95,584 fewer clerical jobs, 80,571 more in finance: IPEDS staffing by occupation on 4,902 colleges, 2012 to 2024

On a fixed panel of 4,902 colleges that reported staffing in every year from 2012 to 2024, office and administrative support went from 463,131 people to 367,547. That is 95,584 positions gone, down 20.6 percent, inside a workforce that grew 9.6 percent overall. Over the same twelve years business and financial operations grew 43.7 percent, from 184,356 to 264,927, and management grew 26.3 percent, from 237,648 to 300,035.

As a share of all staff, the clerical layer fell 3.85 percentage points, from 13.95 percent to 10.10 percent, and it fell in every one of the twelve year-over-year steps. No other category moved half that far. Instruction, the category everyone assumes is the story, barely moved: instructional, research and public service staff went from 42.96 percent of all staff to 42.63 percent, a drift of 0.32 points.

The source is the IPEDS Human Resources component, Employees by Assigned Position, fall 2012 through fall 2024. Each institution sorts its own payroll into occupational categories built on the federal Standard Occupational Classification system and reports headcount by category. For once the federal file is counting you rather than your students.

One thing to settle first. There is no institutional research category in this file. Not one of the 31 official occupational codes contains the words institutional research, analytics, assessment or data, so IR headcount cannot be read out of IPEDS at any institution in any year. Where IR staff do land is covered below.

Roughly a third of a percentage point goes missing each year, in good enrollment years and bad, which is slow enough that nobody inside a single institution registers it as an event. On the panel the share reads 13.95 percent in 2012, 12.30 percent in 2016, 11.49 percent in 2019, 10.71 percent in 2021 and 10.10 percent in 2024. No single year of that looks like a decision, and twelve of them add up to a restructuring.

Note the 2020 point sits inside the pandemic and shows the sharpest single-year moves in several of these series, so read 2019 through 2021 as a disturbed stretch rather than ordinary trend.

The work did not evaporate along with the positions. The three administrative-professional categories that sit above the clerical layer (management, business and financial operations, and computer, engineering and science) together account for 22.50 percent of all staff in fall 2024, more than double office and administrative support at 10.08 percent. Some of what a departmental secretary handled in 2012 is now done by software, some by the faculty member directly, and some by an analyst in a shared services center who costs more and reports somewhere else entirely. IPEDS cannot tell you which share went where. It can tell you the layer that used to absorb routine requests is a fifth smaller than it was.

The variable is OCCUPCAT and its domain currently holds 31 codes. It held 36 from 2012 through 2015, and the six codes retired after that year were all graduate assistant subcategories, none of which feed a category used here. The dictionary entry is explicit about the lineage:

Two properties of that code list will ruin your numbers if you miss them.

First, the categories are nested, not mutually exclusive. Summing all 31 counts the same person several times: code 210 sits inside 200, code 262 sits inside 260 which sits inside 250. Exactly twelve codes form a clean partition (200, 250, 300, 310, 320, 330, 340, 350, 360, 370, 380 and 390) and they sum to code 100, all staff, with a difference of zero in all thirteen years we checked. Every share in this post uses that twelve-code partition.

Second, graduate assistants sit outside the all-staff total, which is the opposite of what most people assume. Add code 400 to the partition and you overshoot code 100 by exactly the graduate assistant count, 406,661 people in fall 2024. If a campus headcount reconciliation is off by roughly the size of your GA population, this is usually why. On the panel, graduate assistants grew 14.4 percent over the twelve years, from 349,378 to 399,835, and none of that shows up in any all-staff figure above.

On the 2018 SOC revision: the code values did not change, and the panel series show no break across 2017, 2018 and 2019.

Zero of the 31 occupational labels mention institutional research, institutional effectiveness, analytics, assessment or data. Four of them contain the word research (200, 215, 220, and 420 graduate assistants research) and all four describe academic research. Code 220 is the tempting one and it is the wrong one: 220 nests inside 200, instructional, research and public service staff, right alongside instruction. It counts principal investigators and research scientists. It is not your office.

For the record, that academic research staff line did grow. On the panel, code 220 went from 80,903 in 2012 to 108,193 in 2024, up 33.7 percent. That is a real finding about sponsored research capacity. It is not a finding about IR capacity, and anyone who quotes it as one is quoting the wrong number. The loaded IPEDS glossary is no help either: it holds 31 rows in total and returns nothing for institutional research, management, computer or business and financial.

So where do IR staff actually sit in this file? Scattered across three categories, with no way to separate them.

The trend figures above deliberately use a fixed panel. For the shape of the workforce as it stands today the whole file is the better base: 3,750,528 employees across 5,837 reporting institutions in fall 2024, the November 1, 2024 payroll snapshot.

Read the part-time column next to the share column, because that is the other half of the story. Instructional, research and public service staff are 42.78 percent of college employees and 44.0 percent part-time, the adjunct economy showing up as one number. That bucket bundles research and public service in with teaching, so it understates the part-time share of instruction proper; the code 210 series further down is the cleaner read. Management is 2.2 percent part-time and business and financial operations 4.6 percent. The categories that grew are the categories that are almost entirely full-time, so headcount growth there converts directly into fixed cost, which is why these lines end up in front of a cabinet at budget time.

National averages hide the thing that matters most to an individual office, which is that the three big sectors staff themselves in genuinely different proportions. The sector figures here all sit on the balanced panel, so the institution counts are lower than the full fall 2024 universe.

Of those three sectors, community colleges are the most instructional by a wide margin: 53.56 percent of staff at 814 public two-year colleges are instructional, research or public service, against 39.13 percent at 794 public four-years. Private nonprofit four-years carry the heaviest management layer, 9.53 percent of staff against 7.61 percent at public four-years and 6.70 percent at community colleges, a gap of 1.9 points over the public four-year sector. And the technical bench at community colleges is roughly a third as dense as at public four-years: computer, engineering and science is 2.71 percent of staff against 8.38 percent.

The twelve-year trends split the same way and harder. Public four-year staffing grew 13.3 percent and private nonprofit four-year 12.8 percent, while public two-year fell 10.9 percent. Community-college instructional staff dropped from 290,262 to 239,406, down 17.5 percent, while public four-year instruction rose 13.0 percent and private nonprofit four-year instruction rose 17.5 percent.

Inside that shrinkage the mix moved. Community colleges grew management 19.4 percent (25,107 to 29,971) and business and financial operations 19.2 percent (12,608 to 15,029), while computer, engineering and science fell 4.9 percent (12,737 to 12,108) and office and administrative support fell 26.1 percent. A sector that lost a sixth of its teaching staff added a fifth to its management and finance layers and cut the technical bench it would need to automate any of the work it just lost people for.

Business and finance grew fastest at public four-years, 105,047 to 157,459 or 49.9 percent, with private nonprofit four-years close behind at 43.9 percent (56,890 to 81,892). Office and administrative support fell in all three sectors without exception: 17.4 percent at public four-years, 22.4 percent at private nonprofit four-years, 26.1 percent at community colleges.

Percentages understate how thin this gets at an individual college, so here it is as a median headcount, this time across every fall 2024 reporter. The median private nonprofit four-year institution employs six people in the entire computer, engineering and science category, on a median total staff of 276. Nearly half of those 1,569 colleges, 47.5 percent, have five or fewer, and 20.8 percent report zero. The median community college has eight, on a median staff of 388, with 38.0 percent at five or fewer.

Only the public four-year sector has anything resembling a bench: a median of 35 across 837 institutions, with 1.6 percent reporting zero. If you work at a small private and it feels like the technical capacity of the institution is you and a couple of colleagues, the median says you are reading it correctly. That is the operating constraint under most analytics-capacity conversations in this sector, and a bigger warehouse does not move it.

Staffing counts tell you where colleges have been. The New Hires file tells you what they did in the last twelve months, which is the closest thing IPEDS offers to a leading indicator. In the year to fall 2024, 3,302 reporting institutions hired 328,967 full-time staff, a rate of 12.2 hires per 100 existing full-time employees.

Instruction is under-represented in that hiring relative to its size: 31.94 percent of the full-time workforce but only 25.16 percent of hires, a rate of 9.6 per 100. Management is lower still, 10.81 percent of the workforce and 7.10 percent of hires, 8.0 per 100, the lowest rate in the file. Every rate in this section has a full-time denominator, since the New Hires file counts full-time hires only.

The line that needs care is office and administrative support. Its hire rate is 14.8 per 100, above average, and it takes 13.78 percent of hires against 11.29 percent of the full-time workforce. That is not growth. This is the same category that lost 20.6 percent of its positions since 2012. High churn and net decline coexist: the clerical layer turns over fast and gets refilled at less than the rate it empties. A hire rate is a churn measure, never a headcount forecast.

Over time the hiring mix moves the same direction as the standing workforce. The instructional share of full-time hires went from 27.45 percent in 2016 to 25.16 percent in 2024, bottoming at 22.66 percent in 2022. Business and finance went the other way, 7.76 percent to 9.49 percent. The technical and management layers held roughly flat: computer, engineering and science was 9.28 percent of hires in 2016 and 9.63 percent in 2024, management 7.39 percent and then 7.10 percent.

Nationally the part-time share of college employees went from 30.6 percent in 2012 to 25.7 percent in 2024, a drop of 4.9 points. On the balanced panel the same measure went from 29.0 percent to 25.5 percent. Instructional staff specifically, code 210, went from 48.5 percent part-time to 47.1 percent. By headcount, colleges got less casualised over these twelve years, which is the opposite of the received story.

Read that as a headcount result and nothing more. IPEDS HR reports people, not FTE. A shift toward fewer but larger part-time appointments would look exactly like this in the data and would mean something different. Nothing here says adjunct workload fell, or that adjunct pay improved, and this post does not claim either.

The administrative categories are the full-time end of the distribution and they got more so: management 3.4 percent part-time in 2012 and 2.2 percent in 2024, business and finance 8.4 to 4.6 percent, computer, engineering and science 9.4 to 6.2 percent, office and administrative support 19.7 to 16.5 percent. The layer that grew is the layer that comes with a benefits load.

Three readings, none of them causal. This file describes headcount by category and cannot tell you why any position moved.

Start with capacity. The layer that historically absorbed routine reporting and record-keeping around an IR office is 95,584 people smaller on this panel than it was in 2012, while the categories that generate requests, management up 26.3 percent and business and finance up 43.7 percent, grew faster than the workforce around them. More people whose job is to ask for numbers, fewer people whose job was to help produce them. That is the structural version of the mismatch we wrote about in why IR capacity has not kept up with data demand, and the hours cost of it is quantified in our IR capacity benchmarks.

Then where your own position sits. If your title is director or AVP you are in code 300, the category with the lowest hire rate in the file, 8.0 per 100 full-time staff. If you are an analyst you are in 310 or 320 depending on your reporting line, and 310 added more positions over the twelve years than any other category on the panel. Treat that as career-planning information.

Last, benchmarking. You cannot benchmark IR staffing from IPEDS, but you can benchmark the administrative-professional block, the instructional share, and the twelve-year trajectory of both against a defensible peer group. That comparison is reproducible, and it is the one a cabinet reacts to. Building the peer group first is the step most people skip, and it is the step that determines whether the comparison survives a follow-up question; the peer benchmarking page is where to start assembling one.

Source: IPEDS Human Resources, Employees by Assigned Position (files EAP2012_RV through EAP2024) and New Hires (files S2012_NH through S2024_NH), loaded as ipeds.employees_assigned and ipeds.staff_new_hires. Sector and institution characteristics come from the IPEDS HD directory, data_year 2024. There is no 2025 IPEDS HR file, so fall 2024 is the current vintage: it is the November 1, 2024 payroll snapshot, not an academic year. Any sentence that puts one of these numbers next to an IPEDS completions figure has to label both vintages, because completions run a year further ahead. The completions file is already at the 2024-25 award year, which is what the six-year degree production breakout uses.

Category definitions came from the IPEDS data dictionary tables rather than from the catalog, because the catalog carries nothing for these tables. Calling get_table_details on employees_assigned and staff_new_hires returns null for every description, business name, value mapping, join hint and example query. The twelve-code partition, the nesting of 220 inside 200, and the position of graduate assistants outside the all-staff total were each verified by arithmetic against the file rather than assumed from documentation.

All trend figures use a balanced panel of 4,902 institutions that reported all-staff headcount in every one of the thirteen years. That is deliberate. The raw reporting universe shrank 22.5 percent over the span, from 7,529 institutions in 2012 to 5,837 in 2024, as colleges closed and merged. Gross national trends on the full file confuse real staffing change with the disappearance of institutions. Fall 2024 cross-section figures use all reporting institutions, with one exception: the sector staffing mix stays on the panel so it lines up with the sector trend table next to it, which is why its institution counts are lower than the ones in the technical bench table.

Two single-institution reporting errors are excluded from every trend figure here. Dell'Arte International School of Physical Theatre reported 318,001 service-occupation staff in fall 2023 and none in either 2022 or 2024. Baptist Theological Seminary at Richmond reported 147,410 employees in fall 2014, of which 147,396 part-time, against 21 the year before and 19 the year after. A scan of every institution-year above 60,000 all-staff headcount returns exactly two rows, and they are these two. Removing them pulls the raw 2023 national total from 3,985,514 down to 3,667,494 and the raw 2014 total from 3,807,317 down to 3,659,889, which puts both years back on a smooth line. If you see a published national series with a bump at 2014 or 2023, that is what you are looking at. The fall 2024 cross-section is untouched by either.

Two consistency checks ran before any of this was written up. Imputation is not driving the results: 13,350 of the 13,366 all-staff rows across 2012 and 2024 carry the reported flag, 99.88 percent. And the full-time and part-time splits reconcile exactly, with eaptot equal to eapft plus eappt on all 134,161 complete 2024 rows, zero exceptions.

IPEDS, Use the Data (NCES), the source portal for the Human Resources component files used here, EAP2012_RV through EAP2024 and S2012_NH through S2024_NH. IPEDS survey components overview (NCES) for the Human Resources component definition and collection cycle. Standard Occupational Classification, U.S. Bureau of Labor Statistics, the federal classification the IPEDS occupational categories are built on, including the 2018 revision referenced in the dictionary entry quoted above.

The full reproducible query log for this post, 21 numbered SQL queries with their outputs and the discarded lines of inquiry, lives in the repository at docs/data-evidence/who-colleges-actually-hire-ipeds-staff-by-occupation/queries.md, with the table grain, column definitions and code domains in schema.md alongside it.