← Introduction to Political Science outline

Week 12 · Chapter

Chapter 12 — Public Opinion, Political Behavior & the Media

Introduction to Political Science Free sample edition

The module's primary reading · ~40–50 minutes · Objective 6 (explain American government and political participation — federalism, separation of powers, and the U.S. Constitution as one case of the survey's concepts; parties, elections, and voting systems and their effects; public opinion, polling, political behavior, and the media — reading election and polling data for what it documents and what it does not settle), Objective 1 (describe the subfields and methods of political science — political theory and philosophy, comparative politics, international relations, American government, and political methodology — and analyze a political text or dataset with the discipline's core tools: concept application, argument analysis, evidence evaluation, the comparative method, and the empirical/normative distinction) · Read it first this week; the lecture, the slides, the tutorial, the quiz, and the workshop all build on it.

In plain language

This week is about what "the public thinks" really means and how anyone finds out. Public opinion (the spread of views across a whole group of people) is almost always measured by asking a sample (a smaller group who actually answer). You will learn when a sample can speak for millions and when it cannot, and how big its margin of error (the plus-or-minus range a poll reports) is. Then you will read one real survey of 5,119 U.S. adults, look at how people form their views and who turns out to vote, and weigh three claims about the media. You will not be told whether lawmakers should follow the polls; you will be shown the strongest case on each side.

Why you already care about this

Picture the dining hall on your campus. Dining services wants to know whether students want it open later on weekends. Every number in this story is illustrative, invented for teaching. Dining services asks the question in two ways. First, it draws 400 students at random from the full enrollment list and asks each of them. Second, it posts a click poll on the dining app, and anyone who opens the app can tap an answer.

The two answers disagree. The random sample says 58 percent of students favor later hours. The click poll says 70 percent. The click poll also has far more answers, 2,000 of them. So which number would you take to the dean? Most people's first instinct is to trust the bigger pile. This chapter shows you why that instinct is wrong, and what to trust instead.

The same question runs through all of politics. A news story says "most Americans" believe something, a campaign says "the people" are behind it, and a council member says the phone calls are running against a plan. Each is a claim about public opinion. This week asks how a poll of a thousand people can speak for millions, and when it cannot.

By the end of this chapter you will be able to:

tell a random sample from a self-selected one, and say why only the first can carry a margin of error;
check a poll for the four facts a trustworthy poll reports;
estimate a margin of error with the approximation, and explain why a real poll's stated margin is the figure you cite;
explain what the margin leaves out, and why a bigger sample fixes precision, not bias;
read a real poll with the data scaffold and end on correlation or cause;
describe socialization, turnout patterns, and three media effects, each with its limits;
state the strongest case for the delegate view and for the trustee view of representation.

Carry this line with you: bigger fixes precision, not bias.

Where we are

Already covered: Week 1 gave you the empirical/normative sort and evidence evaluation. Week 5 made a free press part of liberal democracy. Week 8 added the three-layer read. Week 11 opened the participation unit and read your first real data, the 2024 UK election, with the data scaffold.
This week: public opinion and sampling, the margin of error and what it leaves out, a real Pew Research Center poll, political socialization and turnout, media effects, and the delegate-or-trustee question.
How they connect: Week 11 asked how votes become seats; this week asks how opinions are measured before and between elections, using the same data scaffold.
Where we're going next: Week 13 begins the comparative unit, where you will read governance indices with the same questions you use on a poll this week.

Week 1 taught you that an empirical claim says what is and can be checked against evidence, while a normative claim says what ought to be and is argued from values. A poll is an empirical tool: it reports what people said. Whether officials ought to act on what people said is a normative question, and you will keep the two apart all week. Week 11 gave you the data scaffold, five questions for any piece of data: what is measured, over whom and when, what it shows, what it does not show, and whether it shows correlation or cause. You will run that scaffold on a poll in Section 5. Week 8 gave you the three layers: the design layer is what the rules say, the practice layer is what actually happens, and the verdict layer is what ought to happen.

1. Sample the population: random versus self-selected

Think about tasting a pot of soup. You do not drink the whole pot to know whether it needs salt. You stir it well and taste one spoonful. The spoonful works only because you stirred, so every part of the pot had a fair chance to end up on the spoon. If you skimmed the top without stirring, you would taste only what floated up.

A poll works the same way. The population is the whole group a poll wants to describe, such as all U.S. adults or every student enrolled at one school. The sample is the smaller group of people who are actually asked. Public opinion is the distribution of attitudes across a whole population on a political question, usually measured by asking a sample rather than everyone. Notice the word distribution. Public opinion is not one number; it is how views spread out across people.

Two kinds of sample matter this week.

A random sample is a sample drawn so that every member of the population has a known, non-zero chance of being selected, which is what lets a poll carry a margin of error.
A self-selected sample is a sample whose members choose to take part, as in a click poll or a call-in poll; because no one has a known chance of selection, it carries no margin of error at any size.

The stirred spoonful is the random sample. The skimmed top is the self-selected one.

Worked example — the dining-hours distribution. Dining services' random sample drew 400 students from the full enrollment list, so every enrolled student had a known chance of being picked. Of the 400, 232 favor later weekend hours, 120 oppose, and 48 are unsure. Check the count first: 232 + 120 + 48 = 400. Now turn each count into a share of the people who answered:

Favor: 232 ÷ 400 × 100 = 58 percent.
Oppose: 120 ÷ 400 × 100 = 30 percent.
Unsure: 48 ÷ 400 × 100 = 12 percent.

Check the shares: 58 + 30 + 12 = 100. In words: the student body's opinion, as this sample measures it, is a spread of 58, 30, and 12 percent, not a single verdict. A dean deciding the hours would see a majority in favor and a real minority opposed.

Worked example — the click poll. On the dining app, 2,000 users chose to answer, and 1,400 of them favor later hours. The share is 1,400 ÷ 2,000 × 100 = 70 percent. The click poll has 2,000 ÷ 400 = 5 times the responses of the random sample. It still carries no margin of error at all, because no one had a known chance of selection. The students who open the dining app late on a weekend and bother to tap an answer may be the very students who want later hours. That is the skimmed top of the pot.

Misconception check — the bigger pile. Students often think the click poll is better because it has five times as many answers. That is tempting, because more answers feel like more evidence. But size cannot repair the way people got into the sample. A self-selected sample of 2,000, or of 200,000, still tells you only about the people who chose to answer. A random sample of 400 tells you about the whole student body, within a margin you can state.

Memory hook: stir the pot before you taste it.

2. Check the four facts a trustworthy poll reports

Before you trust a used car, you check a few things: the mileage, the year, the accident record, and who is selling it. A seller who will not show you any of them has told you something already. A poll is the same. A trustworthy poll always reports four facts, so you can judge it yourself.

The sample size: how many people answered.
The field dates: the days on which a poll's interviews were carried out, which tell a reader when the snapshot of opinion was taken.
The sampling method: the procedure a pollster used to choose who was asked, such as a random draw from a list or a standing survey panel, which a trustworthy poll describes.
The stated margin: the margin of error a pollster reports for its own survey, usually a little larger than the rule-of-thumb approximation because of weighting and design; it is the figure a reader cites.

Each fact answers a question you would otherwise have to guess at. The size tells you how precise the poll can be. The dates tell you when the snapshot was taken, since opinion can move. The method tells you whether the sample was random or self-selected. The stated margin tells you how much room to leave around each number.

Worked example — the four-fact check on two polls. Run the check on the dining polls. The random sample reports its size (400), its method (a random draw from the full enrollment list), and, as you will compute in Section 3, a margin of about 5 percentage points; dining services would add its field dates. The click poll reports a size (2,000) and a method (anyone who opened the app), and the method itself tells you that no margin can be stated. In words: the four-fact check does not only find missing facts; it also tells you when a fact rules a poll out.

Misconception check — a number without its facts. Students often think a poll's headline number is the important part and the fine print is optional. That is tempting, because the headline is what gets repeated. But "58 percent favor" means nothing until you know who was asked, when, how they were chosen, and how wide the margin is. A number without its four facts is a rumor with a percent sign.

Memory hook: size, dates, method, margin.

3. Size the margin of error: the approximation and its ladder

Suppose you and a friend each scoop a different spoonful from the same well-stirred pot. Your spoonfuls will taste almost the same, but not exactly. Polls are like that. Two random samples of the same population will land a little apart, by chance alone. The margin of error puts a size on that chance wobble.

The margin of error is the plus-or-minus range around a poll's figure that reflects sampling error alone; at the 95 percent level it is about one over the square root of the sample size, so about 3 percentage points for 1,000 people and about 5 for 400. The 95 percent level is the confidence level behind a margin of error: a statement that the sampling procedure, repeated many times, would land within the margin about 95 times in 100.

This course teaches the margin as an approximation and writes it that way: about one over the square root of the sample size, at the 95 percent level. To get the answer in percentage points, you divide 100 by the square root of the sample size. The square root of a number is the number that, multiplied by itself, gives it: the square root of 400 is 20, because 20 × 20 = 400.

Worked example — the margin for 400 people (an approximation). The square root of 400 is 20, so the margin is about 100 ÷ 20 = 5 percentage points at the 95 percent level. The dining poll found 58 percent in favor. The range for the favor share runs from 58 − 5 = 53 to 58 + 5 = 63 percent. In words: the sampling procedure dining services used would, about 95 times in 100, land within about 5 percentage points of the whole student body's true share. So the dean can say "a majority, somewhere in the fifties or low sixties," not "exactly 58."

Now climb the ladder. Each rung is the same approximation at a larger sample, and the whole ladder is labelled an approximation wherever it appears.

Worked example — the ladder (an approximation). Each rung divides 100 by the square root of the sample size.

For 1,000 people the square root is about 31.62, and 100 ÷ 31.62 ≈ 3.2, about 3 percentage points.
For 1,500 people the square root is about 38.73, and 100 ÷ 38.73 ≈ 2.58, about 2.6 percentage points.
For 2,500 people the square root is 50, and 100 ÷ 50 = 2 percentage points.

The margin-of-error ladder: approximate margin at the 95 percent level, in points, for four sample sizes (an approximation)
Sample size	Square root	Approximate margin, in points
400	20	about 5
1,000	about 31.62	about 3
1,500	about 38.73	about 2.6
2,500	50	about 2

Read down the last column and something odd appears. The sample grows a lot, and the margin shrinks only a little.

Worked example — quadrupling. Take the 400-person sample and multiply it by four: 400 × 4 = 1,600 people. The square root of 1,600 is 40, and 100 ÷ 40 = 2.5 percentage points. That is 5 ÷ 2 = 2.5, half the margin at 400. The extra 1,600 − 400 = 1,200 people bought 5 − 2.5 = 2.5 percentage points. In words: quadrupling the sample halves the margin. That is diminishing returns, and it is why a pollster deciding on a budget rarely pays for ten thousand interviews when fifteen hundred will do.

Real polls do not print the approximation. They report their own stated margin, which usually runs a little larger, because pollsters weight their samples and design them in ways the simple approximation ignores. Pew Research Center's short read on the margin of error gives about 3 percentage points for 1,067 cases.

Worked example — the real poll runs larger. The real poll you read in Section 5, Pew Research Center's survey of 5,119 U.S. adults, states a margin of 1.6 percentage points. The approximation gives about 1.4, so the stated figure is 1.6 − 1.4 = 0.2 percentage points larger. In words: the approximation explains why a poll of a few thousand people has a margin under 2 percentage points; the pollster's own stated figure is the one you cite.

Misconception check — the rule of thumb as the answer. Students often compute the approximation for a real poll and report that number as the poll's margin. That is tempting, because the arithmetic is quick and feels precise. But the approximation explains the size of a margin; it does not replace the pollster's own stated figure, which accounts for weighting and design and is usually a little larger. When a poll states its margin, you cite the stated margin.

Memory hook: four times the people, half the margin.

4. Separate precision from bias: what the margin leaves out

Imagine a bathroom scale that always reads two pounds light. You can weigh yourself a hundred times and average the readings, and you will get a very steady number. It will still be two pounds wrong. Weighing more often made the reading steadier, not truer.

Polls have the same two kinds of problem.

Precision is how tightly a poll's estimate is pinned down by its sample size; a larger random sample gives a smaller margin of error and more precision.
Bias is a lean built into a poll by who answered or how the question was asked, which pushes results away from the population's true views and which a bigger sample does not remove.

Two sources of bias matter most.

Question wording is the exact words a poll uses to ask its question; different wordings can move answers, and that effect is not counted in the margin of error.
Non-response is the failure of some selected people to answer a poll; when those who do not answer differ from those who do, the results lean, and the margin of error does not count it.

Put them together and you get the idea of total error: all the ways a poll's figure can miss the population's true value, including sampling error, question wording, and non-response; the margin of error covers only the sampling part. Pew Research Center's short read makes the same point: the margin covers sampling variability only, and the other sources of error are much harder to put a number on and are rarely reported.

Worked example — what the dining poll's margin does and does not cover. The random sample's margin is about 5 percentage points, so the favor share runs from 53 to 63 percent. That range covers one thing: the chance wobble from which 400 students happened to be drawn. Now suppose the question had read "Would you like the dining hall to stay open later so you can eat after evening classes?" A wording that hands the respondent a reason can pull the answer upward, and nothing in the 5 percentage points accounts for it. Suppose also that many of the 400 never replied, and those who replied were mostly students who eat late. That lean is not in the 5 percentage points either. In words: the margin answers "how much could chance move this?" and stays silent on "was the question fair?" and "who stayed quiet?"

Worked example — size against bias. Return to the click poll. Its 2,000 answers are 5 times the random sample's 400. If size could cure bias, the click poll would be the better poll. It is not, because its lean comes from who chose to answer, and adding more self-chosen answers adds more of the same lean. In words: bigger fixes precision, not bias.

Misconception check — "±3 means accurate to 3 percent." Students often read a margin of about 3 percentage points as a promise that the poll is accurate to within 3 percent. That is tempting, because it is how the number gets reported. But the margin is a 95-percent statement about the sampling procedure, and it covers sampling error only. A badly worded question or a lopsided group of non-responders can push a poll off by more than its margin, and the margin will not warn you.

Misconception check — "a bigger sample fixes a biased poll." Students often think a biased poll can be rescued by asking more people. That is tempting, because the margin really does shrink as the sample grows. But a bigger sample buys precision, a tighter margin around the same leaning answer. It does not remove bias. Bigger fixes precision, not bias.

Memory hook: a steady scale can still read light.

5. Read a real poll: the Pew Research Center release

Now you read a real one. In June 2026 Pew Research Center published Americans and AI 2026: Chatbots, Smart Devices and Views on Impact. Its topic, how people use AI tools, was chosen for this course because it is not a hot-button political fight, so you can practice the reading without a side to take.

Worked example — the four facts on the Pew release. Read the methodology page and fill in each fact.

Sample size: 5,119 U.S. adults.
Field dates: the 17th to the 23rd of February 2026, which is 23 − 17 + 1 = 7 days.
Sampling method: the American Trends Panel, which Pew describes on its methodology page.
Stated margin: plus or minus 1.6 percentage points at the 95 percent level.

Tally: 4 of 4 facts reported. In words: this poll passes the check, and every fact you need is on its methodology page.

Worked example — the approximation against the stated figure. The square root of 5,119 is about 71.55, and 100 ÷ 71.55 ≈ 1.4 percentage points. The stated margin is 1.6, which is 1.6 − 1.4 = 0.2 percentage points larger. You cite the stated 1.6. In words: the approximation explains why a poll of about five thousand people has a margin under 2 percentage points, and Pew's own figure, a little larger as real margins usually are, is the one that goes in your sentence.

Now the headline. Of U.S. adults, 44 percent reported they have used ChatGPT.

Worked example — the data scaffold on the 44 percent. With the stated margin, the range runs from 44 − 1.6 = 42.4 to 44 + 1.6 = 45.6 percent. Now run the five questions.

What is measured? Self-reported use: people telling an interviewer what they have done.
Over whom and when? U.S. adults, one week in February 2026.
What does it show? A share at one point in time: a little under half of U.S. adults said they had used the tool.
What does it not show? Why people use it, whether using it helps or harms them, and whether question wording or non-response moved the share, since neither is in the 1.6.
Correlation or cause? Neither, as it stands. A single share is a snapshot. It shows no cause at all, and it does not even show a correlation until you set it beside another measure.

Kind and layer: "Pew surveyed 5,119 U.S. adults" is empirical, checkable against the methodology page. "A self-selected poll carries no margin of error" is empirical, a fact about how sampling works. "A representative ought to vote as the polls say" is normative, on the verdict layer, and no poll settles it.

In practice (your tools). You can check every figure in this section yourself. Open Pew Research Center's methodology page for Americans and AI 2026 and use your browser's find command (Ctrl+F on Windows, Command+F on a Mac) to search 5,119 and then margin. Copy the figures as the page prints them. In a spreadsheet, you can rebuild the approximation: type =100/SQRT(5119) in a cell and you will see about 1.4, which is why you cite the stated 1.6 and treat the 1.4 as an explanation.

Misconception check — a poll as proof of why. Students often read a poll as an explanation: "44 percent use the tool because it is useful." That is tempting, because a large share feels like it must mean something. But the poll measured what people said they did, over one week. It cannot tell you why they did it. A poll shows a share or a pattern, not a cause.

Memory hook: a poll is a snapshot, not a reason.

6. Trace how opinions form and who votes

Think back to the first time you had a political opinion. Maybe it came from a dinner-table argument, a teacher who made a topic come alive, a friend who cared about an issue, a job that changed how you saw taxes, or a video that stuck with you. You did not invent your views alone. They grew out of the people and places around you.

Political socialization is the lifelong process by which people acquire their political attitudes, values, and habits from the people and institutions around them.
The agents of socialization are the sources that shape a person's political attitudes: family, school, peers and life events, and the media.

Socialization explains why opinion is a distribution rather than a single view. People grow up in different families, attend different schools, live through different events, and follow different media, so they arrive at different places.

Then there is the question of who takes part. Turnout is the share of eligible people who vote in an election, which tends to rise with age, education, and income and runs higher in presidential years than in midterm years. Turnout also varies with registration rules: how and when a person must sign up before voting. These are documented patterns, stated plainly.

Each pattern is a correlation: an association in which two things tend to rise or fall together; a correlation alone does not show that one causes the other.

Worked example — correlation or cause on the education pattern. Start with the pattern: turnout tends to rise with education. Now ask the last scaffold question. Does the pattern show that education causes voting? It does not, by itself. Age, income, and other factors travel with education, so a person with more schooling often differs in several ways at once. A claim that schooling itself raises turnout would need an explicit causal argument, named and defended, not the pattern alone. In words: the pattern is empirical and documented; the cause is a further claim the pattern does not settle.

Kind: "turnout runs higher in presidential years than in midterm years" is empirical. "Turnout ought to be higher" is normative, and so is any claim about which rules should change to raise it. A caveat worth keeping: a fact about who votes does not tell you whether that is a problem or what should be done. Keep the documented pattern and the normative question in separate sentences.

Misconception check — the pattern as the reason. Students often read a turnout pattern as an explanation: "older people vote more because age makes people responsible." That is tempting, because the pattern is real and a story fits it. But the pattern shows only that age and turnout tend to come together. Why they do is a separate question, and several stories could fit the same pattern.

Memory hook: together is not because.

7. Weigh media effects: agenda-setting, framing, echo chambers

Imagine a week when every local news site leads with a broken bridge. By the end of the week, people at the bus stop are talking about the bridge. They do not all agree about who is to blame or what to do, but they are all thinking about it. Something has happened, and it is worth naming carefully.

Media effects are the ways news and other media may shape what the public knows, thinks about, and believes, each claim studied with its critics and its limits. Three claims come up again and again.

Agenda-setting is the finding that media coverage shapes which issues the public thinks are important, rather than what people conclude about them.
Framing is the way a story is presented, such as which cause or consequence it stresses, which can shape how people judge the issue it reports.
An echo chamber is a setting, online or offline, in which people mostly meet views that match their own; whether such settings are widespread and how much they harden opinion is debated.

Worked example — three effects on one invented story. A neighborhood association newsletter in an invented town reports that a bus route will skip two stops. Version one leads with "Riders lose two stops" and quotes a commuter. Version two leads with "Route trimmed to cut travel times" and quotes the transit planner. Sort what each media effect claims here:

Agenda-setting: by covering the route at all, the newsletter makes readers more likely to think the bus route matters. It does not tell them what to conclude.
Framing: the two versions stress different consequences, a loss for riders and a gain in speed, and may shape how readers judge the same change.
Echo chamber: a reader who only ever sees version one, from a group that always opposes route changes, meets one view repeatedly.

In words: the same event can raise a topic, tilt a judgment, or reinforce a view, and those are three different claims.

Each claim has its critics, and you should hear them at full strength.

On agenda-setting, defenders argue the pattern is well documented. Critics respond that people also choose what to follow, so the arrow can run from public interest to coverage as well as the other way.
On framing, proponents argue that wording and emphasis can move judgments in controlled studies. Critics respond that framing effects are often smaller in competitive real-world settings than in controlled studies, because people meet rival frames and bring their own views.
On echo chambers, proponents argue that sorting by views hardens opinion and makes compromise harder. Critics respond that many people meet more varied views than the claim assumes, and that the size of the effect is contested.

A limitation to keep: none of these claims comes with a number this course asks you to memorize. The size of a media effect is a contested, moving figure, so you learn each claim with its critics rather than a single statistic.

Misconception check — "the media tells us what to think." Students often sum up media effects as "the media tells us what to think." That is tempting, because it is a common complaint. But agenda-setting is about what we think about, not what we conclude, and it has documented limits and competing findings. Framing can tilt a judgment, yet its effects are often smaller outside the lab. Say which effect you mean, and name its limit.

Memory hook: about, not what.

8. Weigh representation: delegate or trustee?

Imagine you are elected class representative, and the class wants a group project replaced by an exam. You think the project teaches more. Do you vote the way your classmates asked, or the way you judge is best for them? Every representative faces a version of that question, and polls make it sharper, because now the representative can see the numbers.

A delegate is a representative who votes as the people they represent want, as polls, messages, or meetings show it.
A trustee is a representative who votes on their own judgment of the public's interest, even when that runs against current opinion.

This is this week's open question, and both sides deserve their strongest case.

Proponents of the delegate view argue that representation means following the represented. A representative is elected to carry the voters' wishes, not to replace them, and a well-run poll is a direct way to hear what those wishes are. On this view, a representative who overrides the public is putting personal opinion in the voters' seat.
Proponents of the trustee view argue that representatives can study a question in depth that most voters do not have time to study. They add that a snapshot poll can be badly worded, can miss the people who did not answer, and can shift next month, so following it vote by vote could mean following noise. On this view, voters can judge the representative's record at the next election.

Each side names a real cost in the other. A pure delegate may follow a poll that is badly worded or out of date. A pure trustee may drift away from the people who elected them. Many representatives mix the two, depending on the issue.

Worked example — sorting the debate by kind and layer. An invented town council member reads a random-sample poll of residents that finds most oppose a new bus depot, and the member believes the depot is needed. Sort three claims. "The poll's random sample found most residents oppose the depot" is empirical, on the practice layer, checkable against the poll. "Council members are elected by the town's voters" is empirical, on the design layer, checkable against the town's rules. "The member ought to vote with the poll" is normative, on the verdict layer. In words: the poll can tell the member what residents said; it cannot tell the member what to do.

Pause and think, choosing one of these two:

(a) State the strongest case for the view of representation you find less convincing, delegate or trustee, as its best defender would state it.
(b) Think of a group you belong to that chose a representative, such as a team, a club, or a workplace committee. Say whether you would want that person to act more as a delegate or more as a trustee, and name one cost of your choice.

Misconception check — "the poll settles it." Students often think that if a good poll shows a majority, a representative must follow it, or else that a poll can be waved away as only a number. Both are tempting shortcuts. The poll is empirical evidence of what people said, and a careful reader checks its four facts and its margin. Whether to follow it is a normative question about representation that the poll itself cannot answer.

Memory hook: the poll reports; the representative decides how to weigh it.

Key terms

public opinion — The distribution of attitudes across a whole population on a political question, usually measured by asking a sample rather than everyone. (first used: 1. Sample the population: random versus self-selected)
population — The whole group a poll wants to describe, such as all U.S. adults or every student enrolled at one school. (first used: 1. Sample the population: random versus self-selected)
sample — The smaller group of people who are actually asked a poll's questions, used to estimate the views of the whole population. (first used: 1. Sample the population: random versus self-selected)
random sample — A sample drawn so that every member of the population has a known, non-zero chance of being selected, which is what lets a poll carry a margin of error. (first used: 1. Sample the population: random versus self-selected)
self-selected sample — A sample whose members choose to take part, as in a click poll or a call-in poll; because no one has a known chance of selection, it carries no margin of error at any size. (first used: 1. Sample the population: random versus self-selected)
field dates — The days on which a poll's interviews were carried out, which tell a reader when the snapshot of opinion was taken. (first used: 2. Check the four facts a trustworthy poll reports)
sampling method — The procedure a pollster used to choose who was asked, such as a random draw from a list or a standing survey panel, which a trustworthy poll describes. (first used: 2. Check the four facts a trustworthy poll reports)
stated margin — The margin of error a pollster reports for its own survey, usually a little larger than the rule-of-thumb approximation because of weighting and design; it is the figure a reader cites. (first used: 2. Check the four facts a trustworthy poll reports)
margin of error — The plus-or-minus range around a poll's figure that reflects sampling error alone; at the 95 percent level it is about one over the square root of the sample size, so about 3 percentage points for 1,000 people and about 5 for 400. (first used: 3. Size the margin of error: the approximation and its ladder)
95 percent level — The confidence level behind a margin of error: a statement that the sampling procedure, repeated many times, would land within the margin about 95 times in 100. (first used: 3. Size the margin of error: the approximation and its ladder)
precision — How tightly a poll's estimate is pinned down by its sample size; a larger random sample gives a smaller margin of error and more precision. (first used: 4. Separate precision from bias: what the margin leaves out)
bias — A lean built into a poll by who answered or how the question was asked, which pushes results away from the population's true views and which a bigger sample does not remove. (first used: 4. Separate precision from bias: what the margin leaves out)
question wording — The exact words a poll uses to ask its question; different wordings can move answers, and that effect is not counted in the margin of error. (first used: 4. Separate precision from bias: what the margin leaves out)
non-response — The failure of some selected people to answer a poll; when those who do not answer differ from those who do, the results lean, and the margin of error does not count it. (first used: 4. Separate precision from bias: what the margin leaves out)
total error — All the ways a poll's figure can miss the population's true value, including sampling error, question wording, and non-response; the margin of error covers only the sampling part. (first used: 4. Separate precision from bias: what the margin leaves out)
political socialization — The lifelong process by which people acquire their political attitudes, values, and habits from the people and institutions around them. (first used: 6. Trace how opinions form and who votes)
agents of socialization — The sources that shape a person's political attitudes: family, school, peers and life events, and the media. (first used: 6. Trace how opinions form and who votes)
turnout — The share of eligible people who vote in an election, which tends to rise with age, education, and income and runs higher in presidential years than in midterm years. (first used: 6. Trace how opinions form and who votes)
correlation — An association in which two things tend to rise or fall together; a correlation alone does not show that one causes the other. (first used: 6. Trace how opinions form and who votes)
media effects — The ways news and other media may shape what the public knows, thinks about, and believes, each claim studied with its critics and its limits. (first used: 7. Weigh media effects: agenda-setting, framing, echo chambers)
agenda-setting — The finding that media coverage shapes which issues the public thinks are important, rather than what people conclude about them. (first used: 7. Weigh media effects: agenda-setting, framing, echo chambers)
framing — The way a story is presented, such as which cause or consequence it stresses, which can shape how people judge the issue it reports. (first used: 7. Weigh media effects: agenda-setting, framing, echo chambers)
echo chamber — A setting, online or offline, in which people mostly meet views that match their own; whether such settings are widespread and how much they harden opinion is debated. (first used: 7. Weigh media effects: agenda-setting, framing, echo chambers)
delegate — A representative who votes as the people they represent want, as polls, messages, or meetings show it. (first used: 8. Weigh representation: delegate or trustee?)
trustee — A representative who votes on their own judgment of the public's interest, even when that runs against current opinion. (first used: 8. Weigh representation: delegate or trustee?)

Summary

Public opinion is a distribution of views across a population, measured by asking a sample.
A random sample gives everyone a known, non-zero chance; a self-selected sample carries no margin of error at any size.
A trustworthy poll reports four facts: sample size, field dates, sampling method, and stated margin.
The margin is about one over the square root of the sample size at the 95 percent level: about 5 percentage points for 400, about 3 for 1,000.
Quadrupling the sample halves the margin; a real poll's stated margin is the figure you cite.
Question wording and non-response are not in the margin; bigger fixes precision, not bias.
Pew's 2026 poll of 5,119 U.S. adults carries a stated margin of 1.6 percentage points and shows a share, not a cause.
Socialization shapes views; turnout patterns are correlations, and why they hold is a separate question.
Agenda-setting is about what we think about; framing and echo chambers each come with critics and limits.
Delegate or trustee is a verdict-layer question, and the verdict is yours.

Check your understanding

Work these before looking at the answers — they are ungraded, and the answers (with reasoning) are at the bottom of the page.

A city library draws 900 card holders at random from its full patron list and asks whether it should open on holiday mornings. Using the approximation, about how large is the margin of error, in points? Show the steps.
A local radio station's call-in poll gets 5,000 calls, and 4,100 callers support a new stadium. What share of callers support it, and what margin of error can the station attach to that share?
A survey of 625 randomly chosen adults in a made-up province finds 52 percent support a new rail line. Using the approximation, give the margin and the range. Then name one problem the margin does not cover.
A made-up report says: "Towns with more local news outlets have higher turnout, so opening a news outlet will raise turnout." Which part is a documented pattern, which part is a causal claim, and what kind of claim (empirical or normative) is each?

Answers (worked reasoning — read after attempting)

About 3.3 percentage points. The square root of 900 is 30, because 30 × 30 = 900. Then 100 ÷ 30 ≈ 3.3 percentage points at the 95 percent level. This works only because the library drew a random sample from its full list, so every card holder had a known chance of selection.
The share is 4,100 ÷ 5,000 × 100 = 82 percent of callers. The station can attach no margin of error at all, because callers chose themselves. The 5,000 calls do not help: bigger fixes precision, not bias, and the people who call a radio station about a stadium may differ from those who do not.
The square root of 625 is 25, so the margin is about 100 ÷ 25 = 4 percentage points. The range runs from 52 − 4 = 48 to 52 + 4 = 56 percent. Notice that the range dips below 50, so the survey cannot tell you a majority supports the line. One problem the margin does not cover: the question's wording, or non-response if the people who did not answer differ from those who did.
"Towns with more local news outlets have higher turnout" is the documented pattern, an empirical claim about a correlation. "Opening a news outlet will raise turnout" is a causal claim, also empirical, and the pattern alone does not support it, because other features of those towns could travel with both. The report jumps from "together" to "because," which is the trap from Section 6.

Next up in this module: the Readings & Resources page, the lecture tutorial, the practice set, and the quiz. Next week, the comparative method shows how to test a cause across countries rather than read one from a single poll.

Built with The Course Maker · samples.thecoursemaker.com