Core rule: no receipt, no value
Rate Archive does not estimate a missing tuition value, carry a price forward, borrow from a related campus, or infer a fee from a combined charge. A blank source cell remains blank. An unavailable page remains unavailable. A parser error stops at review rather than replacing the last validated observation.
“Unsupported,” “not yet reviewed,” and “source unavailable” are correct outputs when the evidence does not support a number.
The six-stage evidence chain
- 01
Source catalog
The official publisher, stable institution key, source URL, format, and automation boundary are recorded first.
- 02
Raw capture
The exact response bytes are hashed with SHA-256 and stored under a content-addressed key. Raw evidence stays private.
- 03
Dictionary mapping
Only fields defined by the official release dictionary enter the adapter. Units and dimensions remain explicit.
- 04
Validation
Range, uniqueness, completeness, foreign-key, imputation, and year-to-year continuity checks run before acceptance.
- 05
Revision chain
A newer release appends a new observation and points to the record it supersedes. Earlier receipts remain intact.
- 06
Publication review
A human gate decides which validated observations may become public. Validation alone never implies publication.
Source priority and identity
IPEDS UNITID is the stable key for the initial college archive. A system office, main campus, online division, and branch can share a brand or web source while remaining different entities. Rate Archive does not collapse them. For each logical period and series, an eventual public resolver will prefer an applicable final federal release, then an approved direct institutional source for a newer period not yet present in IPEDS.
Equally preferred sources that conflict beyond tolerance produce a flagged gap. The losing observation is retained with its provenance; source priority changes what may display, not what evidence survives.
Money, time, and dimensions
Money is stored as exact integer cents with an explicit decimal scale. Academic years use start and end dates, not a bare number. Residency, program level, attendance mode, campus, and unit—annual, term, or per credit—belong to the series definition.
The historical IC_AY CHG fields are dictionary-defined as combined published tuition and required fees. They remain a combined price series. Newer COST data separately supplies tuition and mandatory fees; those components stay separate. The archive will not manufacture a historical split that the source does not contain.
Releases, revisions, and imputation
Annual IPEDS archives may repeat up to four academic years. Rate Archive verifies the archive and dictionary hashes, selects an explicitly revised CSV member when the official ZIP supplies one, and retains source imputation/status codes. A later repeated value supersedes the earlier observation even when the value is unchanged, creating an auditable release chain.
The first Stage 1B pass covers IC2000_AY through IC2023_AY. Across those releases, 52,200 target slots reconciled to 50,761 validated release observations and 1,439 explicit gaps, with zero invalid values admitted. Later releases repeated 35,583 logical values; 1,679 of those changed and 33,904 remained the same. Twenty-six latest-series continuity flags remain queued for review.
Publication status
“Validated” means an observation passed automated schema and reconciliation checks. It does not mean the value is public. “Publishable” requires the separate review gate. As of this methodology version, the archive has no publishable tuition observations. Public institution pages contain source metadata only.
Capture limits and respectful automation
Only sources rated ready for unattended capture are scheduled. The pilot checks at most four sources per daily run, waits at least 30 days before a source is eligible again, respects relevant robots rules, limits response size, and does not reuse another product’s hosting resources. PDF, dynamic, blocked, and multi-page sources stay manual or require a specialized reviewed parser.
Reproducibility
Each official archive has an exact URL, byte count, SHA-256 digest, selected member, row count, column count, dictionary receipt, adapter version, and ingest summary. The public source catalog exposes safe source metadata; raw captures and nonpublishable observations stay private.