Methodology
How we count, and what the numbers cannot tell you
Every figure on this site is either a count taken from an official file or an estimate produced by a VoteIQ model and labeled as one. This page says which is which, how each is built, and where each one stops being reliable. If you are deciding whether to cite something here, this is the page to read first.
Where the data comes from
Three sources, and they do different jobs. The NC State Board of Elections supplies the statewide voter file, vote history, and candidate filings. The Wake County Board of Elections supplies the daily early-vote file during an election period. The U.S. Census Bureau supplies a geocoder, which turns an address into a location so a voter can be placed in a neighborhood, and block-group statistics that feed VoteIQ’s internal models.
No Census figure appears in any count on this site. Registration and turnout numbers come from the election boards alone.
What this cannot tell you
How often it updates
During an election period the Wake County file arrives around 6 AM and the refresh finishes by about 7 AM, so “updated each morning” is literal rather than a figure of speech. Off season, the site refreshes weekly, after the NC State Board of Elections publishes its Saturday file.
Every page states the vintage it was built from, so a figure you cite can be reproduced from the same file later.
What this cannot tell you
What counts as registered
Registered means active and inactive registrations together. That is the same basis the Wake County Board of Elections uses for its own published total, which is why our county number matches theirs.
What this cannot tell you
How a voter is assigned to a district
We do not place voters into districts ourselves. Precinct, NC House, NC Senate, US House, county commission, school board, municipality, and council ward all come from the state’s own fields in the voter file. The only judgment we add is which of those to show: council wards are published for Cary and Raleigh, the two Wake municipalities that elect by ward.
What this cannot tell you
How turnout is counted
This is the number most likely to differ from one you have seen elsewhere, so it is worth being precise. Past-election turnout on this site counts voters who are on the Wake roll today and who voted in that election, attributed to the district they live in today. The denominator is the current registered population of the same district.
Read it as: of the people who live here now, how many turned out in that cycle. It is internally consistent, because the numerator and denominator are built on the same present-day boundaries.
What this cannot tell you
These figures are lower than the official results, by design. Voters who have since moved away or been removed from the roll are not counted, because they are not in today’s electorate. If you compare a VoteIQ turnout figure against the official count for the same election, ours will be smaller, and the gap is those departed voters.
District boundaries also moved between cycles. A cycle that predates a redistricting is a reconstruction on the current map, not what was reported at the time.
Coverage is general and primary elections from 2016 onward. Municipal elections are not included in these figures yet.
How small numbers are protected
Any published cell built from fewer than 10 voters is withheld at precinct and other fine scopes, so that no individual can be identified by narrowing a filter far enough. The rule is applied in the database function that builds each aggregate, not in the page that displays it, so it cannot be forgotten downstream.
District-level figures are not suppressed, because a district contains tens of thousands of voters and no cell at that scope comes close to the threshold.
What this cannot tell you
What is counted, and what is modeled
Almost everything on a public page is a count. Exactly one figure is not: the enhanced Asian and Middle Eastern total, which comes from CommunityIQ, a VoteIQ model that estimates likely heritage from a voter’s name. It is built in house from public records. It is not licensed, appended, or purchased from anyone.
The model exists because the voter file’s race field is self-reported and often left blank, so the official Asian count is known to be low. Where a district page shows the enhanced figure, it is labeled as modeled and carries a note saying so. Every other bar on that chart is exactly as self-reported.
VoteIQ also models partisan lean and turnout propensity. Neither appears anywhere on a public page, at any level of aggregation. They exist only inside the gated workspace, for verified candidates, and they are labeled as estimates there too.
What this cannot tell you
What we never publish
No public page carries voter-level data, and this is enforced by the structure of the system rather than by care alone. The public database role has no access to the tables holding voter records, so a public page cannot read one even if someone asked it to. Public pages read aggregates and nothing else.
We also never publish religion at any level of aggregation, and never attach a modeled estimate to a named individual on a public page.
Questions
- How often is Wake County voter data updated?
- During early voting, every morning: the Wake County Board of Elections file lands around 6 AM and the refresh finishes by about 7 AM. Off season, weekly, after the NC State Board of Elections Saturday file.
- What does VoteIQ count as a registered voter?
- Active and inactive registrations together, which is the same basis the Wake County Board of Elections uses for its own published total. Inactive voters have failed a mail check and many have moved, so the roll is larger than the set of people a campaign can realistically reach.
- Why do VoteIQ turnout figures not match the official results?
- They are not the same measurement. VoteIQ counts voters who are on the Wake roll today and voted in a past election, attributed to the district they live in today. Voters who have since moved away or been removed are not counted, so these figures run below the official period turnout by design.
- Does VoteIQ ever show individual voters?
- Never on a public page. Public pages carry aggregates only, and any cell built from fewer than 10 voters is withheld at precinct and other fine scopes so that no individual can be identified from it.
- Which figures are modeled rather than counted?
- One, on public pages: the enhanced Asian and Middle Eastern total, produced by the CommunityIQ name model and always labeled as modeled. Everything else on a public page is a count from an official file.
Something here unclear, or a figure that does not reconcile with a source you trust? Write through our contact form and we will either explain it or fix it. The underlying files are public: you can download them from the boards of elections listed on our about page and check any number here yourself.