(i) confirming that motivational leadership does indeed have a positive influence
on employee job performance in hospitality catering service (a relationship that
has not, hitherto, been measured in a hospitality context); and
(ii) the similar strength of the ML→DSB relationship provides a validation of the
ML→JP link insofar as DSB is a peer-assessment of extra effort designed to
minimise inflated performance self-assessment.
A final set of values which should be considered when interpreting the effect sizes
in the structural model is the squared multiple correlations (SMCs). In a
structural equation model, an SMC value is reported for each endogenous
variable. The SMCs are analogous to R^2 values in regression analysis and are
calculated by raising the path coefficient to the power of 2 – i.e. the SMC is the
square of the structural coefficient. The SMC value represents the amount of
variance explained in that factor by the structural relationships in the model (i.e.
the percentage of variance in the latent dependent variable explained by the
latent independent variable/s). For model 1, the SMC for ML→JP is 0.414^2 =
0.171 and for ML→DSB it is 0.396^2 = 0.157.
In substantive terms, the SMC values reflect the ability of each structural
relationship to fully explain the variance in the endogenous (dependent) variable.
Lower estimates for SMCs signal that the structural coefficient estimates are less
reliable than if the SMC values were higher.
Knowing the SMC values for the two endogenous variables in Model 1 is not
terribly informative – that is, they are simply the squared values of the structural
coefficients. However, in models where an endogenous variable is influenced by
more than one predictor variable, it is not such a straightforward matter to
calculate the SMC. Accordingly, for this research, in models where an
endogenous variable has more than one predictor, SMC values will be displayed
on the model illustrations.
As with the measurement model (CFA 1:3), model SEM 1:1 was found to have a
multivariate C.R. value of 10.9. As this estimate is indicative of departure from
multivariate normality, AMOS’s bootstrapped estimation procedure was
performed. This procedure indicated that all of the parameters are robust under
the prevailing conditions of multivariate non-normality (once again, the full details
of both the C.R. estimate and the subsequent bootstrapped estimates can be
found in Appendix IV).