A registered voter (RV) poll surveys people on the voter rolls; a likely voter (LV) poll narrows that group to the people the pollster expects to actually vote. Because not every registered voter turns out, and the ones who do can lean differently from the ones who don’t, the same race can look a few points different depending on which group was polled. Neither label is “the truth”; they answer slightly different questions.
What does RV mean in a poll? #
A registered-voter sample includes anyone who says they’re registered (or who is matched to the voter file). It’s a broad group. In a presidential year about two-thirds of eligible citizens vote; in a midterm it’s closer to half. The University of Florida Election Lab estimates turnout at about 50% of eligible voters in 2018 and about 46% in 2022. So a midterm RV poll includes a lot of people who won’t cast a ballot.
That’s not a flaw. RV polls are stable, easy to compare over time, and don’t depend on the pollster’s guesses about turnout. They’re the standard for most of an election cycle.
What does LV mean in a poll? #
A likely-voter sample is a subset of registered voters that the pollster believes will vote. There’s no official list of likely voters. Each pollster builds its own likely voter screen or turnout model, which is why LV polls can disagree more than RV polls do.
Most pollsters switch from RV to LV samples around Labor Day of an election year, when voters are paying more attention and turnout patterns are easier to estimate. By the final weeks, most published horse-race polls are LV.
How do pollsters decide who is a likely voter? #
They usually combine several signals.
Stated intention and interest #
Questions like “How certain are you to vote?” on a 1-to-10 scale, or “How much thought have you given to the election?” People overstate their intention to vote, because voting is socially approved, so this alone lets too many people through.
Past voting history #
Better pollsters match respondents to the state voter file, which records whether someone voted in past elections (never who they voted for). Someone who voted in the 2022 midterm and the 2024 general is very likely to vote in 2026. Someone who has never voted in a midterm is less likely.
Knowledge and logistics #
Some screens ask whether the respondent knows where their polling place is, or has already requested a mail ballot. These questions sort engaged voters from casual ones.
Cutoffs vs. probability weights #
The old “Gallup-style” approach gave each respondent a score and kept the top slice matching expected turnout. Many pollsters now give every registered voter a probability of voting and weight by it, so no one is dropped entirely. The probability approach is smoother but depends heavily on the model.
Why do RV and LV polls show different results? #
Because the people who reliably vote differ from those who don’t. Which party benefits from the LV filter has changed over time:
- For decades, likely-voter screens tended to help Republicans by a point or two, because older and higher-income voters, who turned out more reliably, leaned Republican.
- More recently the pattern has blurred. College-educated voters, who vote at high rates, have shifted toward Democrats, while some lower-propensity groups have shifted toward Republicans. In recent cycles the LV screen has helped Democrats in some races and Republicans in others.
The practical lesson: don’t assume which direction the LV shift goes in a given race. Look at whether the pollster published both RV and LV numbers. Many do, and the gap between them tells you how much the race depends on turnout.
How to compare RV and LV polls without getting fooled #
| Situation | What to do |
|---|---|
| A candidate “gains” between a June RV poll and a September LV poll | Check whether the same pollster ran both. The change may be the screen, not a real shift. |
| Two LV polls disagree by 5 points | Look at their turnout assumptions. One may expect a much younger or older electorate. |
| An LV poll in the spring | Treat it cautiously. Turnout models that early are guesses. |
| A poll labeled “adults” | Useful for approval and issues, weakest for predicting an election. |
Three rules of thumb:
- Compare like with like. RV to RV, LV to LV, and ideally the same pollster.
- Expect fewer undecideds in LV polls. Likely voters are more engaged and more likely to have picked a side.
- Trust RV early, lean on LV late. Before Labor Day, RV samples are more stable. In October, LV samples are closer to the electorate that shows up.
Where you’ll see these labels #
Every serious poll release lists the population next to the sample size, usually as something like “n=1,000 LV” or “n=1,200 RV.” Poll listings on aggregator sites include it too.
In Election Tracker, our free iPhone app, every poll card shows the population (likely voters, registered voters or adults) next to the pollster, sample size and field dates. That makes the RV-to-LV switch visible as it happens in September and October: if a Senate race’s average moves, you can scroll down and see whether the new polls are LV samples replacing RV ones. The same labels appear on the Approval tab, where most polls are of adults.
For more on how averages combine polls with different populations, see how polling averages work. For the other checks worth running on any poll, see how to tell if a political poll is reliable.
Why midterms make the difference bigger #
In a midterm, turnout is lower and less predictable than in a presidential year, so the gap between registered and likely voters is wider. Enthusiasm matters more: the side whose voters are angrier or more energized tends to turn out at a higher rate. That’s part of why the president’s party has usually struggled in midterms, and it’s why pollsters’ turnout assumptions get so much scrutiny in October. Our explainer on why turnout is lower in midterm elections covers the causes.
For November 3, 2026, expect most polls published after mid-September to be LV. When a pollster releases both, the RV number shows the full pool of potential support, and the LV number shows the pollster’s best guess of who shows up.
Frequently asked questions #
Which is more accurate, a registered voter or likely voter poll? #
Close to Election Day, a well-built likely voter poll is usually closer to the result, because it reflects the people who vote. Early in the cycle, registered voter polls are often more reliable, since likely voter models depend on turnout guesses. A poorly designed LV screen can be worse than a plain RV sample.
What does “n=800 LV” mean? #
It means 800 respondents were counted, and they are likely voters according to the pollster’s screen. The “n” is the sample size. Some pollsters report the full registered sample and the likely voter subset separately, so check which one the headline number uses.
Do likely voter models ever get turnout wrong? #
Yes. When turnout surges among groups that don’t usually vote, or drops among groups that usually do, models built on past behavior miss it. That’s one reason polling errors tend to be larger in elections with unusual turnout.
What is the difference between “adults” and “registered voters” in a poll? #
An adults sample includes everyone 18 and over, including people who aren’t registered or eligible to vote. It’s the standard for presidential approval and issue polling. A registered voter sample is closer to the electorate, so it’s more useful for elections.
When do pollsters switch to likely voters? #
Most switch around Labor Day of an election year, and some earlier for primaries. By mid-October nearly all horse-race polls from major pollsters are likely voter samples.