Explainer
How surveys are weighted: raking, matching and propensity scores
Survey weighting adjusts a sample so that it matches known characteristics of the population, correcting for people who did not respond and for groups that were deliberately oversampled.[1][2] The most common method, raking, repeatedly adjusts each respondent's weight until the sample matches population figures on variables such as age, sex, race, education and region.[3][4] Pew's research found that the choice of weighting variables matters more than the method, that weighting widens the margin of error, and that it cannot remove most of the bias in opt-in samples.[5][6][7]
No survey sample is a perfect miniature of the population. Some kinds of people are easier to reach, some decline to take part, and pollsters sometimes interview extra members of small groups on purpose. Weighting is the statistical step that corrects for this by counting some respondents more and others less.[1][8] This page explains why it is needed, the main methods, and what research by Pew Research Center shows about its limits. For the earlier steps of a poll, see how opinion polls work.
Why samples need weighting
Without adjustment, polls tend to over-represent people who are easier to reach and under-represent those who are harder to interview, so pollsters weight their data to match the population.[8] Pew describes the combined effect of nonresponse and deliberate oversampling as leaving raw samples as imperfect models of the population, which weighting corrects by aligning the sample with known population characteristics.[1]
Oversampling is intentional. When a survey interviews extra members of a group so it can report on that group separately, the group’s share of the sample is larger than its share of the population, and weighting brings its influence on the overall result back into proportion.[2]
Suppose a poll ends up with too many university graduates. Weighting gives each graduate’s answers slightly less influence, and each non-graduate’s slightly more, until the sample’s education mix matches the population’s. The same is done for age, sex, race and region.[3][4]
What surveys weight on
Public opinion surveys have traditionally been weighted on a core set of demographics: sex, age, race and ethnicity, education and region. These are used because large, high-quality government surveys such as the Census Bureau’s American Community Survey measure them well, giving reliable population targets.[4]
Probability panels add a step at the start. Each member of Pew’s American Trends Panel begins with a base weight reflecting their probability of selection, and the weights are then calibrated to population benchmarks to correct for nonresponse and for people leaving the panel.[9] Pew also occasionally retires panel members from over-represented groups, such as more educated people, to avoid having to weight some panelists too heavily.[10]
The main methods
Raking, also called iterative proportional fitting, is the most common method in public opinion surveys. It adjusts each respondent’s weight over and over until the sample matches known population distributions on the chosen variables.[3] It needs only the population share for each variable on its own, not for every combination of characteristics, and it is the standard method used by Pew and many other public pollsters.[11]
Matching has been proposed as a way to adjust online opt-in samples. Each case in a representative target sample is paired with the most similar respondent from the opt-in sample, and respondents left unmatched are discarded.[12]
Propensity weighting rests on the principle that weighting each respondent by the inverse of their probability of selection removes bias from groups being represented in the wrong proportion. In an opt-in sample those probabilities are unknown and have to be estimated, and the resulting weights can vary widely.[13]
Pew’s 2018 comparison applied raking, matching and propensity weighting alone and in combination, using either core demographics or demographics plus political variables (voter registration, party identification, ideology and evangelical identity).[14] Complex methods never reduced average bias by more than 0.3 points beyond simple raking.[5] Highly variable propensity weights can also widen margins of error.[13]
The cost: a wider margin of error
Weighting helps avoid biased results but makes the margin of error larger. Statisticians call this increase in variability the design effect, and a margin of error that ignores it claims more precision than the survey warrants.[6] AAPOR’s disclosure standards ask probability surveys to report whether their sampling error has been adjusted for the design effect of weighting, clustering or other factors, and to describe how the weights were calculated.[15][16] Margin of error explains the figure itself.
What weighting cannot fix
Pew tested weighting on more than 30,000 online opt-in interviews from 2016. No procedure brought average bias across 24 benchmarks below 6 percentage points, down from 8.4 points unweighted, so even the best approach removed only about 30% of the bias.[7] The choice of variables mattered more than the choice of method.[5] Adding political variables changed substantive results: support for the Affordable Care Act fell from 51% to 46% when they were added to a demographic raking adjustment.[17]
Bigger opt-in samples did not solve the problem. Raking on demographic and political variables, the modelled margin of error shrank from plus or minus 1.8 points at 2,000 interviews to 0.5 points at 8,000, but average bias stayed at 6.3 points.[18] A small margin of error, in other words, says nothing about bias that weighting has not removed.[18][19]
Read next
Next in the crash course
Mean, median and mode: which average to trust
How the mean, median and mode differ, why outliers and skewed data pull the mean, and why income statistics usually report the median.
Questions readers ask
Why do pollsters weight their data?
Without adjustment, polls tend to over-represent people who are easier to reach and under-represent those who are harder to interview. Nonresponse and deliberate oversampling of some groups leave raw samples as imperfect models of the population, and weighting aligns the sample with known population characteristics.[8][1]
What is raking?
Raking, or iterative proportional fitting, is the most common weighting method in public opinion surveys. It repeatedly adjusts each respondent's weight until the sample matches known population distributions on the chosen variables, and it needs only the population share for each variable separately.[3][11]
What do surveys usually weight on?
Traditionally a core set of demographics (sex, age, race and ethnicity, education and region), because large government surveys such as the American Community Survey measure them well.[4]
Sources
· 19 claims from 2 publishers
Each numbered claim is a statement we checked against the sources listed with it. Tap a number in the text to see its sources in place.
- [1]
Pew says nonresponse and deliberate oversampling of some groups leave raw survey samples as imperfect models of the population, and weighting adjusts them by aligning the sample with known population characteristics. confirmedas of 2026-11-14
- The American Trends Panel · Pew Research Center · Primary source · Weighting (retrieved 2026-11-14)
- The American Trends Panel · Pew Research Center · Primary source · Weighting (retrieved 2026-11-14)
- [2]
When a survey deliberately oversamples a group so it can report on it separately, that group's share of the sample is larger than its share of the population, and weighting brings its influence on the overall results back into proportion. confirmedas of 2026-11-14
- The American Trends Panel · Pew Research Center · Primary source · Weighting (retrieved 2026-11-14)
- [3]
Raking, or iterative proportional fitting, is the most common weighting method in public opinion surveys; it repeatedly adjusts each respondent's weight until the sample matches known population distributions on the chosen variables. confirmedas of 2026-11-14
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · Raking (retrieved 2026-11-14)
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · Raking (retrieved 2026-11-14)
- [4]
Public opinion surveys have traditionally been weighted on a core set of demographics (sex, age, race and ethnicity, education and region), which are well measured by large government surveys such as the Census Bureau's American Community Survey. confirmedas of 2026-11-14
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · introduction (retrieved 2026-11-14)
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · introduction (retrieved 2026-11-14)
- [5]
Pew found that choosing the right weighting variables mattered more for accuracy than choosing the statistical method; adding political variables such as party identification cut more bias, while complex methods never did more than 0.3 points better than simple raking. confirmedas of 2026-11-14
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · key findings (retrieved 2026-11-14)
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · key findings (retrieved 2026-11-14)
- [6]
Weighting helps avoid biased results but also makes the margin of error larger; statisticians call this increase in variability the design effect, and a margin of error that ignores it claims more precision than the survey warrants. confirmedas of 2026-11-14
- 5 key things to know about the margin of error in election polls · Pew Research Center · Primary source · 2016-09-08 · section 5 (retrieved 2026-11-14)
- 5 key things to know about the margin of error in election polls · Pew Research Center · Primary source · 2016-09-08 · section 5 (retrieved 2026-11-14)
- 5 key things to know about the margin of error in election polls · Pew Research Center · Primary source · 2016-09-08 · section 5 (retrieved 2026-11-14)
- Margin of Sampling Error/Credibility Interval · American Association for Public Opinion Research · Primary source · Sample Size and MOSE (retrieved 2026-11-14)
- [7]
In Pew's 2018 study of more than 30,000 online opt-in interviews, no weighting procedure brought average bias across 24 benchmarks below 6 percentage points, down from 8.4 points unweighted, so even the best approach removed only about 30% of the bias. confirmedas of 2026-11-14
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · overview (retrieved 2026-11-14)
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · key findings (retrieved 2026-11-14)
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · key findings (retrieved 2026-11-14)
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · key findings (retrieved 2026-11-14)
- [8]
Without adjustment, polls tend to over-represent people who are easier to reach and under-represent those who are harder to interview, so pollsters weight their data to match the population, usually on several demographic measures. confirmedas of 2026-11-14
- 5 key things to know about the margin of error in election polls · Pew Research Center · Primary source · 2016-09-08 · section 5 (retrieved 2026-11-14)
- 5 key things to know about the margin of error in election polls · Pew Research Center · Primary source · 2016-09-08 · section 5 (retrieved 2026-11-14)
- [9]
Pew weights its American Trends Panel in several steps; each panelist starts with a base weight reflecting their probability of selection, and the weights are then calibrated to population benchmarks to correct for nonresponse and people leaving the panel. confirmedas of 2026-11-14
- The American Trends Panel · Pew Research Center · Primary source · Weighting (retrieved 2026-11-14)
- The American Trends Panel · Pew Research Center · Primary source · Weighting (retrieved 2026-11-14)
- [10]
Pew occasionally retires American Trends Panel members from demographically over-represented groups, such as more educated people, to avoid having to weight some panelists too heavily. confirmedas of 2026-11-14
- The American Trends Panel · Pew Research Center · Primary source · Panel management (retrieved 2026-11-14)
- The American Trends Panel · Pew Research Center · Primary source · Panel management (retrieved 2026-11-14)
- [11]
Raking needs only the population share for each weighting variable separately, not for every combination of characteristics, and it is the standard method used by Pew Research Center and many other public pollsters. confirmedas of 2026-11-14
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · Raking (retrieved 2026-11-14)
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · Raking (retrieved 2026-11-14)
- [12]
Matching, a technique proposed for adjusting online opt-in samples, pairs each case in a representative target sample with the most similar respondent from the opt-in sample and discards respondents left unmatched. confirmedas of 2026-11-14
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · Matching (retrieved 2026-11-14)
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · Matching (retrieved 2026-11-14)
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · Matching (retrieved 2026-11-14)
- [13]
Propensity weighting rests on the principle that weighting each respondent by the inverse of their probability of selection removes bias from groups being represented in the wrong proportion; for opt-in samples those probabilities are unknown and must be estimated, and the resulting weights can vary widely and widen margins of error. confirmedas of 2026-11-14
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · Propensity weighting (retrieved 2026-11-14)
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · Propensity weighting (retrieved 2026-11-14)
- How different weighting methods work · Pew Research Center · Primary source · 2018-01-26 · Propensity weighting (retrieved 2026-11-14)
- [14]
Pew's 2018 study tested raking, propensity weighting and matching, alone and in combination, each with basic demographics only and with demographics plus political variables (voter registration, party affiliation, ideology and identification as an evangelical Christian). confirmedas of 2026-11-14
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · overview (retrieved 2026-11-14)
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · overview (retrieved 2026-11-14)
- [15]
AAPOR's disclosure standards ask that probability surveys report their sampling error and say whether it has been adjusted for the design effect of weighting, clustering or other factors. confirmedas of 2026-11-14
- Disclosure Standards · American Association for Public Opinion Research · Primary source · Section III, A, sample sizes (retrieved 2026-11-14)
- [16]
AAPOR's Code of Professional Ethics and Practice asks researchers who release results to disclose who sponsored and conducted the research, the exact question wording, the population, how the sample was drawn, the dates, sample sizes and how the data were weighted, and AAPOR says the Code applies to all public opinion researchers, members or not. confirmedas of 2026-11-14
- Disclosure Standards · American Association for Public Opinion Research · Primary source · Section III, A (retrieved 2026-11-14)
- Disclosure Standards · American Association for Public Opinion Research · Primary source · Section III, A (retrieved 2026-11-14)
- Disclosure Standards · American Association for Public Opinion Research · Primary source · FAQ (retrieved 2026-11-14)
- [17]
Adding political variables to the weighting moved estimates in a Republican direction; support for the Affordable Care Act fell about 5 points, from 51% to 46%, when they were added to a demographic raking adjustment. confirmedas of 2026-11-14
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · key findings (retrieved 2026-11-14)
- [18]
In the same study, larger opt-in samples did not fix the bias; raking on demographic and political variables, the modelled margin of error fell from plus or minus 1.8 points at 2,000 interviews to 0.5 points at 8,000, but average bias stayed at 6.3 points. confirmedas of 2026-11-14
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · key findings (retrieved 2026-11-14)
- For weighting online opt-in samples, what matters most? · Pew Research Center · Primary source · 2018-01-26 · key findings (retrieved 2026-11-14)
- [19]
Sampling error is only one source of survey error; question wording, who is reached, how candid respondents are and likely-voter models can also introduce error, and unlike sampling error these are hard to quantify and rarely reported. confirmedas of 2026-11-14
- 5 key things to know about the margin of error in election polls · Pew Research Center · Primary source · 2016-09-08 · section 5 (retrieved 2026-11-14)
- 5 key things to know about the margin of error in election polls · Pew Research Center · Primary source · 2016-09-08 · section 5 (retrieved 2026-11-14)
Revision history (1)
- Page created.
Created Nov 14, 2026. Last reviewed by an editor on Nov 14, 2026. Next scheduled review: Aug 14, 2027.
Cite this page
"How surveys are weighted: raking, matching and propensity scores." ContentLora, updated Nov 14, 2026. https://contentlora.com/explain/how-surveys-are-weighted
Spotted an error? Suggest a correction or emailcorrections@contentlora.com.