Introduction
The most dangerous chart in an IR office is the ten-year completion trend that looks fine. If it spans the 2020-21 IPEDS collection cycle, it is quietly two different taxonomies stitched end to end: the pre-2020 years are counted under the 2010 CIP and the later years under the 2020 CIP. Most codes survived the revision untouched, so most series are genuinely continuous. The ones that are not, where a code was deleted, moved, split, or its series renamed, develop a break at exactly the seam year, and the break reads as a real trend. This post is about finding those programs and repairing the series before the chart goes in front of a cabinet.
When the switch happened
IPEDS moved to the 2020 CIP with the 2020-21 data collection cycle, and institutions have collected on it since. That gives the seam a precise address: any completion series that starts before the 2020-21 collection and ends after it crosses the revision. If your trend tool pulls straight from the IPEDS completions files, the pre- and post-seam years arrived coded under different editions, and nobody re-based them for you.
What actually moved
The scale of the 2020 revision is documented in the CIP explainer: nearly 70 new four-digit series and more than 300 new six-digit codes. Less quoted, but more important for trend work, is the rest of the change ledger. Codes were deleted outright. Codes were moved between series. And several series were renamed around content that had drifted into them, for example:
Series-level renames between the 2010 and 2020 editions
| 2-digit series | CIP 2010 title | CIP 2020 title |
|---|---|---|
| 12 | Personal and culinary services | Culinary, entertainment, and personal services |
| 31 | Parks, recreation, leisure, and fitness studies | Parks, recreation, leisure, fitness, and kinesiology |
| 43 | Security and protective services | Homeland security, law enforcement, firefighting, and related protective services |
A rename at the series level is cosmetic for 2-digit reporting but a real signal lower down: series 43 growing a homeland security identity in 2020 came with new codes and moved codes inside the series, and any institution with programs there needs the 6-digit check, not just the series title. The full ledgers live on the CIP user site, which publishes lists of new, deleted, and moved codes alongside the crosswalk itself.
Reading the crosswalk
The NCES crosswalk tool maps every 2010 code to its 2020 counterpart and classifies the relationship: "no substantive changes," "new," "deleted," "moved from," and "moved to." You can query it for all codes at once, a single 2-digit series, or one code, which is the mode you want for auditing a specific program. The classification is the whole trick, because each action type has a different repair. The site also carries the 2010 edition side by side, so you can read a code's old definition and its new definition and see exactly what the boundary change did.
The four repair patterns
No substantive changes: splice and move on
If the crosswalk says no substantive changes, the series is continuous. Splice the years, cite the check, and stop worrying. This is the majority of codes, which is why most trends survive.
Moved: substitute the code, footnote the seam
A moved code is the same content at a new address. Re-code the pre-seam years to the new code, and annotate the seam year so a reader knows a re-code happened there. The numbers are comparable; the audit trail is what makes them credible.
Split (one old code, several new): sum, and only if boundaries held
When one 2010 code became several 2020 codes, the pre-seam series can be carried forward by summing the successor codes, but only if the split divided the old code's content rather than redrawing it. If definitions shifted in the split, the honest move is to report the old series as ending and the new codes as starting, with a visible gap at the seam.
Merged (several old codes, one new): sum backwards
The mirror case: to extend the new code's series before the seam, sum the predecessor codes' pre-2020 counts. The merged series is the only version that exists on both sides of the revision.
The field-of-study angle
The rebase matters beyond internal trend charts. The federal field-of-study files, the College Scorecard and PPD earnings and debt publications, are published on 4-digit CIP under the 2020 edition. Any comparison of your internal program history against those published numbers is a comparison of your 2010-coded past against a 2020-coded present, and it needs the same crosswalk pass first. There is a bonus in the other direction: rolling a program up one CIP level is also the standard move for escaping suppression in the field-of-study files, and we measured its yield in why your program has no earnings number. The crosswalk and the roll-up are usually one operation.
A worked example: protective services
Say your institution has reported a criminal justice program for a decade and a dean wants the completions trend. At the 2-digit level, series 43 simply renamed; the series total is continuous. At the program's 6-digit code, the question is whether that specific code was new, deleted, moved, or untouched, and the crosswalk answers it in one query. If the program's code was untouched, the trend ships with a footnote that the series title changed in 2020. If it moved, the pre-seam years get re-coded to the new address before the chart is drawn. Ten minutes of crosswalk work either way, against the alternative of presenting a seam artifact as a curriculum trend.
Where to go next
The anatomy of the code levels is in what is a CIP code, and the practical guide to assigning codes, including the re-check-at-every-revision rule, is in how to find the CIP code for any program. To see how a rebased series sits in the current taxonomy, browse the 2-digit families and 4-digit series in our CIP code index. For the 2010-to-2018 occupation-side revision, which breaks labor market trends the same way, see what is a SOC code. And for when each federal file drops, the federal data release calendar tracks all twelve datasets IR teams depend on.
Rebase a trend in one conversation
Ask for a program's ten-year completion trend and get the crosswalk check, the seam annotation, and the source vintage built in.
Book a demo