operation shown by many of the hotel general managers - even when contact was
supported and endorsed by either the hotel association to which that hotel
belongs – supports this position.
In practical terms, the sample consists of non-supervisory waiting staff working in
table service restaurants in 3 and 4 star hotels in the UK and can therefore be
considered reflective of this constituency.
Evaluating survey non-response
Survey research findings can be influenced by non-response bias that can be
introduced where the responses of persons who did complete the survey differ
substantially from those who do not complete the survey (see e.g. Keegan and
Lucas 2005). For example, if married couples were under-represented in a
particular survey, and married couples have significantly different opinions and/or
experiences from the remainder of the population, then a form of systematic bias
has been introduced to the survey data.
In this survey, bias could potentially have been introduced through the
discriminatory selection of recipients on the part of the person distributing the
survey forms, thus preventing some potential respondents actually receiving the
survey form. The specific concern for this survey had been that hotels might
select employees who enjoy good leader relations in order to produce a
favourable result for that hotel or hotel group. To counter this possibility, the
researcher made it clear to all of the participating hotels that, because the
statistical analysis required a sample size of 200 or more, it would be impossible
to make any inferences about individual hotels or even hotel groups and that they
should not, therefore, be concerned in any way about any findings emerging that
might be unfavourable for their organisation.
While steps have been taken to counter this particular source of non-response
bias, it is nevertheless good practice to evaluate, as far as possible, the nature of
any survey non-response so that any potential bias on the findings can be
anticipated, investigated and ultimately accounted for. The remainder of this
section follows a structured approach to evaluating the nature of any survey non-
response. The effects of any such non-response on the research findings (i.e. the
resulting non-response bias) will be evaluated following the development of the
statistical models using multi-group SEM analyses (see Section 7.15).