New Arc Collective / FAQ
Frequently asked questions
How does New Arc Collective identify Muslim and Arab American voters?
We use two sources in combination. The first, the demographic model, is a licensed model that flags voters as likely Muslim. The second, the religion score, is a licensed model from a different company that rates each voter from 0 to 100 on how likely they are to be Muslim. A voter is in our Core + Likely universe, our primary working list, if the demographic model flags them and the religion score is 5 or higher, or if the religion score alone is 50 or higher. Core, our highest-confidence tier, takes only voters the demographic model flags and the religion score rates 21 or higher. Voters the demographic model flags but the religion score rates under 5 are left out.
Are you identifying individual voters by their religion, race, or ethnicity?
Not as a fact about anyone. We estimate who is likely part of the community, but we never know it for any one person. No one is asked their religion or ethnicity, and our data records nothing about anyone's actual religion. Our sources are statistical models, not records of anyone's religion or ethnicity. Being on our list never establishes any one person's religion or ethnicity, and our lists include some people who aren't Muslim or Arab American and miss some who are.
Do you know who people voted for?
No. How someone marked their ballot is secret, and no one, including New Arc, can see it. Whether someone voted, and whether they voted by mail, early, or on Election Day, is public record kept by the state. We use that record to track turnout, publish only group totals, and help partners avoid contacting people who have already voted. New Arc Collective is a nonpartisan 501(c)(3).
How do you protect voter privacy?
Our matching process runs inside a secure database. Our partner dashboard, public reports, and graphics show only counts and rates for groups, never individual records. In them, we suppress counts below 10, show a rate only when it is based on at least 10 people in a group of at least 50, and show data for a precinct only when at least 20 voters in our universe live there. Individual voter lists are shared only with approved New Arc partner organizations that have signed our Data and Partnership Agreement. We never use party or partisanship data to build our universe or our voter lists.
What are the known limitations of your data?
We are transparent about what our data does and does not capture. Arab Christian voters can't be reliably separated from Arab Muslim voters by name-based models. Requiring a religion score of at least 5 removes many voters the demographic model flags, but some remain in our universe, especially in Macomb and Oakland County suburbs with large Chaldean communities. Black American Muslims and converts are easy to miss, because both models work largely from names. Women are likely undercounted, especially in Minnesota. Minnesota's voter file doesn't record sex, so our gender breakdowns there are estimates. Name-based models struggle to identify Somali voters, so our counts in Somali communities are less certain. People who registered after our voter files were pulled (September 1 in Michigan, September 23 in Minnesota) aren't in our counts. We document these gaps publicly because accurate data requires knowing where the data falls short. New Arc also exists to identify these gaps and, as we grow, work toward better, more representative data for our communities.
What makes New Arc's approach different?
Three things distinguish our approach. First, we use two sources built separately by different companies. Our Core tier counts only voters the demographic model flags and the religion score rates 21 or higher, so it is smaller than either source alone and less likely to over-count. Because the two sources can share the same blind spots, their agreement is a consistency check, not proof. Our broader Core + Likely universe also includes voters the religion score alone rates 50 or higher. Second, in Michigan we build from the state's own voter file, matched by state voter ID for more than 99% of registrants, so our counts track the state's records. In Minnesota, we work from a licensed copy of the statewide voter file. Third, we publish our methodology and its main limitations publicly, and give partners the added guidance they need to use our lists well.
Has this methodology been validated?
Partly. Our two data sources agree far more often than chance would predict. Of the Michigan voters the demographic model flags or the religion score rates 21 or higher, about half meet both tests. Because the two sources can share the same blind spots, that agreement isn't proof. Our testing and review show that both align geographically with where Muslim and Arab American communities actually live, strongest in places like Dearborn and Hamtramck. In Minnesota, in every community we tested, including St. Cloud and Minneapolis, the demographic model flags a larger share of voters than it does across the state as a whole. We have not checked either model against voters' own answers about their religion. Voter files don't record religion, so we can't say what share of our universe is Muslim. We document what has and has not been tested so that users of the data can make informed decisions.
Why are New Arc's numbers different from other estimates?
Mostly because different groups count different lists. Our turnout numbers count our Core + Likely universe of registered voters: about 195,000 in Michigan and about 84,000 in Minnesota. A list built from either of our sources alone comes out a different size than ours, and counts of all adults, or of all Muslim and Arab American residents, run larger than any voter list. Percentages differ too: Michigan's own turnout figures divide by active registered voters, while ours divide by all registered voters. The fair comparison is a list against itself over time.
How current is your data?
Our dashboard updates each morning when new ballot data comes in. Ballot status comes from state records and usually lags the state's count by a day or more; a mailed ballot shows up only after the clerk records it. Our registration data comes from voter files pulled September 1, 2026, in Michigan and September 23, 2026, in Minnesota, so people who registered after those dates aren't in our Muslim and Arab American counts.
Who can access New Arc's voter data?
Voter lists are available only to New Arc partner organizations that have signed our Data and Partnership Agreement and been approved for list access. Partners also get a dashboard that shows counts, rates, and turnout trends for groups, never individuals. We use those group-level counts and rates to create reports, graphics, and analysis we share with the public. Nothing we publish identifies an individual voter.
Who can be a New Arc partner?
We partner with Muslim- and/or Arab-serving nonprofit and nonpartisan community organizations that are working to increase civic and voter engagement in their community. Partners must be registered nonprofit organizations, operate in a state where New Arc is active (currently Michigan and Minnesota) or expanding, and agree to the terms of our Data and Partnership Agreement, which governs how shared data is used and protected.
Why partner with New Arc?
Because the data infrastructure that makes civic engagement more effective has rarely been built for our communities. Approved New Arc partners get access to Muslim and Arab American voter lists, list-building tools, turnout tracking, and a coordinated network of organizations working toward the same goal: increased Muslim and Arab American voter participation. You bring the community relationships and organizing capacity. We bring the data and coordination infrastructure to make that work go further. New Arc exists to fill a gap, build your capacity, and offer resources, not to duplicate or replace the work you are already doing.
Is there a cost to partner with New Arc?
There is no cost to become a New Arc Collective partner. Our data tools, list access, and coordination support are provided to approved partner organizations at no charge. New Arc is funded through foundation grants and donor support specifically so that the organizations doing this work are not priced out of the infrastructure they need to do it well.
