Following the previous page, on how to approach people for interviews, this page covers the techniques of survey interviewing.
When a suitable respondent has been found, and is ready to be interviewed, at last the interview can begin. It's now time to introduce the survey.
Introductions
When a likely respondent has been found, the interviewer now needs to briefly explain the purpose of the survey. In a country where most people have never been interviewed, the interviewer may also need to explain what a survey is, and how it works.
Many survey organizations have a printed introduction on each questionnaire, and require the interviewers to read this aloud. To me, this always sounds very stilted, and it's not a good way of persuading people to take part in the survey - which is the whole purpose of an introduction. Instead, I ask each interviewer to develop his or her own introduction, including:
Here 's an example of an introduction.
"Hello, my name is Bertha Bluggs. I'm working for Audience Dialogue, and we're doing a survey with owners of new houses about the reasons why people choose to live in particular areas. The purpose of the survey is to improve the planning of housing facilities. So I'd like to ask you some questions. This will take between 10 and 20 minutes, and all your answers will be kept strictly confidential. So may I interview you, please? Is it OK now, or can I make an appointment for another day?"
This was from a telephone survey in Australia. For a face-to-face interview, introductions are usually longer than this, and in other countries (such as Cambodia) a much longer and more eloquent introduction would be expected.
After the introduction, when the respondent has agreed to participate (but before the questions begin) the interviewer can give further information about the process. This can include:
The respondent at this point is invited to ask any questions about the survey.
Creating a comfortable setting
Before the interview can begin, the respondent needs to feel at ease, not threatened in any way, and not hurried. If the interview will last more than a few minutes, it's usually best to sit down. In some countries, a research interview is such an unusual activity that it may attract a lot of onlookers. I've seen interviews in developing countries where more than 20 people were watching and listening. This doesn't make it easy for a respondent - but the same person may also feel uncomfortable closeted indoors with the interviewer, particularly one of the opposite sex. If the respondent is a woman, and her children are hanging around, this can be a great distraction. It's often best for the interviewer to take responsibility for shooing away onlookers, explaining to them that the respondent has been selected at random, and that this is a private matter.
If some emergency happens in the middle of an interview - e.g. a child hurting itself - the interviewer should offer to suspend the interview, and return later to finish it.
The actual interview
After you have located a respondent, persuaded him or her to take part in the survey, and explained its purpose, the actual interviewing is perhaps the easiest part of the process.
All the interviewer needs to do is to read out the questions, follow the instructions, and record the answers given.
If all goes well, it is very simple - but some respondents aren't easy to deal with. What do you do when people misunderstand the questions, give irrelevant answers, refuse to answer some questions, and so on?
If an interviewer seems not to understand a question, the first step is to repeat it, but more slowly. The reason for this is that interviewers unconsciously speed up. By the time they are interviewing their 20th respondent, they usually know the questions by heart. There's a tendency to speed up, to gabble the question quickly. But each respondent is hearing the question for the first time, and will need a few seconds to absorb its full meaning.
When a respondent gives an answer that's clearly irrelevant, the interviewer should say something like "I'm not sure if you heard that question properly. Let me repeat it...."
When the respondent doesn't understand the question
If, after a question is repeated, the respondent still doesn't understand or is unable to answer, interviewers are tempted to rephrase the question in simpler language. Most books on surveys (and most survey organizations) say interviewers must never do this, but if a question is badly worded, an interviewer will usually try to get a valid answer.
One solution to this, of course, is to make sure that a question (if it is the multiple-choice type) covers all possible alternatives, that it is unambiguous, and that it is short enough to remember in its entirety.
Research on question wording has found that some types of question - particularly questions about attitudes and beliefs - are very sensitive to changes of wording. In some cases, changing a single word can radically change the spread of answers. For other types of question - about observable facts, such as age group - the answers don't vary much with changes in wording - but these questions are usually understood clearly by respondents.
Despite the wishes of questionnaire writers, interviewers will reword questions - specially in situations where they cannot be checked, such as door-to-door surveys. To avoid this, training must be very thorough, and you must explain to interviewers how the wording can affect the responses. If interviewers are treated as speaking-machines (as in many large market research companies) the survey results will not be as accurate as possible.
In a survey, every multiple-choice question should allow for all possible answers, but sometimes this doesn't happen, and a respondent may give an answer that's not a listed alternative.
In this case, my advice is for the interviewer to write in the answer that the respondent gives, and let the coders worry about how to classify it. (This is covered in more detail in section 1 of the chapter on analysis.)
Probing
If an interviewer asks a question without also giving an exhaustive list of possible answers, the respondent may not answer it in the way intended, or may not give a full answer. This calls for probing - which means asking follow-up questions to clarify the initial answer given by a respondent.
The more vaguely a question is worded, the more likely that the interviewer will need to probe for an answer. Sometimes a question is intentionally worded very loosely. This is perhaps commonest when the survey organizers want to see how many respondents mention a particular thing of special interest, but don't want to put words into people's mouths by listing this thing (whatever it is) as a possible answer.
Probing is better described by example, rather than being explained in detail. Let's take a question, and find out how to get more detail out of the answers. We'll start with a very vague question, "What's your opinion of the Prime Minister?" I'm not recommending this as a real question in a survey, as it's too vague to be useful for most purposes; respondents are likely to give all sorts of different answers which simply cannot be compared. Any question as feeble as this should have been weeded out at the piloting stage of a questionnaire - but if you can probe this one, you can probe anything - specially when the respondent avoids giving detailed answers.
Interviewer: What do you think of the Prime Minister?
Respondent: Oh, he's OK, I guess.
As usual, a vague question gets a vague answer. The interviewer now needs to probe, to make the answer more specific, but without biasing the answer by making specific suggestions to the respondent.
General probes
There are some probe questions that can be used in practically any situation, regardless of the previous answer. These include:
"Can you tell me some more about that?"
"Can you give some more details?"
"What do you mean by that?"
"Mmm-hmm."
"Yes?"
"Can you explain that a bit more?"
"In what way?"
"I see..."
Pausing for several seconds, when an answer seems incomplete.
Similar phrases, which are not acceptable in probing, are "Good" and "That's right." Respondents could take these to mean that the interviewer agrees with and/or is satisfied with the content of the answers; this could bias later answers.
Specific probes
Specific probes (unlike the general probes just listed) are related to the last answer given. For example, a suitable probe questions to follow up the answer "Oh, he's OK, I guess" would be "In what ways is he OK?" Notice the use of the plural "ways" not "way": the assumption is that there is more than one answer to be given. To continue the imaginary dialogue:
Interviewer: In what ways would you say he's OK?
Respondent: Well at least he's better than the other man.
Interviewer: Who's "that other man"?
Respondent: The P.M. before him, I can't remember his name.
Notice that probing has two elements: expanding on the answer, and making it clearer. For example, the interviewer had to ask who the "other man" was. It would have been unwise not to ask, and to assume (for example) that the respondent meant the leader of the opposition.
Interviewer: Can you give any examples of how this Prime Minister is better than the previous one?
Respondent (after a long pause): He seems to stick to his promises a bit better.
The interviewer is getting onto slightly dangerous ground here, by asking for examples, as the question being probed simply asked for an opinion, not the facts that supported it. However as the respondent didn't directly answer the first probe question, a more drastic than usual measure is called for, to get the respondent to give a more direct answer. After a slight departure from the question, it's now time to return to it:
Interviewer: Are there any other opinions you have about the Prime Minister?
Respondent: Well I suppose he's good enough to vote for again, when I think of the one before him.
Interviewer: So to summarize your answer, you think the Prime Minister is OK, better than the previous one because he seems to stick to his promises a lot better, and good enough to vote for again. Is there anything else you'd add to that?
Respondent (very bored by now): No, that's all.
The interviewer ends by asking "Is there anything else?" or "Have I left out anything?"
The obvious danger in probing is creating an attitude where none really existed. It's common to find respondents like the one in this example, whose opinions are not coherently formed. The interviewer could have probed further, and the respondent might have obligingly manufactured a detailed opinion on the spot. However, if another interviewer had come along a month or two later asking the same question, and the respondent had forgotten the answer he gave the first time, the second interviewer's probing could construct a totally different answer, building up from whatever aspect the respondent happened to think of first.
Though the above example may make probing seem difficult, I deliberately chose an over-vague question and an uncooperative respondent. Usually the progress flows much more smoothly. The art lies in knowing what to ask, when. A good tactic, if you are a trainee interviewer, is to memorize some of the stock phrases and try probing your friends and family in normal conversation. The longer they take to notice what's happening, the better you are doing it! After a little practice, you'll find it comes quite naturally.
Another skill to learn is when to stop probing. Sometimes respondents start to feel very twitchy when they realize what is happening, and they can see no end to this barrage of detailed questioning. If you detect signs of defensiveness, explain why you are probing: "Often people can't think straight away of the full answers they'd want to give, so I'm trying to help you make sure you don't miss giving part of the answer." Maintaining the right tone of voice while probing will usually help to prevent a defensive reaction.
When recording the answers given in probing, write down each statement as it is made, word for word. It's the convention to separate each probe by a slash. General comments by the interviewer are often helpful, and are usually enclosed in brackets.) Thus the interviewer in the above example would have written:
At least he's better than other guy / i.e. previous PM, can't remember name/ seems to stick to his promises a bit better/ suppose he's good enough to vote for again, when think of previous one. [Gave up probing here - he didn't seem to have much of an opinion at all.]
Though it may seem tedious to write so much, the danger of summarizing the answer on the spot is that the flavour of the answer may be lost.
Probing for specific media
When a radio or TV network has several channels (as with the ABC in Australia or the Radio the Voice of Vietnam) some audience members who use only one channel will give it the generic name. Many of these do not realize there is more than one station. So if you are asking about specific stations, and a respondent answers with the name of the organization, some probing questions are needed, e.g.
Can you name some of your favourite programs or announcers on that station?
As long as the interviewers are well trained in the differences between the stations, they can usually work out which station a respondent is referring to.
Filling in questionnaires
The main principles of completing questionnaires are:
It's also obvious that interviewers should write legibly. It's easy to say this when you're sitting in an office, but when an interview is being done outdoors, in wind and rain, it's not surprising that completed questionnaires are sometimes hard to read. So it's important, when designing a questionnaire, to allow plenty of space for the interviewer to write open-ended answers. Any money saved in the cost of paper for questionnaires is usually more than wasted in extra coding costs.
A good practice is to give interviewers more questionnaires than they will need - about 20% more. Ask interviewers to recopy any questionnaires that will be hard to read, and to send in both the original questionnaire and the copy (with COPY written on it). Then if any discrepancies are found with the neat copy, the untidy original is there to be referred to. When copying questionnaires, interviewers usually copy the words exactly, but sometimes forget to circle codes, tick boxes, etc.