Polls estimate opinions in a defined population. A careful reader checks whom the poll aims to represent, how respondents were selected, how interviews were conducted, when the survey occurred, how questions were worded, how undecided voters were treated, and what uncertainty is reported. A poll of all adults is not automatically informative about likely primary voters. A voluntary website survey is self-selected, not representative merely because thousands answered.

A margin of sampling error describes uncertainty from observing a sample instead of the entire target population under the stated sampling design. It is usually reported for an estimate at a specified confidence level. It does not mean that every number is certainly within that range, and it does not measure coverage error, nonresponse, question effects, processing mistakes, or a faulty likely-voter model.

Suppose one poll reports Candidate A at 47 percent and Candidate B at 45 percent, with a stated margin of sampling error of plus or minus three points. A has the higher point estimate, but the poll does not provide enough evidence to conclude confidently that A leads in the target population. Do not mechanically double the three points or treat the reported margin as a magic tie rule; the uncertainty of the difference depends on the survey design and relationship between the estimates. The safe interpretation is narrower: sampling uncertainty is large enough that this one poll cannot confidently identify the leader.

Polls can guide decisions about messages, travel, and resources, but they do not award votes or delegates. A change between two polls also does not prove that an advertisement caused the movement. Different samples, methods, news events, turnout assumptions, or genuine opinion change remain possible. The disciplined reader asks what was measured, in which population, with what uncertainty, and whether the evidence supports description or causation.

Video lesson: Political Campaigns: Crash Course Government and Politics #39