Individual Job Performance

According to TTF theory, performance benefits improve a range of performance outcomes15. The specific outcomes that have been examined in prior TTF research include system use/intention to use, job performance, satisfaction with technology, opinions regarding a technology, appropriation changes made, decision efficiency, decision quality, decision strategy employed, joint profit attained, number of ideas generated, quality of solution, task completion time, task accuracy, ability to perform tasks, perceived ease of use, perceived usefulness, perceived playfulness, perceived risk… 15.

High performance implies that there is a mix between improved efficiency, improved effective and/ or improved quality10. It was often confused with productivity16. However, it is actually measured by more global variables such as the quality of outputs, job knowledge, leadership, or judgment16. In the traditional office, job performance is largely established by scanning for employees’ presence and through direct and indirect observations16.

There exists a wide range of employee outcomes in ERP implementation, such as systems satisfaction, job stress, job satisfaction, and individual job performance8. Job performance is a way to measure employee outcomes. Job performance is a good way of performing an employee's work8.

This study defined individual job performance in ERP environment was that the employee feels that with the help of ERP systems he/she can improve a range of performance outcomes.

Task — Technology Fit

Task — technology fit concept derives from the Task-Technology Fit theory (TTF) 11, 10, 9. According to TTF, the task — technology fit represents the degree of matching or alignment between the capabilities of an information system and the demands of the tasks that must be performed15. Base on TTF, this study defined t ask — technology fit is the degree of relevance between the ability of ERP system and the tasks which an employee must perform.

Based on three theories including TTF, TAM and TTF models combined, and the DeLone and McLean IS Success Model (2013)13, this study develops the hypothesis H1. Firstly, according TTF, task — technology fit impact on performance benefit 10 while individual job performance is used as an indicator of performance benefit15. Secondly, TAM and TTF models combined show that task — technology fit have significant effect on actual tool use12. At the same time, the user behavior has an impact on the job performance17. Finally, Peter et al.18 suggested that task compatibility impact on the IS success. Task compatibility is the fit or consistency between the task and the IS that supports that task 13. In this study, task — technology fit was defined similarly to task compatibility. Petter et al. 18 measured the IS success based on the update DeLone and McLean2 IS success model, including information quality, system quality, service quality, intention to use/ use, user satisfaction, and net benefits. In this study, individual job performance was considered an indicator of net benefits. Based on the above arguments, this study developed hypothesis H1.

This hypothesis is also supported by several studies, such as Norzaidi et al. (2009), Bhattacherjee (2001), Teo and Bing (2008), Kositanurit et al. (2006), D'Ambra and Wilson (2004a), D'Ambra and Wilson (2004b), Wongpinunwatana et al. (2000), Goodhue et al. (1997), Goodhue and Thompson (1995), Goodhue (1995), Henseler (2015)25, 24, 10, 9, 23, 22, 21, 5, 20, 14, 19. In particular, Staples and Seddon18 show that task — technology fit had an impact on individual job performance in both kinds of IT use, voluntary and mandatory. In this study, if an enterprise is using an ERP system, employees are required to use it. So Staples and Seddon18 strongly support the hypothesis H1.

H1: The task – technology f it has a positive effect on the job performance of employee in ERP environment.

Perceived Usefulness

In TAM, perceived usefulness is “the degree to which a person believes that using particular system would enhance his or her job performance” 26 (p.320). Perceived usefulness is a dimension of performance expectancy in UTAUT27. In this study, perceived usefulness of ERP systems was understood as the level at which the user believes ERP systems delivers good results for their work.

According TTF, task – technology fit impact on performance benefit10 and perceived usefulness can be used as an indicator of performance benefit 28, 15. Besides, Dishaw et al. 12 suggested TAM and TTF models combined that demonstrated task — technology fit to have a significant impact on perceived usefulness. Based on TTF, TAM and TTF models combined and some studies such as Norzaidi et al. (2009), Tjahjono (2009), Chang (2008), Wu et al. (2007), Klopping and McKinney (2004), Dishaw and S trong (1999), Goodhue (1995)10, 9, 33, 32, 31, 30, 29, 19 hypothesis H2 is stated as follows:

H2: The task — technology fit has a positive effect o n the perceived usefulness in ERP environment.

Based on three theories, namely TAM and TTF models combined, ECM (Expectation–Confirmation Model) 14 and DeLone and McLean IS Success Model (2013)2, this study developed the hypothesis H3. The first, according TAM and TTF models combined, perceived usefulness has effect on attitude toward use, then attitude toward use impact on intention to use, after that intention to use effect on actual tool use12. Next, ECM shows that perceived usefulness impact on IS continuance intention 14. Last, user expectations was proven to have an impact on the IS success2. User expectations is the degree to which the user’s perceptions about the IS are consistent with the actual IS 10. In this study, perceived usefulness was close to the user expectations used in Petter et al. (2013)13. In general, TAM and TTF models combined, ECM and DeLone and McLean IS Success Model (2013)2 have shown that the perceived usefulness of the information system had a positive impact on user behavior. At the same time, the user behavior had an impact on the job performance17.

Furthermore, some studies include Rajan and Baral (2015), Furneaux (2012), Sternad and Bobek (2013), Soto-Acosta et al. (2013), Elkhani et al. (2014), Zhang et al. (2013), Keong et al. (2012), Norzaidi et al. (2009), Youngberg et al. (2009), Calisir et al. (2009), Lee et al. (2010), Chang (2008), Wu et al. (2007), Ramayah and May- Chiun Lo (2007), Seymour et al. (2007) and Amoako- Gyampah and Salam (2004) have also shown that the perceived usefulness of IS/ ERP had a positive impact on user behavior44, 43, 42, 31, 30, 41, 40, 39, 19, 38, 37, 36, 35, 34, 15, 17. In addition, Goodhue demonstrated that the perceived usefulness of IS has a positive impact on the individual job performance in IT context9. Based on the above arguments, this study hypothesized the following:

H3: The perceived usefulness of the ERP has a positive effect on the job performance of employee in ERP environment.

Figure 1 represents the proposed research model that was used for this research.

Measures

All research constructs included in this study had multi-item scales derived from the relevant literature. Each item in the survey employed a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree), and a not applicable (NA) option was available for the respondents to choose. We now elaborate our measures for the constructs. This study accepted scale of the individual job performance in ERP environment (PER) from Goodhue and Thompson10. PER is a first – order construct and reflective measures with 2 items. We measured task – technology fit (TTF) using a11- item scale adapted from Kositanurit et al.5 that captures the eight dimensions including currency (CURR), right data (RDAT), right level of detail (RDET), meaning (MEAN), ease of use (EOU), training (TRAI), authorization (AUT) and system reliability (REL). TTF is a high – order construct and reflective – reflective measures. The perceived usefulness (PU) was measured with 6 items adapted from Calisir and Calisir45 (from the TAM model). PU is a first – order construct and reflective measures.

Data collection

This study was conducted using quantitative means as it aimed to validate the proposed relationships between factors affecting individual job performance in ERP context. The research instrument was used a questionnaire distributed to the end-users (employees). The data were collected from June 2017 to August 2017. The questionnaires were sent by email or postal mail to an initial sample of 5 00 employees who are using ERP system. We collected 265 responses (response rate was 53 %). Baruch46 points out that for surveys addressed to individual, the average response rate is 52.7%. Thus, the response rate of our study seems to be above average. Of the 265 employees, 225 employees from 49 companies gave usable responses at all points of measurement.

Table 1 presents the sample characteristics age, gender, education and average computer experience. The sample consisted of 161 (71.6 %) female and 64 (28.4 %) male. The Table 1 shows that 75.5% of the sampled individuals were fewer than 35. In addition, 72% of the sampled individuals had bachelor degree. Average experience using an ERP system was 2.56 years.

The Issue of Common Method Bias

Because there was only one respondent for each individual, common method bias (CMB) was a potential problem. In this study, we took a number of steps suggested by Podsakoff et al. 47 to reduce the possibility of common method bias. Firstly, we used multiple items for each construct and ensured the neutral wording of the items. Secondly, we assured respondents of the anonymity of their responses and emphasized that there were no right or wrong answers; each of these actions enabled them to answer questions as honestly as possible. Thirdly, we separated the measurement of predictors and criterion variables in the questionnaire to diminish the respondent’s ability and motivation to use his/her prior responses to answer subsequent questions. Finally, we also used the Harman’s single-factor test and the marker variable approach to control for common method variance (CMV). Results are discussed in data analysis and results section.

Measurement model

We estimated the internal consistency reliability, convergent validity, and discriminant validity of each measurement scale to assess the measurement model. We used two criterions for internal consistency were composite reliability (CR) and Cronbach’s alpha. All the reflective constructs in our model show in Table 2 have a Cronbach’s alpha over the cut off of 0.70, as suggested by Hair et al.48. Similarly, a composite reliability (CR) of all the constructs is also higher than 0.7, as suggested by Fornell and Larcker49, implying high internal consistency.

Convergent validity is verified through the t-statistic for each factor loading. In PLS_SEM, we can use an indicator's outer loading. An outer loading should be above 0.7 and the t-statistic for each outer loading significant48. Results of measurement models show that the items including AUT1, AUT2, REL1, REL2 and TRAI have outer loading was above 0.7 but the t-statistic for each outer loading was not significant. As such, the AUT1, AUT2, REL1, REL2 and TRAI were excluded from the TTF scales. Table 2 shows that results of final measurement models. A ll factor loadings are greater than the typical cut off value of 0.748 and significant at the p <0.001 level. In this study, we also used the average variance extracted (AVE) to assess convergent validity. An AVE value of 0.50 or higher indicates that, on average, the construct explains more than half of the variance of its indicators.

To establish discriminant validity, we used the HTMT criterion, Fornell – Larcker criterion and cross loadings. The results of discriminant validity are show in Table 3. Cross-factor loadings are reported in Appendix A.

The square root of the AVE of each construct should be higher than its highest correlation with any other construct 49. Table 3 shows that the square root of AVE exceeds the correlation between other constructs. In addition, all HTMT of constructs are significantly smaller than 1 25. These results imply satisfactory discriminant validity.

Structural model

The structural model was examined to test the hypotheses. The R, which is generated for each regression equation, indicates the explanatory power or variance explained of the latent endogenous variable. Figure 2 shows the structural model result.

The PLS path analysis results show that task – technology fit was significantly related to individual job performance (β = 0.434, p < 0.001) and perceived usefulness (β = 0.607, p < 0.001) supporting hypotheses H1 and H2. Perceived usefulness was significantly related to individual job performance (β = 0.504, p < 0.001) supporting hypotheses H3.

The external variables including task – technology fit and perceived usefulness could explain 70.7 percent variance in individual job performance (R = 0.707). Task – technology fit explained 36.8 percent of variance of perceived usefulness (R = 0.368).

Next, we assessed the predictive relevance of the path model by Q values. All Q values are considerably above zero (Q of individual job performance is 0.631 and Q of perceived usefulness is 0.261), thus providing support for the model's predictive relevance regarding the endogenous latent variables.

The final assessments address the f and q effect sizes. Table 4 summarizes the results of the f and q effect sizes with respect to all the relationships in the model. Target constructs appear in the first row, whereas the predecessor constructs are in the first column. Table 4 shows TTF has a large effect size of 0.405 (0.290) on PER and of 0.583 (0.404) on PU. Similarly, PU has a large effect size of 0.546 (0.345) on PER.

Additionally, the variance inflation factor (VIF) was assessed to check multicollinearity. The collinearity diagnostics given inTable 5 shows that VIF for the independent variables higher than 0.20 (lower than 5) which further suggests that multicollinearity does not exist among the independent variables.

The Issue of Common Method Bias

We used the Harman’s single-factor test and the marker variable approach to control for CMV in PLS analysis. Result of Harman’s single-factor test by EFA shows that one factor only account for 40.537% of the total variance. In this case, CMV is not a serious problem50.

The marker variable approach was conducted by using marker variable. The first stage, we involved survey questionnaire that had a question was “Do you really like black coffee?” – this question was a marker variable. The next stage, we used PLS to test path coefficient of the relationship from marker variable to other variable in proposal model including perceived usefulness, task – technology fit and individual job performance. Analysis results showed that all path coefficients of the relationships from marker variable to perceived usefulness, task – technology fit and individual job performance were less than 0.3 (-0.102, 0.094 and -0.060). This finding suggests that CMV was not a serious problem in this study.

Besides, we also based on VIF to test CMB. Table 5 show that all VIFs resulting from a full collinearity test were lower than 3.3, the model can be considered free of CMB.