Methodology: Muslim Footsteps, 1920–2026

Prepared on 23 October 2026 and updated 26 October 2026 (America/New_York). This page describes how the mosque, officeholder and Sharia-law layers were assembled, which sources were used, how each field was derived, and where the data are known to be weak. The guiding rule was that no founding year, coordinate or denomination was invented. A field is blank whenever no source supplied it.

1. Definition of a "mosque"

The national calibration series (Bagby / ISPU / CAIR U.S. Mosque Surveys of 2000, 2011 and 2020) defines a mosque as a Muslim organization that holds Jum'ah (Friday) prayers and controls the use of its space. No source in this dataset applies that test directly. The working definition here is therefore broader: a site or organization that at least one source describes as a mosque, masjid, Islamic center or Muslim place of worship. The following are excluded:

• Muslim schools, charities, student associations, media organizations, councils, cemeteries and similar bodies that are not places of worship (IRS/HIFLD records classified muslim_org_non_mosque).
• Non-Muslim organizations caught by name patterns (for example "Shiatsu", "Sunnyside", "Messiah", "Shiawassee"), classified false_positive_non_muslim.
• IRS-only records whose name is ambiguous (NTEE X40 "Islamic" without a mosque-like name, or a name-pattern hit without the NTEE code). They were set aside for manual review rather than discarded silently, and are kept in the dataset only when an independent source (OSM, MosqueIndex, Wikidata, Wikipedia) places a mosque at the same location.
• Moorish Science Temple of America organizations, which are not generally counted as mosques in the national surveys. Nation of Islam mosques are kept and labeled as such where a source supplies that denomination.

Because the working definition is broader than the survey definition, and because IRS records sometimes represent more than one organization per site, raw totals in this dataset are not directly comparable with the survey totals. The survey totals are shown on the map as an overlay for calibration only.

2. Sources and outcomes

3. Geocoding

• Coordinates are taken from the best available source in this order: curated record geocoded at street address (U.S. Census Geocoder, benchmark Public_AR_Current); OpenStreetMap feature; MosqueIndex record derived from OSM; Wikidata P625; MosqueIndex record derived from Google; HIFLD/IRS address geocode (HERE for HIFLD, Census batch for new IRS records); city centroid from Nominatim.
• Nominatim was called at no more than one request per 1.1 seconds with an identifying User-Agent, and every response was cached.
• coord_precision records the outcome: point (a mapped feature), address (street-address geocode), approximate (HERE match below address level), or city (city centroid). City-centroid points are plotted but should not be read as locations.

4. Deduplication

Records from all sources were clustered with union-find over a spatial grid. Two records were merged when any of the following held:

1. Within 150 m and name similarity of at least 0.3 (token-based, after removing generic words such as "Islamic", "Center", "Masjid", "Inc").
2. Within 400 m and name similarity of at least 0.75.
3. Within 40 m and from different sources.
4. The same normalized street address (house number, first street token and ZIP code).
5. The same city and identical distinctive name tokens, when at least one record has imprecise (city or approximate) coordinates.
6. Wikipedia list rows matched to an existing record by name and state (similarity of at least 0.75).

Curated historic records are excluded from rules 5 and 6 so that a historic building is not merged with a later congregation of similar name. Some duplicates certainly remain (for example, an IRS organization registered at a mailing address far from the mosque), and some distinct congregations sharing a building may have been merged.

5. Founding year and timeline year

Two separate concepts are kept apart:

• founding_year is populated only from a source that states a date for the mosque: curated citations, Wikidata P571, the Wikipedia list "Year" column, or the OSM start_date tag. Priority is curated > Wikidata > Wikipedia > OSM. founding_year_precision is exact, year, approximate year or decade. founding_year_basis states what the date actually measures (establishment, building completion, opening, construction), because sources differ. Conflicting dates from other sources are recorded in founding_year_notes. Where the Wikipedia or OSM date may refer to the building rather than the congregation, the basis says so.
• timeline_year is the year the map uses. It equals founding_year when one exists (timeline_basis = documented). Otherwise it is the earliest IRS exemption ruling year among the merged IRS records (timeline_basis = inferred: IRS exemption ruling year …). A ruling year is the year the IRS recognized the organization's exemption. It is not a founding year: many congregations operated for years before applying, some re-incorporated, and a mosque may hold its exemption through an umbrella body. Inferred years should therefore be read as "no later than roughly this year, for the organization", and they are styled differently on the map (hollow blue circles rather than solid red).
• Records with neither are undated. They are hidden from the map by default and can be shown, grey, by switching on the undated option in the map legend.

No founding year was estimated, interpolated or inferred from any other field.

6. Cole & Crescentini (GVSU) outcome

The Cole & Crescentini study was given top priority. The following were attempted on 23 October 2026:

• ScholarWorks@GVSU records fsdg/51 ("Diffusion of an Ideology: Mosques and the spread of Islam in North America, 1920 to 2010", East Lakes AAG meeting, 2010–11) and fsdg/52 ("Muslims without Mosques, Mosques without Muslims …") both state "This document is currently not available here." Neither page has a download link or a citation_pdf_url tag. The related Cole paper fsdg/182 ("Building Muslim Community in the United States", 2011) also says the document is not available.
• Guessed viewcontent.cgi URLs returned a Cloudflare 403 challenge. Those article identifiers were guesses, so the 403 is not evidence that a file exists.
• The Wayback Machine returned HTTP 429 on the first attempt; a later retry of the CDX index returned an empty response for both the fsdg/51 page and fsdg viewcontent files, and a direct snapshot request timed out. This may reflect throttling rather than the absence of captures.
• Web searches for a ResearchGate, Academia.edu or author-hosted copy found none.

Result: the study contributed no mosque-level records. Record counts before and after this step were unchanged. Its abstracts supply only qualitative aggregate findings, which are used as calibration: the authors compiled mosque location and establishment date from Muslim-service websites, the Pluralism Project and mosque websites, found two peak periods of establishment (the early 1980s and the late 1990s), and found that new establishment "dropped off markedly after 2001".

7. Calibration figures

Only the 1994–2020 figures share a methodology. The often-repeated figure of "about 100 mosques in 1975" could not be traced to a primary source and is not used. A figure of 3,150 for the present, attributed to Islam Global News, has no stated method and is not used.

Purpose-built mosques (ISPU): 314 in 2000, 632 in 2010, 1,025 in 2020. These are plotted as purple squares on the chart. In this dataset purpose_built is populated only where a cited source states it. The OSM tag building=mosque is reported separately as osm_building_mosque and is only a weak proxy, since mappers use it for any building used as a mosque.

ARDA survey microdata (region-level only) give these aggregate distributions, which were used to sanity-check the shape of the timeline:

• USMAS2020 (n=470) founding decade: before 1960s 12; 1960s 15; 1970s 48; 1980s 79; 1990s 102; 2000s 113; 2010s 88; missing 13. Present facility: purchased 235, built by the community 170, rented 50, a room 6. Tradition: Sunni 442, Shi'i 28.
• USMAS2011 (n=524) founding decade: before 1960 25; 1960s 28; 1970s 69; 1980s 114; 1990s 144; 2000s 124; 2010–11 10. Facility: purchased 277, built 148, rented 71. Sunni 489, Shi'i 35.
• MOSQUE2000 (n=416) founding decade: 1920s 12; 1930s 7; 1940s 1; 1950s 20; 1960s 24; 1970s 102; 1980s 129; 1990s 119; 2000 2. Sunni 405, Shi'i 4.

8. Denomination

denomination is populated only from an explicit source field: curated citation, OSM denomination tag, Wikipedia list "Group" column, MosqueIndex "Tradition" field, or the Wikidata P31 type. Spellings were normalized (for example sunni, suni → Sunni; shiite, shia → Shia; nation_of_islam → Nation of Islam). Values that are not denominations ("muslim", "congregational mosque", a place name) were dropped. It was never inferred from a name.

9. Coverage after 2018 and current status

HIFLD's IRS extract ends around 2018. The October 2026 IRS BMF was added to cover 2019–2026 and to flag HIFLD organizations that are no longer in the BMF (status). The figure for recent years therefore depends almost entirely on IRS ruling years, which lag founding and include re-registrations. OSM and MosqueIndex contribute currently existing mosques but without dates.

10. Known limitations

• Documented founding years cover only a small share of records. The documented-only curve on the chart is a floor, not an estimate.
• Pre-1960 coverage is very thin. Many early prayer spaces (homes, rented halls) left no record in these sources.
• IRS-inferred years are biased late and, around 2000–2020, also reflect changes in IRS processing and the 2010–11 automatic revocations.
• Counts differ from the survey totals because of the broader definition, residual duplicates, musallas and non-Jum'ah prayer spaces listed in OSM, and organizations that have closed. They should not be read as a national estimate.
• This public version omits mosque records that depend on Google-derived MosqueIndex listings (see Section 11), so it under-counts currently existing mosques.

11. What this public version contains

The working dataset holds 4,485 mosque records. Of these, 2,024 depend on MosqueIndex listings that the site derives from Google, which it states are not offered for redistribution. Those records are not published here. This page therefore maps 2,442 mosques (129 with a documented founding year, 1,806 with a year inferred from an IRS exemption ruling, and 507 undated) out of 2,461 redistributable records (19 lack usable coordinates). The omitted records are mostly currently existing mosques that have no founding date, so the public version understates present-day mosque counts more than it affects the historical timeline.

12. Officeholder and Sharia-law layers

• Officeholders. Each pin is a person documented in a sourced research table as a Muslim holding public office, shown at the year the office was begun. Only people with a sourced start date are plotted. The pin sits at the city or county seat of the jurisdiction where known and otherwise at the state capital, so many pins are approximate locations; overlapping pins are spread apart slightly. Where no source states a person's religion, the person is not added. Each pin's popup links to its own source, and all officeholder sources are listed under "Sources" on the map page.
• Sharia-law milestones. Only documented legal actions are plotted, each with court or primary-source citations. The single entry is Oklahoma State Question 755 (2010), a ballot-approved constitutional amendment that courts blocked; it is plotted as a blocked ban, not as an enacted law. No claim that a U.S. city adopted a Sharia-compliant law is plotted, because none of the sources collected supports one.

13. Results of the public build (mapped records only)

• Founding-year precision for all 2,461 redistributable records: exact date 5, year 131, decade 2, none 2,323.
• Mosques mapped: 2,442. Timeline basis: documented 129, inferred from IRS ruling year 1,806, undated 507.
• Denomination is populated for 258 records.

Cumulative counts against calibration

The public dataset sits well below the survey series at every survey date, partly because it omits the records described in Section 11 and partly for the reasons in Section 10. It should be read as a partial, sourced sample and not as a count of U.S. mosques.