Teacher licensure data — methods & sourcing

The datasets, where they came from, how they were cleaned, and how individual privacy is protected.

This page documents the specific data behind the teacher-licensure database and its breakdowns (the licensed-vs-employed comparison, the full breakdown, and the county map). It follows the site’s general methodology & sourcing principles; here are the particulars for this dataset.

The datasets

Everything here is built from public records produced by the New Mexico Public Education Department (PED). Five extracts are used:

Where we got them

The ten-year assignment panel was obtained through an Inspection of Public Records Act request (IPRA 25-384) to PED. The licensure file, the EOY licensed-staff extract, the position directory, and the STARS salary files are PED’s own published extracts and Sunshine Portal downloads. All of it is public record. We keep the original, unmodified files so every published figure can be traced back to its source.

How we cleaned them

• Normalization. NULL and empty strings are read as missing; every text column is whitespace-trimmed. Original values are preserved alongside cleaned ones.
• One row per person. Records are collapsed to a single row per person, keyed on certNum (the license number). A person holding several licenses or endorsements is counted once, so no one is double-counted across the breakdowns.
• Linking across extracts. The extracts slice the same people differently. We join on certNum first; but because 24.6% of assignment-panel rows carry no certNum, we added a name, and name-plus-teacher-license, fallback match — each link recorded so it can be audited. A certNum-only join would silently drop whole teaching careers.
• Lifetime licenses. Identified by the LIFE: license type rather than an expiry-year rule. (An earlier expiry-year heuristic caught only two mistyped rows; keying on the license type gives the true count, 1,422.)
• Grade bands. Special Education and Specialty Area credentials authorize all grades (PK–12), so they are shown as their own category rather than being forced into a single grade band.
• Specialty vs. endorsement. A specialty license (its own subject credential) is kept distinct from an endorsement (a subject add-on to a base teaching license). They are never mixed.

How “employed” is defined

A licensed person is counted as employed if the EOY extract names an employer for them or they have an active (no end-date) position row in the authoritative 2024–25 Position Directory. That yields 40,148 employed, of the 82,266 current license holders.

The two signals overlap heavily (the directory lists 37,169 active, the EOY extract names an employer for 38,434). A small group — 1,806 people — has an employing district named but no school assignment and no directory row; these are counted as employed, on the basis that PED names an employer for them and the missing school is a data-completeness gap, not evidence of non-employment. In the source data, an assigned district and an assigned school are always populated together, so there is no “district listed, school blank” assignment bucket — the only edge case is employer-named-without-assignment.

Counting rules

• License-type, grade-band, and support/administrator breakdowns count people, each once.
• Endorsement breakdowns count endorsements, not people: one teacher can hold several, so subject bars are not additive and sum to more than the number of teachers. Each such chart says so.
• Every chart states its base population and denominator so it stands on its own.

Privacy & disclosure control

This is a release of statewide aggregates only. No individual teacher is named or published, and almost every published cell counts people in the hundreds or thousands.

• No person rows. Person-level records exist only in an internal file and never leave it.
• Keyed pseudonyms. Internally, direct identifiers are replaced by a secret-keyed HMAC-SHA256 pseudonym. Because a license number is publicly enumerable through PED’s license lookup, a plain hash could be brute-forced; a secret-keyed code cannot. The key and the pseudonym↔identity crosswalk are never published.
• Small-cell suppression. Any person-count below five is blanked; any rate is blanked when its numerator or its complement is below five, so neither a tiny group nor a tiny remainder can pin down an individual. District breakdowns are limited to districts with at least 100 staff; employer breakdowns to at least 150.
• Re-identification scan. Before release, candidate aggregate schemas are run through a k-anonymity scan, and the release form is chosen on measured risk.

Limits & caveats

• The licensure file is a current snapshot. It reflects who is licensed now, not a time series, and renewal cycles mean a “licensed” count is not the same as an “actively teaching” count.
• The salary figures cover only the two school years PED has made available through the Sunshine Portal (2011–12 and 2018–19). They are point-in-time snapshots, not a continuous series; intervening years were not published.
• Employment reflects the 2024–25 school year as captured in the EOY extract and the 40th-day directory.
• Cross-extract links recovered by name are high-confidence but not certain; ambiguous names are left unlinked rather than guessed.