Notes on Public Opinion Polls

Public Opinion

  • Public opinion: the public's expressed views about an issue at a specific point in time.
  • Ideology and public opinion are inextricably linked: ideology is the prism through which people view political issues; thus ideology informs opinions across a full range of issues.
  • Public opinion as a socializing agent (Elizabeth Noelle-Neumann): public opinion provides an independent context that can affect political behavior.
  • Despite constant polling and questions about accuracy, public opinion has historically played an important role in American politics.
  • Public opinion is manifested in various forms:
    • Demonstrators protesting on state capitol steps
    • Bloggers posting opinions
    • Citizens communicating directly with officials (e.g., telling city council members about tax plans, calling Congress on legislation)
  • Voting is one of the most important ways public opinion is measured, but another key tool is the public opinion poll—a survey of a population's opinion on an issue at a particular time.
  • Why policymakers care about public opinion: to develop policies that reflect public views; such policies are more likely to gain support from other leaders and help elected officials be reelected by representing constituents.

How Public Opinion Polls Are Conducted

  • Polls play an integral role in national, state, and local political events; poll results appear in news coverage and influence decision-makers’ deliberations.

  • Polls help determine who decision-makers will be (e.g., candidates gauge name recognition, campaign strategy, opponents’ weaknesses, and voter response to their message).

  • After election, policymakers use polls to gauge constituents' opinions and assess their performance in office.

  • The poll process has several steps:

    • Determine the target population: the group whose opinions are of interest (e.g., a candidate's name recognition within a congressional district, limited to registered voters, possibly limiting to likely voters).
    • Design the survey instrument (poll) and then select a representative sample from the population.

Target Population and Sampling (Page 2–3)

  • Population targeting example: a neighbor running for the U.S. House would want opinions from people in the district, registered to vote, and possibly likely voters (those who voted in past congressional elections).
  • Sampling goal: obtain a sample that represents the target population.

Sampling: Random Sampling and Its Rationale (Page 3)

  • Random sampling: a scientific method in which each member of the population has an equal chance of being included in the sample.
  • Purpose: ensure the sample is not biased toward a particular subgroup, preventing overrepresentation.
  • Example illustrating sampling impact:
    • If surveying a class about parking facilities, surveying only the 8:00 a.m. government class may yield a skewed view (these students may have different parking conditions than the overall student body).
    • Random sampling helps avoid such bias by not restricting to a single class or time slot.
  • Practical sampling method (illustrative): obtain a roster (e.g., registrar list), determine sample size, randomly select every nth person, contact them to obtain views.

Challenges in Sampling and Respondent Coverage (Page 4)

  • Even random samples can be problematic due to coverage and response issues.
  • Cell phones and sample frames:
    • Some polls exclude cell phones; many now include cell phone users, but nonresponse remains a problem due to caller ID and the portability of cell phones (owners may be unavailable when called).
    • People willing to respond tend to be more politically engaged than those who ignore calls.
  • Quota sampling: a method to address sampling biases by ensuring the sample reflects the population’s characteristics (e.g., town demographics). Example demographics for a town: 40% White, 35% African American, 20% Latinx, 5% Asian; with a sample of 200 voters, the quota would aim for 80 White, 70 African American, 40 Latinx, and 10 Asian respondents.
  • Pollsters routinely rely on quota sampling and often collect demographic characteristics at the end of the poll.

Stratified Sampling and Four-Region Stratification (Page 5)

  • Stratified sampling: divide the population into regions (often four broad regions) and sample areas within these regions to reflect national composition.
  • This method is considered more reliable than simple quota sampling and is widely used by major polling organizations.
  • U.S. Census data are typically used as the basis for determining the four sampling regions.
  • Stratified sampling underpins much public opinion data used by political and social scientists (e.g., General Social Survey (GSS), American National Election Study (ANES)).

Sampling Error (Page 5–6)

  • Sampling error (margin of error) arises when a poll uses a sample rather than the entire population.
  • Key figures for typical polls:
    • National polls: samples usually range from about
    • 1,000 to 1,500 respondents.
    • For smaller populations (states or congressional districts): typically 300 to 500 respondents.
  • Nonrepresentative samples can skew results; an underrepresentation of a voter block can lead to inaccurate reflections of the population’s views (e.g., concerns raised about 2016 presidential election polling).
  • Sampling error is the difference between poll results from the sample and the true population values; it is an inherent limitation of polling.

Internet Polls and Nonresponse (Page 6)

  • Internet polls are increasingly used by market researchers, public opinion firms, and candidates but come with unique challenges (e.g., repeated responses or gaming the system).
  • Agencies adjust sampling using methods like including cell phones, but problems such as nonresponse and self-selection persist.
  • Margin of error for Internet polls can be larger, reaching up to
    • ±5 percentage points (as opposed to around ±3 points for traditional polls).
  • Example interpretation: If an Internet poll reports that 70% support cannabis legalization with MOE ±5%, the true population value lies roughly in the interval [0.65, 0.75].

Types of Political Polls (Page 7)

  • Tracking polls: measure changes in public opinion over time by repeatedly asking the same questions; useful for tracking long-term trends, short-term campaign effects, and the impact of media strategy.
    • Example: During the 2020 Democratic presidential primaries, tracking polls showed declines in support for some candidates, contributing to their withdrawal from the race.
  • Push polls: designed to skew public opinion or elicit information about candidate strengths/weaknesses; often present respondents with a hypothetical scenario to gauge the importance of an issue in voting decisions.
    • Example prompt: “If you knew that Congresswoman Jackson lives outside the district, how would that affect your vote?”
    • Pros and cons: can help target messages but have a negative reputation because some entities use them to smear opponents without substantiated charges.
  • Exit polls: conducted at polling places on Election Day to project winners quickly after polls close; frequently sponsored by news organizations.
    • Purpose: predict outcomes and provide data about why voters voted the way they did.
    • Role: offer timely election projections and detailed information for media, campaigns, and parties.

Connections, Implications, and Practical Notes

  • Public opinion informs policy decisions and political behavior by shaping what leaders think the public cares about; it can also influence the policy environment and re-election prospects.
  • Poll design choices (population, sampling method, sample size, and mode) affect the reliability and generalizability of findings.
  • Ethical considerations include avoiding manipulation through push polls, ensuring representative samples, and transparency about margin of error and limitations.
  • Practical implications include understanding why polling results may diverge from election outcomes (e.g., sampling frames, nonresponse, weighting, late responses).
  • Real-world relevance: polling data guide campaign strategies, policy prioritization, and media narratives; they also raise questions about the accuracy and integrity of public opinion measurement in a changing communications landscape.

Key Terms and Concepts

  • Public opinion: expressed views about an issue at a particular time.
  • Ideology: the lens through which individuals view political issues; shapes opinions across issues.
  • Margin of error (MOE): the range within which the poll result is expected to reflect the population, typically expressed as ±X percentage points.
  • Sampling error: difference between sample results and true population values; a natural consequence of using samples.
  • Random sampling: each member of the population has an equal chance of selection.
  • Quota sampling: sampling that mirrors population characteristics by enforcing quotas for subgroups.
  • Stratified sampling: dividing the population into regions or strata and sampling within each to improve representativeness.
  • Tracking polls: repeated measurements over time to track changes in opinions.
  • Push polls: polls designed to influence opinion by presenting biased or hypothetical scenarios.
  • Exit polls: polls conducted as voters exit polling places to project outcomes and study voting reasons.

Summary of Numerical References

  • Typical national poll sample sizes:
    • 1{,}000 to 1{,}500 respondents; smaller populations: 300 to 500.
  • Margin of error (MOE):
    • Traditional polls: ±3 percentage points.
    • Internet polls: up to ±5 percentage points.
  • If a poll reports p = 0.70 with MOE = ±0.05, then the population range is approximately [0.65, 0.75].
  • Regional stratification is commonly based on four regions using Census data to frame sampling regions.

Formulas (LaTeX)

  • Margin of error for a proportion (standard formula):
    ext{MOE} = z \, \sqrt{\frac{p(1-p)}{n}}
  • Example interpretation with a 95% confidence level (z ≈ 1.96) and p = 0.50, n = 1000 yields MOE ≈ 0.031 (≈ 3.1%). Note: actual MOE depends on p and n; the expression above is the common basis for calculation.