This development process resulted in two substantive models:
Model 2, describing the relationships between Motivational Leadership (ML),
Work Meaning (ME), Job Performance (JP) and Discretionary Service Behaviour
(DSB); and
Model 5b, describing the relationships between Employee Empowerment (EM),
Motivational Leadership (ML), Employee Positive Attitudes (EPA), Job
Performance (JP) and Discretionary Service Behaviour (DSB).
The model development process is reviewed immediately below and the important
findings from each model iteration are highlighted. Section 8.3 then goes on to
discuss these relationships and their implications for hospitality and leadership
research in greater detail.
The models were developed in an iterative manner with the core hypotheses
being examined first and successive constructs being added to the model
thereafter. In this way, the core hypotheses (H 1 to H 4 ) testing the relationships
between motivational leadership, work meaning and job performance (JP and
DSB) are tested in the absence of additional, and potentially confounding,
constructs.
This approach was justified by two findings. Firstly, the hypotheses relating to
Work Meaning (H 3 and H 4 ) could not (with hindsight) have been tested if the full
model was initially specified. This is because the three employee attitude
constructs (Work Meaning, Affective Organisational Commitment and Job
Satisfaction) did not demonstrate sufficient discriminant validity, meaning that
they could not be included as discrete entities within one model. Secondly, the
effect of work values on employee work attitudes (Model 4) would have been
overlooked had the Work Values construct not been modelled separately from the
Employee Empowerment construct.
Model 1 tested the effect of motivational leadership (the ML construct) on self-
reported job performance (the Job Performance construct, JP) and peer-assessed
job performance (the Discretionary Service Behaviour construct, DSB). This
model established that these three constructs did resolve as valid statistical
artefacts and are related at a statistically significant level. Modifications in the
form of removal of some indicator variables were made to each construct in order
to produce a model that fitted well to the data.