Data collection modes
The choice of data collection mode or modes will depend upon the research themes and objectives (target population, sensitive subject and/or sensitive population), methodological and logistical constraints (existence or not of a sampling frame, survey locations and geographic coverage, respondent characteristics and competencies), budget and time constraints.
Main data collection media and modes
Questionnaires are always filled out either on paper or on a computer screen. However, they can be transmitted to respondents in one of two ways—through written or spoken words—and may or may not involve an interviewer. Questionnaires administered by an interviewer are called “hetero-administered”; those completed by respondents themselves are “self-administered.” A data collection mode is defined by the physical medium, the communication channel, and the presence or not of an interviewer. In computer surveys, questionnaires may be administered online or, at the local level, on a computer or tablet not connected to the internet.
Respondents answer the questionnaire themselves, using either paper and pencil (Paper And Pencil Interview or PAPI) or a computer (computer screen, tablet, smartphone, etc., a method called Computer-Assisted Self-Interview or CASI). In these cases, the procedure is visual. But on the computer it can also be a mix of visual and auditory: respondents can read the questionnaire on the screen while it is also read aloud by the computer. In a few cases, listening alone is required, as in CASI and telephone questionnaires: the questionnaire is read aloud by a computer and the respondent answers either by pressing telephone buttons or by saying “yes” or “no.” This last way of proceeding is not used for INED surveys or surveys by any other French public statistical body.
Here the questionnaire is read to the respondent by an interviewer and his/her answers are recorded either on paper or computer (Computer-Assisted Personal Interview or CAPI). The telephone may also be used (Computer Assisted Telephone Interview or CATI).
Touch tablets and smartphones are rarely used in large-scale quantitative surveys. The ELIPSS programme (Longitudinal Internet Social Science Survey) run by Science Po is one of the few self-administered questionnaire programmes that exists in France; INED is an active ELIPSS participant. Different data collection methods may be combined to make it easier for respondents to participate, to improve sample representativeness and increase response rates. For example, a questionnaire can be filled out online but also on paper. However, not all combinations are possible or desirable because filling out a questionnaire oneself is not the same as responding to questions asked by an interviewer.
Each data collection method is associated with specific questionnaires and follow-up methods. These in turn must be tested (clarity, filter questions, space provided for responses, orientation within the questionnaire, type and frequency of reminders, etc.).
Specific features of longitudinal surveys
Increasing use is being made of longitudinal surveys, which may be either retrospective (e.g., life course surveys) or prospective (cohort follow-up, multi-wave surveys). The various experiments conducted at INED have pointed up the specific methodological and/or ethical features of these surveys, most of which are related to time management (in the questioning process and the survey protocol).
Retrospective surveys: available tools
In the 1980s, at INED and elsewhere, specific tools were developed to collect quantitative life course information. As experience in conducting this type of survey increased, these tools were improved and enhanced. Use of Life Event History Calendars improves respondent recall and facilitates the recording of individual life events. Such tools can be adapted to different types of research questions and have been tested in a wide variety of fields on extremely diverse populations (in both Northern and Southern countries).
Interviewers have to be specially trained to administer this kind of questionnaire, and the interviews tend to be quite long. Their basic structure is simple, readily adaptable and effective in collecting exhaustive, high quality retrospective information. However, when the questionnaire is being developed, it is important to formulate response categories that are not time-dependent; and that correspond to the different periods (or contexts) respondents proceed through in the course of their lives.
Prospective surveys: limiting attrition
Keeping track of respondents from one survey wave to the next is a central component of prospective surveys. One of the main problems in longitudinal follow-up is attrition (when respondents either cease to participate or can no longer be contacted for some reason). It often happens that specific categories drop out of the survey, undermining sample representativeness and therefore the reliability of survey findings. It is therefore extremely important for survey designers to maintain ties and good relations with respondents—by means of relatively frequent (though not excessive) contacts.
There are two points here, one logistic—it is crucial to keep track of any changes in address, telephone number, email address that may have occurred between two survey waves—the other methodological and indeed ethical: it is crucial to keep respondents interested and willing to collaborate. Several types of respondent contact can be used to this effect.
Likewise, to minimize the number of survey dropouts, survey designers usually collect contact information on “intermediaries" who can provide the respondent’s contact information should the tie be broken.
Because the quality of respondent tracking and follow-up impacts so strongly on the quality of the cohort and the data collected, it is important to allocate time and resources for long-term respondent follow-up.
Respondents’ consent and confidentiality of responses:
Since the identity of the person providing information and responses is necessarily known, the imperatives of confidentiality and data protection are particularly important and must be carefully thought out and planned for. Survey protocols must be reported to the CNIL (see above). Furthermore, there are specific rules to abide by when informing respondents about the survey. They have to be clearly informed of survey objectives and stages, what their information is to be used for, and who will have access to it.
Respondents’ informed consent is of course required, but they are also free to change their minds and withdraw that consent at any time. If the project is changed in significant ways, respondent consent must be requested a second time.
The first survey wave, where contact is established, is decisive. If the survey is not self-administered, it is preferable for that first contact to be a face-to-face interview, as this seems to be the best way to maintain high respondent participation rates in the following waves. Further interviews can then be conducted by telephone.