Getting into the field
requires the selection and training of interviewers as well as their deployment
and management.
Interviewers
Whenever possible, only
trained and experienced interviewers should be used. In any case, training is
extremely important. The process of selecting household respondents must be thoroughly
reviewed if professional interviewers will not be used. The questionnaire
should also be covered extensively, including how questions should be read and
certain words emphasized. The importance of not prompting respondents
inappropriately should be stressed and interviewers should be cautioned against
giving their own opinion either explicitly or implicitly through clothing,
facial expressions, or body language. Even experienced interviewers must be
trained extensively with a new questionnaire.
Because they often work
cheaply, university students are often hired to conduct interviews. A word of
caution about this option is in order, however. Students who are interested in
socio-political surveys often are politically active and may be more prone to
communicate their own preferences to respondents in implicit and even explicit
ways.
As mentioned above, one
should make sure that interviewers are from the same background as their
respondents, especially if the survey touches on related matters. On some
occasions, however, it might be desirable to follow a different procedure.
Fieldwork
Fieldwork should be done
under the strict supervision of field supervisors. Most respected companies
will conduct call backs to at least 10 to 15 percent of the households
interviewed, find the person interviewed to confirm they were actually
interviewed, and also go through some of the questions to verify that the
answers recorded were actually their answers.
Probably even more
important than doing this is that the interviewers know that it will be done
and that their payment will depend on getting satisfactory results from the
call backs. As a cross section of society, fieldwork interviewers are no
different from an ordinary cross section of society. And, unfortunately, many a
story has been told by exasperated researchers of finding their field workers
sitting under a tree and filling in questionnaires with fictitious names,
addresses, and answers. Call backs can be done by phone if such service is
widespread in a population. If not, they need to be done in person and probably
before an interview team leaves an area.
Field supervisors should
also check all questionnaires before the team leaves an area to make sure
everything has been filled in completely and correctly, and if not, send the
interviewer back to that person and obtain the necessary information.
Receiving Data
Actual responses will then
need to be entered into a computer readable format. There are several
statistical software packages that provide accessible data entry features and
can also read and manipulate that data once it is entered. SPSS (Statistical
Package for the Social Sciences) is one widely-used programme.
Fieldwork companies
typically will enter the data and even provide a technical report. Even so, the
organisation paying for the survey should obtain its own data set on computer
disk, preferably compatible with SPSS format. The organisation may also want to
do its own analysis, or if it is inexperienced in statistics, contract another
individual or entity to do this. The organisation should, however retain the
freedom to monitor and evaluate what is provided. Just as important, there are
probably much larger sets of data manipulations, cross tabulations, or
correlations that could be run, but that a survey company will not present in
its technical report.
Time
To do a proper job, the
length of time from conceptualization to actual data analysis and report
writing can be unexpectedly long. Even when it is absolutely imperative to go
into the field quickly to capture public reactions to some fast-breaking event,
it is hard to imagine doing a good job in less than six weeks.
Personal interviews of a
national sample could often take the most experienced company several weeks to
complete. Larger projects, such as attempting to test some model of voter
participation, will usually take at least several months or even up to a year
if fastidious academics become involved.
Costs
Project costs will include
administration costs, data entry, and general overheads. The largest portion of
these costs usually is for fieldwork and includes costs of transport, lodging,
and the actual labour of interviewers and field supervisors. The latter are
determined by the number of interviews to be done, the number of telephone
calls or household visits required to realize the number of actual interviews,
the number of hard-to-reach areas to be visited, and the length of each
interview. Thus sample size, stratification, interview clustering, and whether
a probability or a quota sample is used are important both methodologically as
well as financially.
Given the costs associated
with conducting independent surveys, however, it is possible to buy space for
one or two, or sometimes even a dozen or so, questions in on-going market
research surveys. Market research organisations tend to conduct surveys on a
regular basis, and, because a number of their clients may include questions,
the costs of the survey are shared. Costs often can be calculated on a
"per question" basis, sometimes with an initial "buy-in"
charge. Many organisations decide to "piggy-back" their questions
onto ongoing omnibus surveys. This is very efficient when one only wants to put
a few questions to a representative sample, such as checking on current levels
of interest in the next election, or current levels of registration. Also, the
frequency of such omnibus surveys allows one to check these issues on a more
regular basis and to monitor trends over time.
Some questions, like
"Why are people uninterested?" or "Why are they not yet
registered?", lead to many other questions. The more questions to be used,
the higher the costs will be. In addition, it may be desirable to get
respondents focused on issues of voting, elections, and democracy in order to
obtain more thoughtful and considered responses. With an omnibus survey, there
may be no control over whether respondents are answering a question about the
competitiveness of elections immediately after being asked about their monthly
consumption of motor oil. Finally, due to reasons of cost and client interest,
ongoing market research surveys may not be done in remote rural areas or poorer
areas.
