Field Survey: Protocols Development

Introduction

Survey research is the methodological field within the social sciences concerned with the systematic collection and analysis of information from a subset of individuals or groups of persons chosen randomly from a population. Technically speaking it involves the following steps: (1) question wording, (2) structuring of the questionnaire, (3) sampling, (4) interviewing, (5) coding, (6) reporting.

Types of Surveys

There are three main types of surveys, depending on how information is collected: face-to-face, mail, and telephone surveys. Each of these collection methods presents advantages and disadvantages (Groves, 1979; Backstrom & Hursh, 1986). The face-to-face method is the most efficient method for interviewing people from difficult-to-reach groups. Another advantage is that the researcher has more control over respondents. This allows for longer interviews and increases the probability that the interview will be completed. Face-to-face interviews also allow for the use of visual aids. Their biggest disadvantage is their cost and the fact in large cities it is often difficult to get access to where people live. Telephone surveys can avoid this problem; moreover, they are much cheaper: with current computer programs, a single person can randomly select respondents, interview them, and code the answers directly into the computer (Saris, 1991). The drawback of this method is that interviewers have less control over respondents than in face-to-face interviews. This sets limits on the duration of interviews. The cheapest survey method is the mail survey. It is particularly suited for topics that are not very complex. Its main drawbacks are the very low response rate, which can introduce bias in the results, and the absolute lack of control over the person who fills out the questionnaire.

Comparison with Other Data-Gathering Methods

Compared to ethnographic and experimental methods, survey research presents advantages in terms of external validity (ability to extrapolate survey results to the target population) and disadvantages in terms of internal validity (ability to draw causal conclusions from observed associations) (Cook & Campbell, 1972; Kish, 1987). Indeed, the correct implementation of probability sampling methods allows estimation with a given margin of error confidence, of intervals of varying precision for common statistics, like the mean or percentages, for a particular population. This sort of inferential precision is not possible with experimental and ethnographic methods. On the other hand, the general lack of randomization in the assignment of treatments differentiates survey research from experimental research, but not from ethnographic research, and complicates the process of attaching causal meaning to measured associations. The use of statistical controls in survey data analysis mitigates this problem but, since the number of potential statistical controls is infinite, this makes the statistical conclusions of survey research highly sensitive to the theoretical soundness that guides the selection of control variables and which thus determines the inclusion of particular questions in a questionnaire.

Another threat to the internal validity of survey research is that surveys are generally cross-sectional, which raises the problem of determining the causal order characterizing a particular association. Panel surveys and retrospective questions in cross-sectional surveys reduce this problem, but only at the cost of complex and not always easy to estimate statistical models in the case of Panel studies and lower reliability [Page 414]of the answers when one relies on retrospective information.

...