Research Methods in Tourism, Hospitality and Events Management

Errors in sampling frames

The following are some of the main errors which are commonly found with sampling

frames. The details presented below have been developed from the work of Moser

and Kalton (1993) and Frankfort-Nachmias and Nachmias (1996):

1. The information is incorrect.

This can occur because mistakes were made in entering names and addresses

when the list was written. Often, lists of names and addresses are compiled for

purposes other than the needs of a future survey. Moreover, such lists quickly

become out of date, with people moving away.

2. The information is incomplete.

This commonly occurs because the need for a sampling frame was never

envisaged. Many tourist attractions or event organisers are unlikely to know

who visits their site or their event. Similarly, transport carriers may have some

information where travel tickets are posted to an address or via an agent, but not

about travellers who purchase at a rail or coach station. It is unlikely that a

restaurant will have all the details of every customer who purchases a meal and

individuals may visit in a group which means that the information will be

incomplete.

3. The information is duplicated.

This occurs when lists are combined to develop a sampling frame. For example,

a tour operator or hotelier may wish to send postal questionnaires to a sample of

customers who have purchased a holiday or hotel room from the company in the

last five years. Although the names and addresses are known for all the

customers, some will have purchased a holiday or hotel room more than once

and therefore may appear on the same sample frame several times. As such, the

sample drawn from this list may be biased as some holidaymakers or guests will

have a greater chance of being included in the sample. If you assume that

frequent purchasers are more favourable towards the company, then such a

biased sample may ultimately produce results that are not truly representative.

---

In this example, this is on top of the potential problem that going back five

years will undoubtedly produce names and addresses that are out of date, so the

sample may be biased in terms of a higher response rate from those who used

the company more recently and have not moved house.

4. The information is clustered.

The problem here is that perhaps several people live at a single address but the

purpose of the survey is to interview a particular member of that household.

Thus, the precise definition of sampling units (particular individuals in this

case) is not possible from the sampling frame available. Another instance where

this type of problem can occur is at tourist attractions where there is information

known that a certain proportion of visitors come from within a 5-mile radius, a

10-mile radius or from particular towns and cities. As the survey focuses on

individual visitors, this information is helpful in gaining an opinion about the

proportions of those who should be interviewed from these locations, but not

about who the visitors are or their characteristics.

5. The information includes aspects not applicable.

A similar problem is that names and addresses may well still be valid but only

certain types of individual are required for the survey. The survey may need to

contact individuals of a particular gender, age or some other aspect, such as

those holidaymakers who went on a particular excursion as part of their holiday.

Hence, the list includes information which is not applicable but may not include

the extra information needed to identify the particular sampling units (in this

case certain holidaymakers) which are required for the survey.