13.07.2015 Views

Journal of Business logistics, Vol. 28, no. 2, 2007 ... - Global Initiatives

Journal of Business logistics, Vol. 28, no. 2, 2007 ... - Global Initiatives

Journal of Business logistics, Vol. 28, no. 2, 2007 ... - Global Initiatives

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

118 hult, boyer, and ketchen, Jr.Discriminant validity was assessed by calculating the shared variance between pairs <strong>of</strong> constructsand verifying that it was lower than the average variances extracted for the individual constructs(Fornell and Larcker 1981). Shared variance was calculated as:2γ = 1 −ψwhere γ 2 = shared variance between constructs, and with the diagonal element <strong>of</strong> ψ indicatingthe amount <strong>of</strong> unexplained variance. Since η and ε were standardized, γ 2 was equal to the squaredcorrelation between the two constructs. In all but one case (i.e., the behavioral intentions scale), theaverage variances extracted were higher than 50% (cf. Fornell and Larcker 1981) (ranging from43.7% to 70.7%; see Table 2). The shared variances between pairs <strong>of</strong> all possible scale combinationsindicated that the variances extracted were higher than the associated shared variance in all cases (seeTable 2 for shared variances and Table 3 for average variances extracted).In the interest <strong>of</strong> thoroughly examining discriminant validity, we conducted one additional test<strong>of</strong> discriminant validity (Anderson 1987; Bagozzi and Phillips 1982). This test entailed analyzing allpossible pairs <strong>of</strong> constructs in a series <strong>of</strong> two-factor CFA models using LISREL 8.72. Each modelwas run twice – once constraining the φ coefficient to unity and once freeing this parameter. A χ 2 -difference test was performed on the nested models to assess if the Δχ 2 was significantly lower forthe unconstrained models (Anderson and Gerbing 1982). The critical value (Δχ 2 > 3.84) was(Δdf=1)exceeded in all cases. Thus, the seven constructs and their purified 33 indicators were found to bereliable, valid, and robust across samples.Testing for potential common method biasBefore moving on to the hypothesis testing, we conducted an examination <strong>of</strong> potential commonmethod variance problems. We used Harmon’s One-Factor test to examine common method bias(Podsak<strong>of</strong>f and Organ 1986). As such, the 33 remaining items were factor analyzed via SPSS 11.0with a principal component extraction method and varimax rotation to examine if one single factorwould emerge and/or if one general factor would account for most <strong>of</strong> the covariance in the variables.Rotated factors can<strong>no</strong>t be used to test for common method bias since the variance explained is allocatedto a particular rotated factor, <strong>no</strong>t to the first and strongest factor. The unrotated solution allocatesthe most variance to the first factor, the second most to the next factor, and so on. If a largemajority <strong>of</strong> the variance is explained by the first factor it is an indication that common method biasmay be an issue. Using a varimax rotation, the factors had variances explained ranging from 4.45% to21.99%. Without rotation, the first factor explained 34.29% <strong>of</strong> the total 71.01%. Thus, our inferenceis that common method bias does <strong>no</strong>t appear to be an inhibiting factor in this study.Hypothesis Testing and ResultsIn the testing <strong>of</strong> our predictions, we followed the literatures on configuration theory (Doty,Glick, and Huber 1993; Miller 1997) and pr<strong>of</strong>ile deviation analysis (Drazin and Van de Ven 1985;Venkatraman and Prescott 1990). The data were standardized (mean-centered with a mean <strong>of</strong> zeroand a standard deviation <strong>of</strong> 1) to remove the effects <strong>of</strong> different measurement units and potentialmulticollinearity (Gresov 1989; Jaccard and Turrisi 2003).

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!