Does Choosing a Major Narrow Your College List?

Key findings

1. Only 2,554 of 9,976 field-and-state combinations (25.6%) resolve to a college list materially different from every other field in that state.
2. In 51 of 51 jurisdictions, at least one pair of fields resolves to college lists that overlap 80% or more.
3. 368 combinations resolve to a college list identical to another field's in the same state, meaning the two fields cannot be told apart by where they are taught.
4. The median overlap between a field's college list and its closest sibling field in the same state is 45%.
5. Alaska is the least differentiated jurisdiction: 107 fields have evidence and 0 resolve to a distinct college list.
6. California is the most differentiated: 67.5% of its fields resolve to a distinct college list.

For each of the 51 jurisdictions: how many fields have completions evidence, how many resolve to a college list materially different from every other field there, and the closest-overlapping pair of fields in that state.

Methodology

Source data is the U.S. Department of Education IPEDS completions collection for 2024-25, which records credentials actually conferred. Institutions are matched to the federal directory on normalised name and state; any name that resolves to more than one federal ID is excluded rather than guessed.

For each field and state, the set of colleges that conferred credentials in that field is compared against the set for every other field in the same state, using Jaccard overlap (shared colleges divided by the union of both lists).

A combination is counted as resolving to a distinct list when it has at least 5 colleges, an overlap below 0.6 against its closest sibling field, and at least 70% of its colleges also conferred credentials in 2023-24.

The two-year persistence requirement exists because a single year of completions cannot distinguish a running program from one that has already stopped admitting students.

Fields are the ones CampusPin tracks, mapped to exact federal CIP codes. A field may claim its whole 4-digit CIP family only when the official NCES family title matches the field itself, so broad families are never treated as a specific field.

This measure compares every field against every other field in the state, including unrelated ones, because that is what the question asks: if a state's English colleges are also its Biology colleges, then choosing between those two fields genuinely does not narrow the list. It follows that the count here is NOT the number of pages CampusPin publishes for field-and-state combinations, which uses a deliberately different and looser test. A page is compared only against fields in its own subject neighbourhood, since two fields taught at the same colleges still make two different pages. Both figures are published in the dataset metadata so neither can be quoted as the other.