A survey is two separate problems wearing one name. Choosing who to ask is a statistics problem, and it is solved by probability sampling, stratification, clustering and weighting. Working out what an answer means is a cognitive psychology problem, and it is solved by question wording, mode design and pretesting. Most books do one of these well and the other badly, which is why this path alternates between them rather than running the statistics end to end.
If you only want to read published polls critically, start with Herbert Asher's Polling and the public and stop there. It is 222 pages written for citizens rather than statisticians on margin of error, house effects, question wording and who was excluded, and it makes vivid what everything else on the list is trying to prevent. Floyd Fowler's Survey research methods is the compact practical map of the whole process — sampling, mode, questions, interviewers, ethics — in under 200 pages. Our record is an early edition of a book now in its fifth; the structure is stable, but buy current, because the mode chapters have been rewritten around online surveys.
Asking the question
Fowler's Improving survey questions is the narrow follow-on: how to write a question and how to find out whether it worked, through cognitive interviews and field pretests. Short and immediately usable. Bradburn, Sudman and Wansink's Asking questions is the drafting reference, organised by question type — behaviour, attitude, knowledge, sensitive topics — with hundreds of real wordings that failed.
Then the theory those two rest on. Tourangeau, Rips and Rasinski's The psychology of survey response is the intellectual heart of the field: the four-stage model of comprehension, retrieval, judgement and reporting that describes what a respondent actually does. Read it after the practical books, so you recognise the phenomena it explains. Dillman, Smyth and Christian's Internet, Phone, Mail, and Mixed-Mode Surveys is the tailored design method and the standard reference for mode and layout — how a question looks on a screen against on paper, and how to run one survey across several modes without the mode becoming a variable. Our record is the fourth edition, the current one. None of these four assumes any statistics.
Sampling design
The statistics stages do assume something: comfort with probability, expectation and variance. Graham Kalton's Introduction to survey sampling is 96 pages and still the clearest statement of why probability sampling works and what each design buys you. Read it before any textbook.
The core text of the path is Sharon Lohr's, catalogued under the bare display title Sampling — full derivations, real datasets, R and SAS code, and unusually good coverage of nonresponse and nonprobability samples. Our record is the current third edition, 603 pages. Levy and Lemeshow's Sampling of populations is the health and epidemiology counterpart and the better choice if your populations are clinics, households or villages rather than survey panels; our record is an early edition, and the current one adds substantial material on telephone and complex designs.
The standard text and the classical theory
Groves and colleagues' Survey methodology is the standard graduate text, from the Michigan team that largely defined the discipline, and it organises everything else here under one idea: total survey error. Read it once you know both halves it unifies. Our record is the first edition and the second is current.
William Cochran's Sampling techniques is the classic, and still where you go for a derivation nobody else bothers to give — ratio and regression estimators, optimal allocation, systematic sampling. Our catalogue record is the 1953 first edition; the third edition of 1977 is the one universally cited and the one to buy. Särndal, Swensson and Wretman's Model assisted survey sampling is the theoretical summit at 694 pages, unifying the design-based and model-based traditions and serving as the reference for calibration and generalised regression estimation. It is genuinely demanding and is not a first book on anything.
Error, nonresponse, and the data you actually got
Groves's Survey errors and survey costs is the monograph that made total survey error a framework rather than a slogan, by treating each error source alongside what reducing it costs. It is the argument behind every design tradeoff earlier in the path. Heeringa, West and Berglund's Applied Survey Data Analysis is the practical ending: computing variances, fitting regressions and handling weights and strata for complex-design data in real software. This is where most of the errors in published survey analysis are actually made — analysts treating a stratified multistage sample as though it were simple random. Our record is the first edition; a second exists.
The staged order, with lengths and edition notes for each, is on the survey methodology reading path.
