Venkatesh et al.

/User Acceptance of IT

Qarterly RESEARCH ARTICLE

USER ACCEPTANCE OF INFORMATION

TECHNOLOGY: TOWARD A UNIFIED VIEW^

By: Viswanath Venkatesh Abstract

Robert H. Smith School of Business
University of Maryland Information technology (IT) acceptance research
Van Munching Hall has yielded many competing models, each with
College Park, MD 20742 different sets of acceptance determinants. In this
U.S.A. paper, we (1) review user acceptance literature
wenkate@rhsmith.umd.edu and discuss eight prominent models, (2) empiri-
cally compare the eight models and their exten-
Michael G. Morris sions, (3) formulate a unified model that integrates
Mcintire School of Commerce elements across the eight models, and (4) empiri-
University of Virginia cally validate the unified model. The eight models
Monroe Hall reviewed are the theory of reasoned action, the
Charlottesville, VA 22903-2493 technology acceptance model, the motivational
U.S.A. model, the theory of planned behavior, a model
mmorrls@virginia.edu combining the technology acceptance model and
the theory of planned behavior, the model of PC
Gordon B. Davis utilization, the innovation diffusion theory, and the
Carlson School of Management social cognitive theory. Using data from four
University of Minnesota organizations over a six-month period with three
321 19'^ Avenue South points of measurement, the eight models ex-
Minneapolis, MN 55455 plained between 17 percent and 53 percent of the
U.S.A. variance in user intentions to use information
gdavis@csom.umn.edu technology. Next, a unified model, called the
United Theory of Acceptance and Use of Tech-
Fred D. Davis nology (UTAUT). was formulated, with four core
Sam M. Walton College of Business determinants of intention and usage, and up to
University of Arkansas four moderators of key relationships. UTAUT was
Fayetteville, AR 72701-1201 then tested using the original data and found to
U.S.A. outperform the eight individual models (adjusted
fdavis@walton.uark.edu R^ of 69 percent). UTAUT was then confirmed
with data from two new organizations with similar
results (adjusted f^ of 70 percent). UTAUT thus
Cynthia Beath was the accepting senior editor for this provides a useful tool for managers needing to
paper.

MIS Quarterly Vol. 27 No. 3. pp. 425-478/September 2003 425

Venkatesh et at./User Acceptance of IT

assess the likelihood of success for new techno- The current work has the following objectives:
logy introductions and helps them understand the
drivers of acceptance in order to proactively de- (1) To review the extant user acceptance
sign interventions (including training, marketing, models: The primary purpose of this review
etc.) targeted at populations of users that may be is to assess the current state of knowledge
less inclined to adopt and use new systems. The with respect to understanding individual
paper also makes several recommendations for acceptance of new information technologies.
future research including developing a deeper This review identifies eight prominent models
understanding of the dynamic influences studied and discusses their similarities and dif-
here, refining measurement ofthe core constructs ferences. Some authors have previously ob-
used in UTAUT, and understanding the organiza- served some of the similarities across
tional outcomes associated with new technology models.^ However, our review is the first to
use. assess similarities and differences across all
eight models, a necessary first step toward
Keywords: Theory of pianned behavior, inno- the ultimate goal of the paper: the develop-
vation characteristics, technology acceptance ment of a unified theory of individual accep-
model, sociai cognitive theory, unified model, tance of technology. The review is presented
integrated modei in the following section.

(2) To empirically compare the eight models:

We conduct a within-subjects, longitudinal
validation and comparison of the eight
Introduction models using data from four organizations.
This provides a baseline assessment of the
The presence of computer and information tech-
relative explanatory power of the individual
nologies in today's organizations has expanded
models against which the unified model can
dramaticaiiy. Some estimates indicate that, since
be compared. The empirical model compari-
the 1980s, about 50 percent of all new capital
son is presented in the third section.
investment in organizations has been in informa-
tion technology (Westland and Clark 2000). Yet,
(3) To formulate the Unified Theory of Accep-
for technologies to improve productivity, they must
tance and Use of Technology (UTAUT):
be accepted and used by employees in organi-
Based upon conceptual and empirical simi-
zations. Explaining user acceptance of new tech-
larities across models, we formulate a unified
nology is often described as one of the most
model. The formulation of UTAUT is pre-
mature research areas in the contemporary infor-
sented in the fourth section.
mation systems (IS) literature (e.g , Hu et al.
1999). Research in this area has resulted in
(4) To empirically validate UTAUT: An empirical
several theoretical models, with roots in informa-
test of UTAUT on the original data provides
tion systems, psychology, and sociology, that
preliminary support for our contention that
routinely explain over 40 percent ofthe variance in
UTAUT outperforms each of the eight original
individual intention to use technology (e.g., Davis
models. UTAUT is then cross-validated using
et al. 1989; Taylor and Todd 1995b; Venkatesh
data from two new organizations. The empiri-
and Davis 2000). Researchers are confronted
cal validation of UTAUT is presented in the
with a choice among a multitude of models and
fifth section.
find that they must "pick and choose" constructs
across the models, or choose a "favored model"
and largely ignore the contributions from
Forexample, Moore and Benbasal (1991) adapted the
alternative models. Thus, there is a need for a perceived usefulness and ease of use items from Davis
review and synthesis in order to progress toward et al.'s (1989) TAM to measure relative advantage and
a unified view of user acceptance. complexity, respectively, in their innovation diffusion
model.

426 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et aL/User Acceptance of IT

t—
1
• 1 1
Individual reactions to Intentions to use Actual use of
using information information information
technology technology technology

1
Figure 1. Basic Concept Underiying User Acceptance Modeis

Review of Extant User models and defines their theorized determinants

of intention and/or usage. The models hypo-
Acceptance Models thesize between two and seven determinants of
acceptance, for a total of 32 constructs across
Description of Models the eight models. Table 2 identifies four key
and Constructs moderating variables (experience, voluntariness,
gender, and age) that have been found to be
IS research has long studied how and why indivi- significant in conjunction with these models.
duals adopt new information technologies. Within
this broad area of inquiry, there have been several
streams of research. One stream of research
focuses on individual acceptance of technology by Prior Model Tests and
using intention or usage as a dependent variable Model Comparisons
(e.g., Compeau and Higgins 1995b; Davis et al.
1989). Other streams have focused on There have been many tests of the eight models
implementation success at the organizational level but there have only been four studies reporting
(Leonard-Barton and Deschamps 1988) and task- empirically-based comparisons of two or more of
technology fit (Goodhue 1995; Goodhue and the eight models published in the major informa-
Thompson 1995), among others. While each of tion systems journals. Table 3 provides a brief
these streams makes important and unique overview of each of the model comparison
contributions to the literature on user acceptance studies. Despite the apparent maturity of the re-
of information technology, the theoretical models search stream, a comprehensive comparison of
to be included in the present review, comparison, the key competing models has not been con-
and synthesis employ intention and/or usage as ducted in a single study. Below, we identify five
the key dependent variable. The goal here is to limitations of these prior model tests and compari-
understand usage as the dependent variable. The sons, and how we address these limitations in our
role of intention as a predictor of behavior (e.g., work.
usage) is critical and has been well-established in
IS and the reference disciplines (see Ajzen 1991; Technology studied: The technologies that
Sheppard et al. 1988; Taylor and Todd 1995b). have been studied in many of the model
Figure 1 presents the basic conceptual framework development and comparison studies have
underlying the class of models explaining indivi- been relatively simple, individual-oriented
dual acceptance of information technology that information technologies as opposed to more
forms the basis of this research. Our review re- complex and sophisticated organizational
sulted in the identification of eight key competing technologies that are the focus of managerial
theoretical models. Table 1 describes the eight concern and of this study.

MIS Quarterly Vol. 27 No. 3/September 2003 All

428
^H
•

^^M
•
••

1
1
Hi
1

0
1
1

o
o
••lr

(A
Theory ol Rea sone d Acti

0)
1
re Const

H"
inil

o
to
Drawn froi

E
cial p:sychol k is one of the most

c
0

>

r
k_
0

>
o.

<

.0
ositive ( negati "(3

r-
ct
•o
ngs (e'^alu

"a
1
_c
fundamen tal ar luentiall the. ; of human belhavior. tude Tow

0
0

o
.c

0_

3
ro
cu

C
orming e targ<

sro
ro

CO
S
It has bee vior (F•ishi

ro
o x>
)d to 1predic' lavior

CD
anqe of behav'iors

C
JD

"M

^g
<"
Q.
JT"
(see Shepipard etal. 1988 fora revi ew). Davis et al.

tj
ro
(1989) apiDlied IKA to individuc ceptanee of techno-

0
c
o
"(0

Q.
CO

•o
0

o
Q.

c
pune
pen ception at mo (e who are

O3
03
>,

>,
thati[he vairianc eexiplained was la

ro
t

c
0

o
O o

CO
1 UJI

«
LUI

E
a. "«
C

Q.

5
ro

0
"o

Djeetive ^ Ihmk h( hould

Q.

03
ro
E
eonsistent: with studi t hac ployed TRA in the

_i

l_
c
to
.11

O"
CO

avii in 197;
1 r*.
d.

lest ion" (Fi bein a

CN"

1 context of othe r behiaviors.

Ver)katesh et al./User Acceptance of IT

<
1—
1 Technoto ccepitance Mod AM)

"S m
0
0
0
0
0

XI
-D
O
C

TAM is tai lored to IS conte xts, i

(0

was designed to pre-

.2
O3
lich a p
CD

hat usi
O3

tj
0
3
0
c
0

ai

o
Q.
c

to
to

TO
t/3
"to
diet inforn"latior 1 teehnologj
CD
epta nee and usagi ceived U

ro
_ro
would

Q.
nerjob

"B
TO

O
'0

M/S Quarterly Vol. 27 No. 3/September 2003

the job. Unlike TRA , thefinal e jptualization o fTAM
CJ3"'

ivis 19E p. 321

•2
•n
exeludes t he at ! eon SI derto better eexplain lieh a p ion be
O3

hat usi

0
4—

o
CO
ceived E

'r-'. d
o
°
intention plarsin usiy. :ended TAM b;/ in-
a:

0 tr
O) fD
•§ 3
2 "55
d

T-

L' 0
0 JD
0 CD

would free c ' (Davis

Q. I

^
0 to
SB UJJC

"o p ro
eluding su ive nc iddit ional predictor

CM
ro
CD

J
13
n
.c

13
?

5
•D
OJ
ettings (Venkc

>>
intention \i

1-^

Ji •g
and Davis 200C dely applied t(

3
<

<

Z
If

.>
X)
1-

s—
X)

CO
X)

E
g

AJTPB.

"C
E

iclude
ro
"H. M2onl

if
0
diverse set o f t f rs.

2
"(5
1 Motivatioi lodel (MM)

d
0
0

tn
g

A signifiea nt beidy of resea reh ir

;^

/chology has E )Up-

m

at user
"Q.

vill wa
ro

rform j

r—
c
"c i ;

O
0

0)
c
rr

ro

ported gerleral motiv ation 1:heor an explanatio

3

penceived
_2 E

be ins tal in a ivini

35

0
m 3
>

UJ
0

CO
o
o

o
>
o

LJ
behavior :ral st udies have: exa mined motivaljonal
.^•

rinsic Moi that an listinol le aeti\ tsel

f~

c
<n %

CD
T3
3
CO
0"

tn
£

CJ

theory anc1 ada pted it for s|Decifiiceo ntexts. Valler;

E
Q.

ved job pe rmanc

Q.

or pro mot ion;

ro

c
' ^
"0
cn
d

(1997) pre sent;

CO

ro
CM'

fxcelle nt re'^'lew of thefundanrlental al.

ro
9^
CM"

tenets ot tlIIS theoret loal beise. Withiin the informa tion

0

H
c
0

.9

CJ

systems diomai,
a.
m

CD
vis et ia!. (1992) 1 applied motiv at user i'ill wa rform i etiv
ro

°.
c

c
c
0
0/

c
'0

0
apun
to

03

tional theo ry to rstand teef lnology adopti triinsic Moti ntr einforoi lent ot
(11

"S
/~\

n the proc(
TO

0
0

0
0
>

0
to ' ^

CO

to
ro .£
to

CO

ro
03

and use (s ee al jnkate

Q.
ro

tivity pt
ro
CM

peier 1999). ie" (Di

CM"

3i. 199:
d
 I•
I•
•

wp
•

1
••

m\
^1-2

0
tn
^ H (=

C)
V)

o
o

Kip O
c
ry of 1Planned B ehavi r(TPB reC

[^
•n

n
>

1-
(1)
TJ

"o

CO
c
sxtencJed TRA b:/addi } the Cl perci

CD X
ll
X3
^ioral eontrol. In TPR ereeivf vin ral cc

Q-
_2
<
C
0
3
o
"g
ro

o
c

c
o

"c
«•-

c
.}-
o
t—

B
TJ
TJ

E
0 ro
E

a.
al dete inten

ro
orizeci to be an <idditic RA.

.if

CD
0

±j
.0
0
>

c
>•—
o

ro
lehavi or. Ajzen ;i99i
( )resenl

CO
^t

0
0
.^

CO
c/3
CQ

"TO
dies that sijcces ully usi predi'

'

0
c
o
13
;ion arId behavio r in a de var ings.

X3
o

"g f" n\

x:
T3

CO
ro

ro
een s uccessfull^/ appi d to thf nding

^
0

TJ
<
livi'dual aicceptance and I age of !rent i

Z
0
4—

T3
B

^>
XJ
g

E
E
1—

ro
"Q.

CO
bject

T-
—

O
E 0
loqlies (hlarrison et al. 19 '; Math 1;Ta5

Q
0

TJ

r—

.S c ro
'odd 1 995b). A l'elatei model ist Dmpo

'o

0
x:
o
B
SUJ

LL.
Q.

"5
*lanned Be havio DTPB

0
•

o
c Q.
(0
ntieal t eontr

0 CD
;tin g inte ntion, DTP B is ic
0

0
0
0
^ •

>

0
O
o

a
"o
cr

ease • ffieul perf ormi

Q
0
1

1.

u 1-
H- T3 o
<

a.
c

0
0

0
n
Bbut similar to 1lAM, s" att
Oi
T—

Q
c
U)

x;
o
o

TJ
a;
CO

C13

"2
. 188 the coni
_,_:,

rceiv lavi

c
"o

T3
sn

0'
0
c
c 0 c

0
0

jf
•o

f—

subje etive norm , and jfceive al eo

tn

0 o
CO
B
o
'ro

CJ

h
nst

Q- CJ
ntrol eptior \ intei

-
1

0 g o
c
O
O

o ro E

CO
D3
a o c

0
>
Q. Q

>>
C

E "0 —
o
s the underlying beliei tructur o xro:
X3
"ro CJ)

cn
TD

CU

raylor tTod 95b,

d.

Q-
tion cc3ntexts.

o
o
E
bined TAM and TPB' -TAM-

TJ

OQ 1 -

1-
—
TJ
o
<

CO
ro
Tiodel combines the p dictors
it
l~

T5
o

TJ
_0
CL

I
E
CQ

ro
"5.

1— t~
^
K 0

o. "o 2

UJ
"s ^

tJ
X3

? 0
fived IJ sefulness from \ M t o r

Q.
<

z0
>

TJ
TJ
C73

TJ
D ro 0
o
o

TJ
TJ
«t
Q.

E
B
5

H (Ta>'lor and To' o

CD

E
a:

ro

< CO CO
"a.

.a"

"c?
0
0
0
sz

Tl
o
"TO

rceiv lavi
<
0
H-
•D

•a
D.

^— 1—
g
E
CD

Q.
ro

o_ o
O
ntrol
1

<
^

0
0
0
o

TJ
B

TJ
D[0
1—

D.
E

TD

ro

CO
CO
"a.

rceiv "AM.
CD

MIS Quarterly Vol. 27 No. 3/September 2003

Venkatesh et al./User Acceptance of IT

429
Venkatesh et ai/User Acceptance of IT

c
o o
c 05
^ CD ^ B o Q.
if) ro c .9 ~ E
0) * - ^ Q.
O) 3 CO «- ro 01 3 05 !S2 ^ o
E o ^ iyro
0 O CN o r^ CO

•D i t ' - .b c
>S^ o . ro
ro «_ 0 .9
x: o ro £ "to (0 Q.
« g ro • ^
r-..
3 ro 0 0
cn -Q ^ tfl CO tu
2 0
•c- ._ CD 0 -D 5
0 O C LL
3 c "TO CD O C
CD 0)
_3 CL
-n O D
ro 0) •>
Ii CJ

iR 2: ' o CD
E rf ^ ro 3
c to
•5 c >.
" cn 0 (N
fr ro Q. T- ca ro
i o ii
£Z 0
-_
0
0
i_ ro
2^ _ CO
^ ro ^ a ii ^
0
E
.2
E
u
'0 o —.. ->-'
ro CD _: oi «, 2 S5 9 3 J 0 Q-
Si 3 0
0 S; n% 01

CO CO
^
•Q
0
-'-
C
3
"S
(O
0
.2 0 c .E £
o „ •> ro Q-
CO

III '^ (-
01
^
E
o cn
ci
^•

0
«
O

. 2 CD
CO
CD
-CJ = 0
0
•^
_
ro
o p ^ i=
cn

0 "^ 0 t" "K - C . ^ ^ c ro Q. c ro

0) O ra S "c ro i5 ,_ 0 x: 3 o *-
CD 5 D
P CD ED 0 O .E ro O E .-^ O CO
O 0

CO

c
o
O
cn
E 5 c
a. CO
E o
o o ro
O 58 CO

0
c > c
_: P ^ ro — C B CD .i c
t 0 ro " ro ro
O O)
O 0 ro 2!
0 to d 3 ^ — x: •- sz
-9 € 0 0 *- 05 * - E ^
i£
</> _
ro
N Q. 3 o j=
ro CO
ro -c
X 0

£ <^
Q. 0

Ii
= * i 13 05 * -
3 0) ro Q. ro
(••1 <i^
II
0
c
05 ro
CD ^
^
o
CL ^ -co D_ ro CJ P o £ O
iC E
o t
•—
ra
il
CJ E £
X> 0 Q.
CD -O 0 . ^
t (1 to n
o; 0
"I JZ

I ^
-^

o
P E
ro — 0
•53 £
0 .0
'5) 0
ro O D)
0"
fl 0 c
^ 8
^
0 = . 9 CD
I-
E
Q-
0
O O

0
0
3
•>

•o ^
c
I=
ti
2 | C 0 !_' O
>. t 0 T3 •9 a. P c
0 01 g ^
Q. Q.
ro
E Z
o 9; CD O)
ro * 1 P ro ^ *"' Q. £^
S 0 ro .».-
-D P Q. ro c QJ .2 c 3
o >-
s •> ro ^ ^ -
0-

Q
> ro *- 01 w 3
X O
^ o ^ B I 0
^£ T-
p t
I 1°
01 01
c 05
ro t-

430 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance ot IT

1r

dto 1
1

ore 1
Stem 1
05

ing

ing

ing
- ro

sat
0 0
ra 0 0
E > . to ro o
X) X "* 0 CO X3 > ro
^H
•
CO 0 —5 CD X I (fl CO C c ^- 0 CO
^H Ol

•
CJ S? O
ro 1 —
ro cn 1= c ro ro 0 .9 = ro
•a 0 01
TD
0 "TO 0 D- ^ Ol CO ro if" Q. c

usin
iocia
CO
ci.

and
n is
ro ro ro to CD
0
CJ L c o

•
O "' .9 ro
0
fd
0 0 CJ)
• " CO
(0 0 "^ o "£^Ei
0 o D. fli O ro
isat
« .52 o c ^ sz p

nov
c nj C O If p C CO - O
.g .c .c c CT 0 E o 0 E2 0) CJ
.E

•
0
"TO c^ c t/) t/)
o "ra ^ 0 -^ "D Q) •—•

o ro 3 D- c .c.

iir
c CD o > S" P ro
c 2^ p iS 1- ro B cn TS i2 >. c^ • ^
s
.- -E c ro 3 1= o 0
• — E
m
Cu
0 12 g
"-1 O ^
0} Q.
to ^ Csl
0
Q c c ro c CO TO 05
2_
ro CO ro 0 P ro 2J CO
d. 3
>s
^H

•
.c ^ O .c •5- .c CD C
"o
^
^H
H tj
{-
3
CJ c O ;= ro (1 CJ CJ 0 •--•
0
^~
^
tn CJ
0 x: "> E -C J = the 0rgani;zatic £ P p
5 S i: a. 'o o
I «
o CO a
P p :§ "p - ^ 1
gree 1

"*"^ 'j^s r r
gree t

gree t

igibilit

nbasc
ttert han it

gree t
ig the
"" 01
•

-•—' LO
jfinit ions

0
2! o
3
E o i .„ 0 "^ ^
3
<7)
t-
O) 0 Cl fl5 c _•
o

91, p
O) " ^ nsisti
perie
ficult

0 a 0 0 0 > Q.
91).

0 _ro c T3 CD C31 . _ '

0 0 0 0 0 ._
f— *
O5
0 CT) ^ CJ XZ
x: jE c 3 o X cn 0 O
• • * "
p 0 e \- O 0 -r- ?^ - ro f

CD
KTB 0 CO
Ol D
^ C f l (A «•—
Hi c o
ra o CO
> 0 8 05
^^-^ t/) 0 CD
Comp tibi

VolunI irin
Q
u
0 2O ro
>>
™
Image

ResuH
strabil
Relati

Ease
|Core

E >

I1
^ 0 O)
novat ions,

- o
ifital 199
tivfi

C
0 o
1 red id
ould 1

0
eseni

ice. f

eeal£
1991)

JD
to
nal

ro — C_J ^
ca
x: c g £ ~— a
"^ . . ro c
l O >•— ro "TO (o ra " Q . c
CJ) N > 0) c £ 0
C/)
o ro
Ol ro O *^ CJ I
x:
E
ers 1!

. 1998; Kara
cteris
ral 30lS to orgar

ort fo
'e and Benb
Klein 1I9«2)
:udy a \ /ariet

p: u) CJ
ro o ro
P
o cn 1c 2 S5
^ ° Q. ro
E p
i8|
3 CC

o
0) Q to *-•
o u) T=;
O ^ o
0 c f^

B h-
c cn
o ro P "S 0
2 •- CO ro
- 3 CJl
CJl

o 2 ro ^ o
o o 0)' T3
divi
d in S' cioli

inn
II dL

CD
ned
ncL

ro E
Ct

1
It Q
O5
O)
B
(0 ,i t= C
ai
0)
I/;
o
c
novati on Di

e LUOJ

0 0) Q.
o
d BenIDasat

-C to •- - ^
0 •o
c "n
It c o £ = "I ro
ed sini

lidity o
lovatio

QJ "0
D) "TO
1 0 J2 .2 0
O) £ "S. Si X3
o ro S* 0
X3 O to c
LE •^
ro <o DI
.E ro Ul D ro

M/S Quarterly Vol. 27 No. 3/September 2003 431

Venkatesh et al./User Acceptance of IT

432 M/S Quarterly Vol. 27 No. 3/September 2003

 V/N
Theory of Experience was not explicitly included in Voluntarmess was not N/A
Reasoned the original TRA. However, the role of included in the original
Action experience was empirically examined TRA. Although not
using a cross-sectional analysis by Davis tested, Hartwick and
et al. (1989). No change in the salience of Barki (1994) suggested
determinants was found. In contrast, that subjective norm
Karahanna et al. (1999) found that attitude was more important
was more important with increasing when system use was
experience, while subjective norm became perceived to be less
less important with increasing experience. voluntary.
Technology Experience was not explicitly included in Voluntariness was not Gender was not in- N/A
Acceptance the original TAM. Davis et al. (1989) and explicitly included in the cluded in the original
Model (and Szajna (1996), among others, have pro- original TAM. Within TAM. Empirical evi-
TAM2) vided empirical evidence showing that TAM2, subjective norm dence demonstrated
ease of use becomes nonsignificant with was salient only in that perceived useful-
increased experience. mandatory settings and ness was more salient
even then only in cases for men while perceived
of limited experience ease of use was more
with the system (i.e., a salient for women
three-way interaction). {Venkatesh and Morris
2000). The effect of
subjective norm was
more salient for women
in the early stages of
experience {i.e., a
three-way interaction).
V/N
N/A

MIS Quarterly Vol. 27 No. 3/September 2003

Motivational N/A N/A
Model
Venkatesh et ai/User Acceptance of IT

433
434
K
E
•
••

lHIi
1

H^

2
0

0
U
c
^•<
•del Exper 1 Volu

• l

1
•r

>

t ^
0
0

c
c

o
C
ooo;
to
ro

"ro
Experi nee was 1not exp licl tly inel udei Votu ari sS was not 3sh et a orri Ven :esh 1

0
o

1- 0 .
0
0
o
o

C
C

(8
•o

T3
CN
the ori inal TPB or DTP
ro "ro

"5

cd
It has inclu :he original lat attitL was ldth atti-

«
0
I.
c 1-
0

>
C
"T3

a> o
o

tn

o
d
c

Q.

Q
ior ineorp rate d into TPBvi

m
TPB B. As alient foi was lore ient 1
Ver^katesti et at./U$er Acceptance of tT

ro 0
3

c
0

•n
0
CO - ^

UJJ'

,_:,

(0
E CQ
o "o
tn

di
!orri!3 and Venkat h200C Bunc
notei diseussion r wor

1
a

(j
0

H
C

•a

d)
d)
ro

75
"to
f ^

Empiri videnice has

Q.

CD
nonsti rega eeived 1 per( Jived
ro

0
3
>
—

0
0
0
0

o
cn -C

x:
JOLU

r~
if

"ro
c O)
experi nee modejrates tl relatio nshi t tested, Dntrol w viors eont rol was 1

ro
ro
0

c 0
o c

o
0
0
0

(0
o
tn
cn
ro
"o

.E
betwe
•2 5 % E

1 bjective norm a nd bellavii subjf orm was alie for wom

0
0
0

M/S Quarterly Vol. 27 No. 3/September 2003

t_

•o
UJJ
-5.

o
o

c0 c
CJ
o o

intenti lat subj' ive no sugg ste D be more 'St ages of )er- ers. iubje Cti'

>

0
-

C
o

•c

r—
UJJC

tn _3)
CO
tu

ro

CD

DJ
becorr !ss im portant th incr easi impo len system .e., threi Tiore liient 1

0
-
0

^ -

0
o

o
o
o

tn
ro
Q.

CO
B
levels •f ex pener ice. Th IS simi lart use \ ceived to ions).

.sz
men

0
3

XI
"o 0

_0

U)
in

to

tu
ro
o
•(UO!

cn

CO
c o •B

CO
sugge tion of Ka rahann (tal. (1 ntary (Hart- nterj

.—.
ro

0
c
c

1-
•o

o
o
o
CQ
ro
text 1 wiek rki 1994).

ro

0
<
z
Z

o
5

o
iqiu
ined Experi nee was i ncorpo Bd inteithiE N/A

s
<
—
model en-sub jects des ign

ro

ax
.=
dj

111
tn

~l
(exper id ancJ inexpt meed i

>
—'
^-
5

•c g
m ro
Pereei jsefuliness, a Jde to'

Si
o

c
UJ

.o
03
0 |B-

o.
rceived ihavioi

o
E
were £ re sal ient wit icreasling

0
0

>

0
O
LUJOU t

»
E
experi while subjee

ro
0

m
less s< ient with 1ncreasi exper ienc

cn

i_
(Taylo and Todd 1995a
 V/N V/N V/N
Model of PC Thompson et al. (1994) found that com-
Utilization plexity, affect toward use, social factors,
and facilitating conditions were all more
salient with less experience. On the other
hand, concern about long-term conse-
quences became increasingly important
with increasing levels of experience.
V/N
Innovation Karahanna et al. (1999) conducted a Voluntariness was not N/A
Diffusion between-subjects comparison to study the tested as a moderator,
Theory impact of innovation characteristics on but was shown to have
adoption (no/low experience) and usage a direct effect on
behavior (greater experience) and found intention.
differences in the predictors of adoption
vs. usage behavior. The results showed
that for adoption, the significant predictors
were relative advantage, ease of use, trial-
ability, results demonstrability, and visi-
bility. In contrast, for usage, only relative
advantage and image were significant.
V/N
N/A
V/N N/A
Social
Cognitive
Theory

MIS Quarterly Vol. 27 No. 3/September 2003

Venkatesh et al./User Acceptance of IT

435
 ••
•

•
•

436
B
•
•

EI
I•l

B
^CB
H^v
m
1

-O
o

o
c
text of Stuc

1
Newne ss of

• s^
H ^

&
c

o o
(0
' (A
o
c

3
o

Z
(Inc Techn<jlogy ints Lon

ra

£ o
ft

<u a
•n

c
3

••c

o w
(A
c

>•-
E s

CL
n

E
o

Ol
ared | Tedhnot<

.!2
iZ
1 Studiei

at

E
Ana

o _>,

<

Q
0

>
T -
Z}
0

Q. -u
o to
j2 jg
1-
1S-UII
c

"TO
.c

With ibjects

ro
re a>
1 Partioiplants eks Croj riance

1/3

CO
u
o
c
•o
"D

o
"to
mod mpari- were n(

Q.
sect 8 expls

d ra
ra us

1— rco

ro
0
C

CO
CN
_C CN
0 CD

son of int ention the ted1 nolog anal ysi :% and

p
ro <

C/1 C
1-
5

(-^
TS o ^

XJ

and use (

ro
ci^

CO
"c > . ro LO

"TO C/3
two poi ra
CO

0
0
proc esso time lively.
(/)

5
<
0

c
0

1.
3
0S8

GQ
O. 0

TJ

</i
t-

CD
c

CN
CN
c
c

m
tn

O
thi

cn
c

_c

i^een- subject

ro
1 Some famil- CroE riance
ra

"0
o
0

o
I—

mod mpari- iarity wiith the

"TO
sect ion xpl ed by'
ro
Venkatesh et al./User Acceptance of IT

o
b
son ot int ention t technology as
is

Id TPB
c ra

ro
use
-pea

Q.
each p£irtici-
shee pant ha dto

ro
calcijiatoi choose a tecl'

MIS Quarterly Vol. 27 No. 3/September 2003

nology to per-
1 form a 1[ask

<
»
«
c
•D
c

TS
c

U.

d)
.c
O

0
0
1
c
c

CD
1-

ns-uii lOLU-

Z3
"m
tf)
ran With 1 bjects .sz
g

1 Many student riance

ro
1

,^

t-
•o
ro o
Q

1-
•0

Q.
1

CD
0

o
o
10
ro c

TPB mod mpari-

V)

.
were al ready
CL

Q.
ants secti ed by'
ro
ro

ro

0
—

o
0
1-
"O

CJ
LO
in

QJ
son ot
1 intention t familiar With tl VISI
CO

<n
c

PBwa:
0
Q
X rsi 1 -

use. a computing
CQ
CL

center was 60
*^
E

reso urce center

CD

' i.e. ipk

0

E a
Q.
?
:nt.

3
o
"«
0

CD
c
c

c
o

to

o
>

With in-su bjects

"S
£

Survey nance i :ent

O
o
n
C
0

•0
0
o

Q. CM
o
o
E
o

mod mpari- adminisitered

"TO

ro
secti

</3
ed by 1
tf?

ro
ra

{ -
w
nj
Ln

son (ot be haviora

as " Q . 5?
cn

after 10• monti

cn

Id IDT
ro c ro

interition 1to use of use

and use iln the

o
(U
cont f a mar-
kettirial 01
eledTonic

S ro"
to Q.

c
:em

F ^
art card

^
 Ver}katesh et al.AJser Acceptance of IT

Participants: While there have been some mandatory settings that are possibly of more
tests of each model in organizational settings, interest to practicing managers. This re-
the participants in three of the four model search examines both voluntary and man-
comparison studies have been students— datory implementation contexts.
only Plouffe et al. (2001) conducted their
research in a nonacademic setting. This
research is conducted using data collected
from employees in organizations. Empirical Comparison of the
Timing of measurement: In general, most of
Eight Models ^ ^ ^ ^ ^ ^ ^ ^ ^ ^
the tests of the eight models were conducted
well after the participants' acceptance or Settings and Participants
rejection decision rather than during the
active adoption decision-making process. Longitudinal field studies were conducted at four
Because behavior has become routinized, organizations among individuals being introduced
individual reactions reported in those studies to a new technology in the workplace. To help
are retrospective (see Fiske and Taylor 1991; ensure our results would be robust across
Venkatesh et al. 2000). With the exception of contexts, we sampled for heterogeneity across
Davis et al. (1989), the model comparisons technologies, organizations, industries, business
examined technologies that were already functions, and nature of use (voluntary vs.
familiar to the individuals at the time of mea- mandatory). In addition, we captured perceptions
surement. In this paper, we examine techno- as the users' experience with the technology
logies from the time of their initial introduction increased. At each firm, we were able to time our
to stages of greater experience. data collection in conjunction with a training
program associated with the new technology
Nature of measurement: Even studies that introduction. This approach is consistent with
have examined experience have typically prior training and Individual acceptance research
employed cross-sectional and/or between- where individual reactions to a new technology
subjects comparisons (e.g., Davis etal. 1989; were studied (e.g., Davis et al. 1989; Olfman and
Karahanna et al. 1999; Szajna 1996; Taylor Mandviwalla 1994; Venkatesh and Davis 2000).
and Todd 1995a; Thompson et al. 1994). A pretested questionnaire containing items mea-
This limitation applies to model comparison suring constructs from all eight models was
studies also. Our work tracks participants administered at three different points in time:
through various stages of experience with a post-training (TI), one month after implementation
new technology and compares all models on (T2), and three months after implementation (T3).
all participants. Actual usage behavior was measured over the six-
month post-training period. Table 4 summarizes
Voluntary vs. mandatory contexts: Most of key characteristics of the organizational settings.
the model tests and all four model com- Figure 2 presents the longitudinal data collection
parisons were conducted in voluntary usage schedule.
contexts.^ Therefore, one must use caution
when generalizing those results to the
Measurement
A questionnaire was created with items validated
Notable exceptions are TRA (Hartwick and Bark! 1994)
and TAM2 (Venkatesh and Davis 2000) as well as in prior research adapted to the technologies and
studies that have incorporated voluntariness as a direct organizations studied. TRA scales were adapted
effect (on intention) in order to account for perceived from Davis et al. (1989); TAM scales were
nonvoluntary adoption (e.g., Agarwal and Prasad 1997;
Karahanna et al. 1999; Moore and Benbasat 1991). adapted from Davis (1989), Davis et al. (1989),

MIS Quarterly Vol. 27 No. 3/September 2003 437

 Venkatesh et al./User Acceptance of IT

X 0 X 0 X 0 X 0
Training User System User System User System Usage
Reactions Use Reactions/ Use Reactions/ Use Measurement
Usage Usage
Measurement Measurement
1 week 1 month 3 months 6 months

Figure 2. Longitudinal Data Collection Schedule

Table 4. Description of Studies

Functional Sample
Study Industry Area Size System Description
Voluntary Use
Online meeting manager that could be
Product used to conduct Web-enabled video or
la Entertainment 54
Development audio conferences in lieu of face-to-face
or traditional phone conferences
Database application that could be used
Telecomm to access industry standards for particular
1b Sales 65
Services products in lieu of other resources (e.g.,
technical manuals, Web sites)
Mandatory Use
Business Portfolio analyzer that analysts were
2a Banking Account 58 required to use in evaluating existing and
Management potential accounts

Public Proprietary accounting systems on a PC

2b Accounting 38 platform that accountants were required
Administration
to use for organizational bookkeeping

and Venkatesh and Davis (2000); MM scales were perceived voluntariness was measured as a
adapted from Davis et ai. (1992); TPB/DTPB manipulation check per the scale of Moore and
scales were adapted from Taylor and Todd Benbasat (1991), where 1 was nonvoluntary and
(1995a, 1995b); MPCU scales were adapted from 7 was completely voluntary. The tense of the
Thompson et al. (1991); IDT scales were adapted verbs in the various scales reflected the timing of
from Moore and Benbasat (1991); and SCT scales measurement: future tense was employed at T I ,
were adapted from Compeau and Higgins (1995a, present tense was employed at T2 and T3 (see
1995b) and Compeau et al. (1999). Behavioral Karahanna et al. 1999). The scales used to mea-
intention to use the system was measured using sure the key constructs are discussed in a later
a three-item scale adapted from Davis et al. section where we perform a detailed comparison
(1989) and extensively used in much of the (Tables 9 through 13). A focus group of five
previous individual acceptance research. Seven- business professionals evaluated the question-
point scales were used for all of the aforemen- naire, following which minor wording changes
tioned constructs' measurement, with 1 being the were made. Actual usage behavior was mea-
negative end of the scale and 7 being the positive sured as duration of use via system logs. Due to
end of the scale. In addition to these measures. the sensitivity of usage measures to network

438 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

availability, in all organizations studied, the system findings emerged from these analyses. First, all
automatically logged off inactive users after a eight models explained individual acceptance,
period of 5 to 15 minutes, eliminating most idle with variance in intention explained ranging from
time from the usage logs. 17 percent to 42 percent. Also, a key difference
across studies stemmed from the voluntary vs.
mandatory settings—in mandatory settings (study
2), constructs related to social influence were
Results
significant whereas in the voluntary settings (study
1), they were not significant. Finally, the deter-
The perceptions of voluntariness were very high in
studies l a and 1b (la: M = 6.50, SD = 0.22; 1b: minants of intention varied over time, with some
M = 6.51. SD = 0.20) and very low in studies 2a determinants going from significant to nonsigni-
and 2b (la: M = 1.50, SD = 0.19; 1b: M = 1.49. ficant with increasing experience.
SD = 0.18). Given this bi-modal distribution in the
data (voluntary vs. mandatory), we created two Following the test of the baseline/original specifi-
data sets: (1) studies 1a and 1b, and (2) studies cations of the eight models (Tables 5 and 6). we
2a and 2b. This is consistent with Venkatesh and examined the moderating influences suggested
Davis (2000). (either explicitly or implicitly) in the literature—i.e.,
experience, voluntariness, gender, and age
(Table 2). In order to test these moderating influ-
Partial least squares (PLS Graph, Version
ences, stay true to the model extensions
2.91.03.04) was used to examine the reliability
(Table 2), and conduct a complete test of the
and validity of the measures. Specifically, 48
existing models and their extensions, the data
separate validity tests (two studies, eight models,
were pooled across studies and time periods.
three time periods each) were run to examine
Voluntariness was a dummy variable used to
convergent and discriminant validity. In testing
separate the situational contexts (study 1 vs.
the various models, only the direct effects on
study 2); this approach Is consistent with previous
intention were modeled as the goal was to
research (Venkatesh and Davis 2000). Gender
examine the prediction of intention rather than
was coded as a 0/1 dummy variable consistent
interrelationships among determinants of inten-
with previous research (Venkatesh and Morris
tion; further, the explained variance (R^) is not
2000) and age was coded as a continuous vari-
affected by indirect paths. The loading pattern
able, consistent with prior research (Morris and
was found to be acceptable with most loadings
Venkatesh 2000). Experience was operationa-
being .70 or higher. All internal consistency
lized via a dummy variable that took ordinal values
reliabilities were greater than .70. The patterns of
of 0, 1. or 2 to capture increasing levels of user
results found in the current work are highly con-
experience with the system (TI. T2, and T3).
sistent with the results of previous research.
Using an ordinal dummy variable, rather than
categorical variables, is consistent with recent
PLS was used to test all eight models at the three research (e.g., Venkatesh and Davis 2000).
points of measurement in each of the two data Pooling the data across the three points of mea-
sets. In all cases, we employed a bootstrapping surement resulted in a sample of 645 (215 ^ 3).
method (500 times) that used randomly selected The results of the pooled analysis are shown in
subsamples to test the PLS model.'' Tables 5 and Table 7.
6 present the model validation results at each of
the points of measurement. The tables report the
variance explained and the beta coefficients. Key Because pooling across time periods allows the
explicit modeling of the moderating role of exper-
ience, there is an increase in the variance ex-
plained in the case of TAM2 (Table 7) compared
* The interested reader is referred to a more detailed to a main effects-only model reported earlier
exposition of bootstrapping and how it compares to other
techniques of resampling such as jackknifing (see Chin (Tables 5 and 6). One of the limitations of pooling
1998; Efron and Gong 1983). is that there are repeated measures from the

MIS Quarteriy Vol. 27 No. 3/September 2003 439

Venkatesh et al./User Acceptance of IT

Table 5. Study 1: Predicting Intention in Voluntary Settings

TImel (N = 119) Time 2 (N = 119) Time 3 (N = 119)

Models Independent variables R= Beta R^ Beta R^ Beta

TRA Attitude toward using tech. .30 .55'" .26 .51'" .19 .43"*
Subjective norm .06 .07 .08
TAM/ Perceived usefulness .38 .55*" .36 .60'" .37 .61"'
TAM2 Perceived ease of use .22** .03 .05
Subjective norm .02 .06 .06
MM Extrinsic motivation .37 .50*" .36 .47'" .37 .49"*
Intrinsic motivation .22" .22" .24'"
TPB/ Attitude toward using tech. .37 .52"' .25 .50"' .21 ,44-**
DTPB Subjective norm .05 .04 .05
Perceived behavioral control .24'*' .03 .02
C-TAM- Perceived usefulness .39 .56**' .36 .60'" .39 .63"'
TPB Attitude toward using tech. .04 .03 .05
Subjective norm .06 .04 .03
Perceived behavioral control .25"* .02 .03
MPCU Job-fit .37 .54'" .36 .60"* .38 .62"'
Complexity (reversed) .23"' .04 .04
Long-term consequences .06 .04 .07
Affect toward use .05 .05 .04
Social factors .04 .07 .06
Facilitating conditions .05 .06 ,04
IDT Relative advantage .38 .54'" .37 .61"* .39 .63*"
Ease of use .26** .02 .07
Result demonstrability .03 .04 .06
Trialability .04 .09 .08
Visibility .06 .03 .06
Image .06 .05 .07
Compatibility .05 .02 .04
Voluntariness .03 .04 .03
SCT Outcome expectations .37 .47**' .36 .60'*' .36 .60*"
Self-efficacy .20*" .03 .01
Affect .05 .03 .04
Anxiety -.17' .04 .06

Notes: 1. 'p < .05; " p < .01; " ' p < .001.
2. When the data were analyzed separately for studies 2a and 2b. the pattern of results
very similar.

440 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

Table 6. Study 2: Predicting Intention in Mandatory Settings

Time 1 {N = 96) Time 2 (N = 96} Time 3 (N = 96)
Models Independent variables R^ Beta R^ Beta R' Beta
TRA Attitude toward using tech. .26 .27*" .26 .28"* .17 ,40**'
Subjective norm .20" .21" .05
TAM/ Perceived usefulness .39 .42*" .41 .50*" .36 .60'"
TAM2 Perceived ease of use .21* .23" .03
Subjective norm .20* .03 .04
MM Extrinsic motivation .38 .47"* .40 .35 .44"'
Intrinsic motivation .21" .24" .19"
TPB/ Attitude toward using tech. .34 .22' .28 .36'" .18 .43"'
DTPB Subjective norm .25"* .26" .05
Perceived behavioral control .19* .03 .08
C-TAM- Perceived usefulness .36 .42'" .35 .51'" .35 .60"*
TPB Attitude toward using tech. .07 .08 .04
Subjective norm .20* .23*' .03
Perceived behavioral control .19' .11 .09
MPCU Job-fit .37 .42"' .40 .50"' .37 .61*"
Complexity (reversed) .20* .02 .04
Long-term consequences .07 .07 .07
Affect toward use .01 .05 .04
Social factors .18' .23" .02
Facilitating conditions .05 .07 .07
IDT Relative advantage .38 .47*" .42 .52"' .37 .61""
Ease of use .20' .04 .04
Result demonstrability .03 .07 .04
Trialability .05 .04 .04
Visibility .04 .04 .01
Image .18* .27" .05
Compatibility .06 .02 .04
Voluntariness .02 .06 .03
SCT Outcome expectations .38 .46"' .39 .44'" .36 .60'"
Self-efficacy .19" .21"* .03
Affect .06 .04 ,05
Anxiety -.18' -.16' .02

Notes: 1. *p < .05; " p < .01; '**p < .001.

2. When the data were analyzed separately for studies 2a and 2b, the pattern of results was
very similar.

MIS Quarterly Voi 27 No. 3/September 2003 441

Venkatesh et al./User Acceptance of IT

Table 7. Predicting Intention—Model Comparison Including Moderators: Data

Pooled Across Studies (N = 645)
Model Version Independent Variables R^ Beta Explanation
TRA 1 Attitude (A) .36 .41*" Direct effect
Subjective norm (SN) .11
Experience (EXP) .09
Voluntariness (VOL) .04
A x EXP .03
SN X EXP -.17' Effect decreases with Increasing
experience
SN X VOL .17' Effect present only in mandatory
settings
TAM 2a Perceived usefulness (U) .53 .48*" Direct effect
TAM2 Perceived ease of use (EOU) .11
Subjective norm (SN) .09
Experience (EXP) .06
Voluntariness (VOL) .10
EOU X EXP -.20*' Effect decreases with increasing
experience
SN X EXP -.15
SN X VOL -.16* Cannot be interpreted due to
presence of higher-order term
EXP X VOL .07
SN X EXP X VOL -.18" Effect exists only in mandatory
settings but decreases with
increasing experience
2b Perceived usefulness (U) .52 .14* Cannot be interpreted due to
TAM presence of interaction term
incl. Percd. ease of use (EOU) .08
gender Subjective norm (SN) .02
Gender (GDR) .11
Experience (EXP) .07
U X GDR ,31"' Effect is greater for men
EOU X GDR -.20" Effect is greater for women
SN X GDR .11
SN X EXP .02
EXP X GDR .09
SN X GDR X EXP ,17" Effect is greater for women but
decreases with increasing
experience

442 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

Table 7. Predicting Intention—Model Comparison Including Moderators: Data

Pooled Across Studies (N = 645) (Continued)
Model Version Independent Variables R^ Beta Explanation
MM 3 Extrinsic motivation ,38 ,50*** Direct effect
Intrinsic motivation ,20*** Direct effect
TPB/ 4a Attitude (A) ,36 .40*" Direct effect
DTPB TPB Subjective norm (SN) .09
inci, voi Percd. behrl. control (PBC) .13
Experience (EXP) .10
Voluntariness (VOL) .05
SN X EXP -.17** Effect decreases with increasing
experience
SN X VOL ,17" Effect present only in mandatory
settings
4b Attitude (A) ,46 .17*** Cannot be interpreted due to
TPB presence of interaction term
incl. Subjective norm (SN) .02
gender Percd. behri, control (PBC) .10
Gender (GDR) ,01
Experience (EXP) ,02
A x GDR ,22*" Effect is greater for men
SN X EXP -.12
SN X GDR ,10
PBC X GDR .07
PBC X EXP ,04
GDR X EXP ,15* Term included to test higher-
order interactions below
SN X GDR X EXP -.18" Both SN and PBC effects are
higher for women, but the effects
PBC X GDR X EXP -16* decrease with increasing
experience
4c Attitude (A) .47 ,17"" Cannot be interpreted due to
TPB presence of interaction term
incl. Subjective norm (SN) ,02
age Percd. behrl. control (PBC) ,10
Age (AGE) ,01
Experience (EXP) .02
A x AGE -.26*** Effect is greater for younger
workers
SN X EXP -.03
SN X AGE .11
PBC X AGE ,21" Effect is greater for older
workers

Quarterly Vol. 27 No. 3/September 2003 443

Venkatesh et aiJUser Acceptance of IT

Table 7. Predicting Intention—Model Comparison Including Moderators: Data

Pooled Across Studies (N = 645) (Continued)
Model Version Independent Variables R^ Beta Explanation
AGE X EXP .15* Term included to test higher-
order interaction below
SN X AGE X EXP -18" Effect is greater for older
workers, but the effect
decreases with increasing
experience
C- 5 Perceived usefulness (U) .39 .40*** Direct effect
TAM- Attitude (A) ,09
TPB
Subjective norm (SN) .08
Perceived beholder control .16* Cannot be interpreted due to
(PBC) presence of interaction term
Experience (EXP) .11
UxEXP .01
A x EXP ,08
SN X EXP -.17* Effect decreases with increasing
experience
PBC X EXP -.19** Effect decreases with increasing
experience
MPCU 6 Job-fit (JF) .47 ,40*** Direct effect
Complexity (CO) (reversed) .07
Long-term consequences (LTC) .02
Affect toward use (ATU) .05
Social factors (SF) .10
Facilitating condns. (FC) ,07
Experience (EXP) .08
CO X EXP (CO—reversed) -.17* Effect decreases with increasing
experience
LTC X EXP .02
ATU X EXP ,01
SF X EXP -,20"* Effect decreases with increasing
experience
FC X EXP .05
IDT 7 Relative advantage (RA) .40 .49*" Direct effect
Ease of use (EU) .05
Result demonstrability (RD) .02
Trialability (T) .04
Visibility (V) .03
Image (1) .01
Compatibility (COMPAT) .06
Voluntariness of use (VOL) ,11

444 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et at./User Acceptance of IT

Table 7. Predicting Intention—Model Comparison Including Moderators: Data

Pooled Across Studies (N = 645) (Continued)
Model Version Independent Variables R^ Beta Explanation
Experience (EXP) .03
EU X EXP -.16* Effect decreases with increasing
experience
RD X EXP .03
V X EXP -.14" Effect decreases with increasing
experience
1 X EXP -.14* Effect decreases with increasing
experience
Voluntariness of use (VOL) .05
SCT 8 Outcome expectations .36 44**- Direct effect
Self-efficacy .18' Direct effect
Affect .01
Anxiety -.15* Direct effect

Notes: 1. *p < .05; " p < .01; '**p < .001.

2. The significance of the interaction terms was also verified using Ohow's test.

same individuals, resulting in measurement errors cross-sectional tests of TAM (without moderators).
that are potentially correlated across time. The explained variance of TRA, TPB/DTPB,
However, cross-sectional analysis using Chow's MPCU, and IDT also improved. For each model.
(1960) test of beta differences (p < .05) from each we have only included moderators previously
time period (not shown here) confirmed the tested in the literature. For example, in the case
pattern of results shown in Table 7. Those beta of TAM and its variations, the extensive prior
differences with a significance of p < .05 or better empirical work has suggested a larger number of
(when using Chow's test) are discussed in the moderators when compared to moderators sug-
Explanation" column in Table 7. The interaction gested for other models. This in turn may have
terms were modeled as suggested by Chin et al. unintentionally biased the results and contributed
(1996) by creating interaction terms that were at to the high variance explained in TAM-related
the level of the indicators. For example, if latent models when compared to the other models.
variable A is measured by four indicators ( A l , A2, Regardless, it is dear that the extensions to the
A3, and A4) and latent variable B is measured by various models identified in previous research
three indicators ( B l , B2, and B3), the interaction mostly enhance the predictive validity of the
term A " B is specified by 12 indicators, each one various models beyond the original specifications.
a product term—i.e., A l x B l , A l x B2, Al x B3,
A2 X Bi.etc. In looking at technology use as the dependent
variable, in addition to intention as a key predictor,
With the exception of MM and SCT, the predictive TPB and DTPB employ perceived behavioral
validity of the models increased after including the control as an additional predictor. MPCU employs
moderating variables. For instance, the variance facilitating conditions, a construct similar to per-
explained by TAM2 increased to 53 percent and ceived behavioral control, to predict behavior.
TAM including gender increased to 52 percent Thus, Intention and perceived behavioral control
when compared to approximately 35 percent in were used to predict behavior in the subsequent

Quarterly Vol. 27 No. 3/September 2003 445

 Venkatesh et ai/User Acceptance of IT

Table 8. Predicting Usage Behavior

Use,, Use,3 Use3,
Independent Variables R^ Beta R^ Beta R^ Beta
Studies Behavioral intention to use (Bl) .37 .61"* .36 .60"' .39 .58"*
la and 1b
(voluntary) Perceived behavioral controi (PBC) .04 .06 .17*
(N=119)
Studies Behavioral intention to use (Bl) .35 .58*" .37 .61*** .39 .56*'*
2a and 2b
(mandatory) Perceived behavioral control (PBC) .07 .07 .20*
(N = 96)

Notes: 1. Bl, PBC measured at TI were used to predict usage between time periods 1 and 2 (denoted
Use,;); Bl, PBC measured at T2 were used to predict usage between time periods 2 and 3
(Use^a); Bl, PBC measured at T3 were used to predict usage between time periods 3 and 4

2. *p < .05; ' • p < .01; ' " p < .001.

time period: intention from T1 was used to predict TRA and TPB/DTPB, perceived usefulness in
usage behavior measured between T1 and T2 and TAM/TAM2 and C-TAM-TPB, extrinsic motivation
so on (see Table 8). Since intention was used to in MM, job-fit in MPCU, relative advantage in IDT,
predict actual behavior, concerns associated with and outcome expectations in SCT. Second,
the employment of subjective measures of usage several other constructs were initially significant,
do not apply here (see Straub et al. 1995). In but then became nonsignificant over time, in-
addition to intention being a predictor of use, per- cluding perceived behavioral control in TPB/DTPB
ceived behavioral control became a significant and C-TAM-TPB, perceived ease of use in TAM/
direct determinant of use over and above intention TAM2, complexity in MPCU, ease of use in IDT,
with increasing experience (at T3) indicating that and self-efficacy and anxiety in SCT. Finally, the
continued use could be directly hindered or voluntary vs. mandatory context did have an
fostered by resources and opportunities. A neariy influence on the significance of constructs related
identical pattern of results was found when the to social influence: subjective norm (TPB/DTPB,
data were analyzed using facilitating conditions C-TAM-TPB and TAM2), social factors (MPCU),
(from MPCU) in place of perceived behavioral and image (IDT) were only significant in
control (the specific results are not shown here). mandatory implementations.

Having reviewed and empirically compared the

eight competing models, we now formulate a Formulation of the Unified
unified theory of acceptance and use of techno-
logy (UTAUT). Toward this end, we examine com- Theory of Acceptance and
monalities across models as a first step. Tables Use of Technology
5, 6, 7, and 8 presented cross-sectional tests of (UTAUT) ^ ^ ^ ^ ^ ^ ^ ^ ^
the baseline models and their extensions. Several
consistent findings emerged. First, for every
Seven constructs appeared to be significant direct
model, there was at least one construct that was
determinants of intention or usage in one or more
significant in all time periods and that construct
of the individual models (Tables 5 and 6). Of
also had the strongest influence—e.g., attitude in
these, we theorize that four constructs will play a

446 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

Performance
Expectancy

T
Effort
Expectancy
4/ Behavioral
Intention
Use
Behavior
Social
Influence
-if
///
Facilitating
Conditions

f
1
Gender Experience
Voluntariness
of Use

Figure 3. Research Model

significant role as direct determinants of user performance. The five constructs from the dif-
acceptance and usage behavior: performance ferent models that pertain to performance
expectancy, effort expectancy, social influence, expectancy are perceived usefulness (TAM/TAM2
and facilitating conditions. As will be explained and C-TAM-TPB), extrinsic motivation (MM), job-fit
below, attitude toward using technology, self- (MPCU), relative advantage (IDT), and outcome
efficacy, and anxiety are theorized not to be direct expectations (SCT). Even as these constructs
determinants of intention. The labels used for the evolved in the literature, some authors acknowl-
constructs describe the essence of the construct edged their similarities: usefulness and extrinsic
and are meant to be independent of any particular motivation (Davis et al. 1989, 1992), usefulness
theoretical perspective. In the remainder of this and job-fit (Thompson etal. 1991), usefulness and
section, we define each of the determinants, relative advantage (Davis et al. 1989; Moore and
specify the role of key moderators (gender, age, Benbasat 1991; Plouffe et al, 2001), usefulness
voluntariness, and experience), and provide the and outcome expectations (Compeau and Higgins
theoretical justification for the hypotheses. 1995b; Davis et al. 1989), and job-fit and outcome
Figure 3 presents the research model. expectations (Compeau and Higgins 1995b).

The performance expectancy construct within

Performance Expectancy each individual model (Table 9) is the strongest
predictor of intention and remains significant at all
Performance expectancy is defined as the degree points of measurement in both voluntary and man-
to which an individual believes that using the sys- datory settings (Tables 5, 6, and 7), consistent
tem will help him or her to attain gains in job with previous model tests (Agarwal and Prasad

MIS Quarterly Vol. 27 No. 3/September 2003 447

Venkatesh et al./User Acceptance of IT

Table 9. Performance Expectancy: Root Constructs, Definitions, and Scales

Construct Definition Items
Perceived The degree to which a 1. Using the system in my job would
Usefulness person beiieves that using enable me to accomplish tasks more
(Davis 1989; Davis et a particular system would quickly,
al, 1989) enhance his or her job 2, Using the system would improve my job
performance. performance.
3, Using the system in my job would
increase my productivity,
4, Using the system would enhance my
effectiveness on the job.
5. Using the system would make it easier
to do my job.
6, 1 would find the system useful in my job.
Extrinsic Motivation The perception that users Extrinsic motivation is operationalized using
(Davis etai. 1992) wiii want to perform an the same items as perceived usefulness
activity because it is per- from TAM (items 1 through 6 above).
ceived to be instrumental
in achieving valued out-
comes that are distinct
from the activity itseif, such
as improved job perfor-
mance, pay, or promotions
Job-fit How the capabilities of a 1, Use of the system wiil have no effect on
(Thompson etal. system enhance an indi- the performance of my job (reverse
1991) viduai's job performance. scored).
2, Use of the system can decrease the
time needed for my important job
responsibiiities.
3, Use of the system can significantly
increase the quality of output on my job.
4. Use ofthe system can increase the
effectiveness of performing job tasks.
5. Use can increase the quantity of output
for the same amount of effort.
6. Considering all tasks, the general extent
to which use of the system could assist
on the job. (different scale used for this
item).

448 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et at./User Acceptance of IT

Table 9. Performance Expectancy: Root Constructs, Definitions, and Scales

(Continued)
Construct Definition Items
Relative Advantage The degree to which using 1. Using the system enables me to
(Moore and Benbasat an innovation is perceived accomplish tasks more quickly.
1991) as being better than using 2. Using the system improves the quality of
its precursor. the work 1 do.
3. Using the system makes it easier to do
my job.
4. Using the system enhances my
effectiveness on the job.
5. Using the system increases my
productivity.
Outcome Outcome expectations If 1 use the system...
Expectations relate to the consequences 1. 1 will increase my effectiveness on the
(Compeau and of the behavior. Based on job.
Higgins 1995b; empirical evidence, they 2. 1 will spend less time on routine job
Compeau etal. 1999) were separated into per- tasks.
formance expectations 3. 1 will increase the quality of output of my
(job-related) and personal job.
expectations (individual 4. 1 will increase the quantity of output for
goals). For pragmatic the same amount of effort.
reasons, four of the highest 5. My coworkers will perceive me as
loading items from the competent.
performance expectations 6. 1 will increase my chances of obtaining a
and three of the highest promotion.
loading items from the 7. 1 will increase my chances of getting a
personal expectations raise.
were chosen from Com-
peau and Higgins (1995b)
and Compeau et al. (1999)
for inclusion in the current
research. However, our
factor analysis showed the
two dimensions to load on
a single factor.

1998; Compeau and Higgins 1995b; Davis et al. formance expectancies, which focus on task
1992; Taylor and Todd 1995a; Thompson et al. accomplishment, are likely to be especially salient
1991; Venkatesh and Davis 2000). However, from to men. Gender schema theory suggests that
a theoretical point of view, there is reason to such differences stem from gender roles and
expect that the relationship between performance socialization processes reinforced from birth
expectancy and intention will be moderated by rather than biological gender per se (Bem 1981;
gender and age. Research on gender differences Bem and Allen 1974; Kirchmeyer 1997; Lubinski
indicates that men tend to be highly task-oriented et al. 1983; Lynott and McCandless 2000; Moto-
(Minton and Schneider 1980) and, therefore, per- widlo 1982). Recent empirical studies outside the

MIS Quarterly Vol. 27 No. 3/September 2003 449

Venkatesh et al./User Acceptance of IT

IT context (e.g., Kirchmeyer 2002; Twenge 1997) The effort expectancy construct within each model
have shown that gender roles have a strong (Table 10) is significant in both voluntary and
psychological basis and are relatively enduring, mandatory usage contexts; however, each one is
yet open to change over tinne (see also Ashmore significant only during the first time period (post-
1990; Eichingeretal. 1991; Feidmanand Aschen- training, Tl), becoming nonsignificant over
brenner 1983; Helson and Moane 1987). periods of extended and sustained usage (see
Tables 5, 6, and 7). consistent with previous
Similar to gender, age is theorized to play a research (e.g., Agarwal and Prasad 1997, 1998;
moderating role. Research on job-related atti- Davis etal. 1989; Thompson etal. 1991, 1994),
tudes (e.g., Hall and Mansfield 1975; Porter 1963) Effort-oriented constructs are expected to be be
suggests that younger workers may place more more salient in the early stages of a new behavior,
importance on extrinsic rewards. Gender and age when process issues represent hurdles to be
differences have been shown to exist in techno- overcome, and later become overshadowed by
logy adoption contexts also (Morris and Venkatesh instrumentality concerns (Davis et al. 1989;
2000; Venkatesh and Morris 2000). In looking at Szajna 1996; Venkatesh 1999).
gender and age effects, it is interesting to note
that Levy (1988) suggests that studies of gender Venkatesh and Morris (2000), drawing upon other
differences can be misleading without reference to research (e.g., Bem and Allen 1974; Bozionelos
age. Forexample, given traditional societal gen- 1996), suggest that effort expectancy is more
der roles, the importance of job-related factors salient for women than for men. As noted earlier,
may change significantly (e.g., become sup- the gender differences predicted here could be
planted by family-oriented responsibilities) for driven by cognitions related to gender roles (e.g.,
working women between the time that they enter Lynott and McCandless 2000; Motowidio 1982;
the labor force and the time they reach child- Wong et al. 1985). Increased age has been
rearing years (e.g., Barnett and Marshall 1991). shown to be associated with difficulty in pro-
Thus, we expect that the influence of performance cessing complex stimuli and allocating attention to
expectancy will be moderated by both gender and information on the job (Plude and Hoyer 1985),
age. both of which may be necessary when using
software systems. Prior research supports the
H1: The influence of performance ex- notion that constructs related to effort expectancy
pectancy on behavioral intention will will be stronger determinants of individuals' inten-
be moderated by gender and age, tion for women (Venkatesh and Morris 2000;
such that the effect will be stronger Venkatesh et al. 2000) and for older workers
for men and particularly for younger (Morris and Venkatesh 2000). Drawing from the
men. arguments made in the context of performance
expectancy, we expect gender, age, and exper-
ience to work in concert (see Levy 1988). Thus,
Effort Expectancy we propose that effort expectancy will be most
salient for women, particularly those who are older
Effort expectancy is defined as the degree of ease and with relatively little experience with the
associated with the use ofthe system. Three con- system.
structs from the existing models capture the
concept of effort expectancy: perceived ease of
use (TAM/TAM2), complexity (MPCU), and ease H2: The influence of effort expectancy
of use (IDT). As can be seen in Table 10, there is on behavioral intention will be
substantial similarity among the construct defini- moderated by gender, age, and
tions and measurement scales. The similarities experience, such that the effect will
among these constructs have been noted in prior be stronger for women, particularly
research (Davis et al. 1989; Moore and Benbasat younger women, and particularly at
1991; Plouffe etal, 2001; Thompson etal. 1991). early stages of experience.

450 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

Table 10. Effort Expectancy: Root Constructs, Definitions, and Scales

Construct Definition Items
Perceived Ease of Use The degree to which a 1. Learning to operate the system would be
(Davis 1989; Davis et person believes that easy for me.
al, 1989) using a system would be 2. 1 would find it easy to get the system to
free of effort. do what 1 want it to do,
3. My interaction with the system would be
clear and understandable,
4. 1 would find the system to be flexible to
interact with.
5. It would be easy for me to become
skillful at using the system,
6, 1 would find the system easy to use.
Complexity The degree to which a 1, Using the system takes too much time
(Thompson etal. 1991) system is perceived as from my normal duties.
relatively difficult to 2. Working with the system is so
understand and use. complicated, it is difficult to understand
what is going on,
3. Using the system involves too much time
doing mechanical operations (e.g., data
input).
4. It takes too long to learn how to use the
system to make it worth the effort.
Ease of Use The degree to which 1. My interaction with the system is clear
(Moore and Benbasat using an innovation is and understandable,
1991) perceived as being 2. 1 believe that it is easy to get the system
difficult to use. to do what 1 want it to do.
3. Overall, 1 believe that the system is easy
to use.
4, Learning to operate the system is easy
for me.

Social Influence the explicit or implicit notion that the individual's

behavior is influenced by the way in which they
Social influence is defined as the degree to which believe others will view them as a result of having
an Individual perceives that important others used the technology. Table 11 presents the three
believe he or she should use the new system. constructs related to social influence: subjective
Social influence as a direct determinant of behav- norm (TRA, TAM2, TPB/ DTPB, and C-TAM-TPB),
ioral intention is represented as subjective norm in social factors (MPCU), and image (IDT).
TRA, TAM2, TPB/DTPB and C-TAM-TPB, social
factors in MPCU, and image in IDT. Thompson et The current model comparison (Tables 5, 6, and
al. (1991) used the term social norms in defining 7) found that the social influence constructs listed
their construct, and acknowledge its similarity to above behave similarly. None of the social influ-
subjective norm within TRA. While they have ence constructs are significant in voluntary con-
different labels, each of these constructs contains texts; however, each becomes significant when

MIS Quarterly Vol. 27 No. 3/September 2003 451

Venkatesh et al./User Acceptance of IT

Table 11. Social Influence: Root Constructs, Definitions, and Scales

Construct Definition Items
Subjective Norm The person's perception 1. People who influence my
(Ajzen 1991; Davis etal. that most people who are behavior think that 1 should use
1989; Fishbein and Azjen important to him think he the system.
1975; Mathieson 1991; should or should not 2. People who are important to me
Taylor and Todd 1995a, perform the behavior in think that 1 should use the
1995b) question. system.
Social Factors The individual's inter- 1. 1 use the system because of the
(Thompson et al. 1991) nalization of the reference proportion of coworkers who use
group's subjective culture. the system.
and specific interpersonal 2. The senior management of this
agreements that the indivi- business has been helpful in the
dual has made with others, use of the system.
in specific social situations. 3. My supervisor is very supportive
of the use of the system for my
job.
4. In general, the organization has
supported the use of the system.
Image The degree to which use of 1. People in my organization who
(Moore and Benbasat 1991) an innovation is perceived use the system have more
to enhance one's image or prestige than those who do not.
status in one's social 2. People in my organization who
system. use the system have a high
profile.
3. Having the system is a status
symbol in my organization.

use is mandated. Venkatesh and Davis (2000) The role of social influence in technology
suggested that such effects could be attributed acceptance decisions is complex and subject to
to compliance in mandatory contexts that a wide range of contingent influences. Social
causes social influences to have a direct effect influence has an impact on individual behavior
on intention; in contrast, social influence in through three mechanisms: compliance, inter-
voluntary contexts operates by influencing per- nalization, and identification (see Venkatesh and
ceptions about the technology—the mech- Davis 2000; Warshaw 1980). While the latter
anisms at play here are internalization and two relate to altering an individual's belief struc-
identification. In mandatory settings, social ture and/or causing an individual to respond to
influence appears to be important only in the potential social status gains, the compliance
early stages of individual experience with the mechanism causes an individual to simply alter
technology, with its role eroding over time and his or her intention in response to the social
eventually becoming nonsignificant with sus- pressure—i.e., the individuai intends to comply
tained usage (T3), a pattern consistent with the with the social influence. Prior research sug-
observations of Venkatesh and Davis (2000). gests that individuals are more likely to comply

452 MIS Quarterty Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

with others' expectations when those referent Facilitating Conditions

others have the ability to reward the desired
behavior or punish nonbehavior (e.g., French Facilitating conditions are defined as the degree
and Raven 1959; Warshaw 1980). This view of to which an individual believes that an organi-
compliance is consistent with results in the zational and technical infrastructure exists to
technology acceptance literature indicating that support use of the system. This definition cap-
reliance on others' opinions is significant only in tures concepts embodied by three different con-
mandatory settings (Hartwick and Barki 1994), structs: perceived behavioral control (TPB/
particularly in the early stages of experience, DTPB, C-TAM-TPB). facilitating conditions
when an individual's opinions are relatively ill- (MPCU), and compatibility (IDT). Each of these
informed (Agarwal and Prasad 1997; Hartwick constructs is operationalized to include aspects
and Barki 1994; Karahanna et al. 1999; Taylor of the technological and/or organizationai envi-
and Todd 1995a; Thompson etal. 1994; Venka- ronment that are designed to remove barriers to
tesh and Davis 2000). This normative pressure use (see Table 12). Taylor and Todd (1995b)
will attenuate over time as increasing exper- acknowledged the theoretical overlap by
ience provides a more instrumental (rather than modeling facilitating conditions as a core com-
social) basis for individual intention to use the ponent of perceived behavioral control in
system. TPB/DTPB. The compatibility construct from
IDT incorporates items that tap the fit between
Theory suggests that women tend to be more the individual's work style and the use of the
sensitive to others' opinions and therefore find system in the organization.
social influence to be more salient when forming
an intention to use new technology (Miller 1976; The empirical evidence presented in Tables 5,
Venkatesh et al. 2000), with the effect declining 6, 7, and 8 suggests that the relationships
with experience (Venkatesh and Morris 2000). between each of the constructs (perceived
As in the case of performance and effort expec- behavioral control, facilitating conditions, and
tancies, gender effects may be driven by psy- compatibility) and intention are similar. Specifi-
chological phenomena embodied within socially- cally, one can see that perceived behavioral
constructed gender roles (e.g., Lubinski et al. control is significant in both voluntary and man-
1983). Rhodes' (1983) meta-analytic review of datory settings immediately following training
age effects concluded that affiliation needs in- (TI), but that the construct's influence on
crease with age, suggesting that older workers intention disappears by T2. It has been
are more likely to place increased salience on demonstrated that issues related to the support
social influences, with the effect declining with infrastructure—a core concept within the facili-
experience (Morris and Venkatesh 2000). tating conditions construct—are largely captured
Therefore, we expect a complex interaction with within the effort expectancy construct which taps
these moderating variables simultaneously influ- the ease with which that tool can be applied
encing the social influence-intention relation- (e.g., Venkatesh 2000). Venkatesh (2000)
ship. found support for full mediation of the effect of
facilitating conditions on intention by effort
H3: The influence of social influence expectancy. Obviously, if effort expectancy is
on behavioral intention will be not present in the model (as is the case with
moderated by gender, age, TPB/DTPB), then one would expect facilitating
voluntariness, and experience. conditions to become predictive of intention.
such that the effect will be Our empirical results are consistent with these
stronger for women, particularly arguments. For example, in TPB/DTPB, the
older women, particularly in man- construct is significant in predicting intention;
datory settings in the early stages however, in other cases (MPCU and IDT), it is
of experience. nonsignificant in predicting intention. In short.

MIS Quarierly Vol. 27 No. 3/September 2003 453

 Venkatesh et al./User Acceptance of IT

Table 12. Facilitating Conditions; Root Constructs, Definitions, and Scales

Construct Definition Items
Perceived Behavioral Reflects perceptions of 1. 1 have control over using the system.
Control internal and external 2. 1 have the resources necessary to use
(Ajzen 1991; Taylor and constraints on behavior the system.
Todd 1995a, 1995b) and encompasses self- 3. 1 have the knowledge necessary to
efficacy, resource facili- use the system.
tating conditions, and 4. Given the resources, opportunities
technology facilitating and knowledge it takes to use the
conditions. system, it would be easy for me to use
the system.
5. The system is not compatible with
other systems 1 use.
Facilitating Conditions Objective factors in the 1. Guidance was available to me in the
(Thompson etal. 1991) environment that selection of the system.
observers agree make 2. Specialized instruction concerning the
an act easy to do. system was available to me.
including the provision 3. A specific person (or group) is
of computer support. available for assistance with system
difficulties.
Compatibility The degree to which an 1. Using the system is compatible with
(Moore and Benbasat innovation is perceived all aspects of my work.
1991) as being consistent with 2. 1 think that using the system fits well
existing values, needs. with the way 1 like to work.
and experiences of 3. Using the system fits into my work
potential adopters. style.

when both performance expectancy constructs intention). In fact, the effect is expected to
and effort expectancy constructs are present, increase with experience as users of technology
facilitating conditions becomes nonsignificant in find multiple avenues for help and support
predicting intention. throughout the organization, thereby removing
impediments to sustained usage (Bergeron et al.
H4a: Facilitating conditions will not 1990). Organizational psychologists have noted
have a significant influence on that older workers attach more importance to
behavioral intention.^ receiving help and assistance on the job (e.g.. Hall
and Mansfield 1975). This is further underscored
The empirical results also indicate that facilitating in the context of complex IT use given the
conditions do have a direct influence on usage increasing cognitive and physical limitations asso-
beyond that explained by behavioral intentions ciated with age. These arguments are in line with
alone (see Table 8). Consistent with TPB/DTPB, empirical evidence from Morris and Venkatesh
facilitating conditions are also modeled as a direct (2000). Thus, when moderated by experience and
antecedent of usage (i.e., not fully mediated by age, facilitating conditions will have a significant
influence on usage behavior.

H4b: The influence of facilitating con-

To test the nonsignificant relationship, we perform a ditions on usage will be mode-
power analysis in the results section.

454 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et ai/User Acceptance of IT

rated by age arid experience, motivation^ (MM), affect toward use (MPCU), and
such that the effect will be affect (SCT). Table 13 presents the definitions
stronger for older workers, par- and associated scale items for each construct.
ticularly with increasing exper- Each construct has a component associated with
ience. generalized feeling/affect associated with a given
behavior (in this case, using technology). In
examining these four constructs, it is evident that
Constructs Theorized Not to Be they all tap into an individual's liking, enjoyment.
Direct Determinants of Intention joy. and pleasure associated with technology use.

Although self-efficacy and anx/efy appeared to be Empirically, the attitude constructs present an
significant direct determinants of intention in SCT interesting case (see Tables 5. 6, and 7). In some
(see Tables 5 and 6), UTAUT does not include cases (e.g.. TRA, TPB/DTPB. and MM), the
them as direct determinants. Previous research attitude construct is significant across all three
(Venkatesh 2000) has shown self-efficacy and time periods and is also the strongest predictor of
anxiety to be conceptually and empirically distinct behavioral intention. However, in other cases (C-
from effort expectancy (perceived ease of use). TAM-TPB. MPCU. and SCT). the construct was
Self-efficacy and anxiety have been modeled as not significant. Upon closer examination, the
indirectdeterminantsofintention fully mediated by attitudinal constructs are significant only when
perceived ease of use (Venkatesh 2000). Consis- specific cognitions—in this case, constructs
tent W\ih this, we found that self-efficacy and related to performance and effort expectancies—
anxiety appear to be significant determinants of are not included in the model. There is empirical
intention in SCT—i.e., without controlling for the evidence to suggest that affective reactions (e.g.,
effect of effort expectancy. We therefore expect intrinsic motivation) may operate through effort
self-efficacy and anxiety to behave similarly, that expectancy (see Venkatesh 2000). Therefore, we
is. to be distinct from effort expectancy and to consider any observed reiationship between
have no direct effect on intention above and attitude and intention to be spurious and resulting
beyond effort expectancy. from the omission of the other key predictors
(specifically, performance and effort expec-
H5a: Computer self-efficacy will not tancies). This spurious relationship likely stems
have a significant influence on from the effect of performance and effort expec-
behavioral intention.^ tancies on attitude (see Davis et al. 1989). The
non-significance of attitude in the presence of
H5b: Compute anxiety will not have such other constructs has been reported in pre-
a significant influence on beha- vious model tests (e.g., Taylor and Todd 1995a:
viorai intention.'' Thompson et al. 1991), despite the fact that this
finding is counter to what is theorized in TRA and
Attitude toward using technology is defined as an TPB/DTPB. Given that we expect strong relation-
individual's overall affective reaction to using a ships in UTAUT between performance expectancy
system. Four constructs from the existing models and intention, and between effort expectancy and
align closely with this definition: attitude toward intention, we believe that, consistent with the logic
behavior (TRA, TPB/DTPB. C-TAM-TPB). intrinsic developed here, attitude toward using technology

p
To test the nonsignificant relationship, we perform a Some perspectives differ on the role of intrinsic
power analysis in the results section. motivation. For example. Venkatesh (2000) models it as
a determinant of perceived ease of use (effort
To test the nonsignificant relationship, we perfonm a expectancy). However, in the motivational model, it is
shown as a direct effect on intention and is shown as
power analysis in the results section.
such here.

MtS Quarterly Vol. 27 No. 3/September 2003 455

Venkatesh et ai/User Acceptance of IT

Table 13. Attitude Toward Using Technology: Root Constructs, Definitions, and
Scales
Construct Definition Items
Attitude Toward An individual's positive or 1. Using the system is a bad/good idea.
Behavior negative feelings about 2. Using the system is a foolish/wise idea.
(Davis etal. 1989: performing the target 3. 1 dislike/like the idea of using the
Fishbein and Ajzen behavior. system.
1975; Taylor and Todd 4. Using the system is unpleasant/
1995a, 1995b) pleasant.
Intrinsic Motivation The perception that users 1. 1 find using the system to be enjoyable
(Davis etal. 1992) will want to perform an 2. The actual process of using the system
activity for no apparent is pleasant.
reinforcement other than 3. 1 have fun using the system.
the process of performing
the activity per se.
Affect Toward Use Feelings of joy, elation, or 1. The system makes work more
(Thompson et al. 1991) pleasure; or depression. interesting.
disgust, displeasure, or 2. Working with the system is fun.
hate associated by an 3. The system is okay for some jobs, but
individual with a particular not the kind of job 1 want. (R)
act.
Affect An individual's liking of 1. 1 like working with the system.
(Compeau and Higgins the behavior. 2. 1 look forward to those aspects of my
1995b; Compeau etal. job that require me to use the system.
1999) 3. Using the system is frustrating for me.
(R)
4. Once 1 start working on the system, 1
find it hard to stop.
5. 1 get bored quickly when using the
system. (R)

will not have a direct or interactive influence on Behavioral Intention

intention.
Consistent with the underlying theory for all of the
H5c: Attitude toward using techno- intention models'" discussed in this paper, we
logy will not have a significant expect that behavioral intention will have a
Infiuence on behaviorai significant positive influence on technology usage.
intention.^
i-i6: Behavioral intention will have a
significant positive influence on
usage.

10
^ To test the absence of a relationship, we perform a For example, see Sheppard et al. (1988) for an
power analysis in the results section. extended review of the intention-behavior relationship.

456 MiS Quarteriy Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

Empirical Validation of UTAUT correlations. The results from the measurement

model estimations (with reduced items) are not
Preliminary Test of UTAUT shown here in the interest of space.

Using the post-training data (T1) pooled across An examination of these highest loading items
studies (N = 215), a measurement modei of the suggested that they adequately represented the
seven direct determinants of intention (using ail conceptual underpinnings of the constructs—this
items that related to each of the constructs) was preliminary content validity notwithstanding, we
estimated. All constructs, with the exception of will return to this issue later in our discussion of
use, were modeled using reflective indicators. Ail the limitations of this work. Selection based on
internal consistency reliabilities (ICRs) were item loadings or corrected item-total correlations
greater than ,70. The square roots of the shared are often recommended in the psychometric
variance between the constructs and their mea- literature (e.g., Nunnally and Bernstein 1994),
sures were higher than the correlations across This approach favors building a homogenous
constructs, supporting convergent and discri- instrument with high internal consistency, but
could sacrifice content validity by narrowing
minant validity—see Table 14(a). The reverse-
domain coverage," The items selected forfurther
coded affect items of Compeau and Higgins
analysis are indicated via an asterisk in Table 15,
(1995b) had loadings lower than .60 and were
and the actual items are shown in Table 16,
dropped and the model was reestimated. With
the exception of eight loadings, all others were
Tables 17(a) and 17(b) show the detailed model
greater than .70, the level that is generally con-
test results at each time period for intention and
sidered acceptable (Fornell and Larcker 1981; see
usage, respectively, including all lower-level
also Compeau and Higgins 1995a, 1995b; Com-
interaction terms. Tables 17(a) and 17(b) also
peau etal. 1999)—see Table 15, Inter-item corre-
show the model with direct effects only so the
lation matrices (details not shown here due to reader can compare that to a model that includes
space constraints) confirmed that intra-construct the moderating influences. The variance ex-
Item correlations were very high while inter-con- plained at various points in time by a direct
struct item correlations were low. Results of effects-only model and the full model including
similar analyses from subsequent time periods (T2 interaction terms are shown in Tables 17(a) and
and T3) also indicated an identical pattern and are 17(b) for intention and usage behavior, respec-
shown in Tables 14(b) and 14(c). tively.'^ We pooled the data across the different

Although the structural modei was tested on all

the items, the sample size poses a limitation here Bagozzi and Edwards (1998) discuss promising new
because of the number of latent variables and alternatives to this approach for coping with the inherent
associated items. Therefore, we reanalyzed the tension between sample size requirements and the
number of items, such as full or partial aggregation of
data using only four of the highest loading items items.
from the measurement model for each of the
determinants; intention only employs three items PLS does not produce adjusted R^ we used an
and use is measured via a single indicator. alternative procedure to estimate an adjusted R^ PLS
estimates a measurement model that in turn is used to
UTAUT was estimated using data pooled across
generate latent variable obsen/ations based on the
studies at each time period (N = 215). As with the loadings; these latent variable observations are used to
model comparison tests, the bootstrapping estimate the structural model using OLS, Therefore, in
method was used here to test the PLS models. order to estimate the adjusted R^. we used the latent
variable observations generated from PLS and analyzed
An examination of the measurement model from the data using hierarchical regressions in SPSS, We
the analysis using the reduced set of items was report the R' and adjusted R^ from the hierarchical
similar to that reported in Tables 14 and 15 in regressions. This allows for a direct comparison of
variance explained from PLS with variance explained
terms of reliability, convergent validity, discrimi- (both R^ and adjusted R^) from traditional OLS
nant validity, means, standard deviations, and regressions and allows the reader to evaluate the
variance explained-parsimony trade-off.

MIS Quarterly Vol. 27 No. 3/September 2003 457

Venicatesh et al./User Acceptance ofiT

Table 14. Measurement Model Estimation for the Preliminary Test of UTAUT
(a)T1 Results (N = 215)
ICR Mean S Dev PE EE ATUT SI FC SE ANX Bl
PE .92 5.12 1.13 .94
EE .91 4.56 1.40 .31"** .91
ATUT .84 4.82 1.16 .29"* .21** .86
SI .88 4.40 1.04 .30*** -.16* .21** .88
FC .87 4.17 1.02 .18* .31*** .17* .21'* .89
SE .89 5.01 1.08 .14 .33*** .16* .18** .33*" .87
ANX .83 3.11 1.14 -.10 -.38*** -.40*** -.20** -.18** -.36*** .84
Bl .92 4.07 1.44 .38*** .34*** .25*** .35*** .19** .16* -.23" .84
(b) T2 Results (N = 215)
ICR Mean S Dev PE EE ATUT SI FC SE ANX Bl
PE .91 4.71 1.11 .92
EE .90 5.72 0.77 .30*** .90
ATUT .77 5.01 1.42 .25** .20** .86
SI .94 3.88 1.08 .27*** -.19* .21** .88
FC .83 3.79 1.17 .19' .31*** .18* .20" .86
SE .89 5.07 1.14 .24** .35*** .19** .21** .33"* .75
ANX ,79 3.07 1.45 -.07 -.32*** -.35*** -.21" -.17* -.35*" .82
Bl ,90 4.19 0.98 .41'** .27*** .23** .21** .16* .16* -.17* .87
(c) T3 Results (N = 215)
ICR Mean S Dev PE EE ATUT SI FC SE ANX Bl
PE .91 4.88 1.17 .94
EE .94 5.88 0.62 .34*** .91
ATUT .81 5.17 1.08 .21** 24** .79
SI .92 3.86 1.60 .27*** -.15 .20** .93
FC .85 3.50 1.12 .19* .28*** .18* .22** .84
SE .90 5.19 1.07 .14* .30*** .22** .20" .33**' .77
ANX .82 2.99 1.03 -.11 -.30*** -.30*** -.20** -.24" -.32*** .82
Bl .90 4.24 1.07 44*** .24** .20** .16' .16* .16* -.14 .89

Notes: 1. ICR: Internal consistency reliability.

2. Diagonal elements are the square root of the shared variance between the constructs and
their measures; off-diagonal elements are correlations between constructs.
3. PE: Performance expectancy; EE: Effort expectancy; ATUT: Attitude toward using
technology; SI: Social influence; FC: Facilitating conditions; SE: Self-efficacy; ANX: Anxiety;
Bl: Behavioral intention to use the system.

458 M/S Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

Table 15. Item Loadings from PLS (N = 215 at Each Time Period)
Items T1 T2 T3 Items TI T2 T3
U1 82 .81 .80 •SN1 .82 .85 .90

Social Influence
U2 .84 .80 .81 'SN2 .83 .85 .84
U3 .81 .84 ,84 SF1 .71 .69 .76
U4 .80 .80 .84 *SF2 .84 .80 .90

(SI)
U5 .81 .78 ,84 SF3 .72 .74 .77
'U6 ,88 .88 .90 *SF4 .80 .82 ,84
'RA1 ,87 .90 .90 11 .69 .72 ,72
Performance Expectancy

RA2 .73 .70 .79 12 .65 ,76 .70

RA3 .70 .69 .83 13 -71 ,72 .69
RA4 .71 .74 .74 PBC1 72 ,66 .62

Facilitating Conditions
*RA5 .86 .88 .94 •PBC2 84 ,81 .80
(PE)

JF1 ,70 .71 .69 *PBC3 .81 .81 .82

JF2 .67 .73 .64 PBC4 .71 .69 .70
JF3 .74 .70 ,79 -PBC5 .80 .82 .80

(FC)
JF4 .73 .79 .71 FC1 .74 .73 .69
JF5 ,77 .71 .73 FC2 .78 .77 .64
JF6 .81 .78 .81 'FC3 .80 .80 .82
0E1 .72 .80 .75 Cl .72 .72 .70
0E2 71 ,68 .77 C2 .71 .74 .74
0E3 .75 ,70 .70 .77 .70
C3 .78
0E4 .70 ,72 ,67
*SE1 ,80 .83 .84
0E5 .72 ,72 ,70
SE2 .78 .80 .80
Self-Efficacy

0E6 .69 -79 .74

SE3 .72 .79 .74
•0E7 86 .87 .90
(SE)

•SE4 .80 .84 .84

E0U1 .90 .89 .83
SE5 .77 .74 .69
E0U2 .90 .89 .88
,82 .82
'SE6 ,81
•E0U3 .94 .96 .91
'SE7 .87 -85 .86
Effort Expectancy

E0U4 .81 .84 88 .70

SE8 .69 .72
•E0U5 .91 ,90 .90
•ANX1 ,78 .74 .69
Anxiety
(ANX)

•E0U6 .92 ,92 .93

•ANX2 .71 .70 ,72
(EE)

EU1 .84 .80 .84

'ANX3 ,72 ,69 ,73
EU2 .83 88 ,85 'ANX4 ,74 .72 .77
EU3 .89 .84 .80
'EU4 ,91 .91 .92
'BM .88 .84 88
COI .83 .82 .81 £ OS 'BI2 ,82 ,86 .88
CO2 ,83 .78 .80 *BI3 .84 ,88 87
C03 .81 .84 .80
C04 .75 .73 .78
'Al ,80 .83 .85
A2 .67 ,64 .65
Attitude Toward Using

A3 .64 ,64 .71

Technology (ATUT)

A4 .72 ,71 .64

IM1 .70 .78 ,72
IM2 .72 .72 .78
!M3 .73 .75 ,81
'AF1 .79 .77 .84
'AF2 .84 .83 .84
AF3 .71 .70 .69
'Affect 1 .82 .85 .82
Affect2 ,67 .70 .70
Affect3 .62 .68 .64

Notes: 1. The loadings at T l , T2. and T3 respectively are from separate measurement model tests and relate to
Tables 14(a). 14(b), and 14(c) respectively.
2. Extrinsic motivation (EM) was measured using the same scale as perceived usefulness (U),
3. Items denoted with an asterisk are those that were selected for inclusion in the test of UTAUT.

tVttS Quarterly Vol. 27 No. 3/Septerrjber 2003 459

Venkatesh et aL/User Acceptance of IT

Table 16. Items Used in Estimating UTAUT

Performance expectancy
U6: I would find the system useful in my job.
RA1: Using the system enables me to accomplish tasks more quickly.
RA5; Using the system increases my productivity.
0E7: If I use the system, I will increase my chances of getting a raise.

Effort expectancy
E0U3: My interaction with the system would be clear and understandable.
E0U5: It would be easy for me to become skillful at using the system.
E0U6: I would find the system easy to use.
EU4: Learning to operate the system is easy for me.

Attitude toward using technology

Al: Using the system is a bad/good idea.
AF1: The system makes work more interesting.
AF2: Working with the system is fun.
Affecti: I like working with the system.

Social influence
SN1; Peoplewhoinfluencemybehaviorthink that I should use the system.
SN2: People who are important to me think that I should use the system.
SF2: The senior management of this business has been helpful in the use of the system.
SF4: In general, the organization has supported the use of the system.

Facilitating conditions
PBC2; I have the resources necessary to use the system.
PBC3: I have the knowledge necessary to use the system.
PBC5: The system is not compatible with other systems I use.
FC3: A specific person (or group) is available for assistance with system difficulties.

Self-efficacy
I could complete a job or task using the system...
SE1; If there was no one around to tell me what to do as I go.
SE4: If I could call someone for help if I got stuck.
SE6: If 1 had a lot of time to complete the job for which the software was provided.
SE7: If I had just the built-in help facility for assistance.

Anxiety
ANX1: I feel apprehensive about using the system.
ANX2: It scares me to think that I could lose a lot of information using the system by hitting
the wrong key.
ANX3; I hesitate to use the system for fear of making mistakes I cannot correct.
ANX4; The system is somewhat intimidating to me.

Behavioral intention to use the system

BI1; I intend to use the system in the next <n> months.
BI2: I predict I would use the system in the next <n> months.
BI3: I plan to use the system in the next <n> months.

460 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of iT

points of measurement by converting time period nonsignificant determinants (i.e.. self-efficacy,

(experience) into a moderator. The column titled anxiety, and attitude) were dropped from the
Pooled Analysis" reports the results of this model and the model was reestimated; facilitating
analysis. Caution is necessary when conducting conditions was not dropped from the model
such analyses and the reader is referred to because of its role in predicting use." The re-
Appendix A for a discussion of the potential con- estimated model results are shown in Table 17.
straints of pooling. Also reported in Appendix A
are the statistical tests that we conducted prior to In predicting usage behavior (Table 17b), both
pooling the data for the analysis in Table 17. behavioral intention (H6) and facilitating conditions
were significant, with the latter's effect being
As expected, the effect of performance expec- moderated by age (the effect being more impor-
tancy was in the form of a three-way inter- tant to older workers); and, based on Chow's test
action—the effect was moderated by gender and of the beta differences (p < .05), the effect was
age such that it was more salient to younger stronger with increasing experience with the
workers, particularly men, thus supporting H I . technology (H4b). Since H4a, H5a, H5b, and H5c
Note that a direct effect for performance expec- are hypothesized such that the predictor is not
tancy on intention was observed; however, these expected to have an effect on the dependent
main effects are not interpretable due to the variable, a power analysis was conducted to
presence of interaction terms (e.g., Alken and examine the potential for type II error. The likeli-
West 1991). The effect of effort expectancy was hood of detecting medium effects was over 95
via a three-way interaction—the effect was moder- percent for an alpha level of .05 and the likelihood
ated by gender and age (more salient to women of detecting small effects was under 50 pecent.
and more so to older women). Based on Chow's
test of beta differences (p < .05), effort expectancy
was more significant with limited exposure to the Cross-validation of UTAUT
technology (effect decreasing with experience),
thus supporting H2. The effect of social influence Data were gathered from two additional organi-
was via a four-way interaction—with its role being zations to further validate UTAUT and add
more important in the context of mandatory use, external validity to the preliminary test. The major
more so among women, and even more so among details regarding the two participating organi-
older women. The Chow's test of beta differences zations are provided in Table 18. The data were
(p < .05) indicated that social influence was even collected on the same timeline as studies 1 and 2
more significant in the early stages of individual (see Figure 2). The data analysis procedures
experience with the technology, thus supporting were the same as the previous studies. The
H3. Facilitating conditions was nonsignificant as results were consistent with studies 1 and 2. The
a determinant of intention, thus supporting H4a." items used in the preliminary test of UTAUT (listed
As expected self-efficacy, anxiety, and attitude in Table 16) were used to estimate the mea-
did not have any direct effect on intention, thus surement and structural models in the new data.
supporting H5a, H5b, and H5c. Three of the four This helped ensure the test of the same model in
both the preliminary test and this validation, thus
limiting any variation due to the changing of items.
UTAUT was then tested separately for each time
'•^Since these two hypotheses were about nonsignificant
relationships, the supportive results should be inter-
preted with caution, bearing in mind the power analysis
reported in the text. In this case, (he concern about the Since these two hypotheses were about nonsignificant
lack of a reiationship is somewhat alleviated as some relationships, the supportive results should be inter-
previous research in this area has found attitude and preted with caution after consideration of the power
facilitating conditions to not be predictors of intention in analyses reported in the text. The results are presented
the presence of performance expectancy ar^d effort here for model completeness and to allow the reader to
expectancy constructs (see Taylor and Todd 1995b; link the current research with existing theoretical
Venkatesh 2000). perspectives in this domain.

JW/S Quarterly Vol. 27 No. 3/September 2003 461

Venkatesh et al./User Acceptance of IT

Table 17, Preliminary Test of UTAUT

(a) Dependent Variable: Intention
Tl T2 T3 Pooled
(N = 215) (N = 215) (N = 215) (N = 645)
DONLY D +l DONLY D+l DONLY D +l DONLY D+l
R^(PLS) .40 .51 ,41 .52 .42 .50 .31 .76
R^ (hierarchical regrn.) .39 .51 -41 .51 .42 .50 .31 ,77
Adjusted R^ (hierarchical regrn.) .35 .46 .38 .46 .36 ,45 .27 ,69

Performance expectancy (PE) .46'*' .17* .57*'* .15* .59"* .16' .53"* .18'
Effort expectancy (EE) .20* -.12 .08 ,02 .09 .11 .10 ,04
Social influence (SI) .13 .10 .10 .07 .07 ,04 .11 .01
Facilitating conditions (FC) .03 .04 .02 .01 .01 .01 .09 .04
Gender (GDR) .04 ,02 ,04 .01 ,02 .01 .03 .01
Age (AGE) .08 .02 .09 .08 ,01 -.08 .06 .00
Voluntariness (VOL) 01 .04 .03 02 .04 -.04 .02 .00
Experience (EXP) .04 .00
PE « GDR .07 17- .Of^ 02
PE X AGE .13 04 10 01
GDR X AGE .07 .02 02 06
PE '< GDR >^ AGE ,52— .55— 57"- 55—
EE " GDR .17- 08 09 02
EE X AGE 08 04 02 .04
EE X EXP .02
GDR X AGE (included earlier) Earlier Earlier Earlier Earlier
GDR X EXP 02
AGE » EXP .01
EE X GDR X AGE 22" 20-" 18' .01
EE X GDR X EXP -.10
EE X AGE X EXP -02
GDR X AGE X EXP - 06
EE X GDR X AGE " EXP _27"'
SI X GDR .11 00 .02 02
SI X AGE 01 on 01 .02
SI X VOL .02 .01 .1)2 .06
SI X EXP .04
GDR X AGE (included earlier) Earlier rarlier rarli(-;r Earlier
GDR X VOL .01 04 02 01
GDR X EXP (included earlier) Earlier
AGE X VOL .00 02 06 02
AGE X EXP (included earlier) Earlier
VOL X EXP .02
SI X GDR X AGE -.10 .02 .04 .04
SI X GDR X VOL -01 03 02 .01
SI X GDR X EXP 01
SI X AGE X VOL - 17' 02 06 .06
SI X AGE X EXP 01
SI X VOL X EXP 00

462 MiS Quarteriy Voi. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance ot IT

Table 17. Preliminary Test of UTAUT (Continued)

T1 T2 T3 Pooled
(N = 215) (N = 215) (N = 215) (N = 645)
DONLY D+l DONLY D+ I DONLY D +l DONLY D +l
GDR X AGE X VOL J .02 02 01 .00
GDR X AGE X EXP {included Earlier
earlier)
GDR X VOL X EXP .00
AGE X VOL X EXP ! .01
SI X GDR X AGE X VOL 23* • 20 • .04
GDR X AGE X VOL x EXP .02
SI X GDR X AGE " VOL x EXP -.28"'
(b) Dependent Variable: Usage Behavior
R^(PLS) .37 .43 .36 .43 .39 .44 .38 .53
R^ (hierarchical regrn.) .37 .43 .36 .43 .39 ,43 .38 .52
Adjusted R^ (hierarchical regrn.) .36 ,41 .35 .40 .38 .41 .37 .47

Behavioral Intention (Bl) ,61*" .57"* .60**- .58"' .58"' .59"* .59*" .52"-
Facilitaling conditions (FC) .05 .07 .06 .07 18' .07 .10 .11
Age (AGE) .02 .02 .01 .02 (14 13 .08
Experience (EXP) 06
FC X AGE 22- 24- 27-* .02
FC X EXP .00
AGE X EXP .01
FC X AGE X EXP .23"

Notes: 1. D ONLY: Direct effects only; D + l: Direct effects and interaction terms,
2. "Included earlier" indicates that the term has been listed earlier in the table, but is included again for
completeness as it relates to higher-order interaction terms being computed. Grayed out cells are not
applicable for the specific column,

Table 18. Description of Studies i

Functional Sample
Study Industry Area Size System Description

Voluntary Use

3 Financial Research 80 TInis software application was one of the

Services resources available to analysts to conduct
research on financial investment opportunities
and IPOs

Mandatory Use

4 Retail Customer 53 Application that customer service representa-

Electronics Service tives were required to use to document and
manage service contracts

MIS Quarterly Vol. 27 No. 3/September 2003 463

Venkatesh et al./User Acceptance of IT

Table 19. Measurement Model Estimation for the Cross-Validation of UTAUT

(a) T1 Results (N = 133)
ICR Mean S Dev PE EE ATUT SI FC SE ANX Bl

PE .90 4.87 1.20 .88

EE .90 3.17 1.09 .30"' .93
ATUT .80 2.67 0.87 28-*- .16* .80
SI .91 4.01 1.07 .34*" -.21* .20* .89
FC .85 3.12 1,11 .18* .30*" .15* .14 .84
SE .85 4.12 1.08 .19" .37"* .17* ,22** .41"* .82
ANX .82 2.87 0.89 -.17* -.41*" -,22" -.25** -.35*" .80
Bl .89 4,02 1.19 .43"* .31"* .23*" .31*" .22* 24*. -.21*" .85
(b) T2 Results (N=133)
ICR Mean S Dev PE EE ATUT SI FC SE ANX Bl
PE .90 4.79 1.22 .91
EE .92 4.12 1.17 .34"* .89
ATUT .84 3.14 0.79 .32*" ,18** .78
SI ,92 4.11 1,08 .30"" -.20* .21" .91
FC .88 3.77 1,08 ,19* .31*" .13 .11 .86
SE .87 4.13 1.01 .21** .38*" .25" .20" .35*** .80
ANX .83 3.00 0.82 -.17* -.42*" -.32*" -.24** -.26* -.36*** .81
Bl .88 4.18 0.99 .40*** .24" .21" .24** .20* .21" -.22** .88
(c) T3 Results {N = 133)
ICR Mean S Dev PE EE ATUT SI FC SE ANX Bl
PE .94 5.01 1.17 .92
EE .92 4.89 0.88 .34**" .90
ATUT .83 3.52 1.10 .30"* .19" .80
Si .92 4.02 1.01 -.21* .21** .84
FC .88 3.89 1.00 .18* .31*" .14 ,15 .81
SE .87 4.17 1.06 .18" .32*** .21* .20* .35"* .84
ANX .85 3.02 0.87 -.15* -.31"' -.28*" -.22* -,25* -.38*" .77
Bl .90 4.07 1.02 .41*" .20** .21* .19* ,20* .20* -.21*' ,84

Notes: 1. ICR: Internal consistency reliability.

2. Diagonal elements are the square root of the shared variance between the constructs and
their measures; off-diagonal elements are correlations between constructs.
3. PE: Performance expectancy; EE: Effort expectancy; ATUT: Attitude toward using
technology; SI: Social influence; FC: Facilitating conditions; SE: Self-efficacy; ANX: Anxiety;
Bl: Behavioral intention to use the system.

464 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

Table 20. Item Loadings from PLS (N=133 at each time period)
Items T1 T2 T3 Items T1 T2 T3
U6 .91 .92 .91 PBC2 .84 .88 .85
Performance Facilitating
RA1 .90 .89 .88 PBC3 .88 .89 .88
Expectancy Conditions
F?A5 .94 .89 .90 PBC5 .86 .89 .84
(PE) (FC)
OE7 .89 .90 .91 FC3 .87 .78 .81
EOU3 .91 .90 .94 SE1 .90 .84 .88
Effort
E0U5 .92 .91 .90 Self-Efficacy SE4 .88 .82 .81
Expectancy
E0U6 .93 .90 .89 (SE) SE6 .80 .85 .79
(EE)
EU4 .87 .87 .90 SE7 .81 .77 .75
Attitude Al .84 .80 .86 ANX1 .80 .84 .80
Toward AF1 .82 .83 .77 Anxiety ANX2 .84 .84 .82
Using
AF2 .80 .80 .76 (ANX) ANX3 .83 .80 .83
Technology
(ATUT) Affecti .87 .84 .76 ANX4 .84 .77 .83
SN1 .94 .90 .90 BI1 .92 .90 .91
Social
SN2 .90 .93 .88 Intention BI2 .90 .90 .91
Influence
SF2 .89 .92 .94 (Bl) BI3 .90 .92 .92
(SI)
SF4 .92 .81 .79

Note: The loadings at T 1 , T2, and T3 respectively are from separate measurement model tests and
relate to Tables 18(a), 18(b), and 18(c) respectively.

Table 21. Cross-Validation of UTAUT

(a) Dependent Variable: Intention
T1 T2 T3 Pooled
(N = 133) (N = 133) (N = 133) {N = 399)
DONLY D+ l DONLV D+l DONLY D+l DONLY D+l
R-^ (PLS) ,42 .52 .41 52 .42 .51 .36 .77
R' (hierarchical regrn.) .41 .52 .41 .52 .42 .51 36 .77
Adjusted R' (hierarchical regrn.) .37 .48 .36 47 .36 .46 .30 .70

Performance expectancy (PE) .45"* .15 .59"* ,16* .59"* .15* ,53"* .14
Effort expectancy (EE) .22" .02 .06 .06 .04 .01 .10 ,02
Social influence (SI) .02 .04 .04 .04 .02 .01 .02 ,02
Facilitating conditions (FC) .07 .01 .00 .08 .01 .00 .07 .01
Gender (GDR) .02 ,06 .01 .04 .07 .02 .03 .03
Age (AGE) .01 .00 .07 .02 .02 .01 .07 .01
Voluntariness (VOL) .00 .00 01 .06 02 .04 .00 .01
Experience (EXP) .08 .00
PE X GDR 14 \i- .18- .01
PE X AGE Oij 01 02 -04
GDR « AGE 02 04 04 -.02
EE « GDR - 06 02 .04 -09
EE X AGE -02 01 02 -.04
EE K EXP .01

MIS Quarterly Vol. 27 No. 3/September 2003 465

Venkatesh et al./User Acceptance of IT

Table 21. Cross-Validation of UTAUT (Continued)

T1 T2 T3 Pooled
(N =133) (N = 133) (N = 133) (N = 399)
DONLY D+l DONLY D+l DONLY D+l DONLY D+l
GDR ^ AGE (included earlier) Earlier Earlier Earlier Earlier
GDR X EXP .02
AGE X EXP 06
EE X GDR X AGE 21" 18- 16- .04
EE X GDR X EXP 00
EE X AGE X EXP 00
GDR X AGE X EXP 01
EE X GDR X AGE x EXP -.25"*
SI X GDR -06 .00 0? -.07
SI X AGE 02 .01 n-i 02
SI X VOL 01 04 00 00
SI X EXP 02
GDR « AGE (included eariierl Earlier Earlier Earlier Earlier
GDR X VOL .04 .02 -.03 02
GDR X EXP (included earlier) Earlier
AGE X VOL no .01 00 07
AGE "^ EXP (included earlier) Efirlier
VOL X EXP .02
SI X GDR X AGE -.03 - 01 -07 .00
SI X GDR X VOL 04 - 03 04 .00
SI X GDR X EXP 00
SI X AGE X VOL 02 00 01
SI X A G E « EXP 07
SI X VOL X EXP 02
GDR X AGE X VOL 01 06 .01 .04
GDR X AGE X EXP (included Earlier
earlier)
GDR X VOL X EXP 01
AGE X VOL X EXP 00
SI X GDR X AGE X VOL .27"* 21" 16' 01
GDR X AGE X VOL x EXP .00
Si X GDR X AGE X VOL x EXP -.29"-
(b) Dependent Variable: Usage Behavior
R^ (PLS) .37 .44 .36 .41 .36 .44 .38 .52
R^ (hierarchical regrn.) .37 ,43 .36 .41 36 .44 .37 .52
Adjusted R^ (hierarchical regrn.) ,36 .41 .35 .38 .35 .41 .36 .48

Behavioral intention (Bl) .60*" .56"* .59*" .50*" .59"* .56*" .59"* .51"*
Facilitating conditions (FC) 04 .11 .01 .01 .02 .06 .14* .08
Age (AGE) 06 .02 .06 -.03 -.03 -.01 .06 02
Experience (EXP) 10
FC X AGE .17- 2,-. 24" 01
FC X EXP -.06
AGE X EXP -.07
FC X AGE X EXP .22"*

Notes: 1. D ONLY: Direct effects only; D + l: Direct effects and interaction terms.
2. "Included earlier" indicates that the term has been listed earlier in the table, but is included again for
completeness as it relates to higher-order interaction terms being computed. Grayed out cells are not
applicable for the specific column.

466 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

period (N = 133 at each time period). The mea- over any of the original eight models and their
surement models are shown in Tables 19 and 20. extensions. Further, UTAUT was successful In
The pattern of results in this validation (Tables integrating key elements from among the initial set
21(a) and 21(b)) mirrors what was found in the of 32 main effects and four moderators as
preliminary test (Table 17). The last column of determinants ofintention and behavior collectively
Tables 21 (a) and 21 (b) reports observations from posited by eight alternate models into a model that
the pooled analysis as before. Appendix B reports incorporated four main effects and four
the statistical tests we conducted prior to pooling moderators.
the data for the cross-validation test, consistent
with the approach taken in the preliminary test. Thus, UTAUT is a definitive model that synthe-
Insofar as the no-relationship hypotheses were sizes what is known and provides a foundation to
concerned, the power analysis revealed a high guide future research in this area. By encom-
likelihood (over 95 percent) of detecting medium passing the combined explanatory power of the
effects. The variance explained was quite com- individual models and key moderating influences,
parable to that found in the preliminary test of UTAUT advances cumulative theory while re-
UTAUT. taining a parsimonious structure. Figure 3 pre-
sents the model proposed and supported.
Table 22 presents a summary of the findings. It
should be noted that performance expectancy
Discussion appears to be a determinant of intention in most
situations: the strength of the relationship varies
The present research set out to integrate the with gender and age such that it is more signi-
fragmented theory and research on individual ficant for men and younger workers. The effect of
acceptance of information technology into a uni- effort expectancy on intention is also moderated
fied theoretical model that captures the essential by gender and age such that it is more significant
elements of eight previously established models. for women and older workers, and those effects
First, we identified and discussed the eight decrease with experience. The effect of social
specific models of the determinants of intention influence on intention is contingent on all four
and usage of information technology. Second, moderators included here such that we found it to
these models were empirically compared using be nonsignificant when the data were analyzed
within-subjects, iongitudinai data from four organi- without the inclusion of moderators. Finally, the
zations. Third, conceptual and empirical simi- effect of facilitating conditions on usage was only
larities across the eight models were used to significant when examined in conjunction with the
formulate the Unified Theory of Acceptance and moderating effects of age and experience—i.e.,
Use of Technology (UTAUT). Fourth, the UTAUT they only matter for older workers in later stages
was empirically tested using the original data from of experience.
the four organizations and then cross-validated
using new data from an additional two organi- Prior to discussing the implications of this work, it
zations. These tests provided strong empirical is necessary to recognize some of its limitations.
support for UTAUT, which posits three direct One limitation concerns the scales used to mea-
determinants of intention to use (performance sure the core constructs. For practical analytical
expectancy, effort expectancy, and social influ- reasons, we operationalized each of the core
ence) and two direct determinants of usage constructs in UTAUT by using the highest-loading
behavior (intention and facilitating conditions). items from each of the respective scales. This
Significant moderating influences of experience, approach is consistent with recommendations in
voluntariness, gender, and age were confirmed as the psychometric literature (e.g,, Nunnally and
integral features of UTAUT. UTAUT was able to Bernstein 1994). Such pruning of the instrument
account for 70 percent of the variance (adjusted was the only way to have the degrees of freedom
R^) in usage intention—a substantial improvement necessary to model the various interaction terms

MIS Quarteriy Voi. 21 No. 3/September 2003 467

 Venkatesh et al./User Acceptance of IT

Table 22. Summary of Findings

Hypothesis Dependent Independent
Number Variables Variables Moderators Explanation
HI Behavioral Performance Gender, Age Effect stronger for men and
intention expectancy younger workers
H2 Behavioral Effort Gender, Age, Effect stronger for women.
intention expectancy Experience older workers, and those with
limited experience
H3 Behavioral Social Gender, Age, Effect stronger for women.
intention infiuence Voluntariness, older workers, under conditions
Experience of mandatory use, and with
limited experience
H4a Behavioral Facilitating None Nonsignificant due to the effect
intention conditions being captured by effort
expectancy
H4b Usage Facilitating Age, Effect stronger for older
conditions Experience workers with increasing
experience
H5a Behavioral Computer None Nonsignificant due to the effect
intention self-efficacy being captured by effort
expectancy
H5b Behavioral Computer None Nonsignificant due to the effect
intention anxiety being captured by effort
expectancy
H5c Behavioral Attitude None Nonsignificant to the effect
intention toward using being captured by process
tech. expectancy and effort
expectancy
H6 Usage Behavioral None Direct effect
intention

at the item level as recommended by Chin et al, for each of the constructs with an emphasis on
(1996). However, one danger of this approach is content validity, and then revalidating the model
that facets of each construct can be eliminated, specified herein (or extending it accordingly) with
thus threatening content validity. Specifically, we the new measures. Our research employed stan-
found that choosing the highest-loading items dard measures of intention, but future research
resulted in items from some of the models not should examine alternative measures of intention
being represented in some ofthe core constructs and behavior in revalidating or extending the
(e,g., items from MPCU were not represented in research presented here to other contexts.
performance expectancy). Therefore, the mea-
sures for UTAUT should be viewed as preliminary From a theoretical perspective, UTAUT provides
and future research should be targeted at more a refined view of how the determinants of intention
fully developing and validating appropriate scales and behavior evolve over time. It is important to

468 /W/S Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et at./User Acceptance of IT

emphasize that most of the key relationships in the importance of gender roles and the possibility
the model are moderated. For example, age has that "psychological gender" is the root cause for
received very little attention in the technology the effects observed. Empirical evidence has
acceptance research literature, yet our results demonstrated that gender roles can have a pro-
indicate that it moderates all of the key relation- found impact on individual attitudes and behaviors
ships in the model. Gender, which has received both within and outside the workplace (e.g., Baril
some recent attention, is also a key moderating et al. 1989; Feldman and Aschenbrenner 1983;
influence; however, consistent with findings in the Jagacinski 1987; Keys 1985; Roberts 1997; Sachs
sociology and social psychology literature (e.g., et al. 1992; Wong et al. 1985). Specifically, gen-
Levy 1988), it appears to work in concert with age, der effects observed here could be a manifesta-
a tieretofore unexamined interaction. For exam- tion of effects caused by masculinity, femininity,
ple, prior research has suggested that effort and androgyny rather than just "biological sex"
expectancy is more salient for women (e.g., (e.g., Lubinski et al. 1983). Future work might be
Venkatesh and Morris 2000). While this may be directed at more closely examining the importance
true, our findings suggest tfiis is particularly true of gender roles and exploring the socio-psycho-
for the older generation of workers and those with logical basis for gender as a means for better
relatively tittle experience with a system. Wtiile understanding its moderating role.
existing studies have contributed to our under-
standing of gender and age influences indepen- As is evident from the literature, the role of social
dently, the present research illuminates the inter- influence constructs has been controversial.
play of these two key demographic variables and Some have argued for their inclusion in models of
adds richness to our current understanding of the adoption and use (e.g., Taylor and Todd 1995b;
phenomenon. We interpret our findings to sug- Thompson et al. 1991). while others have not
gest that as the younger cohort of employees in included them (e.g., Davis et al. 1989). Previous
the workforce mature, gender differences in how work has found social influence to be significant
each perceives information technology may only in mandatory settings (see Hartwick and
disappear. This is a hopeful sign and suggests Barki 1994; Venkatesh and Davis 2000). Other
that oft-mentioned gender differences in the use of work has found social influence to be more
information technology may be transitory, at least significant among women in early stages of
as they relate to a younger generation of workers experience (e.g., Venkatesh and Morris 2000).
raised and educated in the Digital Age. Still other research has found social influence to
be more significant among older workers (e.g.,
Morris and Venkatesh 2000). This research is
The complex nature ofthe interactions observed, among the first to examine these moderating influ-
particularly for gender and age, raises several ences in concert. Our results suggest that social
interesting issues to investigate in future research, influences do matter; however, they are more
especially given the interest in today's societal likely to be salient to older vi/orkers, particularly
and workplace environments to create equitable women, and even then during early stages of
settings for women and men of all ages. Future experience/adoption. This pattern mirrors that for
research should focus on identifying the potential effort expectancy with the added caveat that social
magic number" for age where effects begin to influences are more likely to be important in
appear (say for effort expectancy) or disappear mandatory usage settings. The contingencies
(say for performance expectancy). While gender identified here provide some insights into the way
and age are the variables that reveal an inter- in which social influences change over time and
esting pattern of results, future research should may help explain some of the equivocal results
identify the underlying influential mechanisms— reported in the literature. By helping to clarify the
potential candidates here include computer contingent nature of social influences, this paper
literacy and sociai or cultural background, among sheds light on when social influence is likely to
others. Finally, although gender moderates three play an important role in driving behavior and
key relationships, it is imperative to understand when it is less likely to do so.

MIS Quarterly Vol. 27 No. 3/September 2003 469

Venkatesh et al./User Acceptance of IT

UTAUT underscores this point and highlights the the extensive moderating influences examined
importance of contextual analysis in developing here, a research study that examines the general-
strategies for technology implementation within izability of these findings with significant repre-
organizations. While each ofthe existing models sentation in each cell (total number of cells: 24;
in the domain is quite successful in predicting two levels of voluntariness, three levels of experi-
technology usage behavior, it is only when one ence [no, limited, more], two levels of gender, and
considers the complex range of potential moder- at least two levels of age [young vs. old]) would be
ating influences that a more complete picture of valuable. Such a study would allow a palnwise,
the dynamic nature of individual perceptions about inter-cell comparison using the rigorous Chow's
technology begins to emerge. Despite the ability test and provide a clear understanding of the
of the existing models to predict intention and
nature of effects for each construct in each cell.
usage, current theoretical perspectives on indivi-
Related to the predictive validity of this class of
dual acceptance are notably weak in providing
models in general and UTAUT in particular is the
prescriptive guidance to designers. For example,
role of intention as a key criterion in user accep-
applying any ofthe models presented here might
tance research—future research should investi-
inform a designer that some set of individuals
gate other potential constructs such as behavioral
might find a new system difficult to use. Future
expectation (Warshaw and Davis 1985) or habit
research should focus on integrating UTAUT with
research that has identified causal antecedents of (Venkatesh et al. 2000) in the nomological net-
the constructs used within the model (e.g.. work. Employing behavioral expectation will help
Karahanna and Straub 1999; Venkatesh 2000; account for anticipated changes in intention
Venkatesh and Davis 2000) in order to provide a (Warshaw and Davis 1985) and thus shed light
greater understanding of how the cognitive pheno- even in the early stages of the behavior about the
mena that were the focus of this research are actual likelihood of behavioral performance since
formed. Examples of previously examined deter- intention only captures internal motivations to
minants of the core predictors include system perform the behavior. Recent evidence suggests
characteristics (Davis et al. 1989) and self-efficacy that sustained usage behavior may not be the
(Venkatesh 2000). Additional determinants that result of deliberated cognitions and are simply
have not been explicitly tied into this stream but routinized or automatic responses to stimuli (see
merit consideration in future work include task- Venkatesh et al. 2000).
technology fit (Goodhue and Thompson 1994) and
individual ability constructs such as "g"—general
cognitive ability/intelligence (Colquitt et al. 2000). One of the most important directions for future
research is to tie this mature stream of research
into other established streams of work. For
While the variance explained by UTAUT is quite example, little to no research has addressed the
high for behavioral research, further work should link between user acceptance and individual or
attempt to identify and test additional boundary organizational usage outcomes. Thus, while it is
conditions of the model in an attempt to provide often assumed that usage will result in positive
an even richer understanding of technology adop- outcomes, this remains to be tested. The unified
tion and usage behavior. This might take the form model presented here might inform further inquiry
of additional theoretically motivated moderating into the short- and long-term effects of information
influences, different technologies (e.g., collabora- technology implementation on job-related out-
tive systems, e-commerce applications), different comes such as productivity, job satisfaction,
usergroups (e.g., individuals in different functional organizational commitment, and other perfor-
areas), and other organizational contexts (e.g., mance-oriented constructs. Future research
public or government institutions). Results from should study the degree to which systems per-
such studies will have the important benefit of ceived as successful from an IT adoption per-
enhancing the overall generalizability of UTAUT spective (i.e., those that are liked and highly used
and/or extending the existing work to account for by users) are considered a success from an
additional variance in behavior. Specifically, given organizational perspective.

470 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et ai/User Acceptance of IT

Conclusion Aiken, L. S., and West, S. G. Multiple Regres-

sion: Testing and Interpreting Interactions.
Following from the Unified Theory of Acceptance Sage, London, 1991.
and Use of Technology presented here, future Ajzen, 1. "The Theory of Planned Behavior,"
research should focus on identifying constructs Organizational Behavior and Human Decision
that can add to the prediction of intention and Processes (50:2), 1991. pp. 179-211.
behavior over and above what is already known Ashmore, R. D. Sex, Gender, and the Individual,"
and understood. Given that UTAUT explains as in Handbook of Personality, L. A. Pervin (ed.),
much as 70 percent of the variance in intention, it Guilford Press. New York, 1980, pp. 486-526.
is possible that we may be approaching the Bagozzi, R. P., and Edwards, J. R. "A General
practical limits of our ability to explain individual Approach to Construct Validation in Organi-
acceptance and usage decisions in organizations. zational Research: Application to the Measure-
In the study of information technology implemen- ment of Work Values," Organizational Re-
tations in organizations, there has been a search Methods (1:1), 1998, pp. 45-87.
proliferation of competing explanatory models of Bandura, A. Social Foundations of Thought and
individual acceptance of information technology. Action: A Social Cognitive Theory, Prentice
The present work advances individual acceptance Hall, Englewood Cliffs, NJ, 1986.
research by unifying the theoretical perspectives Baril, G. L.. Elbert, N., Mahar-Potter, S., and
common in the literature and incorporating four Reavy, G. C. "Are Androgynous Managers
moderators to account for dynamic influences Really More Effective?," Group and Organiza-
including organizational context, user experience, tion Studies (14:2), 1989, pp. 234-249.
and demographic characteristics. Barnett, R. C, and Marshall, N. L. "The Relation-
ship between Women's Work and Family Roles
and Their Subjective Well-Being and Psycho-
Acknowledgements logical Distress," in Women, Work and Health:
Stress and Opportunities, M. Frankenhaeuser,
The authors thank Cynthia Beath (the senior V. Lundberg, and M. A. Chesney (eds.).
editor), the associate editor, and the three anony- Plenum, New York, 1981, pp. 111-136.
mous reviewers. We also thank Heather Adams, Bem, S. L. "The BSRI and Gender Schema
Wynne Chin, Deborah Compeau, Shawn Curley. Theory: A Reply to Spence and Helmreich,"
Ruth Kanfer, Phillip Ackerman, Tracy Ann Sykes, Psychological Review (88:4), 1981. pp. 369-
Peter Todd, Shreevardhan Leie, and Paul Zantek. 371.
Last, but not least, we thank Jan DeGross for her Bem. D. J., and Allen, A. "On Predicting Some of
tireless efforts in typesetting this rather long the People Some of the Time: The Search for
paper. Cross-Situational Consistencies in Behavior,"
Psychological Review (81:6). 1974. pp. 506-
520.
References Bergeron, F., Rivard, S., and De Serre, L. "Inves-
tigating the Support Role of the Information
Agarwal, R., and Prasad, J. "A Conceptual and Center," MIS Ouarterly (14:3), 1990, pp. 247-
Operational Definition of Personal Innovative- 259.
ness in the Domain of Information Technology," Bozionelos, N. "Psychology of Computer Use:
Information Systems Research (9:2). 1998, pp. Prevalence of Computer Anxiety in British
204-215. Managers and Professionals." Psychological
Agarwal. R., and Prasad. J. "The Role of Innova- Reports (78:3), 1996, pp. 995-1002.
tion Characteristics and Perceived Voluntari- Chin. W. W. "The Partial Least Squares Ap-
ness in the Acceptance of Information Techno- proach for Structural Equation Modeling," in
logies," Decision Sciences (28:3), 1997. pp. Modern Methods for Business Research,
557-582. George A. Marcoulides (ed.). Lawrence

MIS Quarterly Vol. 27 No. 3/September 2003 471

Venkatesh et al./User Acceptance of IT

Erlbaum Associates, New York, 1998, pp. 295- Eichinger, J.. Heifetz, L.J., and Ingraham, C.
336. "Situational Shifts in Sex Role Orientation:
Chin. W. W., Marcolin, B. L., and Newsted, P. R. Correlates of Work Satisfaction and Burnout
"A Partial Least Squares Latent Variable Among Women in Special Education," Sex
Modeling Approach for Measuring Interaction Roles (25:7/8), 1991, pp. 425-430.
Effects: Results from a Monte Carlo Simulation Feldman. S. S., and Aschenbrenner, B. "Impact
Study and Voice Mail Emotion/Adoption Study," of Parenthood on Various Aspects of Mascu-
in Proceedings of the International Conference linity and Femininity: A Short-Term Longitu-
on information Systems, J. I. DeGross, A. Srini- dinal Study," Deveiopmentai Psyci^ology{^9:2),
vasan, and S. Jarvenpaa (eds.), Cleveland, 1983, pp. 278-289.
OH, 1996, pp. 21-41. Fiske, S. T., and Taylor, S. E. Sociai Cognition.
Chow, G. C. "Tests of Equality Between Sets of McGraw-Hill, New York, 1991.
Coefficients in Two Linear Regressions," Fishbein, M., and Ajzen, I. Belief, Attitude. Inten-
Econometrica (28:3), 1960, pp. 591-605. tion and Behavior: An introduction to Theory
Colquitt, J. A., LePine, J. A., and Noe, R. A. and Research. Addison-Wesley, Reading, MA.
"Toward an IntegrativeTheory of Training Moti- 1975.
vation: A Meta-Analytic Path Analysis of 20 Fornell, C , and Larcker, D. F. "Evaluating Struc-
Years of Training Research," Journal of Applied tural Equation Models with Unobservable
Psyciiology (85:5), 2000, pp. 678-707. Variables and Measurement Error: Algebra
Compeau, D. R., and Higgins, C. A. "Application and Statistics," Journai ofMari<eting Research
of Sociai Cognitive Theory to Training for Com- (18:3), 1981, pp. 382-388.
puter Skills," Information Systems Research French, J. R. P., and Raven, B. "The Bases of
(6:2), 1995a, pp. 118-143. Social Power," in Studies in Sociai Power. D.
Compeau. D. R., and Higgins, C. A. "Computer Cardwright (ed). Institute for Social Research,
Self-Efficacy: Development of a Measure and Ann Arbor. Ml, 1959, pp. 150-167.
Initial Test," MIS Quarterly (19:2), 1995b, pp. Goodhue, D. L. "Understanding User Evaluations
189-211. of Information Systems," Management Science
Compeau, D. R., Higgins, C. A., and Huff, S. (41:12), 1995, pp. 1827-1844.
"Social Cognitive Theory and Individual Reac-
Goodhue, D. L., and Thompson, R. L. "Task-
tions to Computing Technology: A Longitudinal
Technology Fit and Individual Performance,"
Study," MIS Quarteriy (23:2), 1999, pp. 145-
MIS Quarterly 09:2). 1995. pp. 213-236.
158.
Hall, D., and Mansfield. R. "Relationships of Age
Davis, F. D. "Perceived Usefulness, Perceived
and Seniority with Career Variables of Engi-
Ease of Use, and User Acceptance of Informa-
neers and Scientists," Journai of Applied
tion Technology," MIS Quarteriy (13:3), 1989,
Psychology {&Q:2), 1995, pp. 201-210.
pp. 319-339.
Harrison, D. A.. Mykytyn, P. P., and Riemen-
Davis, F. D., Bagozzi, R. P., and Warshaw, P. R.
schneider, C. K. "Executive Decisions About
"Extrinsic and Intrinsic Motivation to Use Com-
Adoption of Information Technology in Small
puters in the Workplace," Journal of Applied
Business: Theory and Empirical Tests," Infor-
Social Psychoiogy (22:14), 1992, pp. 1111-
mation Systems Research (8:2), 1997, pp.
1132.
171-195.
Davis, F. D., Bagozzi, R. P., and Warshaw. P. R.
Hartwick, J., and Barki, H. "Explaining the Role of
"User Acceptance of Computer Technology: A
User Participation in Information System Use,"
Comparison of Two Theoretical Models," Man-
Management Science (40 A), 1994, pp. 40-465.
agement Science (35:8), 1989, pp. 982-1002.
Efron, B., and Gong, G. "A Leisurely Look at the Helson, R., and Moane, G. "Personality Change
Bootstrap, the Jackknife, and Cross-Validation," in Women from College to Midlife." Journal of
The American Statistician (37:1), 1983, pp. 36- Personality and Sociai Psychology (53:1),
48. 1987, pp. 176-186.

472 MIS Quarterty Vol. 27 No. 3/September 2003

 VenkatBsh et al./User Acceptance of IT

Hu, P. J., Chau, P. Y. K,, Sheng, O. R. L., and the Theory of Planned Behavior," Information
Tam. K. Y, "Examining the Technology Accep- Systems Research (2:3), 1991, pp. 173-191.
tance Model Using Physician Acceptance of Miller, J. B. Toward a New Psychology of
Je\eme6\dnelechno\ogy" Journal of Manage- Women, Beacon Press, Boston, 1976.
ment Information Systems (^6:2), 1999, pp, 91- Minton, H. L., and Schneider, F. W. Differential
112. Psychology, Waveland Press, Prospect
Jagacinski. C. M. "Androgyny in a Male-Domi- Heights, IL, 1980.
nated Field: The Relationship of Sex-Typed Moore, G. C , and Benbasat, I, "Development of
Traits to Performance and Satisfaction in Engi- an Instrument to Measure the Perceptions of
neering, Sex Roles (17:X), 1987. pp. 529-547. Adopting an Information Technology Innova-
Karahanna, E., and Straub, D. W. "The Psycho- tion," Information Systems Research (2:3),
logical Origins of Perceived Usefulness and 1991, pp. 192-222.
Ease of Use." Information and Management Moore, G. C , and Benbasat, I. "Integrating Diffu-
(35:4), 1999, pp, 237-250. sion of Innovations and Theory of Reasoned
Karahanna, E., Straub, D. W., and Chervany, N. Action Models to Predict Utilization of Infor-
mation Technology by End-Users," in Diffusion
L. "Information Technology Adoption Across
and Adoption of Information Technology, K.
Time: A Cross-Sectional Comparison of Fre-
Kautz and J. Pries-Hege (eds.). Chapman and
Adoption and Post-Adoption Beliefs," MIS
Hall, London, 1996, pp. 132-146.
Quarterly {23:2), 1999, pp, 183-213.
Morris, M. G,, and Venkatesh, V. "Age Dif-
Keys, D. E. "Gender, Sex Role, and Career
ferences in Technology Adoption Decisions:
Decision Making of Certified Management Ac-
Implications for a Changing Workforce," Per-
countants," Sex Roles (13:1/2), 1985, pp. 33-
sonnel Psychology (53:2), 2000, pp. 375-403.
46,
Motowidio, S. J. "Sex Role Orientation and
Kirchmeyer, C. "Change and Stability in Mana-
Behavior in a Work Setting," Journal of Per-
ger's Gender Roles," Journal of Applied
sonality and Social Psychology (42:5), 1982,
Psychology {87:5), 2002, pp. 929-939.
pp. 935-945,
Kirchmeyer, C, "Gender Roles in a Traditionally
Nunnally, J, C , and Bernstein, I, H, Psychometric
Female Occupation; A Study of Emergency,
Theory (3'" ed.), McGraw-Hill, New York, 1994,
Operating, Intensive Care, and Psychiatric
Olfman, L., and Mandviwalla, M. "Conceptual
Nurses, Journal of Vocational Behavior (50:1),
Versus Procedural Software Training for
1997, pp. 78-95. Graphical User Interfaces: A Longitudinal Field
Leonard-Barton, D., and Deschamps, I. "Mana- Experiment," MIS Quarterly (18:4), 1994, pp.
gerial Influence in the Implementation of New 405-426,
Technology," Management Science (34:10), Plouffe, C. R., Hulland, J. S., and Vandenbosch,
1988. pp. 1252-1265, M. "Research Report: Richness Versus Parsi-
Levy, J. A. "Intersections of Gender and Aging," mony in Modeling Technology Adoption Deci-
The Sociological Quarterly (29:4), 1988, pp, sions—Understanding Merchant Adoption of a
479-486. Smart Card-Based Payment System," Infor-
Lubinski, D,, Tellegen, A. and Butcher, J. N, mation Systems Research (12:2), 2001, pp.
"Masculinity, Femininity, and Androgyny 208-222.
Viewed and Assessed as Distinct Concepts," Plude, D., and Hoyer, W. "Attention and Perfor-
Journal of Personality and Social Psychology mance: Identifying and Localizing Age Defi-
(44:2), 1983, pp. 428-439, cits," in Aging and Human Performance, N,
Lynott, P, P., and McCandless, N. J. "The Impact Charness (ed.), John Wiley & Sons, New York,
of Age vs. Life Experiences on the Gender Role 1985, pp. 47-99.
Attitudes of Women in Different Cohorts," Jour- Porter, L. "Job Attitudes in Management: Per-
nal of Women and Aging {12:2),2000, pp. 5-2'i. ceived Importance of Needs as a Function of
Mathieson, K. "Predicting User Intentions: Com- Job Level," Journal of Applied Psychology
paring the Technology Acceptance Model with (47:2), 1963, pp. 141-148.

MIS Quarterly Vol. 27 No. 3/September 2003 473

Venkatesh et al./User Acceptance of IT

Rhodes. S. R. "Age-Related Differences in Work Transactions on Engineering Management

Attitudes and Behavior: A Reviewand Concep- (29:1), 1982, pp. 28-45.
tual Analysis," Psychological Bulletin {93:2), Triandis, H. C. Interpersonal Behavior, Brooke/
1983, pp. 328-367. Cole, Monterey, CA, 1977.
Roberts, R. W. "Plaster or Plasticity: Are Adult Twenge, J. M. "Changes in Masculine and
Work Experiences Associated with Personality Feminine Traits Over Time: A Meta-AnatysJs,"
Change in Women?." Journal of Personality Sex Roles (35:5/6), 1997, pp. 305-325.
(65:2), 1997, pp. 205-232. Vallerand, R. J. "Toward a Hierarchical Model of
Rogers, E. Diffusion of Innovations. Free Press, Intrinsic and Extrinsic Motivation," in Advances
New York, 1995. in Experimental Social Psychology (29), M.
Rogers, E. M., and Shoemaker, F. F. Com- Zanna(ed.), Academic Press, New York, 1997.
munication of Innovations: A Cross-Cultural pp. 271-360.
Approach, Free Press, New York, 1971. Venkatesh, V. "Creating Favorable User Percep-
Sachs, R., Chrisler, J. C , and Devlin, A. S. tions: Exploring the Role of Intrinsic Moti-
"Biographic and Personal Characteristics of vation," MIS Quarterly (23:2), 1999, pp. 239-
Women in Management,"Jouma/of \/ocaf/ona/ 260.
Behavior(41 A), 1992, pp. 89-100. Venkatesh, V. "Determinants of Perceived Ease
Sheppard, B. H., Hartwick, J., and Warshaw, P. R. of Use: Integrating Perceived Behavioral
"The Theory of Reasoned Action: A Meta- Control, Computer Anxiety and Enjoyment into
Analysis of Past Research with Recommen- the Technology Acceptance Model," Informa-
dations for Modifications and Future Research," tion Systems Research (11:4), 2000, pp. 342-
Journal of Consumer Research (15:3), 1988, 365.
pp. 325-343. Venkatesh, V., and Davis, F. D. "A Theoretical
Straub, D., Limayem, M., and Karahanna, E. Extension of the Technology Acceptance
Model: Four Longitudinal Field Studies," Man-
"Measuring System Usage: Implications for IS
agement Science (45:2), 2000, pp. 186-204.
Theory Testing," Management Science (41:8),
Venkatesh, V., and Morris, M. G. "Why Don't Men
1995, pp. 1328-1342.
Ever Stop to Ask For Directions? Gender,
Szajna, B. "Empirical Evaluation ofthe Revised
Social Influence, and Their Role in Technology
Technology Acceptance Model," Management
Acceptance and Usage Behavior," MIS
Science (42:1), 1996, pp. 85-92.
Quarterly (24.1), 2000, pp. 115-139.
Taylor, S., and Todd, P. A. "Assessing IT Usage:
Venkatesh, V., Morris, M. G., and Ackerman, P. L.
The Role of Prior Experience," MIS Quarterly
"A Longitudinal Field Investigation of Gender
(19:2), 1995a, pp. 561-570.
Differences in Individual Technology Adoption
Taylor, S., and Todd, P. A. "Understanding Infor-
Decision Making Processes," Organizational
mation Technology Usage: A Test of Com- Behavior and Human Decision Processes
peting Models," Information Systems Research (83:1), 2000, pp. 33-60.
(6:4), 1995b, pp. 144-176.
Venkatesh, V., and Speier, C. "Computer Tech-
Thompson, R. L., Higgins, C. A., and Howell, J. M. nology Training in the Workplace: A Longi-
"Influence of Experience on Personal Computer tudinal Investigation ofthe Effect ofthe Mood,"
Utilization: Testing a Conceptual Model," Organizational Behavior and Human Decision
Journal of Management Information Systems Processes (79:1), 1999, pp. 1-28.
(11:1), 1994, pp. 167-187. Warshaw, P. R. "A New Model for Predicting
Thompson. R. L., Higgins, C. A., and Howell, J. M. Behavioral Intentions: An Alternative to Fish-
"Personal Computing: Toward a Conceptual bein," Journal of Marketing Research (17:2),
Model of Utilization," MIS Quarterly (15:1), 1980, pp. 153-172.
1991, pp. 124-143. Warshaw, P. R., and Davis, F. D. "Disentangling
Tornatzky, L. G., and Klein, K. J. "Innovation Behavioral Intention and Behavioral Expec-
Characteristics and Innovation Adoption-Imple- tation," Journal of Experimental Social Psych-
mentation: A Meta-Analysis of Findings," IEEE ology (21:3). 1985, pp. 213-228.

474 MIS Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

West, S. G., and Hepworth, J. T., "Statistical Gordon B. Davis, Honeywell Professor of
Issues in the Study of Temporal Data: Daily Management Information Systems in the Carlson
Experiences," Journai of Personaiity (59:2), School of Management at the University of Minne-
1991. pp. 609-662. sota, is one of the founders of the academic
Westland, J. C . and Clark, T. H- K. Globai Bec- discipline of information systems. He has lectured
tronic Commerce: Theory and Case Studies, in 25 countries and has written 20 books and over
MIT Press, Cambridge, MA. 2000. 200 articles, monographs, and book chapters. He
Wong, P. T. P.. Kettlewell, G., and Sproule, C. F. participated in and helped form the major aca-
"On the Importance of Being Masculine: Sex demic associations related to the field of man-
Role, Attribution, and Women's Career agement information systems. He has been
Achievement," Sex Roies (12:7/8), 1985, pp. honored as an ACM Fellow and an AIS Fellow,
757-769. and is a recipient of the AIS LEO award for
lifetime achievement in the field of information
systems. His research interests include concep-
About the Authors tual foundations of information systems, informa-
tion system design and implementation, and
Viswanath Venkatesh is Tyser Fellow, associate management of the IS function. He has a Ph.D.
professor, and director of MBA Consulting at the from Stanford University and honorary doctorates
Robert H. Smith School of Business at the Univer- from the University of Lyon, France, the University
sity of Maryland, and has been on the faculty there of Zurich, Switzerland and the Stockholm School
since 1997. He earned his Ph.D. at the University of Economics, Sweden.
of Minnesota's Carlson School of Management.
His research interests are in implementation and Fred D. Davis is David D. Glass Endowed Chair
use of technologies in organizations and homes. Professor in Information Systems and Chair of the
His research has been published in many leading Information Systems Department at the Sam M.
journals including MIS Quarteriy, Information Walton College of Business at the University of
Systems Research, Management Science. Com- Arkansas. Dr. Davis earned his Ph.D. at MIT's
munications of the ACM, Decision Sciences. Sloan School of Management, and has served on
Qrganizationai Behavior and i-iuman Decision the business school faculties at University of
Processes, Personnel Psychology. He is currently Michigan, University of Minnesota, and University
an associate editor at MiS Quarterly and infor- of Maryland. He has taught a wide range of infor-
mation Systems Research. He received MIS mation technology (IT) courses at the under-
Quarteriy's'Reviewerof the Year" award in 1999. graduate, MBA, Ph.D., and executive levels. Dr.
Davis has served as associate editor for the
Michael G. Morris is an assistant professor of scholarly journals Management Science, MiS
Commerce within the Information Technology area Quarteriy, and information Systems Research. He
at the Mclntire School of Commerce, University of has published extensively on the subject of user
Virginia. He received his Ph.D. in Management acceptance of IT and IT-assisted decision making.
Information Systems from Indiana University In His research has appeared in such journals as
1996. His research interests can broadly be clas- MIS Quarterly, information Systems Research,
sified as socio-cognitlve aspects of human Management Science. Journal of Experimental
response to information technology, including user Social Psychoiogy, Decision Sciences, Organiza-
acceptance of information technology, usability tionai Behavior and i-iuman Decision Processes.
engineering, and decision-making. His research Communications of the AIS. and Journal of MiS.
has been published in MiS Quarterly, Qrganiza- Current research interests include IT training and
tionai Behavior and Human Decision Processes, skill acquisition, management of emerging IT, and
and Personnel Psychoiogy, among others. the dynamics of organizational change.

MIS Quarterty Vol. 27 No. 3/September 2003 475

Venkatesh et al./User Acceptarjce of IT

Appendix A

Cautions Related to Pooling Data and Associated Statistical

Tests for Preliminary Test of UTAUT (Studies 1 and 2)

The most critical concern when pooling repeated measures from the same subjects is the possibility of
correlated errors. West and Hepworth describe this problem as follows:

A perusal of the empirical literature over the past 30 years reveals some of the statistical
approaches to repeated measures over time that have been used by personality
researchers. The first and perhaps most classic approach is to aggregate the
observations collected over time on each subject to increase the reliability of
measurement (1991, p. 610).

However, In many cases, these observations will often violate the assumption of independence, requiring
alternate analyses (see West and Hepworth for an extended discussion of alternate approaches).

When error terms can be shown to be independent. West and Hepworth note that" traditional statistical
analyses such as ANOVA or MR [multiple regression] can be performed directly without any adjustment
of the data." (pp. 612-613). The best way to determine whether it is appropriate to pool data Is to conduct
a test for correlated errors. In order for it to be acceptable to pool the data, the error terms should be
uncorrelated. A second approach uses bootstrapping to select subsamples to conduct a between-subjects
examination of the within-subjects data, in order for it to be acceptable to pool the data, this second test
should yield identical results to the test of the model on the complete data set. Below we report the results
from the specific findings from the correlated errors test and the pattern of findings from the second test.

Correlated Errors Test

We computed the error terms associated with the prediction of intention at T1. T2, and T3 in studies 1
(voluntary settings) and 2 (mandatory settings). Further, we also calculated the error terms when pooled
across both settings—i.e., for cross-sectional tests of the unified model at T l , T2, and T3 respectively.
These computations were conducted both for the preliminary test and the cross-validation (reported below).
The error term correlations across the intention predictions at various points in time are shown below. Note
that aii error correlations are nonsignificant and. therefore, not significantly different from zero in ali
situations.

Between-Subjects Test of Within-Subjects Data

While the results above are compelling, as a second check, we pooled the data across different levels of
experience (Tl, T2, and T3) and used PLS to conduct a between-subjects test using the within-subjects
data. We used a DOS version of PLS to conduct the analyses (the reason for not using PLS-Graph as in
the primary analyses in the paper was because it did not allow the specification of filtering rules for
selecting observations in bootstraps). Specifically for the between-subjects test, we applied a filtering rule
that selected any given respondent's observation at only one of the three time periods. This approach en-

476 M/S Quarterly Vol. 27 No. 3/September 2003

 Venkatesh et al./User Acceptance of IT

Table A1. Correlations Between Error Terms of Intention Construct at Various Time
Periods
Study 1 (Voluntary) T1 T2 T3
TI
T2 .04
T3 .11 .09
Study 2 (Mandatory) T1 T2 T3
TI
T2 .07
T3 .08 .13
Study 1 and 2 (Pooled) T1 T2 T3
TI
T2 .06
T3 .09 .10

sured that the data included to examine the interaction terms with experience did not include any potentiai
for systematic correlated errors. Using 50 such random subsampies. the model was tested and the resuits
derived supported the pattern observed when the entire data set was pooled across experience.

Taken together, the analyses reported above support the pooling of data (see Table 17) across ievels of
experience and eiiminate the potential statistical concerns noted by West and Hepworth in the analysis of
temporal data.

MIS Quarterly Voi 27 No. 3/September 2003 477

Venkatesh et al./User Acceptance of IT

Appendix B

statistical Tests Prior to Pooling of Data for

Cross-Validation of UTAUT (Studies 3 and 4)

As with the test of the preliminary model, prior to pooling the data for the cross-validation studies (studies
3 and 4), we conducted statistical tests to examine the independence of observations (as detailed in
Appendix A). The table below presents the error term correlation matrices for intention for studies 3
(voluntary) and 4 (mandatory) as well as pooled across both settings at T I , T2, and T3 respectively.

As in the preliminary test of UTAUT, the error correlation matrices above suggest that there is no constraint
to pooling in the cross-validation study of the model. As before, a between-subjects test of the within-
subjects data was tested using PLS (as described in Appendix A), and the results of that test corroborated
the independence of observations in this sample. In light of both sets of results, we proceeded with the
pooled analysis as reported in the body of the paper (see Table 21).

Table B1. Correlations Between Error Terms of Intention Construct at Various Time
Periods
Study 3 (Voluntary) T1 T2 T3
TI
T2 .01
T3 .07 .11
Study 4 (Mandatory) T1 T2 T3
TI
T2 .04
T3 .02 .08
Study 3 and 4 (Pooled) T1 T2 T3
TI
T2 .03
T3 .05 .10

MIS Quarterly Vol. 27 No. 3/September 2003