Survey Research
Ask a defined group clear questions. Turn their responses into evidence about experiences, beliefs, attitudes, and reported practices.
What you will learn
- Explain what survey evidence can reveal in media and communication research.
- Distinguish survey designs over time and compare modes of administration.
- Plan the main stages of a study and monitor fieldwork and data quality.
- Read a dataset, select an appropriate analysis, and report findings transparently.
Understand: Read Sections 1–4 to connect the question, design, and research process.
Conduct: Use Sections 5–8 to plan collection, fieldwork, and quality checks.
Interpret: Work through Sections 9–12, calculate the example results, and draft your study plan.
What survey research helps you reveal
Survey research gathers information from defined units through a planned process of asking questions, recording responses, and analysing the resulting data. It commonly studies individuals, households, or organisations.
A survey can describe how people use media, what they believe, how they evaluate an institution, or how experiences differ across groups. A sample survey selects part of a population; a census seeks information from every eligible unit. The design and achieved responses shape the conclusions each can support.
Think of a survey as a structured listening process. The questionnaire supplies the questions, while the wider research design determines whom you approach, how you invite them, how answers are recorded, and how you interpret the results. Groves and colleagues [1] explain these connected tasks.
You want to understand how undergraduate students report checking unfamiliar news before sharing it. A survey could examine the frequency of checking, the sources consulted, confidence in making judgements, and differences across settings.
Later in this lesson, an invented two-college example will show how recruitment, response, coding, percentages, and interpretation fit together.
Why surveys matter in communication research
Audience research uses surveys to examine media habits and motives. Political communication researchers ask about public concerns, trust, and participation. Cultivation research can examine people’s perceptions of social reality, while research on the spiral of silence can investigate perceived opinion climates and willingness to speak.
Survey evidence can also connect with other methods. For agenda-setting research, a survey of public concerns can be compared with a separate analysis of the media agenda. Explain how the measures, population, and time periods support that comparison.
Connect research purposes with measures
A survey becomes useful when its measures answer a clear question. Begin with the concepts you want to understand and specify what an answer would help you establish.
Describe
Estimate a characteristic or distribution. For example: how frequently do respondents report checking unfamiliar news?
Compare and examine relationships
Investigate differences among groups or associations between variables. For example: how does reported checking relate to confidence in evaluating news?
Surveys can inform explanatory research when the design and argument support it. An experiment embedded in a survey can randomly assign message versions and compare responses. A one-time observational comparison provides evidence of association that needs to be interpreted in its context.
Distinguish the information you are asking for
On a small screen, swipe across the table.
| Information | What it concerns | Illustrative focus |
|---|---|---|
| Reported behaviour | What people say they did within a defined period. | Checking a story before sharing it during the past month. |
| Belief | What people consider true. | Whether a familiar source is usually accurate. |
| Knowledge | Answers evaluated against a justified criterion. | Recognising the purpose of a correction notice. |
| Attitude | An evaluation, preference, or disposition. | Trust in a specified news organisation. |
| Attribute or circumstance | A characteristic relevant to the study. | Year of study, language used, or access to a device. |
A conceptual definition explains what a construct means. An operational definition states how you will examine it. For example, define news-checking behaviour, then specify the activity, recall period, question, response categories, and coding.
Confidence, knowledge, and checking frequency offer different information. Each can contribute to a project when its role is clear. State whether a variable is an outcome, a predictor, a grouping factor, or relevant background information.
Recognise the measurement level
Nominal categories distinguish kinds, such as a respondent’s selected news source. Ordinal categories add an order, such as “never” through “very often”. Interval measures assume equal differences; ratio measures also have a meaningful zero, as with a count of days of use.
Code numbers according to their meaning. A code of 3 for a news source labels a category; a count of 3 days measures an amount. Individual rating items are commonly ordinal. The treatment of a multi-item score requires a rationale and evidence suited to that score.
Choose a snapshot or study change
The timing of a survey determines which kinds of change you can examine. Decide whether you need a description during one period, a population trend, or evidence about changes within the same people.
Cross-sectional surveys
A cross-sectional survey collects information for a specified period or occasion. Fieldwork may last days or weeks, but the design provides a snapshot. It can describe respondents and examine relationships among their answers during that period.
Repeated surveys: trend, cohort, and panel
Survey the same defined population on successive occasions.
Wave 1: one sampleWave 2: another sampleExample: undergraduates surveyed each academic year.Seek repeated responses from the same sampled people.
Wave 1: P01 · P02 · P03Wave 2: P01 · P02 · P03Example: follow students as their checking practices develop.Define a group through a shared event or starting period.
Students entering in 2025Study that cohort over timeA cohort study can follow the same people as a panel.A repeated cross-sectional survey can show whether a population’s reported checking frequency changes. A panel can examine whether the same students change and how their trajectories differ. These questions require different data structures and analyses.
Plan for attrition when respondents may leave a panel. Maintain appropriate contact arrangements, record participation across waves, and examine how continued participation relates to the measures of interest.
Plan the survey as one research process
Survey design brings together two connected tasks: obtaining relevant measures and reaching the people those measures concern. Develop them alongside one another.
Eight stages to organise the work
- Focus the study: state the research questions and intended contribution.
- Build an information base: review theory, earlier measures, and relevant existing data.
- Define the population and frame: specify the units and how they can enter the study.
- Plan selection and recruitment: justify the design, required responses, and contact strategy.
- Develop the instrument and mode: link questions and procedures with the concepts.
- Pretest and pilot: examine understanding, routing, timing, recruitment, and data handling.
- Conduct and monitor: implement the protocol and record fieldwork outcomes.
- Prepare, analyse, and report: document decisions and present the evidence.
Agree on responsibilities and resources before launch. Specify who will manage recruitment, train interviewers, answer participant queries, secure the data, and approve necessary changes. Prepare an analysis plan early enough to identify the variables and comparisons the instrument must support.
Match the survey mode to the setting
The mode determines how questions reach respondents and how answers return. Choose it through the population’s circumstances, the task, available contact information, accessibility, and resources.
On a small screen, swipe across the table.
| Mode | What it can offer | What to plan for |
|---|---|---|
| Face-to-face | Interviewer support, rapport, and opportunities to present materials. | Travel, scheduling, training, privacy, and consistent administration. |
| Telephone | Interviewer-guided collection without a visit. | Contact coverage, call arrangements, listening burden, and an appropriate spoken format. |
| Web or mobile | Fast distribution, programmed routing, and convenient self-completion. | Connectivity, device display, recruitment, accessibility, privacy, and quality checks. |
| Paper self-completion | Participation without internet access and a familiar reading format. | Distribution, return, legible layout, routing, and accurate data entry. |
| Mixed-mode | Different routes for reaching and supporting the intended participants. | Comparable measurement, mode records, duplicate prevention, and analysis of possible mode differences. |
For example, an online questionnaire with an accessible paper option may help reach students whose connectivity varies. Keep the meaning and routing of the measures aligned and record which mode each respondent used.
Mode effects can arise when presentation or the presence of an interviewer changes responses. Differences between modes may also reflect which people participate through each route. Consider both processes before interpreting a difference.
Mixed-mode collection combines ways of administering a survey. Mixed-methods research deliberately integrates quantitative and qualitative components. State which design you are using and how its parts connect.
These decisions can be developed using the AAPOR survey practice guidance [3]. An online survey can use a probability-based recruitment design or an opt-in approach; explain recruitment separately from the delivery mode.
Two essentials: whom to reach and what to ask
A survey needs a defensible route to participants and a suitable instrument. This section introduces the decisions that connect Sampling and Questionnaire Design with the overall study.
Sampling: define the reach of the evidence
The target population is the group your question concerns. The sampling frame is the operational source from which units can be selected. The selected sample, invited units, eligible cases, and achieved respondents describe different stages; keep their numbers and definitions clear.
In probability sampling, eligible units have known, non-zero selection probabilities. Equal probabilities are one possibility. In nonprobability designs, participation or selection follows other processes, such as convenience access, referrals, or voluntary opt-in. Explain the actual procedure and the assumptions behind any wider inference.
Plan sample size through precision, intended comparisons, expected response, design effects, analytic requirements, and resources. Sample size planning concerns how well a design can estimate or detect a quantity; the underlying effect or relationship is a separate matter.
The Sampling chapter develops selection methods, frame assessment, sample-size reasoning, and implications for inference in detail.
Questionnaire design: connect concepts and answers
Link each question with a research objective. Define the concept, recall period, relevant respondents, response format, and intended coding. Use language suited to the participants and test interpretation in the languages offered.
For the running example, an item might ask: “During the past 30 days, how often did you check an unfamiliar news story before sharing it?” The design also needs a route for respondents who shared no such stories during that period.
Open questions let people answer in their own words; closed questions offer specified options. Plan the format, order, filtering, and coding through the information needed. Pretest understanding and pilot the full process, including recruitment, display, routing, timing, and data export. Pew Research Center’s guide [4] explains important wording decisions.
The Questionnaire Design chapter develops item construction, response options, order, scales, translation, and pretesting more fully.
Conduct fieldwork with a clear protocol
Fieldwork turns the plan into encounters with participants. Consistent procedures and careful records help you understand what happened between selection and the final dataset.
Prepare recruitment and administration
Test the invitation, consent information, accessibility, contact arrangements, survey links, routing, and data export. Confirm the instrument version, fieldwork dates, responsibilities, and a procedure for resolving problems.
Where interviewers are involved, train them to follow the script, record answers accurately, use agreed neutral clarification, and protect privacy. Rehearse difficult situations. Keep a record of any approved departures from the protocol.
Plan reminders and alternative contact times or modes that fit the study. Keep participation voluntary and provide a clear way to decline. In an educational setting, recruitment should help students decide freely about participation.
Record what happened to the selected cases
Maintain disposition records: completed interviews, partial responses, refusals, noncontacts, ineligible cases, and cases whose eligibility remains uncertain. Record the meaning of each status and apply it consistently.
Separate unit nonresponse, where a selected eligible unit supplies no usable interview under the study’s rule, from item nonresponse, where a participant leaves particular questions unanswered. A substantive “don’t know” response may carry information relevant to the question.
Use an appropriate outcome-rate definition and disclose the denominator. AAPOR’s Standard Definitions [5] distinguishes response, cooperation, and other rates. For an unrestricted public link, the number of people who could have encountered the invitation is often unknown; report the recruitment process and the outcomes you can establish.
Protect participation and data
Obtain required ethics approval or the appropriate institutional determination. Explain purpose, voluntary participation, recording, foreseeable risks, data use, and applicable withdrawal arrangements. Collect identifiers only where justified and store contact details separately when feasible.
Review who can access response data and how long it will be retained. For linked panel records, explain the linkage and protect the key. Plan how findings will be shared without exposing participants.
Understand total survey error
Survey quality depends on several connected processes. A total survey error perspective examines how the population, participation, measurement, and processing stages can affect the final estimate.
Investigate nonresponse through relevant evidence
Nonresponse bias can arise when participation is related to the characteristics being estimated. A study of news checking could overrepresent highly interested students if they are more willing to respond.
Examine available frame information, participation patterns, follow-up evidence, or suitable benchmarks. A response rate describes an important fieldwork outcome; its implications depend on who participated and what the survey estimates. AAPOR’s response-rate overview [7] discusses this relationship.
Examine measurement quality
Validity concerns evidence for the intended interpretation and use of a measure. If you want checking behaviour, examine whether the questions capture the specified practices and recall period.
Reliability concerns consistency under conditions relevant to the measure. Repeated-measure evidence assumes an appropriate interval and consideration of real change; multi-item consistency addresses a different aspect. Explain which evidence fits the instrument.
Recall difficulties and social desirability can shape answers. Clear reference periods, suitable administration, privacy, and pretesting can help address these issues. Record what the design can establish and where uncertainty remains.
Code, check, analyse, and report
A clear analysis begins with a documented dataset. Preserve the original responses, record transformations, and connect each result with its variables and denominator.
Build a traceable data file
Use a codebook with variable names, question text, response codes, measurement levels, missing-value rules, and derived-score instructions. In a cross-sectional individual survey, one row commonly represents one respondent. Repeated-wave data also need a stable participant ID and a wave identifier.
Check duplicates, eligibility, ranges, skip patterns, missingness, and data-entry or export problems. Distinguish an unanswered item from an inapplicable item or a valid “don’t know” response. Document exclusions and derived categories.
Use quality indicators through prespecified or justified rules. Review unusual patterns and conduct sensitivity checks where needed. Retain a decision log so the effect of cleaning can be assessed.
Describe the achieved sample first
Report participation and the distributions relevant to the question. Show the number of usable responses for each result. Frequencies and percentages suit categorical measures; appropriate summaries of numerical variables can include means, medians, spread, and distributions.
Weighting assigns different contributions to responses to reflect features of the design or justified adjustments. State the weight’s purpose, construction, and assumptions. Report unweighted case counts, distinguish weighted estimates, and use analysis methods that reflect the survey design.
Select comparisons through the question and data
On a small screen, swipe across the table.
| Question | Possible analysis | Design and measurement check |
|---|---|---|
| How common is a response? | Frequency, percentage, and a justified uncertainty estimate. | Define the denominator, weighting, and scope of inference. |
| Do two groups differ on a numerical score? | A suitable two-group comparison, such as Welch’s t-test. | Check independence, measurement, distribution, and survey design. |
| Do several groups differ on a score? | A suitable ANOVA or alternative. | Consider variance assumptions, dependence, and planned comparisons. |
| Are categorical responses associated? | Cross-tabulation and an appropriate chi-square, exact, or design-adjusted test. | Check cell counts and the sampling design. |
| Are two measures related? | Pearson correlation for an appropriate linear relationship; Spearman correlation for suitable ranked or monotonic relationships. | Inspect the data, measurement, outliers, and relationship form. |
| Do the same people change? | A paired or repeated-measures method suited to the outcome. | Preserve respondent linkage and account for attrition. |
Report the size and uncertainty of differences or relationships alongside any significance tests. Interpret them through the design and the practical meaning of the measure. Multivariable analysis can examine adjusted associations when the model and data support it.
Make the report inspectable
Describe the sponsor, purpose, population, frame, selection, recruitment, modes, languages, dates, instrument, participation, processing, weighting, analysis, and limitations. Share the questionnaire and permitted documentation. AAPOR’s Transparency Initiative [8] provides disclosure guidance.
Use accurate records and acknowledge sources. Keep decisions about corrections, exclusions, and analysis visible in the research record.
A worked survey example
Follow this example from participation to a result. The numbers are invented for teaching, and all estimates below are unweighted descriptions of the example respondents.
Question: How frequently do undergraduate students report checking unfamiliar news before sharing it, and how do the responses compare between two colleges?
Design: A cross-sectional survey during a defined two-week period. The protocol specifies the eligible population, selection and recruitment procedure, survey modes, and privacy arrangements.
Measure: A checking-frequency item uses a defined recall period. For this example, the 120 complete respondents all shared an unfamiliar news story during that period and supplied a usable answer.
Step 1: describe participation
Suppose 200 selected invitees are all known to be eligible. There are 120 complete responses, 80 refusals or noncontacts, no partial interviews, and no cases with unknown eligibility.
Complete-response rate in this simplified setup: 120 ÷ 200 × 100 = 60%.
The calculation depends on the stated case outcomes. In a real study, use the appropriate definition for partial responses, eligibility, and other dispositions.
Step 2: describe the responses
The research plan groups “often” and “very often” as frequent checking. The other three substantive frequency categories form the comparison group. Explain this grouping in the codebook and preserve the original responses.
On a small screen, swipe across the table.
| Response | College A | College B | Total |
|---|---|---|---|
| Often / very often | 48 | 36 | 84 |
| Never / rarely / sometimes | 12 | 24 | 36 |
| Usable responses | 60 | 60 | 120 |
Reveal the calculations
College A: 48 ÷ 60 × 100 = 80%.
College B: 36 ÷ 60 × 100 = 60%.
Overall: 84 ÷ 120 × 100 = 70% of the example respondents.
Difference: 80% − 60% = 20 percentage points.
Keep the case counts alongside the percentages. A percentage-point difference describes the subtraction of two percentages; a relative percentage change uses a different denominator.
Step 3: write the claim at the right level
Among the 120 respondents in this teaching example, 84 (70%) reported checking unfamiliar news often or very often before sharing it. The corresponding proportions were 80% in College A and 60% in College B, a difference of 20 percentage points.
To assess wider relevance, examine the sampling design, population sizes, participation, measurement, and any required weighting. The college comparison describes responses during the study period.
What changes if four respondents skip a different item?
The checking-frequency result still has 120 usable answers. A result for the skipped item may have 116, depending on its eligibility and coding rules. Report the denominator for that result and explain the missing responses.
Learn with Dr. USP’s survey lecture
Use the lecture alongside the chapter. Pause at a section, apply the idea to your project, and return to your protocol with a clearer decision.
Watch the complete survey research lecture on YouTube ↗
A guided route through the lecture
The links begin near the relevant discussion. Questionnaire and sampling passages introduce the connections explored more fully in their dedicated chapters.
Rehearse a survey protocol
- Frame: choose one media-research question and define its population and concepts.
- Design: select a time design and mode; state the essential recruitment and measurement decisions.
- Pilot: with people who agree to a classroom rehearsal, test the invitation, understanding, routing, and data export.
- Prepare: create a small invented dataset with a deliberate missing value and one coding error; document how you handle them.
- Report: write a methods paragraph, one results table, and a statement of the claim’s scope.
A live research project should follow its institutional ethics and review requirements.
Explain your answer before opening the discussion
1. What makes a survey useful for a research question?
The question calls for information people can meaningfully provide, the recruitment and measurement design suits the intended population, and the analysis can connect those answers with a defensible conclusion.
2. When would a panel be useful?
Use a panel when the question concerns change within the same people or units. Plan linkage, recontact, attrition assessment, and repeated-measures analysis.
3. What do mode and recruitment each tell you?
Mode explains how questions are administered. Recruitment explains how units enter the study. Together they help you assess access, participation, measurement, and the scope of inference.
4. Why examine who did not respond?
Participation may be related to the characteristic being estimated. Use suitable evidence to assess whether the achieved respondents differ in ways relevant to the result.
5. How can answers change without showing poor reliability?
People’s circumstances, knowledge, or attitudes can change. Interpret consistency under conditions appropriate to the measure, including the time interval and the possibility of real change.
6. What information should accompany a percentage?
State what was measured, the response categories or grouping, the denominator, weighting status, population or respondent group, period, and any justified uncertainty estimate.
A glossary and a reading route
Use these terms to explain the survey you intend to conduct. Connect each definition with a practical decision.
- Target population
- The group or set of units the research question concerns.
- Sampling frame
- The operational source through which units can be identified or selected.
- Survey mode
- The way questions are administered and responses recorded.
- Operationalisation
- Specifying how a concept will be examined or measured.
- Panel
- A design seeking repeated information from the same sampled units.
- Attrition
- Loss of participating units across waves.
- Disposition
- A recorded outcome for a sampled or approached case.
- Unit nonresponse
- Failure to obtain a usable interview from a selected eligible unit under the study’s rule.
- Item nonresponse
- Missing answers to particular questions within an interview.
- Weight
- A value determining a response’s contribution to an estimate, based on a stated design or adjustment.
- Total survey error
- A framework for examining error across selection, participation, measurement, and processing.
Core lecture
Dr. Uma Shankar Pandey. How to Do Survey Research: USP Lectures. Watch the lecture ↗. The lesson develops the lecture’s overall research process with worked examples and methodological clarifications.
Sources and further reading
- Groves, R. M., Fowler, F. J., Jr., Couper, M. P., Lepkowski, J. M., Singer, E., & Tourangeau, R. (2009). Survey methodology (2nd ed.). Wiley. Publisher’s record ↗. A guide to the complete survey process.
- National Longitudinal Surveys, U.S. Bureau of Labor Statistics. What are longitudinal data? Read the guide ↗. Compare within-unit change with repeated cross-sectional evidence.
- American Association for Public Opinion Research. Best practices for survey research. Read the guidance ↗. Explore decisions across planning, collection, and analysis.
- Pew Research Center. Writing survey questions. Read the methods guide ↗. A companion resource for the Questionnaire Design chapter.
- American Association for Public Opinion Research. (2023). Standard definitions: Final dispositions of case codes and outcome rates for surveys (10th ed.). Definitions and report ↗. Use an outcome-rate definition suited to the design.
- Pew Research Center. U.S. survey methodology. Read the overview ↗. Explore the total survey error perspective.
- American Association for Public Opinion Research. Response rates calculator: Response rates—An overview. Read the explanation and resources ↗. Examine participation outcomes and survey quality.
- American Association for Public Opinion Research. Transparency Initiative. Explore disclosure guidance ↗. Plan the information that should accompany findings.
Take one research question further
Explore Cultivation Theory to connect survey measures with perceptions of social reality. Use Content Analysis when your project also needs evidence about the messages people encounter. Questionnaire Design and Sampling develop the specialised decisions introduced in this lesson.