| < Day Day Up > |
|
ValuNet, a non-Internet-based EDI service that facilitates electronic submission for trade declarations in Hong Kong, was chosen as the EDI service under study. The service was the first large-scale EDI implementation launched by Tradelink, a major EDI service provider, with the Hong Kong Government as the largest shareholder, in Hong Kong in April 1997.
Multiple-item indicators were used for measuring the constructs in the research model in order to ensure proper operationalization and robust psychometric properties for the constructs. All factors except the dependent variable were measured by multi-item indicators. Most of the items were measured using seven-point Likert-type scales. The items used were derived or adapted, wherever possible, from previous studies. Suitable modifications or refinements of items were made in order to make them specifically relevant to EDI.
Various perceived benefits of EDI were identified and validated from previous EDI studies (Banerjee & Golhar, 1994; Cox & Ghoneim, 1996; Drury & Farhoomand, 1996; Fearon & Philip, 1998; Premkumar & Ramamurthy, 1995). In this study, perceived benefit was classified into two categories: direct and indirect. Five most commonly recognized benefits of each category were selected according to their definitions. Respondents were required to evaluate the level of agreement with these benefits that can be achieved after adopting EDI.
The measure of perceived cost was adapted from Premkumar et al. (1994). Three items were used in measuring this construct. It assessed the perceived cost in terms of setup, training and running as barriers for adopting EDI. These cost components were also mentioned in previous studies (Bouchard, 1993; Cox & Ghoneim, 1996; Drury & Farhoomand, 1996; Premkumar et al., 1994; Tuunainen, 1998).
IT knowledge was operationalized, adapted from Fichman and Karmerer (1997) and McGowan and Maday (1998). Many studies measured IT knowledge using experience of IT staff. Instead of using IT staff to evaluate IT knowledge, it is more appropriate to use end-users because of the lack of IT staff and IT departments in SME. Thus, IT knowledge was operationalized by the extent of the employees' experience with EDI-related technology.
Top management attitude towards EDI was developed based on the guideline of Dacin and Brown (1997). EDI services providers, EDI technology and EDI services themselves may influence the attitude of adopting EDI. Respondents were asked to assess top management attitudes toward (1) EDI, (2) Tradelink and (3) ValuNet.
Based on the guidelines of Neo et al. (1994), Yap et al. (1994) and King et al. (1994), the measurement for government incentives and enforcement was developed especially for this study. The Government may assert influence from two different aspects: promotion and imposition. The influence was assessed by using three items of which respondents were asked to evaluate the degree of these influences on the adoption decision.
The measure of trading partners' influence was adapted from Iacovou et al. (1995), Neo et al. (1994), O'Callaghan et al. (1992) and Premkumar et al. (1997). Respondents were asked to evaluate the influence of trading partners from recommendation and imposition aspects.
The dependent variable, EDI adoption, was measured by using a single dichotomous response to determine whether an organization adopted EDI or not. Using a single item for the dependent variable is common in IT adoption studies (e.g., Premkumar et al., 1997; Chau & Tam, 1997). An organization was classified as an early adopter if it had subscribed to ValuNet and had lodged at least one trade declaration using ValuNet in the past three months at the time of the study. It was a late adopter otherwise (since EDI implementations are mandatory for these SME).
Three thousand SME, as defined as companies with fewer than 100 employees, were randomly selected from a database that contains names of organizations that have submitted trade declarations to the Hong Kong Government. As the unit of analysis in this study was the organization, subjects for this study were required to be senior informed respondents within the organization. Therefore, a key informant method was adopted with either the owner or the top manager of the organization being asked to complete the survey.
Names of the randomly selected SME were compared with the database of subscribers to Tradelink's ValuNet services in order to identify early and late adopters at the time of the study based on whether the SME had adopted the system at that time. Two sets of questionnaires were designed for early and late adopters, respectively, since some wordings and tenses could create misunderstandings for the two groups. The appropriate sets of questionnaire (i.e., the sets for early vs. late adopters) were then sent to these SME. Therefore, although two sets of questionnaires were used, a one-to-one correspondence between the two sets of questionnaires was kept as much as possible for purposes of comparison and data analysis [1]. Table 1 summarizes the measures for early adopters. A cover letter, which explained the purpose of this study, and a prepaid reply envelope were sent together with the questionnaires. The respondents were guaranteed confidentiality of their responses. Also, follow-up phone calls were made one week after the questionnaires were made to check if the questionnaires had been returned and if not, to encourage responses.
Perceived Direct Benefits At the time your organization decided to adopt ValuNet, to what extent did you agree that EDI in general, and ValuNet in particular, could help achieve each of the following benefits: (Strongly disagree - Strongly agree)
|
Perceived Indirect Benefits At the time your organization decided to adopt ValuNet, to what extent did you agree that EDI in general, and ValuNet in particular, could help achieve each of the following benefits: (Strongly disagree - Strongly agree)
|
Perceived Cost At the time your organization decided to adopt ValuNet, to what extent did you agree that each of the following was an obstacle to the adoption? (Strongly disagree - Strongly agree)
|
IT Knowledge At the time your organization decided to adopt ValuNet, how did you evaluate your organization in each of the following areas? (Very poor - Very good)
|
Top Management Attitude At the time your organization decided to adopt ValuNet, how did you evaluate your organization in each of the following areas? (Very poor- Very good)
|
Government Incentives and Enforcement At the time your organization decided to adopt ValuNet, how much influence did each of the following activities have on your adoption decision ?(No Influence At all - Strong Influence)
|
Trading Partners' Influence At the time your organization decided to adopt ValuNet, how much influence did each of the following activities have on your adoption decision? (No Influence At all - Strong Influence)
|
Seven hundred and seventy-one questionnaires were returned, giving an initial response rate of 25.7% (771/3000). Ninety-three responses containing missing data were excluded. Also, 34 respondents were identified as having more than 100 employees, so they were also excluded. The final sample contained 644 usable responses. The response rate was 21.5% (644/3000).
Among the 644 responses, the numbers of responses from early and late adopters were 245 and 399 respectively. A large proportion (82.9%) of organizations did not have formal IT departments and around seventy percent of the organizations had three or fewer staff members, regardless of department, who were responsible for IT functions. Nearly 89% of the organizations had personal computer(s), while only 19% of the organization had local area networks. Around 25% of the organizations had performed a formal justification analysis for ValuNet adoption. Also, a large number of the organizations (85%) did not have a formal or separate IT budget.
Although the response rate in this study (21.5%) is not very high, it is comparable to many empirical studies in the SME and EDI research areas. In an attempt to assess response bias, a late-response bias test was conducted by comparing responses from those who responded immediately with those who responded after follow-up steps were taken, as suggested by Armstrong and Overton (1977) and adopted in prior IS studies (e.g., Thong et al., 1996; Hu et al., 1999). Potential bias was assessed by applying appropriate statistical tests to check the difference between the two groups of respondents on the following two groups of measures: demographic data (number of employees and annual turnover) and responses to the questionnaire items on the constructs in the research model. No significant difference, using the chi-square test, was found in terms of number of employees (chi-square = 3.124, d.f. = 4, p = 0.537) and annual turnover (chisquare = 5.659, d.f. = 9, p = 0.774). As for responses to questionnaire items for the seven constructs, as shown in Table 2, no significant difference was found. This suggests that non-response biases, if any, should not be serious.
Factor | Early Respondents (N = 301) | Late Respondents (N = 343) | Significance |
---|---|---|---|
Perceived Direct Benefits (PDB) | 4.990 | 4.815 | t = 1.636 p = 0.117 |
Perceived Indirect Benefits (PIB) | 4.100 | 4.092 | t = 0.055 p = 0.955 |
Perceived Cost (PCOST) | 4.211 | 4.241 | t = -0.226 p = 0.885 |
IT Knowledge (ITKN) | 3.345 | 3.302 | t = 0.276 p = 0.782 |
Top Management Attitude (TMA) | 3.650 | 3.639 | t = 0.056 p = 0.955 |
Government Incentives and Enforcement (GINF) | 4.701 | 4.887 | t = -1.638 p = 0.115 |
Trading Partners' Influence (TPINF) | 3.180 | 3.144 | t = 0.255 p = 0.831 |
Cronbach's alpha is the most commonly used measure to assess reliability. The reliability values for the various constructs are shown in Table 3. The Cronbach's alphas for the constructs ranged from 0.721 to 0.971, indicating an adequate level of reliability of all constructs.
Factor | Items | Mean | Std dev | Cronbach's alpha |
---|---|---|---|---|
Perceived Direct Benefits (PDB) | 5 | 4.897 | 1.215 | 0.905 |
Perceived Indirect Benefits (PIB) | 5 | 4.096 | 1.329 | 0.936 |
Perceived Cost (PCOST) | 3 | 4.227 | 1.535 | 0.869 |
IT Knowledge (ITKN) | 3 | 3.322 | 1.346 | 0.811 |
Top Management Attitude (TMA) | 3 | 3.644 | 1.310 | 0.897 |
Government Incentives and Enforcement (GINF) | 3 | 4.800 | 1.485 | 0.721 |
Trading Partners' Influence (TPINF) | 4 | 3.161 | 1.662 | 0.971 |
Validity was examined by assessing both convergent and discriminant validity via factor analysis. The rotated matrix of factor analysis with Varimax rotation is shown in Table 4. Testing for convergent validity requires checking whether or not all items are correctly loaded on the appropriate constructs. The results show that all items correctly converge on the appropriate single construct (due to the high loadings), which means that the items have convergent validity. Testing discriminant validity requires checking the cross loading of items on multiple factors. All items loaded highly on their associated construct but not others (with factor loadings lower than 0.5), thus exhibiting adequate discriminant validity.
Factor 1 | Factor 2 | Factor 3 | Factor 4 | Factor 5 | Factor 6 | Factor 7 | |
---|---|---|---|---|---|---|---|
PDB1 | 0.799 | ||||||
PDB2 | 0.751 | ||||||
PDB3 | 0.834 | ||||||
PDB4 | 0.823 | ||||||
PDB5 | 0.770 | ||||||
PIB1 | 0.770 | ||||||
PIB2 | 0.847 | ||||||
PIB3 | 0.830 | ||||||
PIB4 | 0.864 | ||||||
PIB5 | 0.875 | ||||||
PCOST1 | 0.883 | ||||||
PCOST2 | 0.902 | ||||||
PCOST3 | 0.834 | ||||||
ITKN1 | 0.687 | ||||||
ITKNP2 | 0.866 | ||||||
ITKNP3 | 0.859 | ||||||
TMA1 | 0.869 | ||||||
TMA2 | 0.883 | ||||||
TMA3 | 0.794 | ||||||
GINF1 | 0.701 | ||||||
GINF2 | 0.848 | ||||||
GINF3 | 0.812 | ||||||
TPINF1 | 0.931 | ||||||
TPINF2 | 0.949 | ||||||
TPINF3 | 0.943 | ||||||
TPINF4 | 0.942 | ||||||
Eigen value | 7.148 | 4.265 | 2.684 | 2.084 | 1.920 | 1.322 | 1.063 |
Percentage of Variance | 27.492 | 16.402 | 10.322 | 8.016 | 7.385 | 5.086 | 4.090 |
The hypotheses were tested using logistic regression with the adoption state of the organization (early or late adopter) as the dependent variable. The results are shown in Table 5. Of the seven independent variables, five were significant. Perceived direct benefits, perceived cost, IT knowledge, government incentives and enforcement, and trading partners' influence are significant in determining adoption (early or late). Organizations that perceive high direct benefits and low costs from adopting EDI, have more IT knowledge and are more sensitive to the environment would tend to adopt EDI early.
Factor | Coefficient | Wald Statistic | Significance |
---|---|---|---|
Perceived Direct Benefits (PDB) | 0.218 | 4.213 | 0.040 [*] |
Perceived Indirect Benefits (PIB) | 0.090 | 0.806 | 0.369 |
Perceived Cost (PCOST) | -0.351 | 21.010 | 0.000 [**] |
IT Knowledge (ITKN) | 0.626 | 39.816 | 0.000[**] |
Top Management Attitude (TMA) | 0.138 | 1.843 | 0.175 |
Government Incentives and Enforcement (GINF) | 0.192 | 6.878 | 0.009[**] |
Trading Partners' Influence (TPINF) | -0.804 | 97.249 | 0.000[**] |
-2 Log Likelihood: Chi-square=583.646; d.f. = 636; p = 0.932 | |||
Model Chi-square: Chi-square=271.942; d.f. = 7; p=0.000 | |||
[*]p<.05
[**]p<0.01 |
Examining the overall validity of the research model, the -2 Log Likelihood was found to be insignificant (chi-square value = 583.65, p = 0.932), suggesting the significant validity of the model in discriminating between early adopters and late adopters. The classification accuracy of the model, 78.11%, which is much better than the chance accuracy at 52.86%, further confirms the usefulness of the model.
[1]The two sets of questionnaires, however, are not completely equivalent since early adopters were asked to respond based on the perceptions as of the time a decision was made to adopt the system whereas the perceptions of late adopters were those at the time of the study. This data collection limitation is further elaborated in the Limitations section.
| < Day Day Up > |
|