
orr more form mative indicatoors, and/or due d to variaation in
unnknown indiccators (error term). Whilee this ambigguity is In Fig. 1, C1 represennts the enddogenous forrmative
innherent to a foormative latennt variable, a formative com mposite composite variaable (error terrm ξ4=0), whhich is shapedd like a
vaariable allowss parameters to be expliciitly estimatedd in the
hexagon to distiinguish it from
m a formativee latent variablle. The
abbsence of the error term [100]. Thus, hypothesized anteecedent
exogenous refllective anteceedent variablle (ξ1) withh three
reelationships annd indicator weights can only be emppirically
inddicators influeences C1 only through refflective LVs η1,η η2
tested with a foormative compposite variablee. On the otheer hand,
reesults related to endogenouus formative composite vaariables annd η3, which act as C1’s formative inndicators. Theerefore,
caannot be exten nded to endoggenous formaative latent vaariables, paath coefficientts (γ1-3) andd measuremeent error (ξ1--3) are
w
which cannot beb tested emppirically undeer any circum mstances esttimated at thee formative indicator level. Indicator weights
w
[55]. (β1-3) represennt the contribuutions of the formative
f inddicators
As the form mative indicaators have ann important role in to the compositte variable. WWe will test thhis solution with
w an
asssessing the relationships in the struuctural modell, it is emmpirical study in the CRM ssystem develoopment contexxt.
immportant to bee able to estiimate their measurement
m e
error. A
tyype II 2nd ordeer measuremennt model [23]], namely, a 1st order III. EMPIIRICAL STUDY
Y
reeflective 2nd orrder formativee model, doess not suffer from this The research model was ddeveloped bassed on an exttensive
prroblem conceerning the laack of estim mation of item m-level revview of all relevant
r acaddemic literatu
ure related to CRM
m
measurement errror with formmative construucts [10]. sysstem developm ment. Due to the interdiscip
plinary naturee of the
Estimating measurement
m error is morre problematiic with phhenomena undder investigatioon, the theoretical review covered
c
enndogenous 1sts order form mative measurres. Referencce [15] rellevant theoriees within maarketing and IS research. More
inntroduced a “sspurious modeel” with multiiple common causes, speecifically, tw
wo research streams in marketing,
m n
namely,
w
which represennts a conceptuaal example off an attempt too tackle CRRM/ SFA adooption studies in sales mannagement and CRM-

International Scholarly and Scientific Research & Innovation 6(12) 2012 1649 scholar.waset.org/1307-6892/8748
 World Academy of Science, Engineering and Technology
International Journal of Computer and Information Engineering
Vol:6, No:12, 2012

peerformance litterature in maarketing, were reviewed to identify

i CRRMDS would change. Conssequently, thee formative 2nd n
order
faactors related to CRM systeem developmeent. Distinct areasa of composite variaable is a coherrent descriptio on which depiicts the
ISS research - risk and projectp manaagement theo ory, IT mu ultidimensionaal nature of C CRMDS.
in
nnovation reseearch, and IT T capability literature – weere also This construct responds to recent callls to develop new
reeviewed to develop
d a higgher-order forrmative meassure of hoolistic operatiionalizations, representing g combinatio ons of
CR RM system development,
d iits antecedentts and conseq
quences, facctors affecting
g IS developpment, by acaademics in rissk and
w
which is presen
nted in Fig. 2. prooject manageement literatuure [18], IT innovation reesearch
[166], and IT cap
pability literatu ure [26]. In th
his view, different IT
ressources in CRRM system deevelopment projects do no ot work
in isolation butt rather in coombinations. Based on reesource
complementarity y arguments, a higher-ordeer conceptuallization
woould better refflect reality thhan stand-alon ne IT resourcees [25].
Suuch an operattional measure has been laacking in thee CRM
context.
Based on theoretical review w, we expect CRMDS to directly
d
inffluence the well-establisshed IT prroject perforrmance
International Science Index, Computer and Information Engineering Vol:6, No:12, 2012 waset.org/Publication/8748

meeasures [31], which

w includee process perfformance (SP PP), i.e.
meeeting budgeet and scheedule estimaates; and product p
performance (SPD),
( i.e. CRM systtem quality [37].
Fuurthermore, wee expect SPP to influence SPD S [3].
Fig. 2 Research
R model
IV. SAMPLE AND MEASUREMEN
NT MODEL
Based on a reeview of IT caapability stud
dies in IS reseaarch, IT
The populatioon under invesstigation was client firms (aat SBU
reesources inclu uding technollogical, humaan, and relattionship
levvel) in Finlan nd using CR RM technolog gy, excluding g small
reesources [35], are expected to be anteced dents of functiional IT
buusinesses. Th he populatioon sample consisted off 526
caapabilities succh as the firrm’s capabiliity to deliverr CRM
orgganizations. The
T final sam mple size afterr screening was
w 161
syystems. More specifically, we adopted operational
o m
measures
usaable responses, which met tthe minimum requirement criteriac
froom [34], whicch included IT T infrastructurre (INF), IS planning
forr PLS-SEM su uggested by [22].
soophistication (ISP), IS personnel skill s (PS), internal
i
Reflective meeasures were subject to rigorous reliabiliity and
paartnership quality (IPQ), and external partnership quality
vaalidity testing, which showeed satisfactory y results. Forrmative
(EEPQ).
meeasurement model asseessment wass based on n the
We developeed the formatiive measure CRMC delivery system
reccommendation ns by [4], [3 30], who provided guideliines to
(CCRMDS), whiich is theoretiically derived from the innovation
ad
ddress multico ollinearity, moodel identificaation, and reliiability
deelivery systemm concept [27 7] in innovattion diffusion theory
an
nd validity asseessment. Baseed on these tests, CRMDS proved p
litterature. Acccording to reeference [16], effective delivery
d
to be robust construct. As the next step, structural model
syystems are chaaracterized byy organizationaal factors suchh as top
asssessment was performed with w PLS-SEM M, which allow wed us
m
management su
upport, trainiing, links to consulting seervices;
to estimate the feasibility
f of C
Cadogan & Leee’s [5] techniique.
annd process model
m factors ensuring fit with the paarticular
technology and d organization n. Reference [25] develop ped the
V. RESULTS AND DISCUSSION
cooncept ERP (enterprise ressource plannin ng) delivery system,
w
which includeed project m management resources (PMR), The primary criteria for structural model
m assessm
ment in
coonsultant resoources (CR), training reso ources (TR) and a top PLLS-SEM are the explainned variancees of endog genous
m
management suupport (TMS). Based on a review of maarketing constructs (R2 values),
v and tthe strength of
o standardizeed path
stuudies [7], [17], [40] andd industry ex xpert interview ws, we coefficients (β)) coupled witth significancce testing (t-v values)
addded user involvement (UII) as a fifth diimension into o a new [200-21].
cooncept CRM delivery
d systemm (CRMDS). The key consstruct CRMDS S was in an endogenous po osition,
We conceptu ualized the CR RMDS as a 1sts order reflecttive, 2nd whhich is prob blematic withh formative constructs due d to
orrder formativee construct. Although
A the five dimensions of diffferent nomo ological netw
works of anttecedents [23 3]. To
CR RMDS do no ot represent an
a exhaustive list of comp ponents, ovvercome this issue, we adopted the sollution suggessted by
th
hey have receiived the most theoretical su upport to justiify their Caadogan and Lee [5] for testting anteceden nt relationshipps of a
in
nclusion into the formatiive CRMDS construct. As A the forrmative comp posite variablee at the indicaator level insttead of
diimensions werre not identifified through a census, they form a thee construct lev
vel. These relationships co ould be estimaated by
fo
ormative comp posite variable, not a formmative latent variable.
v ad
dopting a tech hnique knownn as the hierrarchical com mponent
Fo ollowing refeerence [23], C CRMDS is clearly
c formaative in mo odel [28], [39]].
naature: its dim mensions will not necesssarily co-varry, the Following these
t guideelines, we tested anteecedent
caausality flowss from the dim mensions to thhe construct, and the rellationships to the CRMDS S construct (in grey) at 1sts order
diimensions arre not intercchangeable as a the mean ning of construct level (PMR,
( CR, TR, TMS, UI),, which resulted in a

International Scholarly and Scientific Research & Innovation 6(12) 2012 1650 scholar.waset.org/1307-6892/8748
 World Academy of Science, Engineering and Technology
International Journal of Computer and Information Engineering
Vol:6, No:12, 2012

to
otal of twenty y-five anteceddent paths. While
W the forrmative paarticular, occu upying endoogenous posiitions in stru uctural
coomposite has an error term m fixed at zero
o, measuremen nt error quation modelss.
eq
caan also be estiimated at form
mative indicattor level in a Type
T II In order to improve
i the generalizabili
g ty of the forrmative
hiigher-order co
onstruct. For demonstrativee purposes, th he final composite variaable CRMDS S, future studdies could alsso test
m
model, includin
ng significant paths only, is presented in Fig.
F 3. CRRMDS with predetermined
p indicator weiights. These weights
w
could be predettermined as eequal weightin ngs [5] or baased on
theeoretical considerations [222], for examplee.
In conclusion n, it is importtant to find a solution to measure
m
an
ntecedent relattionships of hhigher-order formative
fo consstructs.
Mo odel parsimo ony, when thheoretically ju ustifiable, hass been
sup pported by scholars
s [6]. According to o reference [32],
[ a
higgher-order co onstruct can m more parsimo oniously explaain the
sin
ngle cumulatiive effect, aas opposed to t multiple distinct
d
efffects of indiviidual facets, oon outcome measures.
m Theerefore,
thee developmen nt of higher-orrder constructts in IS develo opment
is desirable. Arg guably many oof these constrructs are likelly to be
International Science Index, Computer and Information Engineering Vol:6, No:12, 2012 waset.org/Publication/8748

forrmative in nature: IS development is an inherently

mu ultidimensionaal phenomenoon. Consequen ntly, techniqu
ues that
alllow for esttimation of factors affeecting higherr-order
conceptualizatio ons of IS deveelopment shou uld also be avvailable
Fig. 3 Finall structural mod
del
forr future studiees.
In this particcular empiricaal example, theory
t suggessts that
The anteceddent relationshhips show thaat IPQ and ISP had CRRMDS dimensions are inheerently intertw wined and shou uld not
siggnificant path
hs to CRMDS S through its formative
f ind
dicators. be assessed in isolation; therefore, the co omposite variable is
Fu urthermore, thhe total explaained variancce in CRM delivery
d theeoretically justified [266]. CRMDS incorporatees the
syystem quality (R²=0.272) co ould be calcullated in the PL LS path mu ultidimensionaal phenomenoon of CRM delivery system m from
m
model by concceptualizing C CRMDS as a 1st order co onstruct fiv
ve separate co onstructs into a single consstruct. Althou ugh the
reepresenting alll five project--level IT resouurces PMR, CR,C TR, diffferent facets that consstitute CRMD DS may vaary in
TM MS and UI. im
mportance from one CRM M project to another, it would
The Type III higher-ordeer construct CRMDS C impo ortantly argguably be diffficult to achhieve CRM system s develoopment
alllowed for paarameter estim mation at the formative in ndicator succcess in the absence
a of anny given dimmension identified in
level. The paraameters for thee 1st order refflective measu urement existing literatu
ure. The solutiion suggested by Cadogan & Lee
m
models, which represented tthe formative indicators off the 2nd [5]] represents a novel approoach, which has h the poten ntial to
orrder formativ ve constructt, remained stable in various inccrease our knnowledge regaarding factors affecting forrmative
emmpirical tests.. Cadogan & Lee’s [5] co onceptualizatioon also composite variaables, such as C CRMDS in th his study.
prroved to be usseful in measuring the anteecedent relatio onships
REFEERENCES
off each formaative indicatoor, and their contribution to the
fo
ormative consttruct as indicaator weights. These
T measurrements [1] Bagozzi, Richhard P. (2011) M Measurement and meaning in infoormation
systems and organizational
o ressearch: Methodollogical and philoosophical
allso remained d stable acrross differen nt structural model foundations. MIS
M Quarterly, Vool. 35, No: 2, 2611-292.
sccenarios. Baseed on these considerations,, this study provides
p [2] Barclay, D. W.W – Higgins, C. A A. – Thompson, R.R (1995) The parrtial least
emmpirical supp port for thee functionalitty of the solutions squares approach to causal moodeling: Personaal computer adop ption and
use as illustrattion. Technology Studies, Vol. 2, No:
N 2, 285–309.
suuggested by [55], following which
w the anteecedent relatio
onships [3] Barki, Henri – Rivard, Suzannne – Talbot, Jeaan (2001) An inttegrative
off formative vaariables are mmeasured at thee formative in ndicator contingency model
m of softwaare project risk management. Journal of
level, not the fo
ormative consttruct level. Management Information
I Systeems, 17:4, 37-69..
[4] Bollen, Kenneeth A. (2011) E Evaluating effectt, composite, andd causal
However, Cadogan
C & Lee’s
L [5] mod del is subjectt to the indicators in structural
s equationn models. MIS Quarterly,
Q Vol. 355, No: 2,
geeneral limitattions associated with form mative measurrement. 359–372.
Fo ormative indiicator weightss vary acrosss different em mpirical [5] Cadogan, J.W W., & Lee, N. (2010). Improp per use of enddogenous
formative varriables. Accepteed at: Journal off Business Research, 18
daata sets and d research contexts,
c lead
ding to lim mitations October 2010,, 1-34.
reegarding the generalizabilityy of empiricall findings [1]. [6] Cenfetelli, Roonald T. – Basseellier, Geneviève (2009) Interprettation of
Reflective measures
m are thhus considered d more usefull from a formative meaasurement in infoormation systems research. MIS Quarterly,
Q
Vol. 33, No: 4,4 689–708.
th
heory develop pment perspective [22]. Based B on theoretical
[7] Chen, Injazz J. – Popovich, Karen (2003) Understanding customer c
raationale, howeever, CRM ssystem develo opment is clearly a relationship management
m (CR RM). People, process
p and techhnology.
fo
ormative, multidimension
m nal constru uct. Under these Business Process Management Journal, Vol. 9, No: N 5, 672-688.
ciircumstances and
a based on this empiricaal study, Cado ogan & [8] Chin, Wynne W. (1998) Issuues and opinionn on structural equation
modeling. MIS S Quarterly, Vol. 22, March 1998,, 7−16.
Leee’s [5] reco ommended teechnique is currently thee most [9] Coltman, Tim m – Devinney, Tim mothy M. – Midgley, David F. – Venaik,
fruuitful approacch to investigaate formative variables, Ty ype II in Sunil (2008) Formative versuus reflective measurement modeels: Two

International Scholarly and Scientific Research & Innovation 6(12) 2012 1651 scholar.waset.org/1307-6892/8748
 World Academy of Science, Engineering and Technology
International Journal of Computer and Information Engineering
Vol:6, No:12, 2012

applications of formative measurement. Journal of Business Research, intervening variable. Information Systems Research, Vol. 6, No: 3, 191-
Vol. 61, No: 12, 1250–1262. 219.
[10] Diamantopoulos, Adamantios (2006) The error term in formative [32] Pavlou, Paul A. – El Sawy, Omar A. (2006) From IT leveraging
measurement models: Interpretation and modeling implications. Journal competence to competitive advantage in turbulent environments: The
of Modelling in Management, Vol. 1, No: 1, 7–17. case of new product development. Information Systems Research, Vol.
[11] Diamantopoulos, Adamantios – Riefler, Petra – Roth, Katharina P. 17, No: 3, 198–227.
(2008) Advancing formative measurement models. Journal of Business [33] Petter, Stacie – Straub, Detmar – Rai, Arun (2007) Specifying formative
Research, Vol. 61, No: 12, 1203–1218. constructs in information systems research. MIS Quarterly, Vol. 31, No:
[12] Diamantopoulos, Adamantios – Siguaw, Judy A. (2006) Formative 4, 623-656.
versus reflective indicators in organizational measure development: A [34] Ravichandran, T. – Lertwongsatien, Chalermsak (2005) Effect of
comparison and empirical illustration. British Journal of Management, information systems resources and capabilities on firm peformance: A
Vol. 17, No: 4, 263–282. resource-based perspective. Journal of Management Information
[13] Diamantopoulos, Adamantios – Winklhofer, Heidi M. (2001) Index Systems, Vol. 21, No: 4, 237-276.
construction with formative indicators: An alternative to scale [35] Ross, Jeanne W – Beath, Cynthia Mathis – Goodhue, Dale L. (1996)
development. Journal of Marketing Research, Vol. 38, No: 2, 269–277. Develop long-term competitiveness through IT assets. Sloan
[14] Edwards, Jeffrey R. (2001) Multidimensional constructs in Management Review, Vol. 37, Fall 1996, 31-42.
organizational behavior research: An integrative analytical framework. [36] Rossiter, John R. (2002) The C-OAR-SE procedure for scale
Organizational Research Methods, Vol. 4, No: 2, 144-192. development in marketing. International Journal of Research in
[15] Edwards, Jeffrey R. – Bagozzi, Richard P. (2000) On the nature and Marketing, Vol. 19, No: 4, 305–335.
direction of relationships between constructs and measures. [37] Wallace, Linda – Keil, Mark – Rai, Arun (2004) How software project
Psychological Methods, Vol. 5, No: 2, 155–174. risk affects project performance: An investigation of the dimensions of
International Science Index, Computer and Information Engineering Vol:6, No:12, 2012 waset.org/Publication/8748

[16] Fichman, Robert G. (2004a) Going beyond the dominant paradigm for risk and an exploratory model. Decision Sciences, Vol. 35, No: 2, 289-
information technology innovation research: Emerging concepts and 321.
methods. Journal of the AIS, Vol. 5, No: 8, 314–355. [38] Wilcox, James B. – Howell, Roy D. – Breivik, Einar (2008) Questions
[17] Fjermestad, Jerry – Romano, Nicholas C. (2003) Electronic customer about formative measurement. Journal of Business Research, Vol. 61,
relationship management. Revisiting the general principles of usability 1219-1228.
and resistance – an integrative implementation framework. Business [39] Wold, Herman (1982) Soft modeling: the basic design and some
Process Management Journal, Vol. 9, No: 5, 572-591. extensions. In: Systems under indirect observations: Part I, ed. by K.G.
[18] Gemino, Andrew – Reich, Blaize Horner – Sauer, Chris (2008) A Jöreskog & H. Wold, 1-54. North-Holland: Amsterdam.
temporal model of information technology project performance. Journal [40] Zablah, Alex R. – Bellenger, Danny N. – Johnston, Wesley J. (2004b)
of Management Information Systems, Vol. 24, No: 3, 9-44. Customer relationship management implementation gaps. Journal of
[19] Hair, Joe F. – Ringle, Christian M. – Sarstedt, Marko (2011). PLS-SEM: Personal Selling and Sales Management, Vol. 24, No: 4, 279-295.
Indeed a silver bullet. Journal of Marketing Theory and Practice, Vol.
19, No: 2, 139–151.
[20] Hair Joe F. – Sarstedt, Marko – Ringle, Christian M. – Mena, Jeanette A.
(2011) An assessment of the use of partial least squares structural
equation modeling in marketing research. Accepted in: Journal of the
Academy of Marketing Science, 17 May 2011.
[21] Henseler, Jörg – Ringle, Christian M. – Sinkovics, Rudolf R. (2009) The
use of partial least squares path modeling in international marketing.
Advances in International Marketing, Vol. 20, 277–320.
[22] Howell, Roy D. – Breivik, Einar – Wilcox, James B. (2007)
Reconsidering formative measurement. Psychological Methods, Vol. 12,
No: 2, 205-218.
[23] Jarvis, Cheryl Burke – MacKenzie, Scott B. – Podsakoff, Philip M.
(2003) A critical review of construct indicators and measurement model
misspecification in marketing and consumer research. Journal of
Consumer Research, Vol. 30, September 2003, 199–218.
[24] Jöreskog, Karl G. (1973) A general method for estimating a linear
structural equation system. In: Structural Equation Models in the Social
Sciences, ed. by A.S. Goldberger & O.D. Duncan, 255-284. Seminar
Press: New York.
[25] Karimi, Jahangir – Somers, Toni M. – Bhattacherjee, Anol (2007a) The
impact of ERP implementation on business process outcomes: A factor-
based study. Journal of Management Information Systems, Vol. 24, No:
1, 101-134.
[26] Karimi, Jahangir – Somers, Toni M. – Bhattacherjee, Anol (2007b) The
role of information systems resources in ERP capability building and
business process outcomes. Journal of Management Information
Systems, Vol 24, No: 2, 221-260.
[27] Leonard-Barton, Dorothy (1988) Implementation as mutual adaptation
of technology and organization. Research Policy, Vol. 17, No: 5, 251-
267.
[28] Lohmöller, Jan-Bernd (1989) Latent variable path modeling with partial
least squares. Physica: Heidelberg.
[29] MacKenzie, Scott B. – Podsakoff, Philip M. – Jarvis, Cheryl Burke
(2005) The problem of measurement model misspecification in
behavioral and organizational research and some recommended
solutions. Journal of Applied Psychology, Vol. 90, No: 4, 710-730.
[30] MacKenzie, Scott B. – Podsakoff, Philip M. – Podsakoff, Nathan P.
(2011) Construct measurement and validation procedures in MIS and
behavioral research: Integrating new and existing techniques. MIS
Quarterly, Vol. 35, No: 2, 293–334.
[31] Nidumolu, Sarma (1995) The effect of coordination and uncertainty on
software project performance: Residual performance risk as an

International Scholarly and Scientific Research & Innovation 6(12) 2012 1652 scholar.waset.org/1307-6892/8748