total = 90). Thirty two valid responses (16 per cent) in total were received: 18
survey forms (16.4 per cent) from the south east hotel group and 14 survey
forms (15.5 per cent) from the UK-wide hotel group. Four incomplete
questionnaires, 2 from each group, were also received.
Dillman (2006: 323) notes the challenges that can be faced by researchers
attempting to administer respondent-completed surveys within organisations (as
opposed to surveys of individuals or households). A number of organisational
studies researchers have discussed how surveys targeting respondents within
organisations often demonstrate low response rates (see e.g. Hager et al. 2003,
White and Luo 2005) and Dillman (2006: 323) helps to quantify typical response
rates by describing an analysis of 183 business surveys carried out in 1991 where
the mean response rate was found to be 21 per cent.
The 16 per cent return rate from the pilot fell a little short of the 21 per cent
mean return rate described by Dillman. However, a second wave of pilot surveys
was not sent owing to concerns about exhausting the goodwill of the participating
hotels by asking them to participate in three and not two survey waves (the
planned two waves were pilot plus full survey, a second pilot wave would have
taken the total to three). Further, the time between distribution and receipt of
the pilot survey forms was 6 weeks – once again, wishing not to exhaust the
goodwill of the participating hotels (in particular the UK-wide group with a further
24 hotels to be included in the full survey) by engaging them for an extended
period, priority was given to moving onto the full survey implementation.
Following the guidelines for preliminary (i.e. pre-test) assessment of data for SEM
analysis provided by Hair et al. al (2006: 780 - 781), the data from these 32
questionnaires were used to evaluate the suitability of the indicator variables by
assessing distributional normality, mean values and (using both exploratory and
confirmatory factor analysis) the factor structures and item loadings. The
characteristics of the data are as follows.
Regarding distributional normality, each indicator variable was checked for
univariate normality and only one variable (EM6) violated the bounds of either 2
for univariate skewness or 7 for univariate kurtosis described by Curran et al.
(1996). Specifically, the skewness value for EM6 was 2.1, however, as this did
not transgress the boundary by a great deal and because the sample size was
quite small, it was considered that removing EM6 at this stage may be a hasty