|
6.5.2 Common errors in interviews
Errors can occur throughout the survey process.
Chapter 5
suggested how
questions can be poorly written, which leads to error in terms of inaccurate
or poor responses.
Chapter 4
highlighted potential pitfalls with sampling,
which again can lead to unrepresentative results occurring. Moreover, the
correct selection of the method of data collection for the purpose of the
market research is crucial to the success of the project. When it comes to
interviewing, there are, again, several areas where errors can occur. These
are now discussed.
1. Interviewer cheating
Perhaps the most dramatic of all interview errors is cheating by the
interviewer. The worst form of this is the falsification of data – that is,
making up the responses. Either fictitious respondents who were never
interviewed are recorded or not all the interview was completed and
the interviewer ‘makes up’ the gaps. In either case, the action is
deliberate and dishonest. To control and discourage this, the research
director must put in place a variety of checks on the interviewers. For
site surveys at tourist attractions, for example, the interviewer can be
observed periodically and his/her questionnaires checked throughout
the day. For interviews in the respondent’s home, where names and
addresses are required, interviewers can explain that a small sub-
sample of respondents will be contacted to verify that an interview
took place. Such checks can avoid this type of problem arising.
2. Interviewer influence: Type 1
Occasionally, in an attempt to develop a rapport with the respondent,
or because the interviewer is embarrassed asking certain questions, the
interviewer will pass a personal opinion. Where this happens, bias may
be introduced into the responses. A respondent may not wish to
disagree with the interviewer or be uncertain about their own opinion
and consequently ‘side’ with the interviewer. Such bias is clearly
inappropriate and should be a focus of training in the survey
requirements. Regardless of the nature of the research or interview
technique, the interviewer must never pass their own opinion.
3. Interviewer influence: Type 2
Another type of interviewer bias occurs in relation to the appearance
and unspoken manner of the interviewer when interacting with the
respondent. Age, gender, apparent class, ethnic origin, personality,
clothes and hairstyle may all have some influence on how respondents
react. Similarly, the age, gender, etc., of the respondent may in turn
affect the interviewer’s approach. In western countries, women are
often more successful as interviewers, particularly where it is
necessary to gain entry into households to collect data. Day-time
home-based interviews of females by male interviewers may
inevitably suffer from low response rates. In very large surveys, these
influences may be limited, but how a potential interviewer will be
viewed by respondents should be given consideration in the selection
procedure.
4. Errors when asking questions
Sometimes, quite unwittingly and with the best intentions, interviewers
will change the wording of a question. This may be done to better fit
the way the interviewer would ask such a question or be reworded,
thinking it would elicit a better reply from the respondent. In
structured interviews, such practice cannot be permitted, as
considerable variability in the responses may occur. In the one of the
authors’ own experience (Brunt, 1990), where a face-to-face site
survey of a tourist attraction was being undertaken, an open question
required respondents to state where they had been on day trips in the
previous four weeks. For speed of coding, a range of different types of
day-trip site were printed on the questionnaire for the interviewers’
own use. Because it was a difficult question, requiring recall, one
interviewer speeded up the process by using the questionnaire as a
prompt card and asked respondents to indicate yes or no to each of the
categories printed on it. Consequently, the results from this interviewer
were completely out of step with other interviewers who asked the
question in the way intended. This type of problem, yet again, stresses
the need for clear instructions and thorough training.
5. Problems probing
Even structured questionnaires often have a number of ‘open’
questions where interviewers must ‘probe’ for the fullest possible
answer. However, where several interviewers are used to collect data
there is a real possibility of inconsistency between them. Clearly,
where the aim of the survey is to compare the responses to such
questions and make judgements, inconsistent answers from variable
probing are worthless. To cope with this type of problem, there are a
variety of solutions, as listed below:
extensive interviewer training
a probing ‘pro-forma’, i.e. probing on strictly prescribed lines
a mixture of question types: open questions followed by closed
questions of the same topic to facilitate cross-checking
no open questions at all
a separate data collection method using unstructured interviews in
support of structured ones.
6. Errors recording the answers
With closed questions where the interviewer is required to tick a box, errors
here are down to carelessness. This of course assumes the interviewer
understands how to complete the questionnaire. Other errors of this type
occur in response to open questions where either the interviewer must write
down what is said or interpret the open answer and allocate it to a particular
box of pre-coded categories. The potential for error in the former case is
due to carelessness, laziness in writing or trying to abbreviate the answer,
and in doing so misses the point. In the latter case, the interpretation of
replies and trying to make them ‘fit’ a particular category can cause bias.
Wherever such circumstances arise, it is better for interviewers to record
answers verbatim, as opposed to paraphrasing or only writing down what
appears relevant. For interviewers to listen, interpret and then write an
answer down in brief whilst maintaining rapport is difficult. Accuracy and
consistency can easily be lost. It is much more preferable for interpretation
to occur at a later stage. This solution further avoids potential problems by
having a variability between interviewers in terms of their vocabulary and
ability to summarise.
Do'stlaringiz bilan baham: | 
