What made the famous 1936 magazine straw poll fail despite millions of responses?
Choose an answer, then check it.
Political Science & Government · Foundations
Public Opinion
On this page 9 sections
In 30 seconds
Public opinion is the distribution of attitudes a population holds about political questions, and polling is the measurement instrument used to estimate it. A well-run survey draws a sample designed to represent the population, asks carefully worded questions, and reports results with a stated margin of error The stated range around a survey estimate within which the true population value most likely falls. Full entry →. Because every step can introduce distortion, reading a poll means reading its methods, not just its headline number.
Why this matters
Poll numbers shape news coverage, campaign strategy, and officials' sense of what the public will accept, yet most people never learn how the numbers are produced. A student who understands sampling, wording effects, and uncertainty can tell a credible estimate from a junk statistic, notice when two polls asking different questions are falsely treated as contradictory, and resist the temptation to read a two-point shift inside the margin of error as a dramatic swing. These are transferable skills for interpreting any statistical claim about people.
The college version
From attitudes to estimates
No one can interview every adult in a country, so pollsters estimate. The founding insight of scientific polling is that a properly drawn sample of roughly a thousand people can represent many millions, provided every member of the population has a known chance of selection. Early twentieth-century failures taught the field this lesson memorably: a magazine straw poll with millions of mailed-in responses famously predicted the wrong presidential winner in 1936 because its sample skewed toward wealthier households, while a much smaller scientific sample got it right. Size matters less than representativeness. Modern pollsters reach respondents by phone, online panels, and mixed methods, then apply weighting Statistical adjustment that aligns a sample's composition with known population characteristics. Full entry → to align the sample with known population characteristics such as age, education, and region, compensating for groups that respond at lower rates.
Uncertainty is part of the number
Every sample-based estimate carries sampling error, conventionally summarized as a margin of error at a stated confidence level. A poll reporting 52 percent support with a margin of three percentage points is claiming the true value most likely sits between 49 and 55. Two candidates separated by one point in such a poll are best described as statistically indistinguishable, and a movement from 52 to 50 across two polls may be noise rather than news. Margins apply to subgroups too, and because subgroup samples are smaller, their error bands are wider, a detail headlines routinely ignore. Beyond sampling error lie harder-to-quantify problems: nonresponse bias Distortion that arises when people who refuse or miss a survey differ systematically from those who take it. Full entry → when the people who decline to participate differ systematically from those who answer, and coverage error when the method cannot reach parts of the population at all.
Wording, order, and the fragility of answers
Reported opinion is sensitive to the instrument. Support for a policy can move substantially depending on whether a question names costs, invokes authority figures, or frames the issue as assistance or as spending. Question order matters as well, since earlier items prime considerations that color later answers. Response options shape results too; offering an explicit undecided choice changes distributions. None of this means opinion is fake. It means many people hold loosely structured attitudes that harden only when an issue becomes salient, so a single question captures one angle of a multi-sided object. Sophisticated consumers of polling compare multiple questions across multiple organizations, watch trends measured with identical wording over time, and treat any lone surprising result as a hypothesis awaiting replication rather than a settled fact.
What polls can and cannot tell government
In a representative system, officials consult opinion but do not simply obey it, and polling itself gives reasons for caution. Measured majorities can be unstable, issue attention is unevenly distributed, and intensity is hard to capture: a poll counts a mild preference and a passionate commitment as the same single response. Election polls add further complications, since they must model who will actually vote, a prediction layered on top of measurement. Used carefully, surveys remain the best available window into what a large public thinks, far better than the loudest voices or a legislator's mailbag. The discipline lies in reading them as estimates with stated methods and uncertainty, and in demanding that anyone citing a poll disclose who conducted it, when, among whom, and with what questions.
Trends, aggregation, and reading polls like a professional
Single surveys make headlines; trends make knowledge. Because each poll carries its own error, analysts prefer repeated measurements with identical wording, where a consistent drift across many readings signals genuine attitude change even when no single pair of numbers differs significantly. Averaging across pollsters helps too, since house effects, the small systematic leans produced by each organization's methods, partially cancel when combined, though averaging cannot rescue a field of surveys sharing one blind spot. Professional consumers also distinguish the population surveyed: all adults, registered voters, and likely voters are different groups whose results should never be casually compared. A practical reading order for any poll story: find the population, the dates, the mode, the wording, and the error margin, and only then look at the percentages everyone else started with.
Eli explains
The same idea, in plain words
Explain it like I’m 10
Suppose you cooked a giant pot of soup and wanted to know how salty it is. You would not drink the whole pot; you would stir well and taste a spoonful. Polling works the same way. If the stirring is done right, about a thousand people can stand in for a whole country. The taste test is never perfect, so pollsters attach a plus-or-minus number that says how far off the spoonful might be. And the way you ask the question changes the answer you get, the way asking whether soup is too salty versus perfectly seasoned nudges people toward different replies. Good polls tell you exactly how they stirred, whom they tasted, and what they asked.
Picture it like this
A poll is a spoonful from a stirred pot: a small, well-mixed sample standing in for something far too large to consume whole.
Where the picture stops working
Soup does not refuse to be tasted, but people decline surveys, and the ones who answer can differ from the ones who do not. Soup also does not change flavor because of how the question was phrased, while human answers do.
Worked example
Two polls about a fictional transit tax seem to conflict. Poll A, from a university center, asks whether respondents support a one-cent sales tax to expand regional bus service, and finds 58 percent in favor, margin of three points. Poll B, sponsored by an anti-tax group, asks whether taxes should rise during a period of high prices, and finds 61 percent opposed. A student comparing them notes the questions measure different objects, checks both samples and sponsors, and looks for a third poll with neutral wording. The lesson: before declaring opinion divided, confirm the surveys actually asked the same question of comparable samples.
Key takeaway
A poll is an estimate built from a sample, a questionnaire, and adjustments, so its headline number is only as trustworthy as the disclosed methods and stated uncertainty behind it.
Quick check
3 questions here, of 5 in this lesson’s practice set. Answers stay hidden until you check.
A survey reports candidate support at 47 percent with a margin of error of four points. Which conclusion respects that uncertainty?
Choose an answer, then check it.
Why do pollsters apply weighting to raw survey responses?
Choose an answer, then check it.
Study tools & related lessonsYou’ll learn to · Common mistakes · Easily confused · Key vocabulary · Related
You’ll learn to
- Define public opinion as a measurable distribution of attitudes rather than a single voice.
- Explain how probability sampling allows a small sample to represent a large population.
- Interpret a margin of error and identify changes that fall inside it.
- Analyze how question wording and order can shift reported results.
Common mistakes
Judging a poll by its sample size alone.
A huge unrepresentative sample can fail spectacularly; representativeness and method matter more than raw count.
Reading movement inside the margin of error as a real shift.
Changes smaller than the error band are statistically indistinguishable from no change.
Comparing results from differently worded questions as if identical.
Wording and order shape answers, so compare trends measured with the same instrument.
Treating one surprising poll as definitive.
Outliers happen by chance; wait for replication across organizations before updating conclusions.
Easily confused
sampling error vs. nonresponse bias
Sampling error is quantifiable randomness from using a subset; nonresponse bias is systematic distortion from who chooses to participate, and it is not fixed by sample size.
opinion measurement vs. opinion formation
Measurement estimates existing attitudes; formation concerns how media, events, and social influence create them, a process studied in the media lesson.
Key vocabulary
- probability sample
- A sample in which every population member has a known chance of selection, allowing valid statistical inference.
- margin of error
- The stated range around a survey estimate within which the true population value most likely falls.
- weighting
- Statistical adjustment that aligns a sample's composition with known population characteristics.
- nonresponse bias
- Distortion that arises when people who refuse or miss a survey differ systematically from those who take it.
- question wording effect
- A shift in measured opinion produced by how a survey item is phrased rather than by real attitude change.
- salience
- The prominence of an issue in people's minds, which affects how firm their expressed opinions are.
Sources & references
- 6.1 The Nature of Public Opinion — American Government 3e (OpenStax) — OpenStax, Rice University
- Election Administration and Voting Survey — U.S. Election Assistance Commission
EliExplains lessons are original prose written from the open, credible references above. See Copyright & Licensing.
Researched 2026-08-25
Educational content only. It is not medical, legal or professional advice. Found an error? Tell us.