Analysis and Results

Each multi-item scale was factor- analyzed to evaluate dimensionality, and reliability analysis was performed to determine if each item contributed to scale reliability. We omitted items if they did not load with the majority of the other scale items or if they failed to improve internal consistency. A final factor analysis confirmed the unidimensionality of each scale. The factor loadings appear in Table 15-1. As a consequence, we summed the HJC and DSI items to form scales. Because the three online buying items were not measured by the same response format, they were submitted to a principal- components analysis and factor scores were computed via the SPSS regression method to represent that scale. Descriptive statistics of the three scales appear in Table 15-2.

Table 15-2: Scale characteristics

	Global Innovativeness (HJC)	Domain Specific Innovativeness (DSI)	Online Buying
Number of Scale Items	5	4	3
Minimum Score	13	4	-1.3
Maximum Score	35	20	4.6
Mean Score	26.7	12.5
Standard Deviation	3.80	4.09	1.0
Skewness (SE)	-.625 (.14)	-.311 (.14)	1.42 (.14)
Kurtosis (SE)	.573 (.28)	-.668 (.28)	3.74 (.28)
Coefficient Alpha	.72	.88	.72

Correlations among the study variables appear in Table 15-3. These correlations show that all three variables are related , at least for the men. Thus, the first three hypotheses are supported, but a moderating effect of gender is suggested.

Table 15-3: Pearson correlations among the study variables

	HJC	DSI
	All Respondents (n = 303)
DSI	.17**
Buying	.20**	.68**
	Men Only (n = 135)
DSI	.30**
Buying	.32**	.67**
	Women Only (n = 168)
DSI	.04
Buying	.07	.67**
** p < .05 level, two-tailed; ** p < .01 level, two-tailed HJC = Hurt, Joseph, and Cook (1977), Global Innovativeness Scale DSI = Goldsmith and Hofacker (1991), Domain Specific Innovativeness Scale

To test the fourth hypothesis, we followed the procedure recommended by Baron and Kenny (1986, p. 1176). Refer to Figure 15-2:

A variable functions as a mediator when it meets the following conditions: (a) variations in levels of the independent variable significantly account for variations in the presumed mediator (i.e., Path a ), (b) variations in the mediator significantly account for variations in the dependent variable (i.e., Path b ), and (c) when Paths a and b are controlled, a previously significant relation between the independent and dependent variables is no longer significant, with the strongest demonstration of mediation occurring when Path c is zero.

Thus, we first regressed online innovativeness on global innovativeness, then regressed online buying on online innovativeness, and then regressed online buying on both global and online innovativeness. These results appear in Table 15-4. The regression analyses confirm the correlation analysis. Domain-specific innovativeness (online innovativeness) (path A) is significantly related to global innovativeness (b = .17) and to online buying (b = .68) (path B). Global innovativeness is significantly related to online buying (b = .20) (path C). But when the mediating effect of online innovativeness is removed statistically, the link from global innovativeness to online buying is highly reduced (b = .09). Baron and Kenny (1986, p. 1176) discuss this result:

Table 15-4: Regression results

When Path c is reduced to zero, we have strong evidence for a single, dominant mediator. If the residual Path c is not zero, this indicates the operation of multiple mediating factors. Because most areas of psychology, including social, treat phenomena that have multiple causes, a more realistic goal may be to seek mediators that significantly decrease Path c rather than eliminating the relationship between the independent and dependent variables altogether. From a theoretical perspective, a significant reduction demonstrates that a given mediator is indeed a potent, albeit not both a necessary and a sufficient condition for an effect to occur.

Thus we can conclude that online innovativeness partially mediates the influence of global innovativeness on online buying, supporting H4. Why this might be the case is discussed later.

While no specific hypothesis was proposed for gender effects related to online innovativeness, a post hoc comparison was conducted to determine if gender played a role. Thus, the next step in the analysis was to repeat these regressions for men and women separately to assess whether gender moderated the relationships. Note that gender was uncorrelated with all three of these variables, and ' moderation implies that the causal relations between two variables changes as a function of the moderator variable' (Barron and Kenny, 1986, p. 1174). The moderating effect is shown if the paths are statistically different in size when compared between the two groups (see Baron and Kenny, 1986). The results appear in Table 15-4. When we compare the unstandardized regression coefficients for the effect of global innovativeness (HJC) on amount of purchasing (BUY), we see that the coefficient for men (.081) is larger than that for women (.02) suggesting that the relationship is stronger for men than for women. To assess this, a new variable was computed as the product of gender (0 = women, 1 = men) times HJC after mean centering the HJC scores to mitigate multicollinearity in the analysis (Cronbach, 1987; UCLA, 2003). Buying was then regressed on mean-centered HJC, gender, and their product. Since the regression coefficient for the interaction term was significant (t = 2.05, p = .041), we can reject the hypothesis that the male and female regression coefficients were equal and conclude that the relationship is stronger for the men than for the women. An identical analysis comparing the men's and women's regression coefficients for the effect of DSI on buying (.173 versus .156) (path B) failed to reject the null hypothesis that these coefficients are equal. Thus, it appears that online buying innovativeness does mediate the influence of global innovativeness on online buying, at least for men suggesting that most online innovators currently are men.