what is being witnessed by this respecification is the transference of some of the
direct ML→DSB effect via the EPA construct.
Figure 7-23 SEM 5b:3 with the EPA→DSB path
The inclusion of the EPA→DSB path does bring another effect into play, however:
there is now an indirect (fully mediated) effect of EM on DSB via EPA
(EM→EPA→DSB; = 0.161). Once more, this path can be rationalised on the
basis that employees working in close proximity experience the same levels of
empowerment.
In summary, the difference made by introducing the previously unhypothesised
EPA→DSB path are that:
the total effect of ML→DSB is down from 0.403 to 0.298;
o 0.298 is made up of: direct = 0.245 and indirect = 0.053
some of the variance in DSB now explained by:
o EM→EPA→DSB ( = 0.161) and
o by EPA→DSB ( = 0.255)
the total variance explained in DSB has risen from 0.16 to 0.20.
Statistically, SEM 5b:3 is a better model owing to its improved global fit
measures, however, in conclusion, the fact that the new EPA→DSB path appears
simply to re-direct some (0.059) the ML→DSB effect through EPA combined with
the conceptual difficulties in theoretically sustaining the notion that one person’s
attitudes can influence another’s behaviour, leads to the rejection of the