Skip to content
Back to search results

Survey — Survey Research

Introduced by
Francis Galton, Charles Booth, and early social statisticians; formalised by Paul Lazarsfeld in the 1940s
Year
Late 19th century; systematic social-science use from 1940s
Type
Quantitative (primarily) or mixed-methods data-collection instrument
Family
Data collection
Font size

What does it do?

A survey is a systematic data-collection method in which a standardised set of questions is posed to a sample of respondents to measure attitudes, behaviours, demographics, or other constructs.

Surveys can be administered via paper, telephone, online platforms, or face-to-face. They are among the most widely used instruments in social, behavioural, health, and educational research because they can reach large, geographically dispersed samples at relatively low cost.

The intuition behind the method

Think of a survey as a snapshot machine: it captures the state of a defined population at a particular moment by asking everyone the same questions in the same way. The key value is standardisation — every respondent receives identical stimuli, so responses can be aggregated and compared. Unlike an in-depth interview, the survey trades depth of understanding for breadth of coverage. A well-designed survey instrument, properly piloted and administered to a representative sample, allows researchers to draw inferences about a much larger population from a manageable number of completed responses.

History

Systematic social surveys emerged in the late nineteenth century through the poverty studies of Charles Booth (1889) and Seebohm Rowntree (1901) in England. Paul Lazarsfeld at Columbia University formalised survey methodology as a social-science tool during the 1940s, establishing sampling theory and attitude-measurement techniques. Rensis Likert introduced his eponymous response scale in 1932. The growth of telephone sampling in the 1970s–1980s and the internet in the 1990s–2000s transformed survey administration; Don Dillman's tailored design method became the dominant practical framework for maximising response rates across modes.

When to use it

Use a survey when you need to describe the distribution of attitudes, behaviours, or characteristics across a defined population, or to test relationships among variables in a large sample. Surveys are especially suited to cross-sectional prevalence studies, needs assessments, programme evaluations, and hypothesis-testing in social or health research. Do not use a survey when the research question requires deep exploration of lived experience (use phenomenology or in-depth interviews instead), when the constructs of interest resist self-report (use observation or physiological measures), or when the target population cannot be reached reliably through any survey mode. Avoid surveys for causal inference without longitudinal or experimental designs.

Limitations and considerations

  • Self-report data are subject to social desirability bias, recall errors, and acquiescence bias.
  • Fixed response options may not capture the full complexity or nuance of respondents' views.
  • Cross-sectional surveys establish association, not causation; longitudinal designs are needed for temporal inference.
  • Low response rates — common in online surveys — introduce non-response bias and threaten generalisability.
  • Question wording effects can substantially distort results if items are poorly crafted.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Reaches large, geographically dispersed samples efficiently at relatively low cost.
  • Standardised questions ensure comparability of responses across respondents and over time.
  • Flexible in mode (paper, phone, online, face-to-face) to suit different populations and budgets.
  • Amenable to statistical generalisation when probability sampling is used.
  • Can simultaneously measure multiple constructs and their inter-relationships.
  • Established psychometric tools (Cronbach's alpha, factor analysis) support rigorous validity and reliability assessment.

Applications

  • National censuses and population health surveys measuring disease prevalence, risk behaviours, and healthcare utilisation.
  • Educational assessment studies examining student engagement, learning outcomes, or faculty satisfaction across institutions.
  • Organisational and HR research on employee attitudes, job satisfaction, and workplace climate.
  • Market research and consumer behaviour studies gauging product preferences and purchase intentions.
  • Political polling and election forecasting tracking voter attitudes and intentions.
  • Programme evaluation studies assessing participant satisfaction and self-reported outcomes following an intervention.

How it works

How it works

1

Specify the research questions, target population, and constructs to be measured. Determine whether the study is cross-sectional (one time-point) or longitudinal (multiple waves). Decide on the sampling strategy — probability sampling (e.g., simple random, stratified, cluster) or non-probability sampling (e.g., convenience, purposive) — and calculate the required sample size to achieve the desired power and precision.

2

Develop the questionnaire by selecting or creating items for each construct. Match item format to construct: Likert-type scales for attitudes, semantic differentials for perceptions, multiple-choice or ranking items for preferences, and open-ended items when exploratory depth is needed. Attend to question wording (avoid double-barrelled, leading, or ambiguous items), response-scale labelling, and questionnaire length. Validate items against established scales where possible.

3

Administer the draft questionnaire to a small convenience sample (typically 10–30 people from the target population) to detect unclear wording, skip-logic errors, excessive completion time, and technical problems (for online surveys). Revise items based on cognitive interviewing feedback and inspect inter-item correlations or factor loadings as preliminary evidence of construct validity.

4

Deploy the finalised survey to the selected sample using the chosen mode (online, postal, telephone, or face-to-face). Maximise response rate through personalised cover letters, reminders, and incentives where appropriate. Track response rates and compare early versus late responders or respondents versus known population characteristics to assess non-response bias.

5

Clean data by handling missing values and out-of-range responses, then score composite scales. Conduct descriptive statistics for each variable, assess reliability (e.g., Cronbach's alpha) for multi-item scales, and apply inferential methods suited to the research questions — frequencies and chi-square for categorical outcomes, t-tests and ANOVA for group comparisons, regression or SEM for explanatory models.

Common pitfalls

  • Ignoring non-response bias — failing to compare respondents with non-respondents on available characteristics means the achieved sample may systematically misrepresent the population.
  • Overly long questionnaires — surveys exceeding 15–20 minutes completion time see sharply higher dropout and lower data quality; prioritise the constructs essential to the research questions.
  • Using unvalidated scales — creating ad hoc items without piloting or psychometric testing produces measures of unknown reliability and validity; prefer established instruments where they exist.
  • Conflating association with causation — reporting survey correlations as causal relationships without appropriate design controls or longitudinal data is a common misinterpretation.
  • Neglecting item order effects — question sequencing can prime respondents; sensitive items placed early can alter responses to later items.

Frequently asked questions

How large does my sample need to be?

Sample size depends on the desired margin of error, confidence level, expected effect size, and analysis plan. For descriptive surveys estimating a proportion, the standard formula is n = z²p(1-p)/e², where z is the z-score for the confidence level, p is the estimated proportion, and e is the margin of error. For analytical surveys testing relationships, an a priori power analysis based on the planned statistical test is preferable. Online calculators (e.g., G*Power) automate these calculations. Convenience-sample sizes common in student research (100–300) are often underpowered for subgroup analyses.

What response rate is acceptable?

There is no universal threshold. Online surveys commonly achieve 20–40%, mail surveys 50–60%, and telephone surveys have declined sharply. The more important question is whether non-response is random or systematic. Even a 20% response rate may be acceptable if you can demonstrate that respondents do not differ from non-respondents on key variables; conversely, a 70% response rate is still problematic if the missing 30% are systematically different.

Can I mix open-ended and closed questions?

Yes, and this is common practice. Closed items yield quantifiable data; open-ended items add contextual depth and allow unexpected responses. However, open items increase completion time substantially and generate text data that must be coded qualitatively. Limit open items to the constructs where fixed options genuinely cannot capture the range of responses.

When should I use a Likert scale versus a visual analogue scale?

Likert scales (typically 5 or 7 labelled response points) are easy to administer and score, and their psychometric properties are well understood. Visual analogue scales (continuous 0–100 lines) offer finer discrimination and are common in pain and health research. Likert scales suit most social and behavioural research; visual analogue scales are preferable when detecting small changes within individuals over time is the priority.

How is a survey different from a structured interview?

In a survey the respondent reads and self-completes the questionnaire without an interviewer present. In a structured interview the same fixed questions are delivered verbally by an interviewer. Surveys reach larger samples at lower cost; structured interviews achieve higher completion rates and can clarify confusing items but are more resource-intensive and introduce interviewer effects.

Related methods

Sources

  1. Dillman, D. A., Smyth, J. D., & Christian, L. M. (2014). Internet, Phone, Mail, and Mixed-Mode Surveys: The Tailored Design Method (4th ed.). Wiley.

    ISBN: 978-1118456149
    Copied.
  2. Fowler, F. J. (2013). Survey Research Methods (5th ed.). Sage.

    ISBN: 978-1452259000
    Copied.