structures (see e.g. Hair et al. 2006 : 1 22 - 127 ). All constructs performed as
hypothesised except the Work Meaning construct. Following the removal of the
problematic ME2 and ME4 indicators, the Work Meaning construct was found to be
unidimensional, as hypothesised.
The exploratory Service Quality (SQ) construct which included a non-response
option (for respondents who never experienced the specific type of service failure
described in a particular item statement) was examined to evaluate the volume of
non-response and to assess the likely impact of such non-response on the ability
to perform SEM analyses using this construct. For the seven SQ items, 5 of the
32 respondents used the non-response option. The distribution of non-responses
was as follows: SQ1 = 5, SQ2 =3, SQ3 = 4, SQ4 = 1, SQ5 = 0, SQ6 = 4, SQ7 =
- This was a suitably low rate of non-response and, together with the good item
loading levels (the range was 0.670 to 0.895) and the unidimensionality of the
scale, provided a rationale for maintaining the SQ construct in the survey.
Aside from the removal of the four item statements described above, one further
modification was made to the survey questions on respondents’ demographic
characteristics. Question 10 was included in the pilot survey in form shown
below:
However, although all respondents ticked either full- or part-time, 27 of the 32
respondents failed to indicate their tenure status. Accordingly, Q10 was reworded
as two separate questions for the full survey.
Summary of the pilot findings
The pilot survey found that (subject to the minor modification to demographic
question 10) the questionnaire works well for a ‘live’ sample of respondents. Four
item statements were removed owing to low item loadings and (in the case of
ME2 and ME4) problems with unidimensionality.