The goodness-of-fit indices of the one-factor model and the two-factor model are shown in
Table 3. The four-factor model and the second-order factor model were not retained, since the
latent covariance matrix of these models turned out to be not positive definite due to a
correlation greater than one between the latent variables of assurance on the one hand and
empathy and reliability on the other hand. This indicates that the different constructs are not
all statistically distinguishable, as was suggested by Nehles et al. (2008). However, since the
two-factor solution did not provide a significantly better fit to the data than the one-factor
model (∆χ² = 3.03, ∆df = 1, not significant) we decided to select the more parsimonious onefactor
model (Kernis & Goldman, 2006). The goodness-of-fit indices of this model satisfy the
criteria set by Hu and Bentler (1998) and Browne and Cudeck (1993). We thus consider HR
department service quality to be a one-dimensional construct in our analyses. This construct
may have different facets, but they are not empirically distinguishable. The factor loadings of
the HR department service quality construct are all statistically significant (p