Pew Research Center

Pew Research Center built its religion research program on a method it uses across all its polling: large probability based surveys asked the same way across years so change can be measured rather than guessed at. The Religious Landscape Study is the flagship, a survey of more than thirty five thousand American adults fielded in 2007, 2014 and again in 2023 to 2024, each wave sorting respondents into named Christian traditions, Evangelical, Mainline and Historically Black Protestant, Catholic, Orthodox and several smaller groups, named non Christian traditions and three separate categories of religious unaffiliation, atheist, agnostic and nothing in particular, rather than lumping every unaffiliated respondent into one row.

The Center's global work extends the same method outward. Its 2012 Global Religious Landscape report and its 2025 study of how the picture changed between 2010 and 2020 estimate the religious composition of more than two hundred countries and territories, the figures this atlas draws on for national and state level religious affiliation facts wherever a country or an American state carries one.

Because Pew is organized specifically as a nonpartisan fact tank that takes no policy position of its own, its religion numbers are cited here as a demographic snapshot rather than as an argument for any conclusion about belief. Where a country's religious composition is disputed or a state's affiliation figures shift between survey waves, the atlas records what the Center's own methodology section says about margin of error and survey design rather than treating either wave's number as final.

Pew Research Center's most cited numbers, the ones this atlas draws on for national and global religious composition, come from a single standing instrument called the American Trends Panel. Built from 2014 onward, the panel now holds roughly ten thousand adults recruited at random from across the United States, not volunteers who signed up on their own. Early recruitment used landline and cell phone random digit dialing; since 2018 the center has recruited by address based sampling instead, mailing an invitation and a small incentive to a random sample of households drawn from the Postal Service's own delivery file, which covers nearly the whole country. Panelists who agree to join answer self administered surveys online roughly once a month, in English or Spanish, and anyone without home internet access is given a tablet and a wireless connection rather than being dropped from the sample for lacking one. What began as a supplement to telephone polling has, as phone response rates fell and phone surveys grew more expensive to run, become the center's primary way of collecting data. Raw answers are not published as counted; every wave is weighted in several steps, starting from each panelist's own probability of having been selected at all and then calibrated against known population benchmarks to correct for the people who declined to join and the ones who drift away over time. That is the machinery behind the Religious Landscape Study and the Global Religious Landscape estimates this atlas cites again and again: not a single poll but a maintained panel, rebuilt from the sampling frame up, with its weighting and its coverage published alongside every release for anyone who wants to check the arithmetic themselves.