Research Methods in Tourism, Hospitality and Events Management

Stratified random sampling

Where information about the population of a survey is known, it is possible to divide

that population into smaller sub-samples or ‘strata’. It is common in our field for this

to be done on the basis of socio-demographic characteristics such as gender, age,

socio-economic group or a categorisation of previous purchasing behaviour. A

particular sampling unit such as a potential individual respondent can be placed in

the sub-sample only once. After each sub-sample has been determined, the individual

sampling units are randomly selected from each sub-sample.

In effect, this process is similar to probability sampling where every member of a

population has an equal chance of inclusion in the survey and selection is conducted

randomly. Here, the population is divided into strata, and within each stratum every

member has an equal chance of inclusion and the selection is random.

The main purpose of this method of sampling is to achieve a more reliable sample.

However, a great deal of knowledge about the population is required, over and above

the extent of the population, if stratification is to be workable. If, for example, within

a survey population there are clearly identifiable groups where there is a high level

of similarity (homogeneity) within the groups and there are many differences

(heterogeneity) between the groups, then stratification may be worthwhile. This may

particularly be the case if the distinct groups vary in size. For instance, if a company

has a small minority who complain about their product, then they may not be

sufficiently recognised in a large survey. If, however, they are characterised as a

group by several easily identifiable factors, then the problem could be much greater

---

than is apparent. This could, for example, be a group of holidaymakers who are

identified by their age or social class, or those who stayed at a specific resort.

There remains the problem, though, of whether such groups can be identified ‘prior’

to the main survey to enable this type of sampling method to be performed.

Computer analysis of survey results allows for such groups to be selected out from

the sample and to be analysed specifically and separately from the rest. Thus, prior

selection and stratification, as well as not often being achievable (lack of prior

knowledge), are not always necessary.

Perhaps the main use of this type of sampling method in the field of tourism,

hospitality or events is where the residence of individuals is known and can be

stratified. To demonstrate this, take, for instance, a survey of season ticket holders to

and from an island for a ferry operator. Season tickets are purchased by commuters

to the island and to the mainland. Prior to the survey, the population of all season

ticket holders was found to be 20,000. Of these, 15,000 live on the mainland and

5,000 live on the island. If it is decided that a sample of 1,000 should be interviewed,

then appropriate proportions could easily be calculated, i.e. 750 mainlanders and 250

islanders.

Illustration 4.4 An example of stratified random sampling from an

undergraduate dissertation

‘The transformational leadership style: an assessment of the application

of emotional contagion and its impacts on employee motivation and

performance’ by Chloe Locke, BSc Events Management, supervised by

Richard Parkman and Steven Jakes, Plymouth University.

Aim

An assessment of emotional contagion within gender groups and

preferred leadership style within service sector, business sector and

manual labour occupations.