+1443 776-2705 panelessays@gmail.com

After reading the text and the supplemental article on Identity Tourism, discuss what you think about the concept. Is it ok to be who you want to be online? Is this just another example of cultural appropriations? Are avatars creating a universal online acceptance that face-to-face society won’t allow? Are people being their authentic selves online or are they impostures? 

Answer questions and read article

How important is the “social” in
social networking? A perceived
value empirical investigation

Mihail Cocosila
Faculty of Business, Athabasca University, Athabasca, Canada, and

Andy Igonor
JR Shaw School of Business, Northern Alberta Institute of Technology,

Edmonton, Canada

Purpose – The purpose of this paper is to report on a value-based empirical investigation of the
adoption of Twitter social networking application. The unprecedented popularity of social networking
applications in a short time period warrants exploring theory-based reasons of their success.
Design/methodology/approach – A cross-sectional survey-based study to elicit user views on
Twitter was conducted with participants recruited through the web site of a North-American university.
Findings – All facets of perceived value considered in the study (utilitarian, hedonic and social) had
a significant and relatively strong influence on consumer intent to use Twitter. Quite surprisingly for a
social networking application, though, the social value facet had comparatively the weakest contribution
in the use equation.
Research limitations/implications – User value perception might have been influenced by the
features of the actual social networking application under scrutiny (i.e. Twitter in this case).
Practical implications – To maximize the chances of success of new social networking applications,
developers and marketers of these media should focus on the hedonic and utilitarian sides of their
perceived value.
Social implications – Additional efforts are necessary to better understand the reasons and factors
leading to a comparatively lower social value perception of a social networking application, compared
to its hedonic and utilitarian values.
Originality/value – Overall, the study opens the door for investigating user perceptions on popular
social networking applications in an effort to understand the unparalleled success of these services in a
short time period.
Keywords Perceptions, Social media, Technology adoption, Social networking, Perceived value,
Twitter, User satisfaction
Paper type Research paper

1. Introduction
Social networking applications recorded an unprecedented success in just few of the
recent years. For instance, people in the USA have been spending 22 per cent of
the time they are online on social media sites while nine million users in Australia have
been spending almost nine hours per month, on average, using top social media
applications (Wikipedia, 2012). Despite these astonishing figures, the social networking
domain is still little understood. Definitions and borders of the social networking (also
called social media) phenomenon are still under debate. However, scholars seem to
agree that content generated by users is the key feature of any social networking
application. For instance, some conceptualization attempts define social media as “a
group of Internet-based applications that build on the ideological and technological

Information Technology & People
Vol. 28 No. 2, 2015
pp. 366-382
© Emerald Group Publishing Limited
DOI 10.1108/ITP-03-2014-0055

Received 19 March 2014
Revised 19 August 2014
6 October 2014
Accepted 30 October 2014

The current issue and full text archive of this journal is available on Emerald Insight at:

This research was supported by a grant from Athabasca University.



foundations of Web 2.0, and that allow the creation and exchange of User Generated
Content” (Kaplan and Haenlein, 2010).

The exponential growth of the number of users and of the frequency of use of these
applications together with the still not enough understood influence on various
domains of human interaction (Gruzd et al., 2012; Shneiderman et al., 2011) attracted a
justified interest from both the business community and the academia. While business
decision makers are exploring ways to turn this phenomenon into profits, academia
is seeking to investigate through theory-based approaches the reasons for this
tremendous success (Abedniya and Mahmouei, 2010; Culnan et al., 2010; Spaulding,
2010; Zhao and Lu, 2012).

As it is well-known from information systems (IS) research that user perceptions are,
further than business or technical aspects, the key factor determining the success or
failure of any new information technology (IT) application (Venkatesh et al., 2002), an
interesting topic of research is to look at social networking applications from a
technology acceptance point of view in order to understand their social penetration.
Out of several possibilities of theoretical investigation examined by previous research
(Al-Debei et al., 2013; Gruzd et al., 2012; Hargittai, 2007; Sinclaire and Vogus, 2011;
Turel and Serenko, 2012; Xu et al., 2012), applying perceived value models seems to be
particularly interesting since individuals are presumably using an IT application only
if this has value for them. Perceived value concept is, thus, broader than business
value that captures measurable benefits mainly (Culnan et al., 2010). Furthermore, value
perception allows a more granular investigation as it captures and ranks individuals’
views from several perspectives. For these reasons perceived value approach has been
used in IS research as a possible way to understand the adoption of other new (but fast
becoming widely popular) IT artifacts like cell phone value-added services (Turel et al.,
2007, 2010).

The research reported here focuses on one of the most popular and frequently used social
networking applications, Twitter, available at Twitter.com, in an effort to understand its
user success through a theoretical lens. An empirical research investigating the role of a
multi-sided perception of Twitter’s value in the use of this application was conducted with
participants familiar with Twitter that were recruited through the web site of a university in
North America. This paper reports on that research as follows: next two sections describe
the theoretical background and the proposed research model. Following that, research
methodology and main results are presented. A discussion section concludes the paper.

2. Theoretical background
Investigating factors of adoption of new information technologies or applications has been
a traditional area of research in IS. In addition to the popular models and theories
validated in various studies (for a detailed review, see Venkatesh et al., 2002 study), a
relatively newer approach has been to examine the adoption of an IT from a value
perspective. This path was adapted from other disciplines like consumer behaviour or
economics where value is used to explain why people buy some things or opt to make
some expenses (Turel et al., 2010). Value is considered to source from the actual interaction
with (or expected use of) a product or service and to reflect an overall perception upon
their importance for an individual. Thus, following a rationale borrowed from consumer
behaviour, value is captured as a perceived value concept through individual views on the
difference between “what is received and what is given” (Zeithaml, 1988).

Although value in marketing was traditionally associated with the perception of the
utility of a product or service, more recent research using perceived value in other


A perceived


disciplines, including IS, acknowledged this construct to be multi-sided (Lee et al., 2002;
Turel et al., 2007, 2010). Although the multi-dimensionality seems to better capture the
complexity of the concept, there are no unanimous opinions on the actual facets of
perceived value. A review of literature stemming from consumer behaviour research
shows some of the most popular components of perceived value of an object or service
to be the following (Bolton and Drew, 1991; Kim et al., 2007; Sweeney and Soutar, 2001;
Sheth et al., 1991):

• functional or utilitarian (i.e. perception of utility associated with the use);
• emotional or hedonic (i.e. state of mood associated with the use);
• monetary or value-for-money (i.e. utility compared to the cost usage involves);

• social (i.e. self-perception of social status associated with the use).

Due to its complexity and multi-sided approach, perceived value is a possible lens to
investigate the adoption of social networking applications that became overwhelmingly
popular in recent years. Twitter micro-blogging service, allowing users to post 140-
character long messages on their daily activities or opinions (Zhao and Rosson, 2009), is a
typical example of success and, hence, an interesting avenue worth researching. Since its
launch in 2006 this service grew exponentially thus reaching in early 2011 about 130
million postings (or “tweets”) per day and even 3,000 per second during major events
worldwide (Wakefield, 2011).

Among various attempts to understand the success of this social media platform
from various angles, it would be interesting to investigate the role of the value users
perceive in Twitter on their adoption intention as it is well-known in IS research that
user perceptions are a key ingredient of the adoption equation. The use of a multi-sided
value approach would allow also seeing whether social reasons are the main motivator
of social media use, as opinions expressed on the internet often assume. Therefore, this
study proposes the following research question:

RQ1. What is the order of importance of the key facets of perceived value that
influence the adoption of Twitter social networking application?

3. Research model
To investigate the perceived value of Twitter, this study proposes a multi-faceted
perceived value construct sourcing from consumer behaviour and IS research. This
multi-dimensional value perception should have a positive influence on the intention to
use the social medium since people would use a service if they perceive it as valuable
for various reasons (Ho and Ko, 2008). Taking into the account of the above, the
following hypothesis is proposed:

H1. The overall perceived value of Twitter social networking application will have a
positive effect on the behavioural intention to use this application.

A consistent body of research identified three facets of perceived value, as discussed in
the section above: utilitarian, hedonic and social (Brown and Venkatesh, 2005; Kim and
Han, 2009; Kim et al., 2005). Some studies also include the monetary side borrowed from
consumer behaviour as a distinct facet (Turel et al., 2010) or as a component of the
utilitarian side (Rintamäki et al., 2006). As the use of social media, including Twitter,
does generally not imply a fee or monetary expense, this research will, hence, consider



only the utilitarian, hedonic and social sides of perceived value as being significant.
Therefore, users would perceive a value in this social media application if using it is
observed to help accomplish some utility needs, to be a source of entertainment, and to
meet social goals. To measure these aspects, following the example of similar work
(Turel et al., 2007, 2010), perceived value is conceptualized as a second-order construct
with three facets. Accordingly, the following hypotheses are formulated:

H2-1. The utilitarian dimension of perceived value of Twitter social networking
application will have a positive effect on the overall perceived value of
this application.

H2-2. The hedonic dimension of perceived value of Twitter social networking
application will have a positive effect on the overall perceived value of
this application.

H2-3. The social dimension of perceived value of Twitter social networking
application will have a positive effect on the overall perceived value of
this application.

As virtually all discussions in the media presume with consistency that the success of
social networking applications reside mostly in their social implications, both in an
individual and business context (Lorenzo-Romero et al., 2011), social facet of the
perceived value is expected to be predominant in the adoption equation. Accordingly,
in the attempt to identify at a granular level the key social aspects in the adoption
equation, similar to previous research (Kwon and Wen, 2010), this research looks with
magnifying lenses at the social dimension of perceived value. Previous research
indicated enhancement of status (i.e. impression the individuals give to others) (Brown
and Venkatesh, 2005; Rintamäki et al., 2006) and of self-esteem (i.e. one’s concept of self)
(Rintamäki et al., 2006) as possible factors influencing perceived social value. These are
conceptualized in the present study as image, that is an adaptation from Venkatesh and
Davis (2000) and expresses individuals’ perception of their status in the social network.

Another social aspect of using the IT put in light by previous research is group
integration (i.e. socialization by belonging to groups) (Lee et al., 2002). Theoretical
reasoning shows this feeling is captured partially through image and partially through
perceived social presence. This latter is defined as individuals’ ability “to project
themselves socially and affectively into a community” of users (Rourke et al., 1999), was
taken into account in earlier IS research on traditional media (Yoo and Alavi, 2001) and
is thought to be important for social media (Xu et al., 2012).

It is believed that the above two factors would capture the enhancement of status
sourcing from the individual perspective of the social environment. In addition to the
above, to complete the picture, we suggest that two other factors may determine
perceived social value by capturing the enhancement of self-concept because of the
influence of the social environment on an individual: critical mass and social norm.
Perceived critical mass, understood as a minimum level of users adopting an IT
innovation after which “its further rate of adoption becomes self-sustaining” (Van Slyke
et al., 2007), was shown to be an important factor of the adoption of the new IT (Hsu and
Lu, 2004; Kumar and Benbasat, 2006). Since perceived critical mass depends on the
number of users already using the system (hence this is an indicator of the social
“success” of a system), it is considered as an antecedent of the perceived social value.
Social norm (or subjective norm) is the social influence regarding the use of a new
system. This represents “the degree to which an individual perceives that important


A perceived


others believe he or she should use the new system” (Venkatesh et al., 2003) and is an
essential side of the social aspects of using a new IT (Dickinger et al., 2008). Taking into
the account all of the above, the following hypotheses are formulated:

H3-1. Image of users of Twitter social networking application will have a positive
effect on the social dimension of the overall perceived value of this application.

H3-2. Perceived social presence of users of Twitter social networking application
will have a positive effect on the social dimension of the overall perceived
value of this application.

H3-3. Perceived critical mass of users of Twitter social networking application will
have a positive effect on the social dimension of the overall perceived value of
this application.

H3-4. Social norm exerted on users of Twitter social networking application will
have a positive effect on the social dimension of the overall perceived value of
this application.

The theoretical model and associated hypotheses are captured in Figure 1.

4. Methodology
Model and hypotheses were tested through a cross-sectional experiment comprising an
online survey. To ensure reliable psychometric properties, survey questions measuring
the items of the latent variables were adapted from measures previously validated in
consumer behaviour and IS research, as reported in top publications (Cyr et al., 2009;
Kim and Han, 2009; Turel et al., 2007; Van Slyke et al., 2007; Venkatesh and Davis,
2000). Survey measures and their related constructs are presented in the Appendix.








Second Order

H2 -1

H2 -2

H2 -3

H3 -1 H3 -2 H3 -3 H3 -4



Figure 1.
Theoretical model
and hypotheses



Participants were recruited through announcements posted on the main web page of
the Faculty of Business of a North-American university. Including conditions required
interested participants to be at least 18 years old and be familiar with Twitter without
necessarily having an account with this service. Participation was anonymous and
respondents were not compensated for completing the survey.

The survey was offered online to all individuals who self-reported meeting the
including conditions and were willing to participate. As the research targeted a dynamic
IT domain, in order to ensure homogeneity of the data collected the survey instrument
were available online for three months in the first half of the year 2012. This survey was
part of a larger data collection process conducted in the same setting.

5. Main findings
A total of 134 valid responses were recorded at the end of the three-month data
collection. A demographic analysis indicated that respondents were 39.0 years old on
average, 60.8 per cent female and 39.2 per cent male. Participants reported having an
average experience with Twitter of 1.6 years and checking the service 20.1 times a
week, on average. A per cent of 61.2 of the respondents reported having a Twitter
account, 28.4 per cent not having an account, while the rest of 10.4 per cent preferred
to not answer this question. Participants having an account reported posting 10.2
messages per week, on average. They were following (i.e. subscribing to the posts of)
84.8 accounts and were having 82.7 followers, on average. All 134 participant valid
responses were subjected to the data analysis.

Data collected were first subjected to a test for non-response bias. Following the
example of previous research, this test was performed by comparing the key
demographics of early to those of late responders (Dimoka et al., 2012; Sun et al., 2009).
Averages for age, gender, Twitter experience and activity did not significantly differ
between the two groups and, therefore, non-response bias was not considered an issue.

A second step in the assessment of the quality of the data collected were to test on the
existence of common method variance (CMV). Tests to, possibly, identify CMV appeared
necessary since all variables in the model were measured through self-reported data
collected in the same one-step survey (Sharma et al., 2009). A Harman’s one-factor test was
conducted following Podsakoff et al. (2003). All measured items of the theoretical model
were subjected to an exploratory factor analysis with no rotation in SPSS. The solution
produced four factors with eigenvalues larger than 1.0, with the smallest one being 1.183.
These four factors accounted for 76 per cent of the variance, with the first factor alone
explaining 57 per cent, hence variables in the model load on more than one factor. A second
test to detect possible CMV was conducted according to Pavlou et al. (2007). Thus, a visual
inspection of the correlations of the model variables indicated the highest value to be 0.84
(Table IV), hence below the threshold of 0.90. Results of these two tests allow some
confidence that there is no systematic CMV bias in the data (Turel and Serenko, 2012).

Main data analysis were done with partial least squares (PLS) modelling method as this is
suitable for small sample size exploratory models (Bontis, 1998), including those containing
formative indicators (Thomas et al., 2005). Perceived overall value was measured as
second-order latent variable using a repeated indicators approach (Lohmoller, 1989).

5.1 Measurement model evaluation
Evaluation of the measurement model was done with SmartPLS (Ringle et al., 2005).
A first run of the software indicated the necessity to eliminate two items (pertaining to
utilitarian value and social value, respectively) out of the total of 29 of the entire


A perceived


model due to poor item-to-construct loading values. After re-running the program, all
reflective constructs for both samples displayed composite reliability and Cronbach’s α
values above 0.7 and average variance extracted (AVE) values above 0.5, as indicated
in Table I. All first-order factor loadings were above 0.7, all items were significant at a
level better than 0.05 (since t-valueW1.96), and all item errors were generally small, as
Table II shows. Thus, based on the results captured in Tables I and II, the measurement

Construct Composite reliability Cronbach’s α AVE

Behavioural intention 0.992 0.984 0.984
Perceived critical mass 0.921 0.884 0.744
Hedonic value 0.965 0.954 0.845
Image 0.919 0.868 0.791
Perceived social presence 0.939 0.918 0.754
Social value 0.966 0.947 0.904
Social norm 0.965 0.928 0.933
Utilitarian value 0.979 0.968 0.939

Table I.
Reliability measures
for first-order

Component Factor loading SE t-Statistic

BI1←behavioural intention 0.992 0.004 239.790
BI2←behavioural intention 0.992 0.005 218.917
HV1←hedonic value 0.935 0.014 65.185
HV2←hedonic value 0.934 0.018 52.779
HV3←hedonic value 0.854 0.049 17.494
HV4←hedonic value 0.937 0.016 59.081
HV5←hedonic value 0.935 0.016 58.848
I1←image 0.901 0.023 40.036
I2←Image 0.920 0.019 48.708
I3←image 0.845 0.064 13.130
PCM1←perceived critical mass 0.877 0.029 30.815
PCM2←perceived critical mass 0.926 0.022 41.786
PCM3←perceived critical mass 0.791 0.058 13.730
PCM4←perceived critical mass 0.850 0.047 18.192
PSP1←perceived social presence 0.904 0.025 35.654
PSP2←perceived social presence 0.825 0.071 11.677
PSP3←perceived social presence 0.830 0.048 17.403
PSP4←perceived social presence 0.912 0.018 50.444
PSP5←perceived social presence 0.868 0.038 23.069
SN1←social norm 0.969 0.008 118.585
SN2←social norm 0.962 0.013 74.850
SV1←social value 0.931 0.022 41.677
SV2←social value 0.963 0.012 78.133
SV4←social value 0.958 0.012 79.605
UV2←utilitarian value 0.975 0.009 109.164
UV3←utilitarian value 0.980 0.005 180.789
UV4←utilitarian value 0.952 0.014 68.311
Notes: PSP, perceived social presence; SN, social norm; I, image; PCM, perceived critical mass;
UV, utilitarian value; HV, hedonic value; SV, social value; BI, behavioral intention. 1-5, item number

Table II.
Item loading and
significance levels
for first-order



model was considered to have appropriate reliability and convergent validity (Bontis,
2004; Fornell and Larcker, 1981; Jarvenpaa et al., 2004).

The following test consisted of examining the matrix of loadings and cross-loadings
for first-order constructs produced by SmartPLS. As this matrix shows (Table III), the
measurement model has appropriate discriminant validity because items load more on
the latent variables they pertain to than on the other constructs (Gefen and Straub,
2005). This conclusion is reinforced by a visual inspection of the matrix in Table IV that
displays the square root of AVEs for all first-order constructs on the diagonal and the
construct correlations off diagonal. Since diagonal numbers are larger than all off
diagonal numbers on the respective rows and columns, the condition for appropriate
discriminant validity is met (Gefen and Straub, 2005).

To test for possible multicollinearity problems, a variance inflation factor (VIF) was
calculated for all relevant constructs by regressing each independent variable on the
remaining antecedents of an endogenous variable. Since the VIFs for the four
antecedents of social value were below 2.5 and those for the perceived overall value
components did not exceed 4.5, hence below the threshold value of 5 (Hair et al., 2009),
multicollinearity is not considered an issue for the measurement model. Confidence
in appropriate discriminant validity is further increased by all AVE values being
above 0.5 (Table I) and all inter-construct correlations being below 0.9 (Table IV)
(Pavlou et al., 2007).


critical mass

value Image

social presence




BI1 0.992 0.589 0.845 0.539 0.572 0.642 0.480 0.788
BI2 0.992 0.563 0.828 0.507 0.550 0.612 0.441 0.774
HV1 0.806 0.635 0.935 0.538 0.637 0.608 0.496 0.846
HV2 0.807 0.592 0.934 0.557 0.618 0.655 0.551 0.828
HV3 0.717 0.507 0.854 0.391 0.490 0.534 0.283 0.735
HV4 0.793 0.581 0.937 0.565 0.527 0.739 0.537 0.729
HV5 0.749 0.534 0.935 0.468 0.517 0.673 0.475 0.677
I1 0.417 0.567 0.479 0.901 0.524 0.730 0.660 0.530
I2 0.570 0.577 0.506 0.920 0.625 0.697 0.645 0.561
I3 0.417 0.541 0.490 0.845 0.505 0.549 0.518 0.466
PCM1 0.419 0.877 0.459 0.572 0.622 0.546 0.439 0.526
PCM2 0.498 0.926 0.559 0.515 0.601 0.488 0.380 0.586
PCM3 0.693 0.791 0.647 0.545 0.568 0.514 0.424 0.661
PCM4 0.381 0.850 0.471 0.538 0.503 0.444 0.467 0.456
PSP1 0.599 0.630 0.581 0.540 0.904 0.542 0.488 0.643
PSP2 0.418 0.501 0.423 0.427 0.825 0.421 0.301 0.519
PSP3 0.532 0.560 0.579 0.468 0.830 0.492 0.378 0.592
PSP4 0.454 0.610 0.534 0.629 0.912 0.617 0.534 0.554
PSP5 0.453 0.592 0.513 0.602 0.868 0.519 0.491 0.574
SN1 0.449 0.497 0.506 0.673 0.518 0.647 0.969 0.455
SN2 0.449 0.460 0.484 0.659 0.473 0.584 0.962 0.477
SV1 0.548 0.535 0.654 0.705 0.571 0.931 0.598 0.504
SV2 0.630 0.588 0.665 0.730 0.596 0.963 0.639 0.577
SV4 0.624 0.535 0.676 0.701 0.553 0.958 0.584 0.582
UV2 0.777 0.606 0.786 0.561 0.625 0.549 0.477 0.975
UV3 0.754 0.635 0.794 0.563 0.632 0.547 0.453 0.980
UV4 0.758 0.648 0.833 0.581 0.670 0.599 0.471 0.952

Table III.
Loadings and

cross-loadings for
first-order constructs


A perceived


5.2 Structural model evaluation
As the measurement model evaluation indicated appropriate reliability and construct
(i.e. convergent and discriminant) validity levels for all first-order constructs, evaluation
of the structural model came next. Results of this evaluation are depicted in Table V and
Figure 2.

Table V and Figure 2 indicate that five out of the eight hypotheses made were
confirmed. Perceived overall value is a key antecedent of the intention to use the
social networking application explaining 72.8 per cent of the variance of the latter.
All three facets of perceived value are significant components in the second-order
construct (at a level p-valueo0.001) with moderately high values of the path
coefficients: between 0.26 and 0.51. Analysis of the total effects of first-order
constructs on the behavioural intention provided by SmartPLS and captured in
Table VI confirms that hedonic value is the most important value facet in the
adoption equation: its total effect coefficient is 0.438 compared with 0.275 for
the utilitarian value and 0.226 for the social value.

All demographic characteristics collected about the sample respondents were tested
as possible control variables by assessing their path coefficients to the endogenous
variables of the model. Each variable was added to the model and SmartPLS was rerun
every time. No changes in the measurement model were noticed. Structural changes
were noticed in only two situations:

(1) having a Twitter account had a path coefficient of 0.106 (significant at the 0.01
level) to the behavioural intention and increased the R2 of this latter from 0.728
to 0.739; and


BI 0.992
PCM 0.581 0.863
HV 0.843 0.621 0.919
I 0.528 0.631 0.551 0.889
PSP 0.566 0.669 0.608 0.621 0.868
SV 0.632 0.582 0.700 0.749 0.603 0.951
SN 0.464 0.496 0.513 0.690 0.514 0.639 0.966
UV 0.788 0.650 0.831 0.587 0.663 0.584 0.482 0.969
Notes: PSP, perceived social presence; SN, social norm; I, image; PCM, perceived critical mass;
UV, utilitarian value; HV, hedonic value; SV, social value; BI, behavioural Intention

Table IV.
Square-root AVEs
and correlation
coefficients for first-
order constructs

Hypothesis Path Path coefficient SE t-Statistic p-Value

H1 Perceived value→behavioural intention 0.853 0.036 23.761 o0.001
H2-1 Utilitarian value→perceived value 0.322 0.014 23.016 o0.001
H2-2 Hedonic value→perceived value 0.513 0.017 30.757 o0.001
H2-3 Social value→perceived value 0.264 0.014 19.119 o0.001
H3-1 Image→social value 0.460 0.134 3.420 o0.001
H3-2 Perceived social presence→social value 0.157 0.118 1.335 0.184
H3-3 Perceived critical mass→social value 0.089 0.109 0.816 0.416
H3-4 Social norm→social value 0.197 0.124 1.583 0.116

Table V.
Path coefficients,
standard errors and
significance levels



(2) the frequency of checking Twitter had a path coefficient of 0.180 (significant
at the 0.05 level) to the perceived social value and increased the R2 of this latter
from 0.617 to 0.647.

To further investigate the possible influence of not having a Twitter account,
responses from this category of participants were removed from the initial sample
and the entire analysis described above was repeated for the resulting sample of
96 participants. Following that, a path-by-path comparison was done by evaluating
a t-statistic of the absolute difference between the corresponding path coefficients in
the two cases – i.e., full sample and reduced sample (Ahuja and Thatcher, 2005;
Chin, 2000):

t ¼ …