Methodology

How we run the Pasquines Polls: who we ask, how we invite them, and how we weight their answers. Our polls ask registered voters in Puerto Rico, Guam, the US Virgin Islands, American Samoa, and the Northern Mariana Islands about the races on their ballots, political status, and the issues in their communities.

Approved by

William-José Vélez González

Editor in Chief

Last updated October 29, 2026

All polls

At a glance

Who

Registered voters

Residents of the five inhabited United States territories who say they are registered to vote. Answers from people who are not registered are left out of the results.

How

Online and opt-in

Anyone with the link can answer our online form. We invite people through our emails, our social media accounts, and paid social media ads, so the poll is not a random sample.

Weighting

By past vote

We weight each answer by how the respondent voted in a recent territory-wide race, so the poll matches the official result.

Error

No margin of error

A margin of error only applies to random samples. Since people choose to take our polls, we report how many people answered instead.

How we apply it

Go through the five steps from question to published result, then try the weighting yourself.

1. Ask

One online form serves all five territories. Everyone answers the shared questions: whether they are a registered voter, where they live, their age, gender, and education, and the year they first voted. Then each person answers the questions for their own territory: the races on the ballot, political status, their opinion of local and national figures, and how they voted in a recent territory-wide race.

The shared questions and Puerto Rico’s questions are in English and Spanish at the same time. The questions for Guam, the US Virgin Islands, American Samoa, and the Northern Mariana Islands are in English.

2. Invite

The form is open to anyone with the link for a set period. We share it in our emails, on our social media accounts, and in paid social media ads.

People choose whether to answer, so the poll is opt-in, not a random sample of voters.

3. Sort

We keep the answers of people who say they are registered voters and tally each territory on its own.

We count every response submitted. The form does not stop someone from answering more than once.

4. Weight

People who answer an opt-in poll are rarely a mirror of the voters. To correct for that, we weight each answer by how the respondent voted in a recent territory-wide race, so the poll’s past vote matches the official result.

Weight = official share of the vote ÷ share of respondents

People who did not vote, or preferred not to say, keep a weight of 1. If no respondent voted for a candidate, that candidate is left out and the other official shares are scaled to add up to 100%. For 2026, we use the 2024 governor’s race in Puerto Rico, the 2022 governor’s race in Guam, the US Virgin Islands, and the Northern Mariana Islands, and the 2024 delegate race in American Samoa.

5. Report

We publish results by territory with the number of respondents and the weighted percentages. Starting with the 2026 poll, every figure in our reports is weighted. In 2024, our main results were weighted and some charts were not.

Approval voting questions let people pick every option they approve of, so their results add up to more than 100%. We do not report a margin of error (see Poll basics below), and the responses to our past polls are on the Polls page.

Try it

How weighting changes a result

An example with made-up numbers. In the last election, candidate A won part of the vote and candidate B the rest. In our example poll, 70% of A’s voters and 30% of B’s voters support a measure.

60%75%

A voters, weight 0.80B voters, weight 1.60

Raw60%
Weighted54%

A’s voters are 75% of respondents but won 60% of the vote, so each of their answers counts 0.80 times and each B voter’s counts 1.60 times. Support for the measure goes from 60% raw to 54% weighted.

Poll basics

A few ideas that help when reading any poll, ours included.

Try it

Why the size of a sample matters

An example with made-up numbers: 55% of all voters support a measure. Each dot is one poll of randomly chosen voters. Pick a sample size and run 20 polls.

With 100 people, 95 of every 100 random polls land within 9.8 points of the true 55%. That range is the margin of error.

Sample

A few stand in for many

A poll asks some people to learn about everyone. The closer the people who answer are to all voters, the better the poll reflects them.

Error

Margin of error

For a random sample, it is the range around the result where 95 of every 100 such polls would land. Four times as many people only cuts it in half.

Weighting

Fixing known gaps

If a group answers more than its share, each of its answers counts less, and the reverse. Weighting only fixes the gaps it measures, like past vote.

Approval

Pick all you approve

Approval voting questions let people pick every option they approve of. Results add up to more than 100% and show how broad each option’s support is.

Read with care

What our polls can and can’t tell you

  • Opt-in: people choose to answer. Weighting by past vote corrects one known difference between them and all voters, not every one.
  • Small samples: in some territories, few people answer. Results for a territory or a group within it can move a lot with a few answers, so check how many people answered.
  • Repeat answers: the form does not stop someone from answering more than once, and we do not remove responses.
  • Language: the questions for Guam, the US Virgin Islands, American Samoa, and the Northern Mariana Islands are in English only, which can leave out residents more comfortable in another language.
  • Wording: as in any survey, how a question is worded and the order of its options can shape the answers.

Questions about our methods? Write to us.

[email protected]