Polling response rates have collapsed because most people no longer answer calls from unknown numbers. Phones now filter, label and screen those calls, robocalls taught everyone to ignore them, and survey requests arrive after every purchase and doctor’s visit. Pew Research Center’s phone surveys reached a 36% response rate in 1997 and 6% by 2018, and many polls today complete interviews with only a few percent of the people they try to reach.
A low response rate doesn’t automatically make a poll wrong. It makes polls more expensive and raises the risk that the few people who answer differ from everyone else. Here’s why rates fell, how pollsters adapted, and how to read polls with that in mind.
How far have response rates fallen? #
Pew’s telephone surveys are the best-documented example. Its response rate went from 36% in 1997 to about 9% in 2012 and 6% in 2018. Pew now does most of its US survey work through an online panel recruited by mail from a random sample of addresses, rather than through phone calls.
Other pollsters have moved the same way, toward text messages, online panels and combinations of methods.
Why do people stop answering polls? #
Robocalls and spam #
When Pew reported its 2018 numbers, it cited an estimate of 3.4 billion robocalls a month in the US. After years of scams and sales calls, a call from an unknown number is something most people ignore by habit.
Phones screen the call before you see it #
Carriers label suspicious calls as “Spam Likely,” and those labels sometimes land on legitimate pollsters. Apple added “Silence Unknown Callers” to the iPhone in 2019, which sends unknown numbers straight to voicemail. iOS 26, released in September 2025, added Call Screening, which asks an unknown caller for information before you decide whether to pick up. That’s useful for dodging scams, and one more step between a pollster and a respondent.
Fewer landlines #
Most US adults now live in households without a landline. Landlines skew toward older people, so a poll that relied on them would badly miss younger voters.
Cell phone rules raise costs #
The federal Telephone Consumer Protection Act restricts using automated dialers to call cell phones without consent. A 2021 Supreme Court ruling, Facebook v. Duguid, narrowed what counts as an autodialer, but many pollsters still have live interviewers dial cell numbers by hand to stay safely within the law. That makes each completed interview slower and costlier.
Survey fatigue, trust and privacy #
People get feedback requests after nearly every transaction, so a 15-minute political survey is a hard sell. People who distrust the media, universities or government are more likely to refuse, and many are wary of telling a stranger about their politics, income or education.
How have pollsters adapted? #
Multiple ways to reach people #
Modern polls often combine several modes:
- Text-to-web, a text message with a link to an online survey, which reaches younger people better than calls.
- Probability-based online panels, recruited by mail from random addresses and paid small incentives to take regular surveys.
- Opt-in online panels, cheap and fast but made up of people who volunteered, which introduces its own bias.
- Live phone calls, still used, often alongside other methods.
- Automated phone surveys (IVR), mainly to landlines.
Each mode reaches different people and can produce different answers, which we explain in what a polling mode effect is. Our piece on whether pollsters call cell phones covers the phone side in detail.
Weighting the sample #
Pollsters adjust raw samples to match the population on age, sex, race, education and region, and increasingly on past vote or party. After 2016, when state polls underweighted voters without college degrees, education weighting became standard among serious pollsters. The details are in how pollsters weight survey data.
Does a low response rate make polls wrong? #
Not by itself. Pew’s own research has found that response rates are an unreliable predictor of accuracy, and the 2018 midterm polls performed well by historical standards despite single-digit rates.
The real risk is non-response bias: if the people who answer differ from those who don’t in ways weighting can’t capture, the poll misses. That’s the leading explanation for the 2020 polling error, when Trump supporters were less likely to take surveys even within the same demographic groups. We explain it in what non-response bias is.
Low response also raises costs. Some outlets now poll less often, and cheaper, less rigorous polls fill the gap. That makes it more important to know who conducted a poll and how.
How to read polls when few people answer #
- Rely on averages. Individual polls carry more method-driven noise than they used to. An average across several pollsters, using different methods, is more stable.
- Check the population. Adults, registered voters and likely voters are different groups.
- Check the method. Look in the release for the mode and the weighting variables. Education weighting is a minimum.
- Watch for one pollster dominating. A firm that releases polls constantly can tilt an average if every poll is counted.
Election Tracker handles the last point in its averaging: each race’s 30-day average counts repeated polls from the same pollster once, at their strongest sample, and weights newer polls more heavily. Every poll card shows the pollster, sample size, population and field dates, and the pollster’s release opens in an in-app browser so you can check the mode and weighting.
Frequently asked questions #
If response rates are so low, how can a poll have a 3-point margin of error? #
The margin of error comes from the size of the final sample, not the response rate. About 1,000 completed interviews gives roughly plus or minus 3 points. But the margin of error assumes the sample is effectively random, and low response makes that assumption harder to meet. Our guide to margin of error explains what it covers and what it doesn’t.
How do pollsters reach young people who never answer calls? #
Mostly through text-to-web invitations and online panels, sometimes recruited by mail. Weighting then adjusts for any remaining shortfall in young respondents.
Are online polls better than phone polls? #
Neither is automatically better. Probability-based online panels can be as accurate as good phone polls; opt-in panels are cheaper but less reliable. Many of the most accurate pollsters combine methods. See whether online opt-in polls are reliable.
Should I answer a call from a pollster? #
That’s up to you. Legitimate pollsters identify themselves and their organization, don’t ask for money, and don’t ask for Social Security or bank numbers. If a “survey” asks for payment information or pushes damaging claims about a candidate, it isn’t a real poll.
Why do polls still miss if pollsters weight their samples? #
Because weighting only fixes the traits pollsters can measure against a known benchmark. If, within a group, the people who answer vote differently from those who don’t, weighting can’t see it. That’s why polls can miss in the same direction across many pollsters in the same year.