Survey Design for a
Social Attitudes Study

RESEARCH QUESTION

What the Study Set Out to Test

This survey design example follows a sociology research team that plans a household survey on trust in institutions and attitudes toward immigration across a region. The team has funding and a questionnaire draft, but no sampling plan, and a funder that asks how precise the estimates will be for each of four sub-regions.

The need is a design decided before fieldwork: who is sampled, how many interviews are required, how the questionnaire is tested, and how nonresponse will be handled. The survey is meant to describe attitudes in the population, so the plan is built around representativeness and the precision of those descriptions, not around testing a hypothesis.

STUDY DESIGN AND DATA

A Cross-Sectional Survey Planned Before Fieldwork

A cross-sectional household survey with a target of about 1,200 completed interviews. The inputs are a list of small areas with household counts, the draft questionnaire with attitude items rated on agreement scales, and the precision the funder wants for each sub-region. No respondent data exist yet, so the engagement produces the design, not an analysis of responses.

METHOD AND ASSUMPTIONS

Why Stratify, Cluster and Plan for Nonresponse

The sampling frame is checked first for coverage, since households missing from the list cannot be reached by any later correction. Small areas are stratified by sub-region so each has its own precision target. Within each sub-region a sample of areas is selected first and households are then sampled within the selected areas, which keeps fieldwork feasible. Respondents in the same area tend to be more alike than respondents chosen at random, so a clustered sample carries less information per interview: the sample size is set for the precision wanted and then inflated by a design effect, using an assumed intraclass correlation that is stated and varied.

The plan sets out a pretest, with cognitive interviews and a small pilot run by the research team, to find items that respondents read differently from the intended meaning. Because some households will not respond, the plan states in advance how base weights are adjusted for nonresponse, how weights are trimmed and how the resulting loss of precision is reported. Rating items are planned to be summarized as distributions and, for scales, with a reliability check, not treated as measured on an interval scale without justification.

OUTPUT AND INTERPRETATION

What the Design Delivers

The output has a sampling plan with the frame, strata, cluster sizes and selection probabilities, and a sample size table that shows the margin of error for each sub-region under several design effects and response rates. The funder can see directly what a smaller or larger sample would deliver, and the confidence interval widths are stated for the proportions the study cares about most.

The plan also sets out the weighting steps, the response rate to be computed with the standard AAPOR definitions, and the sensitivity checks planned once data arrive, such as comparing weighted and unweighted estimates. The write-up states what the survey can describe for each sub-region and what it cannot, for example that small differences between sub-regions will be too imprecise to interpret.

REPORTING

What the Researcher Received

A written design report that the team can attach to a funding application or ethics submission, with the response rate definitions aligned to AAPOR and statistical wording for the sampling and weighting methods, matched to the output, for the team to check and incorporate. The research question, questionnaire content and conclusions remain with the research team.