How do state‑level estimates of Muslim populations (e.g., New York, Michigan, California) differ between mosque-based counts and household surveys, and why?
Executive summary
Estimates of state‑level Muslim populations differ substantially depending on whether a researcher counts people connected to mosques or uses household surveys and demographic projections; mosque‑based tallies often undercount non‑attached Muslims or overcount by extrapolating attendance patterns, while household surveys and demographic models (Pew, RLS, US Religion Census) produce divergent totals because of sampling, weighting and methodological choices [1] [2] [3]. In practice this means states like New York, Michigan and California can appear to host very different Muslim totals depending on the method: mosque counts emphasize visible institutional footprints, surveys and projections capture broader self‑identified populations [4] [5] [6].
1. Why mosque‑based counts and household surveys disagree: two different animals
Mosque counts measure institutional presence—numbers of mosques, reported congregants at key services (e.g., Eid), or leaders’ estimates—and then extrapolate from those figures; household surveys sample residents and ask about religion or infer identity through demographic techniques, and demographic projections combine births, deaths, migration and switching models [7] [1] [2]. Mosque surveys therefore capture “mosque‑connected” Muslims (Bagby’s work estimated roughly 2.6 million mosque‑connected people in a past study) but miss the many Muslims who rarely attend or are unaffiliated, whereas household surveys aim to include non‑attenders but face sampling limitations and rely on self‑reporting and weighting choices that can shift totals [1] [7] [2].
2. State examples — New York, Michigan, California: what the methods show
States with dense urban Muslim communities—New York, Michigan and California—consistently show high mosque concentrations and thus high mosque‑based counts, but household surveys produce different magnitudes: New York and California are repeatedly cited among the states with largest Muslim populations by mosque counts and mosque density studies [4] [5], while Pew’s national projection that yields about 3.45 million Muslims nationwide implies state splits different from the 2020 U.S. Religion Census’s larger 4.5 million estimate and from mosque extrapolations used by community surveys [3] [6]. In short, mosque inventories highlight concentrated urban infrastructures in Queens, Dearborn and Los Angeles areas (visible in local reporting), but household surveys and projections tend to allocate a broader base of Muslims across suburbs and smaller metros that mosque lists miss [4] [8].
3. Mechanisms producing systematic gaps
Three mechanisms drive systematic gaps: attendance variance and social desirability bias (many Muslims report attending mosque only a few times a year—Pew finds 74% do so a few times a year—which depresses weekly counts but inflates reported affiliation) [1]; duplicate service use and multiple mosques per community (one person may connect to more than one mosque, complicating per‑mosque extrapolations) [4]; and sampling/weighting differences in surveys (household panels and large Religious Landscape Studies use different modes, oversamples and post‑stratification that change state allocations) [9] [2]. These create both undercounts (non‑attenders, language barriers, survey nonresponse) and overcounts (extrapolating Eid attendance or using organizational incentives) [7] [10].
4. Hidden incentives and competing agendas in the data
Organizations behind each method have incentives: mosque surveys often come from community research centers and faith‑based coalitions that use counts for service planning or advocacy and may emphasize higher “mosqued” numbers (ICNA/AMRDC using Eid extrapolations produced higher estimates) [7], while independent research centers like Pew prioritize transparent demographic modeling and conservative projections; think tanks or funders (as noted in Justice For All’s summary citing AJCommittee‑funded work) may also produce lower survey‑based estimates, reflecting methodological choices and potential institutional agendas [11] [2]. Media and policy actors then pick the figure that best fits narratives about growth, integration or resource needs [10].
5. What this means for researchers, policymakers and the public
The takeaway is pragmatic: use mosque counts to understand institutional infrastructure and local service demand (where mosques, halal stores and social centers concentrate), but rely on household surveys and demographic models to estimate total populations for policy, funding and electoral analysis—and always interrogate mode, sample size and weighting [5] [9] [2]. No single number is definitive; comparisons across New York, Michigan and California require triangulation between mosque inventories, Pew‑style projections and the U.S. Religion Census—and transparency about each method’s limits [3] [1].