An exit poll samples people after they vote. It can estimate how participating groups voted and which issues they report considering. Its design must cover precincts, voting times, early and mail ballots, languages, and refusals. Exit polls are estimates, not official ballot counts, and they describe voters reached by the design rather than nonvoters or all eligible citizens.

A tracking poll repeats comparable measurements over time. Many use a rolling window: each update adds new interviews and drops the oldest. The smoothing makes a trend easier to see, but adjacent estimates share much of the same data and are not independent daily snapshots. A small one-day move may be chance variation. Changes in mode, wording, likely-voter rules, weights, or respondent composition can create a break that looks like persuasion after a campaign event.

Read the design before naming a cause. A movement that follows a debate may be real, but timing alone does not isolate the debate from news, sample turnover, or altered procedures. Comparable repeated measures and uncertainty provide a stronger basis for interpretation.

Charts can clarify or distort any of these estimates. Identify the population, field dates, unit, denominator, response category, sample size, and uncertainty before reading the shape. A truncated vertical axis can make a small change look dramatic; a stacked bar can hide movement in middle categories. Line charts imply continuity between observations even when surveys occurred only on selected dates.

Percentage points and percent change also differ. A rise from 20 percent to 25 percent is five percentage points and a 25 percent increase relative to the starting value. Net approval subtracts disapproval from approval, so it is not the approval percentage. Subgroup bars based on unequal sample sizes can have very different precision even when drawn with equal width.

Imagine an election-night graphic reporting that 62 percent of interviewed voters under thirty supported a candidate. The proper claim is about the exit-poll estimate for participating young voters reached and weighted by that design. It does not describe all young residents, establish youth turnout, or prove that age caused the vote. The official result still comes from counted ballots.

Give every specialized poll one plain-language job: push polls seek persuasion, exit polls sample people who voted, and tracking polls estimate change through repeated measurements. Then read the source note and axes. A polished graphic cannot repair weak sampling, and a scientific estimate should not be stretched beyond the population and uncertainty that produced it.

Video lesson: Science Behind the News: Opinion Polls & Random Sampling