AP United States Government and Politics

## Learning Outcomes

- Understand what makes one public opinion poll more reliable than another
- Identify the difference between a representative sample and a random sample
- Explain how margin of error works and what it does not measure
- Distinguish sampling error from nonresponse bias, coverage error, and wording effects
- Compare polls of all adults, registered voters, and likely voters
- Evaluate polling data in AP Gov multiple-choice, quantitative analysis, and argument questions

## AP United States Government and Politics Syllabus

For the AP United States Government and Politics, you are expected to understand polling methodology quality, with a focus on the following syllabus points:

- Know that public opinion polls are used to measure attitudes, beliefs, and likely electoral behavior
- Identify the features of scientific polling, including random sampling and representative samples
- Explain why sample size, margin of error, and timing matter when judging poll quality
- Understand how question wording and question order can affect poll results
- Explain the difference between polling all adults, registered voters, and likely voters
- Recognize why cell-phone inclusion, weighting, and transparency matter in modern polling
- Distinguish sampling error from nonresponse bias and coverage error
- Interpret poll results carefully, especially in close races and subgroup data

## Test Your Knowledge

_Attempt these questions before reading this article. If you find some difficult or cannot remember the answers, look more closely at that area during your revision._

1. What is the difference between a representative sample and a random sample?
2. What does a margin of error of ±3 percentage points actually mean?
3. Why can a poll with a large sample still produce poor results?
4. How might question wording change a respondent’s answer?
5. Why might a likely-voter poll show different results from a poll of all adults?

## Introduction

Polling is one of the main tools used to measure public opinion in the United States. On the AP exam, you are not expected to become a statistician. You are expected to judge whether a poll is trustworthy, identify its limits, and explain why polling results may differ.

That matters because polls affect campaigns, media coverage, and policy debates. A poor poll can distort the public picture of an election or issue. A strong poll gives a more accurate estimate of what a population thinks. The key AP skill is evaluation: not just reading the headline number, but asking how the poll was built.

> **Key Term: public opinion poll**
> A survey used to measure the views, attitudes, or preferences of a population.

### What makes a poll scientifically strong

A high-quality poll starts with the right goal: measuring the views of a larger population by studying a smaller group. That only works if the smaller group reflects the larger one.

> **Key Term: representative sample**
> A sample selected to mirror the larger population the poll is intended to describe.

Students often confuse this with randomness.

> **Key Term: random sample**
> A sample chosen so that members of the target population have a chance of being selected, helping reduce bias.

A **representative sample** is the goal. A **random sample** is one major way to get there. A sample can be random in method but still fail to represent the target population if important groups are missed or respond at very different rates.

For AP Gov, this distinction matters. If a question asks why a poll is reliable, the best answer is often that it uses a **scientific polling sample** with random selection aimed at creating representativeness. But if a question asks what matters most in the end, the answer is usually whether the sample actually reflects the target population.

Modern pollsters use several methods:

- random-digit dialing for phone polls
- address-based sampling
- probability-based online panels

Phone polling should generally include both landlines and cell phones. If a poll uses only landlines, it may miss younger adults and some lower-income or mobile-only households.

> **Key Term: coverage error**
> Error that occurs when some members of the target population have little or no chance of being included in the sample.

A poll can fail before the first question is even asked if the wrong people are reachable.

For example, imagine a poll about a Senate race that only calls landlines on weekday afternoons. That method may underrepresent younger workers, people with long commutes, and cell-phone-only households. Even if the sample size is large, the poll may still lean in the wrong direction because its coverage is weak.

> **Test Tip:** On AP multiple-choice questions, eliminate answers that treat a large sample size as the only mark of quality. A bigger sample reduces sampling error, but it does not fix a bad sample design.

### Margin of error, sampling error, and what students often get wrong

Many students see the **margin of error** as the main sign of quality. It is important, but it is only one part of the story.

> **Key Term: sampling error**
> The random uncertainty that occurs because a poll surveys a sample rather than the entire population.
>
> **Key Term: margin of error**
> A statistic that summarizes sampling error, usually reported as a plus-or-minus range around a poll result.

If a poll says Candidate A has **52% support with a margin of error of ±3%**, that means support is plausibly around **49% to 55%**. On the AP exam, you should read this as a range, not as an exact point.

A simple confidence-interval idea is:

- poll estimate ± margin of error

So:

- 48% with ±3% means about 45% to 51%
- 52% with ±3% means about 49% to 55%

If two candidates are within the margin of error, the race is close. You should not call it a clear lead.

Here is a worked example.

A poll of likely voters reports:

- Candidate X: 49%
- Candidate Y: 47%
- Margin of error: ±3%

Step 1: Calculate X’s range

- 49% - 3% = 46%
- 49% + 3% = 52%

Step 2: Calculate Y’s range

- 47% - 3% = 44%
- 47% + 3% = 50%

Step 3: Compare the ranges

- X could be as low as 46%
- Y could be as high as 50%

Because the ranges overlap, you should interpret the race as **competitive**, not as safely led by X.

Students often make three mistakes here:

- assuming margin of error tells you everything about poll quality
- assuming a candidate ahead by 1 or 2 points is definitely winning
- assuming margin of error applies equally to all subgroups

That last point is important. If the full sample is 1,200 but only 150 respondents are Latino voters, the subgroup estimate is less precise than the overall poll.

> **Key Term: confidence interval**
> The range of values around a poll estimate that is likely to contain the true population value, often expressed using the margin of error.

But the biggest AP distinction is this: margin of error measures **sampling error only**. It usually does not capture all the other ways a poll can be wrong.

### Other sources of polling error: wording, order, nonresponse, and weighting

A poll can have a respectable margin of error and still be poor quality. That is because many errors are not about random chance.

One major problem is **question wording**.

A neutral question might ask:

- Do you support or oppose the death penalty?

A more emotional question might ask:

- Would you support the death penalty for someone who murdered a member of your family?

Those questions measure related views, but not in the same way. The second wording invites a more personal and emotional response.

> **Key Term: question wording**
> The specific phrasing of a poll question, which can influence how respondents understand and answer it.

Question order matters too. If respondents are first asked several questions about crime, immigration, or inflation, that may shape how they answer a later question about presidential approval. Earlier questions can prime later responses.

Another common problem is nonresponse.

> **Key Term: nonresponse bias**
> Bias that occurs when people who respond differ in meaningful ways from people who do not respond.

A low response rate by itself does not prove a poll is worthless. But if nonrespondents differ systematically from respondents, the poll can lean off course. For example, if distrustful voters are much less likely to answer political surveys, their views may be undercounted.

This helps explain why a large sample size is not a magic fix. If you collect 10,000 responses but the missing people all share an important political trait, the poll can still be biased.

Pollsters try to correct these imbalances through weighting.

> **Key Term: weighting**
> Statistical adjustment of poll results so the sample better matches the target population on traits such as age, race, gender, education, and region.

Suppose young adults make up 20% of the electorate but only 10% of the people who answered the poll. Pollsters may give more weight to those young respondents so the final results better reflect the population.

Weighting can improve quality, but it has limits:

- it only adjusts for traits the pollster measures
- it cannot fully fix serious coverage error
- heavy weighting can make estimates less stable

So when you evaluate a poll, do not stop at sample size and margin of error. Also ask:

- Who was surveyed?
- How were they reached?
- Who might be missing?
- How were results weighted?
- What exactly was asked?
  > **Test Tip:** If an AP question asks why two polls taken at similar times gave different results, look for differences in target population, wording, likely-voter screens, weighting, or contact methods.

### Adults, registered voters, likely voters, and why polls differ

One of the most tested distinctions in polling is who the poll is actually measuring.

> **Key Term: likely voters**
> Respondents judged by a pollster to be most likely to vote in an upcoming election.

Polls may survey:

- **all adults**
- **registered voters**
- **likely voters**

These are not interchangeable.

A poll of **all adults** includes everyone in the adult population, including nonvoters and unregistered people. That can be useful for measuring broad public opinion on issues like health care or immigration.

A poll of **registered voters** is narrower. It excludes adults who are not registered.

A poll of **likely voters** is narrower still. It tries to identify who is actually going to turn out on Election Day. That often makes likely-voter polls more useful for predicting election outcomes, but it also introduces another judgment call: how the pollster defines likely voters.

Some likely-voter screens use:

- self-reported interest in the election
- certainty of voting
- past voting history
- registration status
- knowledge of polling place or election timing

Different screens can give different results. A pollster with a strict likely-voter model may exclude low-propensity voters who do turn out. A looser model may include people who will never vote.

This is why the textbook notes that tracking polls of likely voters and those merely eligible to vote can differ significantly. The change is not necessarily because one pollster is dishonest. It may be because they are measuring different populations.

A good AP-style comparison:

- **all adults** = broadest opinion
- **registered voters** = narrower electoral population
- **likely voters** = best election estimate, but most dependent on assumptions

### How to evaluate a poll on the AP exam

On the exam, polling questions often appear in multiple-choice or quantitative-analysis form. You may be given a short poll description, a chart, or two poll results and asked which is more reliable or what limitation matters most.

Use this checklist:

- Who conducted the poll?
- Was it a neutral organization or a campaign?
- What was the target population: adults, registered voters, or likely voters?
- How large was the sample?
- What was the margin of error?
- Were cell phones included?
- What were the field dates?
- What was the exact wording?
- Was the sample representative?
- Was weighting used?

Transparency matters because without these details, you cannot fully judge quality.

For example, campaign-sponsored polls are not automatically false, but they deserve closer scrutiny because the sponsor has an interest in the result. Neutral organizations are usually more trusted because they have less incentive to shape the narrative.

Timing matters too. Public opinion can change quickly after a debate, scandal, Supreme Court ruling, or economic shock. An older poll may simply be outdated rather than badly designed.

Finally, remember the difference between a poll being “wrong” and a poll being “useless.” Polls are estimates, not prophecies. A poll can be methodologically sound and still miss late shifts in turnout, enthusiasm, or public reaction.

When AP questions ask for the **best** evaluation, the strongest answer usually combines method and limitation. For example:

- “The poll is more reliable because it used a representative random sample of likely voters, included cell phones, and reported a small margin of error.”

That is better than:

- “The poll is reliable because it surveyed many people.”

## Key Point Checklist

_This article has covered the following key knowledge points:_

- Poll quality depends first on whether the sample represents the target population
- A representative sample is the goal, while random sampling is a key method for achieving it
- Margin of error measures sampling error, not every kind of polling error
- A close race within the margin of error should be treated as competitive
- Large sample size improves precision but does not fix bias
- Coverage error happens when important groups have little or no chance of inclusion
- Nonresponse bias occurs when respondents differ from nonrespondents in politically relevant ways
- Question wording and question order can change poll results
- Weighting adjusts the sample to better match the population, but it is not a perfect fix
- Polls of all adults, registered voters, and likely voters may produce different results
- Likely-voter polls depend on turnout assumptions and screening methods
- Strong AP analysis of polls requires evaluating method, transparency, and limitations together

## Key Terms and Concepts

- public opinion poll
- representative sample
- random sample
- coverage error
- sampling error
- margin of error
- confidence interval
- question wording
- nonresponse bias
- weighting
- likely voters


Measuring Public Opinion - Polling methodology quality