Research Methods in Tourism, Hospitality and Events Management

4.4.5 Non-probability/non-random sampling

In situations where the population is not known or there is insufficient time to build a

sampling frame, some researchers may turn to non-probability sampling methods.

The most commonly used (and criticised) method is called ‘quota sampling’. For

---

example, in street surveys interviewers may be required to question certain types of

people and are given a ‘quota’ to complete of each type. The population may be

divided in terms of apparent age, social class or other controlling variables upon

which the interviewers must make a subjective judgement in the selection of

potential respondents. In the opinion of the authors, this leaves too much discretion

to the interviewer, enabling a significant amount of bias to enter the sample.

4.5 Sample size

Moser (1958) stated that anyone who ever has to advise on sample designs will know

that the first question he is asked is – how big a sample do I need? Many years later,

this statement remains true. However, often the question being asked could be more

accurately re-phrased as ‘what is the minimum number I can get away with?’

Selecting the size of the sample is clearly an important question in the design of any

survey. All surveys where a sample of the population is used, will involve a level of

‘sampling error’. This is the difference between what has been found in the results

based on the sample and the actual values that would have been found out if

everybody in the population had been included in the survey.

It is tempting to assume that by simply enlarging the sample size, problems of error

will disappear. However, any sample which is biased in its sampling methods or

selection of respondents will remain biased whatever its size. Furthermore, until the

sample size becomes a high proportion of the population (say, over 20%), the actual

size of the sample is more important than the sample’s proportion of the population.

Clearly though, an increase in sample size in a survey where the sample is drawn by

fair means will reduce the level of sampling error. Thus, the sample size often

represents a compromise between what is achievable in a pragmatic sense and what

is desirable statistically in relation to the concept of sampling error. These two sides

of determining sample size will now be evaluated in turn.

Illustration 4.6 Sampling extract from an undergraduate dissertation

‘Tourist satisfaction: the mythology of King Arthur’s Tintagel’ by

Danielle Chapman, supervised by Natalie Semley, Plymouth University,

July 2016.