7.15.1 Demographic variables
Section 6.5.2 above compares the descriptive statistics for the sample with
available known population values published in the current People1st Industry
Profile characteristics for hotels and restaurants (People 1st 2011a, 2011b,
2011c). Several differences between the sample and the population values were
described in Section 6.5.2 and these are summarised in Table 7 - 36. Other
respondent characteristics for which no published population values were
identified are employment tenure (temporary / permanent) and length of
employment.
Multi-group analysis can be used to assess whether or not categorical variables
(e.g. gender, age) have an influence on the relationships between constructs in
an SE model. For example, it may be hypothesised that male employees may
behave differently in response to certain leadership styles in comparison with
females, or that a particular intervention may impact younger employees’
attitudes to a greater or lesser degree in comparison with older employees.
Category Description of the sample in comparison with:
(^) Hotels and Restaurant
staff Waiting staff^
Gender Similar to population values Males over-represented
Part-time / Full-time Similar to population values Part-time under-represented
Respondent origin Non-UK employees over-represented
Age <24 over-represented <24 slightly under-represented
Table 7-36 Sample demographics comparison with known population values
In the context of assessing the ways in which survey non-response may be
influencing the relationships between variables in the models, multi-group SEM
can be used to identify whether or not group membership has a statistically
significant influence on model parameter estimates. Where effect sizes are found
to be moderated by group membership, the nature of these between-group
differences can be considered alongside survey non-response to provide a
subjective evaluation of the ways in which the survey findings may be influenced
by the characteristics of the sample. Findings such as these can generate useful
insights into how the survey findings might generalise to the population of
interest.