FAQ
Template page for common questions about the site, methods, and data
How can I trust the statistics on this website?
Accuracy is a formost value for Civic Data Atlas. Several things should add to users trust about the reiability of district statistics: 1) founder bio 2) not vibe coded 3) extensive checks 4) documntation of methadology
Who created Civic Data Atlas?
Civic Data Atlas was founded by Jack Landry, a policy analyst and researcher whose work invoves extensive analysis of Census data. Before creating Civic Data Atlas, he worked on serveral projects that involved computing city-council district statisitics on an ad-hoc basis.
How was AI used in the creation of this website? Was it “vibe coded”?
The website’s overall architecture was heavily dependent on AI assistance—without it, it would not look nearly as good. However, the code to create district-level data was coded by hand. In fact, the founder was making local district-level data on an ad-hoc basis well before widespread AI use in coding. Writ large, we have a deep understanding of what all the code involved in the district-stat calculations is actually doing, in addition to a battery of validation tests to check for potential errors. Having a good understanding of the code powering the website’s aesthetics is not very important, as errors are self-apparent—when something doesn’t load or looks ugly, it can be quickly caught and fixed. District statistic calculation errors are more insidious because you can simply produce wrong data without realizing it—hence the importance of a good understanding of the underlying code as well as validation tests.
How do you validate the data against errors?
We take many precautions to guard against calculation errors, beyond the sampling error and geographic mismatch error that are intrinsic to producing statistics at the local district level. For each city, we reproduce the decennial census district-level population and racial breakdowns, comparing the results to what is published by the city itself. This provides an important check that we are correctly mapping census boundaries to district boundaries and that there are no mismatched districts. Because of the imprecision of city-provided district shapefiles, some cities have very small discrepancies between our calculations and city-provided data. We do the same thing for the ACS data measuring the same underlying concepts, though because these data sources and time periods are different, they will not match exactly. We flag any large discrepancies for additional review.
How often is the data updated?
The main demographic data found on district pages is updated once a year when the Census Bureau publishes new data. The last release was on March 5, 2026.
We attempt to continually update district representative info. For each city, we continually check city-council roster websites for updates. If you see anything out of date, please contact us and we will make a correction.
Why does some data differ from demographic data provided after redistricting?
As part of the redistricting process, most cities report individual districts’ population and racial breakdown. See examples for Los Angeles, New York City, and Chicago. Data for these city reports comes from the 2020 decennial census, while individual district page data comes from the 2024 5-year American Community Survey (ACS). These two sources can differ for four different reasons. First, the 2024 5-year ACS is conducted between 2020 and 2024, while the census was conducted in 2020 and refers to the population as it existed on April 1, 2020. The population and racial composition of the district could have shifted between 2020 and more recent years. Second, the ACS is a survey with sampling error, which could make the ACS-reported figures a few percentage points different from the census by random chance, not because the population actually shifted. Sampling error is described in more detail on the methodology page, and the ACS margin of error is reported on all district pages. Third, the ACS survey comes with geographic aggregation error: it cannot bound survey responses from precisely within district boundaries, which can lead to errors. Geographic aggregation error for race variables is usually very small and directly testable. The methodology page has more details, and each district geographic aggregation error page has specific error figures. Finally, some states edit census data for redistricting so that people incarcerated in prisons and jails are counted at their home addresses rather than the correctional institution’s address. In other words, the census-reported population and racial breakdown can be slightly different from the composition of people actually living in the district because it counts incarcerated populations with home addresses in the district as living there. The ACS data has no such correction, which can result in differences from the Census data.
Why do the “City at a Glance” statistics differ from those on the district pages when comparing a district to the citywide totals?
The “City at a Glance” statistics use population statistics from the latest 1-year ACS data, while all city-wide statistics when doing district comparisons use 5-year ACS data. District-level data can only be produced using 5-year ACS data, while citywide data is available with both 1-year and 5-year versions. To compare districts to the entire city on an apples-to-apples basis, we must use the citywide 5-year data. But to see the most current city-wide statistics, we prefer to use 1-year data, which better reflects present conditions (though in practice, the differences are very small).
How do you handle prison or jail populations?
If a district includes a prison or jail, the people incarcerated there will be reflected in some of the district’s demographic statistics. Any table where the universe is labeled as total population or population of a certain age range, such as population 25 years and over for education statistics, will include the incarcerated population. The incarcerated population will be excluded from many tables where their inclusion would not make sense. For instance, a district’s median rent comes from surveys of housing units, excluding “group quarters,” which includes prison and jail populations. More broadly, any table whose universe is “households,” “housing units,” or “noninstitutionalized” will not include people incarcerated in prisons or jails.
When people incarcerated in prisons or jails are included in district-level estimates, they are counted where the jail or prison is located, not at their home address where they would be located if they were not incarcerated. This differs from how some states treat the incarcerated population for the redistricting process: they count the incarcerated population as residing at their home addresses rather than where the prison or jail is located.
How are other special populations counted, including students in college dorms, juvenile detention residents, nursing-home residents, military barracks residents, and people in shelters or group homes?
These special populations, along with people incarcerated in prisons and jails, are all considered to be living in “group quarters.” This means they are counted in tables whose universe is “total population,” but they will be excluded from tables whose universe is “households” or “housing units.” Tables where the universe is labeled “civilian noninstitutionalized population” exclude incarcerated people in prisons and jails, people in juvenile detention, nursing-home residents, people in mental hospitals and similar institutions, and active-duty military living in barracks or military quarters. College dorm residents are included, as are some group-home and shelter residents that are not categorized as living in institutions.
What does “Universe” mean in the data tables?
In the data tables, “universe” refers to the population the statistic applies to. For instance, when the universe is the total population, it means it counts everyone living in the district; if the universe is “Population 25 years and over” (the universe for the education table), it means it counts everyone above age 25—those younger are not included.
Sometimes universes refer to “households” rather than individuals. For instance, the Household Size table’s universe is “Occupied housing units.” This means that the statistic counts all occupied housing units—each household is counted once, regardless of how many people live in the household. This is an important distinction. For instance, if a district had 50% one-person households and 50% two-person households, most people would live in a two-person household, but households would be evenly split between one-person and two-person households.
Sometimes a universe is at the household level, but refers to a specific individual, the Primary Resident in Household. These statistics refer to one person per household, even when in some cases not all people in the household would have the same answer. For instance, the Years in Current Residence table’s universe is Primary Resident in Household—some household members could have a different number of years living there. The definition of primary resident is the person who owns or leases the residence—if co-owned or multiple names are on the lease, either person could be the primary resident. If no one formally owns or rents the property, any adult member of the household could be the primary resident.
Finally, some universes refer to the “Civilian noninstitutionalized population.” This generally means the entire population of the district except those living in correctional facilities and nursing homes—facilities that the ACS considers “institutions.”
What does “Geography” mean in the data tables?
All district statistics are produced by aggregating up from smaller census-defined geographies to district boundaries. Those smaller census-defined geographies are either census block groups or census tracts. Census block groups are smaller than census tracts, and are generally preferred because smaller geographies can better match district boundaries. However, not all statistics of interest are released by the census at the block group level; some are only released at the tract level. For these statistics, we are forced to use tract-level data. Each table shows what geographic level the statistic is aggregated from: block group level or tract level.
What happens when a Census tract or block group crosses a district boundary?
When a census tract or block group crosses over a district boundary, we incorporate data from the entire tract or block group, as it is not possible to restrict the census data to the exact district boundaries. For each tract or block group that extends past the district boundary, we account for its contribution to the overall district average by weighting the fraction of the population of the tract or block group that is within the district boundary (using 2020 decennial census block-level data, which is disaggregated at high levels of geographic detail that perfectly align with district boundaries).
What does the “±” (Margin of Error) mean on the data tables?
The American Community Survey is a survey conducted by the U.S. Census Bureau. Unlike the decennial Census, it surveys a representative sample of households—it does not attempt to survey everyone. Since a sample-based survey doesn’t count everyone, it comes with sampling error—random variation that comes from the fact that a random sample won’t select the same people with the exact same characteristics every time. The tables give more information about the random variation by allowing the user to toggle the margin of error. Margin of error is a way of quantifying that random variation. The margin of error means that if the ACS survey was reconducted with new people surveyed, the result should be within the margin of error 90% of the time. For instance, if a district page reports that 50% of households are renters with a ±2% margin of error, if the ACS was repeated, 90% of the time, the survey would result in a renter percentage within 2 percentage points of 50%—48% to 52%.
Can I bulk-download the data for my own use?
We are happy to provide all district-level data for non-commercial uses—please contact us.
What are your plans for the future?
We have many ideas for expanding the data and usefulness of the site going forward. High-level ideas include adding more features to district pages, like comparing districts to one another rather than just the city writ large. Historical data could be added, showing how districts have changed over time. Data could be sourced from more places than the Census’s American Community Survey, and far more ACS stats could be published than what is currently displayed. And of course, more cities could be added. We also welcome your ideas—please contact us.