Exploring the Items for Measuring e-Procurement Usage Construct: An Exploratory Factor Analysis

The usage of Information system (IS) in any institution is vital and this human technology interaction is the heart of many world-changing endeavors. The aim of this research is to perform instrument validation through exploratory factor analysis (EFA) of using eprocurement. The questionnaire used in this study is adapted from two different studies: Venkatesh et al., (2012) and Norzaidi (2008). It consists of five sub-constructs; after the questionnaire was distributed, 115 responses were collected to do the EFA. EFA was done for each construct separately. The results show that all of the seven constructs have one component or dimension, the factor loading for every item in each construct is >0.6, Bartlett’s Test of Sphericity was <0.05 for all the constructs, which is Significant (P-value < 0.05). Kaiser-Meyer-Olkin Measure of Sampling Adequacy was higher than 0.6 for all the constructs, and this means that the sample size is adequate. Cronbach’s Alpha test was higher than 0.7 for the entire constructs’ items, which means that these items are all reliable. This study found a valid and reliable instrument for measuring the usage of e-procurement to Malaysian contractors’.


Introduction
E-procurement is the use of integrated information technology for part of or all the procurement functions, from the beginning to end, i.e. from searching, sourcing, negotiating, ordering, and receipt to post-purchase review (Croom & Jones., 2005;Trkman & McCormack, 2010). E-procurement has been recently receiving much attention from businesses, industries, and governments as it reportedly becomes a powerful tool to improve effectiveness and efficiencies as well as the service quality of its adopters.
Malaysian government procurement is perceived as a major function of government and a substantial amount of money is allocated annually for the procurement of goods and services (Thai, 2001;Maniam et al., 2009). The Malaysian government spends more than RM150 billion every year in procuring goods, works and services and this gives a sign of riskiness of public procurement being exposed to corruption (Ministry of Finance, 2011). AG Procurement automates the workflow of procurement/resource management processes, which reduces the cycle time of purchases, decreases stocking requirements, and lowers inventory management costs. Finally, e-Procurement applications enable enterprises to manage long-term relationships with suppliers. These relationships can be leveraged to create an enterprise-wide buying environment with the most favorable conditions.

Significance of the Study
Understanding factors as part of the cause of the system usage has become greater doubt in the field of Information System (IS). E-procurement is the use of integrated information technology for part of or all the procurement functions, from the beginning to end, i.e. from searching, sourcing, negotiating, ordering, and receipt to post-purchase review (Croom & Jones., 2005;Trkman & McCormack., 2010). E-Procurement is critical to construction because it involves a number of partners on each project who all have the need for inventory management in order not to delay the project or to tie space and money on excess inventory while also complying with specifications and other variables (Pheng & Meng, 1997). Beyond the obvious transaction cost savings and access to suppliers, e-Procurement can offer product standardization, quality assurance, inventory management and the opportunity to manage material flows down the value chain. The e-Procurement also automates the workflow of procurement/resource management processes, which reduces the cycle time of purchases, decreases stocking requirements, and lowers inventory management costs. Thus, there is a need to measure the usage of e-procurement in Malaysian construction, and its impact to contractors' performance, which represents the aim of this study to find a validated instrument measuring the usage and performance of the system.

Materials and Method
The data collection employed in this study is from a self-administered survey questionnaire. The questionnaire is adapted from two different studies: Vankatesh et al., (2012) and Norzaidi (2008). The questionnaire was adapted and customized to suit the field of this study and was distributed to the respondents who were responsible and in charge of the company's eprocurement that comprise of company officials -general manager, assistant general manager, project manager, site manager, engineer, and executive. The survey was composed from 5 constructs (after the demographical data concerning the respondents): The first four constructs were related to UTAUT model; first construct: Performance Expectancy (5 items using the scale of 7). Second construct: Effort Expectancy (5 items using the scale of 7). Third construct: Social Influence (5 items using the scale of 7). Fourth construct: Facilitating Condition (6 items using the scale of 7). Fifth construct: E-procurement usage (4 items using the scale of 7). Taherdoost (2019) in his study suggests the use of seven-point rating scale and if there is a need to have respondents to be directed on one side, then he claims that six-point scale is the most suitable. Accordingly, this study applied the interval scale of 7, in which the respondents selected a statement among several statements from 1-7 which is considered to reflect the perceived quality of the subject. Number 1 stands for strongly disagree, while, number 7 stands for strongly agree. According to Awang et al. (2010;2012;2014;2015) and Awang et al. (2018), the researcher should apply a Likert Scale without a label because this measure would give an interval type of data that is continuous and fit the data presumption for parametric analysis. As per Awang (2010;2012;2014;2015) and Hoque et al. (2017;, if the analyst adjusted instruments from past studies and altered accordingly, at that point the scientist needs to direct both pre-test and pilot-test for these "changed items" so as to approve them before it tends to be utilized in the final study. Content validity, face validity, and criterion validity were done as a pre-test for this questionnaire, content validity was done through content experts, and face validity was done through English language experts, criterion validity was done through a statistical expert, after these validation tests were completed, the researcher distributed the instrument to 3 respondents, in order to gather their comments, and check the consistency in their responses. After all the required changes according to pre-test results had been done, the researcher distributed the questionnaire to gather a minimum of 100 responses to be able to run the exploratory factor analysis (EFA) as according to many researchers for example: Awang (2010Awang ( , 2012Awang ( , 2014Awang ( , 2015, Hoque et al. (2017Hoque et al. ( , 2018, Noor et al. (2015), Awang et al. (2018) and Yahaya et al. (2018) claim that EFA should be done for each construct to explore changes in dimensionality of items from past studies due to changes in the characteristics of population from the past.

Results and Discussion
The EFA for the First Construct: Performance Expectancy Performance Expectancy construct was measured by using 5 items namely pe1 till pe5 (Table  1). Every statement of items was measured using Interval Scale ranging between 1 and 7, where 1 stands for strongly disagree and 7 stands for strongly agree. The mean score and standard deviation derived for every single item which measured the constructs are shown in Table 1. pe4 Using e-procurement increases my productivity 5.58 0.936 pe5 Overall, I would find e-procurement to be advantageous 5.45 1.011 EFA using Principal Component Analysis as an extraction method performed for these 5 items to measure performance expectancy of using e-procurement. The results in Table 2 shows Bartlett's Test of Sphericity which is Significant since it's <0.05. Kaiser-Meyer-Olkin Measure of Sampling Adequacy higher than 0.6 which is for the first construct 0.855, and this means that the sample size is adequate (Awang, 2010;2012;2014;2015;Hoque et al., 2017;and Noor et al., 2015). Accordingly, the current data are acceptable. 0.000 The scree plot in Figure 1 shows that only one component emerged from the EFA, accordingly all items in this construct will belong to one component. The results in Table 3 the components or dimension for each item is shown in this table, as it's clear all items are belonging to one component. The factor loading for every item should be >0.6 in order to be retained (Awang, 2010;2012;Awang et al. (2018) and Yahaya et al., 2018). Thus all items will be retained. The results in Table 4 show there are one dimension or component emerged from the EFA procedure based on the computed Eigenvalue >1.0. The total variance explained for measuring this construct is 78.183%. The total variance explained is acceptable since it exceeds the minimum 60% (Awang, 2010(Awang, , 20122014;2015;Noor et al., 2015;Hoque et al., 2017;and Yahaya et al., 2018).

The Internal Reliability for the Instrument Measuring Performance Expectancy
The last test that should be done is the internal reliability of each construct. As Table 5 shows that Cronbach's Alpha test is 0.929, higher than 0.7, which means that these items are reliable. The EFA for the Second Construct: Effort Expectancy This construct was measured using 5 items listed in Table 6 as ee1 to ee5, and each item was measured using Likert-scale of 7, where 1 stands for strongly disagree and 7 stands for strongly agree, the mean response, standard deviation and item statement, for each item, are listed in Table 6. EFA using principal component analysis as an extraction method performed for these 5 items to measure effort expectancy of using e-procurement. The results in Table 7 shows Bartlett's Test of Sphericity which is Significant since it's <0.05. Kaiser-Meyer-Olkin measure of sampling Adequacy higher than 0.6 which is for the 2nd construct 0.847, and this means that the sample size is adequate (Awang, 2010;2012;2014;2015;Hoque et al., 2017Hoque et al., , 2018Noor et al., 2015). Accordingly, the current data are acceptable. 0.000 The scree plot in Figure 2 shows that only one component is emerged from the EFA, accordingly all items in this construct will belong to one component. The results of the components or dimension for each item is shown in Table 8, as it's clear all items are belonging to one component. The factor loading for every item should be >0.6 in order to be retained (Awang, 2010;2012;2014;2015;Awang et al., 2018 andYahaya et al., 2018). Thus all items will be retained.  Table 9 show there are one dimension or component emerged from the EFA procedure based on the computed Eigenvalue >1.0. The total variance explained for measuring this construct is 73.217%. The total variance explained is acceptable since it exceeds the minimum 60% (Awang, 2010;2012;2014;2015;Noor et al., 2015;Hoque et al., 2017Hoque et al., , 2018and Yahaya et al., 2018).

The Internal Reliability for the Instrument Measuring Effort Expectancy
The last test that should be done is the internal reliability of each construct. As Table 10 shows that Cronbach's Alpha test is 0.898, higher than 0.7, which means that these items are reliable. The EFA for the Third Construct: Social Influence This construct was measured using 5 items listed in Table 11 as si1 to si5, and each item was measured using Likert-scale of 7, where 1 stands for strongly disagree and 7 stands for strongly agree, the mean response, standard deviation and item statement, for each item, are listed in Table 11. EFA using principal component analysis as an extraction method performed for these 5 items to measure social influence of using e-procurement. The results in Table 12 shows Bartlett's Test of Sphericity which is significant since it's <0.05. Kaiser-Meyer-Olkin measure of sampling Adequacy higher than 0.6 which is for the third construct 0.787, and this means that the sample size is adequate (Awang, 2010;2012;2014;2015;Hoque et al., 2017Hoque et al., , 2018Noor et al., 2015). Accordingly, the current data are acceptable. 0.000 The scree plot in Figure 3 shows that only one component is emerged from the EFA, accordingly all items in this construct will belong to one component. Figure 3: The Scree Plot for the third construct The results in Table 13, the components or dimension for each item is shown in this table, as it's clear all items are belonging to one component. The factor loading for every item should be >0.6 in order to be retained (Awang, 2010;2012;2014;2015;Awang et al., 2018 andYahaya et al., 2018). Thus all items will be retained. The results in Table 14 show there are one dimension or component emerged from the EFA procedure based on the computed Eigenvalue >1.0. The total variance explained for measuring this construct is 62.123%. The total variance explained is acceptable since it exceeds the minimum 60% (Awang, 2010;2012;2014;2015;Noor et al., 2015;Hoque et al., 2017Hoque et al., , 2018and Yahaya et al., 2018).

The Internal Reliability for the Instrument measuring Social Influence
The last test that should be done is the internal reliability of each construct. As Table 15 shows that Cronbach's Alpha test is 0.809, higher than 0.7, which means that these items are reliable. The EFA for the Fourth Construct: Facilitating Condition This construct was measured using 5 items listed in Table 16 as fc1 to fc5, and each item was measured using Likert-scale of 7, where 1 stands for strongly disagree and 7 stands for strongly agree, the mean response, standard deviation and item statement, for each item, are listed in Table 16. EFA using principal component analysis as an extraction method performed for these 5 items to measure facilitating condition of using e-procurement. The results in Table 17 shows Bartlett's Test of Sphericity which is Significant since it's <0.05. Kaiser-Meyer-Olkin measure of sampling Adequacy higher than 0.6 which is for the fourth construct 0.810, and this means that the sample size is adequate (Awang, 2010;2012;2014;2015;Hoque et al., 2017Hoque et al., , 2018Noor et al., 2015). Accordingly, the current data are acceptable. 0.000 The scree plot in Figure 4 shows that only one component is emerged from the EFA, accordingly all items in this construct will belong to one component. The results in Table 18 the components or dimension for each item is shown in this table, as it's clear all items are belonging to one component. The factor loading for every item should be >0.6 in order to be retained (Awang, 2010;2012;2014;2015;Awang et al., 2018 andYahaya et al., 2018). Thus, all items will be retained. The results in Table 19 show there are one dimension or component emerged from the EFA procedure based on the computed Eigenvalue >1.0. The total variance explained for measuring this construct is 67.955%. The total variance explained is acceptable since it exceeds the minimum 60% (Awang, 2010;2012;2014;2015;Noor et al., 2015;Hoque et al., 2017Hoque et al., , 2018and Yahaya et al., 2018).

The internal reliability for the instrument measuring Facilitating Condition
The last test that should be done is the internal reliability of each construct. As Table 20 shows that Cronbach's Alpha test is 0.880, higher than 0.7, which means that these items are reliable. The EFA for the Fifth Construct: e-procurement Usage This construct was measured using 4 items listed in Table 21 as pu1 to pu5, and each item was measured using Likert-scale of 7, where 1 stands for strongly disagree and 7 stands for strongly agree, the mean response, standard deviation and item statement, for each item, are listed in Table 21. EFA using principal component analysis as an extraction method performed for these 4 items to measure e-procurement usage of using e-procurement. The results in Table 22 shows Bartlett's Test of Sphericity which is Significant since it's <0.05. Kaiser-Meyer-Olkin measure of sampling Adequacy higher than 0.6 which is for the fifth construct 0.793, and this means that the sample size is adequate (Awang, 2010;2012;2014;2015;Hoque et al., 2017Hoque et al., , 2018Noor et al., 2015). Accordingly, the current data are acceptable. 0.000 The scree plot in Figure 5 shows that only one component is emerged from the EFA, accordingly all items in this construct will belong to one component. The results in Table 23 the components or dimension for each item is shown in this table, as it's clear all items are belonging to one component. The factor loading for every item should be >0.6 in order to be retained (Awang, 2010;2012;2014;2015;Awang et al., 2018 andYahaya et al., 2018). Thus all items will be retained. The results in Table 24 show there are one dimension or component emerged from the EFA procedure based on the computed Eigenvalue >1.0. The total variance explained for measuring this construct is 70.973%. The total variance explained is acceptable since it exceeds the minimum 60% (Awang, 2010;2012;2014;2015;Noor et al., 2015;Hoque et al., 2017Hoque et al., , 2018and Yahaya et al., 2018;Khalid, 2020).

The internal reliability for the instrument measuring e-procurement Usage
The last test that should be done is the internal reliability of each construct. As Table 20 shows that Cronbach's Alpha test is 0.880, higher than 0.7, which means that these items are reliable.

Conclusion
The current research adds a remarkable contribution to the measurement of the 5 constructs, mainly in the e-procurement context. The results show that every of the five constructs have one component or dimension, the factor loading for every item exceeds the minimum threshold of 0.6, with high Cronbach's Alpha value, meet Bartlet Test achievements (significant) for all the constructs, Kaiser-Meyer-Olkin Measure of Sampling Adequacy was higher than 0.6 for all the constructs, and factor loading exceeds the minimum threshold of 0.6. Hence, this study found a valid and reliable instrument for measuring the usage of eprocurement to Malaysian contractors'. Therefore, this instrument can be used to measure the usage of e-procurement in the targeted organizations in this study.
However, it is recommended during field study that some precautions need to be done to ensure that the data collection are free from response bias and common method variance and therefore, would generate significant findings.