survey

survey, social survey At first, a survey was any systematic collection of facts about a defined social group, and the term is still used in this way. The term survey is therefore not necessarily synonymous with ‘questionnaire survey’, since other methods of data collection (such as observation of behaviour) may be employed in a survey. In practice, however, most sociological surveys are based on written questionnaires. More precisely, the term usually refers to data collections that employ both interviewing and sampling to produce quantitative data-sets, amenable to computer-based analysis. Sampling and interviewing are employed in many other research designs. It is the combination of the two that has led to the social survey, or sample survey, becoming the most important single type of social research, used by all the social sciences, market research, and opinion polls.

Surveys can be used to provide descriptive statistics for national, regional, or local populations; to examine the clustering of social phenomena; to identify the social location and characteristics of subgroups for more intensive follow-up case-study research; and to analyse causal processes and test explanations. In recent years sociological survey analysis has been greatly extended to include the sophisticated multivariate modelling techniques that are common in econometrics. One of the main attractions of the sample survey for both policy research and theoretical research is its transparency and accountability: methods and procedures can be made visible and accessible to other parties, unlike research designs that depend heavily on the contribution of individual researchers. The key disadvantage is that surveys normally use structured questionnaires, which constrain an enquiry to paths fixed at the start of fieldwork. Other criticisms which are sometimes levelled at surveys are that numerical variables rarely provide adequate operationalizations of sociological constructs; the highly asymmetric power relation between researcher and interviewee is detrimental to the quality of the data collected; they provide a false aura of objectivity which makes their results vulnerable to political manipulation. Many of these criticisms can be overcome by good survey design and implementation.

Surveys can collect information on individuals, roles, social networks, social groups such as households or families, organizations such as schools, workplaces, or companies. In most cases the information is provided by individuals, but the information collected may be about any social unit of interest, with larger and more complex units requiring multiple interviews to avoid the information limitations or bias of a single informant. Surveys are used to study poverty, social stratification, social mobility, political orientations and participation, work and employment, and virtually all the issues addressed by sociologists and other social scientists.

Survey interviews may be personal, postal, or conducted by telephone. Telephone surveys are particularly common in the United States, where most households have telephones, and the size of the country makes face-to-face interviewing of a nationally representative sample of the population prohibitively expensive (see P. J. Lavrakas , Telephone Survey Methods, 1987
). Major surveys, especially national surveys, are now carried out by specialist fieldwork agencies or national research institutes that have the necessary resources for questionnaire design, sample design, sample selection from available registers or other sampling frames, fieldwork planning and supervision, training and debriefing of interviewers, coding of completed questionnaires, consistency checks, and editing of the resulting data-tape. Such agencies often become centres for methodological research on sampling, survey techniques, and design.

CAPI (Computer Assisted Personal Interviewing) and CATI (Computer Assisted Telephone Interviewing) techniques are increasingly used. These involve the interviewer in coding the answers given by each respondent directly on to a data-tape or file, using a personal or laptop computer, at the time of interview. This saves both time and money in the overall survey process, but does mean that particular care must be taken over matters of questionnaire design at the outset, to ensure that there are no filtering or other errors in the final interview schedule.

Most academic research consists of ad hoc surveys, carried out on a one-off basis, to address defined theoretical and other issues. Ad hoc surveys often employ the smallest sample size necessary to achieve representativeness, typically 2,000 respondents for a national survey, relying heavily on statistical inference to generalize the results to the target population. National opinion polls are carried out on a regular basis, but again employ the smallest sample size necessary to achieve representativeness, and necessitate the use of tests of statistical significance. With the change of emphasis from administrative records and registers to interview surveys as the basis of official statistics, a great variety of regular surveys are also carried out by national governments. These involve quite different orders of magnitude to the typical ad hoc survey, with national samples of 5,000–250,000 per year; samples as large as this begin to make tests of statistical significance as redundant as they are with census data, except when data subsets are analysed. In effect, the variety of surveys is now so wide that it ceases to be a homogeneous category of social research. Regular surveys may involve repeat cross-sectional surveys at defined intervals, such as annually every spring, or be carried out on the basis of continuous year-round interviewing so as to smooth out any seasonal variations in the activities covered. The USA Current Population Survey (CPS) and some equivalent Labour Force Surveys (LFS) employ rotating sample designs which offer many of the advantages of data from panel studies for measuring changes over time in the phenomena under study.

Surveys make demands on respondents and require their active co-operation to be successful. They require that respondents adopt the role of interviewee, in effect the role of citizen and commentator on their own lives, and the lives of those around them. This interviewee role has developed over decades in Western industrial societies, and there is increasing recognition that it is not universally understood or accepted in other cultures. For example, in some cultures it would be impolite to express open disagreement with the perceived or expected views of an interviewer, thus invalidating the invitation for respondents to express their own views. Surveys also make information demands that can be difficult to meet in societies where literacy and personal record-keeping are less widespread–so that even dates of birth may be difficult to recall accurately. New techniques of data collection are being developed for surveys in Third World countries and societies with different cultures and social conventions.

There are numerous textbooks on how to design and conduct surveys. Catherine Marsh's The Survey Method (1982) stands out as an elegant defence of the technique against critics who object that surveys are invariably superficial and merely descriptive.