NamePolitics

Methodology

Everything here comes from public records. 5,015 names have a page, and 4,847 of them have a position on the spectrum. Last built September 21, 2026.

The headline score

The number on each page is an average of contributor ideal points from DIME, Adam Bonica’s database of campaign finance and ideology, released into the public domain. It covers 14.4 million individual contributors between 1979 and 2014, including state and local races, not just federal ones. An ideal point is estimated from the whole network of who gave to whom, so a person who funds only the furthest-left candidates scores differently from one splitting money between moderates, even though both are Democrats.

We checked that score against an entirely separate construction of our own, chaining federal contributions from 2016 to 2026 through to the roll-call voting records of the members who received them. The two agree at a correlation of 0.94 across the names they share, despite covering different decades and different methods.

Two different yardsticks

A name page measures the same name twice, against two different things, and it is worth knowing which is which. The headline ranking compares a name with other names. The typical name scores about −0.22, because people who donate to campaigns lean left as a group, so a name can sit left of zero on the raw scale and still be middle of the pack among names. The party bars further down compare a name with an even fifty-fifty split instead. That is why a name can be described as middle of the pack at the top of the page and as drawing 60% Democratic donors below it. Both are true; they are answering different questions.

Party registration

The party bars come from four places. Donor splits use Federal Election Commission bulk records for 2016 to 2026, with a contributor counted as Democratic or Republican when at least 80% of their attributed money went one way. North Carolina and Florida each publish a statewide voter file with actual party registration, and Ohio publishes one recording which primary ballot each voter last chose, since Ohio has no party registration at all. The Florida file is the state's August 2022 extract disk, used from an archived copy of that disk that was checked against the state's own layout.

Do the sources agree?

Seven separate records measure these names, and asking whether they agree is the only real test of whether a name signal exists at all. Each cell below is the correlation between two sources across the names both measure well, with the share of names they place on the same side underneath. Every source is measured against its own average, so positive means left of that population's own centre and nothing more.

DIME ideal pointsFEC donors527 donorsWA donorsNC votersFL votersOH voters
DIME ideal points—0.9087%0.7473%0.8584%0.4364%0.4463%0.7178%
FEC donors0.9087%—0.8782%0.8482%0.2656%0.3560%0.6372%
527 donors0.7473%0.8782%—0.5767%-0.0642%0.0947%0.3658%
WA donors0.8584%0.8482%0.5767%—0.4459%0.4057%0.7977%
NC voters0.4364%0.2656%-0.0642%0.4459%—0.9488%0.7574%
FL voters0.4463%0.3560%0.0947%0.4057%0.9488%—0.7673%
OH voters0.7178%0.6372%0.3658%0.7977%0.7574%0.7673%—

Two things stand out. The three voter files agree strongly with each other, and North Carolina against Florida is the strongest pair on the board, at 0.94 across 1,397 names. Two electorates that share almost nothing demographically rank names the same way, which is the clearest evidence here that the signal is real rather than an artefact of one state.

The donor records agree with each other and not with voters. Federal contributions, 527 filings and Washington state money correlate at 0.84 to 0.87, even though they are different filing systems covering different decades and different offices. Against registered voters the same records fall to 0.26 and below, and the 527 pool against North Carolina is essentially zero. That is not noise in one dataset. It is a real difference between people who give money and people who register to vote, reproduced four times over.

What this cannot tell you

It is not a prediction about you. These are averages over large groups. Plenty of conservative people are called Rachel.

Donors are not voters. People who give money to campaigns are older and wealthier than the country. Where a name can be checked against real voter registration, donors and registered voters point in opposite directions for roughly a quarter of names, and the disagreement is systematic rather than random. Young male names like Cody, Seth and Logan read Democratic among donors and Republican among voters in both states we can measure. That is why name pages show each population separately instead of blending them into one number.

Names carry demographics. A name encodes roughly when you were born, often your sex, and frequently your ethnic background, and each of those predicts politics on its own. In North Carolina, where the voter file records race, controlling for it removes about 60% of the variation between names. Birth decade removes about 1% and sex about 3%. Where a name page shows an adjusted figure, that is what remains after comparing people against others of the same age, sex and race.

Three states is not the country. The only voter files here are North Carolina, Florida and Ohio, because almost every other state charges money, requires an approved application, or blocks automated access. The encouraging part is how well they agree. Across the thousand or so names common in both, North Carolina and Florida point the same way 88% of the time, and the strength of the lean correlates at 0.94, even though the two electorates look nothing alike. Ohio, whose party field is a primary-ballot proxy, agrees with each of them about 70% of the time.

Nothing individual is published. Every figure is an aggregate, no name appears without clearing a minimum count, and no underlying record is redistributed.

Sources

Back to the lookup