The analysis uses a subset of the original questionnaire items to test our hypoth- eses.
Intention.towards.repeat.use (E-Gov Continuance Intention): the dependent vari- able measured on a dichotomous scale, refers to the respondents’ intent to continue to use the Internet for contacting government.
Performance: This independent variable was measured by the combination of two factors that compose the government’s performance: outcome perceived by the user and performance related to expected time. Perceived outcome was captured by a four-point scale question—“The outcome of your last contact with the government?”
that attempts to find the outcome in terms of success in accomplishing the task.
Performance related to expected time was measured with another four-point scale question—“Did your last contact with the government take about the amount of time you expected, more time than you expected, or less time than you expected?”
The responses for these two questions were added to get the overall measure for performance.
Satisfaction: A four-point interval scale question—“How satisfied were you with your LAST experience contacting the government?”—was used as a measure of this second independent variable. Although satisfaction is a single-item scale, previ- ous research has shown considerable precedence for using single-item measures in the context of large-scale satisfaction studies (e.g., LaBarbera & Mazursky, 1983;
Performance
Satisfaction
E-Gov Continuance
Intention (+)
Level of
Confidential Information
Gender Race (+)
(+) (-)
(-)
Figure 1. Research model
Repeated Use of E-Gov Web Stes
Kekre, Krishnan, & Srinivasan, 1995). For example, Yi’s research (1990) compared reliabilities between several multi- and single-item scales and showed that the reli- ability of single-item scales is acceptable.
Confidentiality: The level of confidential information was measured based on the amount of personal information shared by the user; a low level of personal or confidential information shared and a high level of personal information shared.
This was measured by proxy based on responses to the question—“When you made contact with the government last time, what was the purpose of that contact?”—on a dichotomous scale. Categories such as getting information and expressing opinion were considered “low level of confidential information shared,” since they require a small amount of personal information to be given out during a transaction. On the other hand, responses to actions such as filing taxes or registering a car were used to measure “high level of confidential information shared,” since these transactions needed a higher level of personal information to be filled out.
Demographics: We used gender and race as moderators of the relationship between independent and dependent variables. It is important that gender has a critical role when considering e-gov usage in both the organization and household contexts, because they may have different styles for using the Internet and Web sites. Race was also included as a moderator for similar reasons. Race was divided into “white”
and “non-white.”
Analysis.and.Results.
SPSS 11.5 was used for testing relationships of satisfaction, performance and per- ceived confidentiality on intention to continue to use e-gov. Since the dependent variable “intention to use” is dichotomous, this study was based on logistic regres- sion analysis. For testing moderating effect of demographic variables “gender” and
“race,” we followed Jaccard’s method on interaction effect in logistic regression (Jaccard, 2001). This method has been frequently cited as robust way in social science fields for testing the interaction effect. In the model, we also have a second-order dependent variable, “satisfaction,” for hypothesis.
In addition, we conducted linear regression for testing the effect of performance of an e-gov Web site where satisfaction is the continuous variable as described earlier.
0 Cha, Herath, Park, & Rao
The.Effect.of.Independent.Variables..
on.Intention.to.Repeat.Use.
To test the effect of independent variables on Intention to repeat use, we conducted logistic regression analysis. This model was found to be significant (χ2= 27.216, P<0 .000) and is reported in Table 1. Table 1 also summarizes the results that show a significant effect of satisfaction, performance, and level of confidential information on intention towards repeated use. The results indicate that intention to continue to use e-government increases as a user’s satisfaction increases. For every unit increase in satisfaction, the intention to continue to use increases by 1.2 times.
As previous research in various fields, including information systems, marketing and organization, have argued (Anderson & Sullivan, 1993; Gupta, Guimaraes, &
Raghunathan, 1992; Bhattacherjee, 2001), the significant results imply that the posi- tive relationship that exists between satisfaction and intention to use is robust.
Performance (0.491, P<0.001) and level of confidential information (-0.549, P<0.01) were also found to affect the continuance intention significantly. The result shows that a unit increase in performance increases the likelihood of continuous use by 1.634 times. Users who need to share higher levels of personal information use government Web sites less than users who need to share lower levels of personal information. These results are consistent with the study carried out by Ribbink, Riel, Liljander, and Streukens (2004) where assurance was shown to affect loyalty.
An interesting find in this analysis is that satisfaction is a weaker predictor of per- sistent intention than confidentiality. In other words, confidentiality was found to be a more important factor for e-government users to use the service repeatedly. These findings show that users who use government Web sites are more concerned about the security of their personal information than satisfaction or perceived performance of the e-government site.
Effect of Satisfaction, Performance and Level of Confidential Information on Intention Towards.Repeat.Use
Variable B S.E. Wald df Sig. Exp(B)
Satisfaction Performance Level of Confidential Information Shared
0.183*
0.491**
-0.549**
6.064 19.248 9.391
11
1
0.014 0.000 0.002
1.201 1.634 0.577 Chi-square= 27.216, df=3, P<0 .000; Accuracy=65%, sample size=1,474
Table 1. Logistic regression results
Repeated Use of E-Gov Web Stes
The.Effect.of.Exogenous.Variables.on..
E-Government.Satisfaction
This analysis was conducted by linear regression. Due to different dependent variables from the first part of the analysis, the sample size was changed. Table 2 tabulates the results of the effect of confidentiality and performance on satisfaction using linear regression. Unlike confirmation-expectation theory (Oliver, 1980), where satisfaction is determined by disconfirmation between expectation and perceived performance, we apply perceived performance of service quality as an antecedent of satisfaction according to SERVPERF (Cronin & Taylor, 1994, 1992). That is, satisfaction is determined by “performance-only measures,” which means that the user’s percep- tions of the performance of e-government services, as opposed to the difference between the consumers’ perceptions about the performance and their performance expectations. We follow this method since it provides a useful tool for measuring overall service quality attitudes in a simple way (Cronin & Taylor, 1994).
For the baseline model, R2 =0.297 and F2, 549=26.477 shows a decent fit. Our results indicate that performance of e-government’s services as perceived by users is an antecedent of satisfaction, even in the e-government context (b=.297, P< 0.01) and shows significant effect on satisfaction. However, the level of confidential informa- tion shared by the user does not act as an antecedent for satisfaction but directly affects the continuance intention (b=-0.001, P>0.1).