Charted: Class I Rail Headcount Fell ~29% Under PSR — Terminal Dwell Rose First

Precision-scheduled railroading sold itself as a service story: fixed schedules, longer trains, fewer hump yards, tighter asset turns. The payroll ledger tells a colder chronology. Class I employment fell about 29% from the 2015 peak to the 2022 trough — and average terminal dwell rose first.

Using STB Class I employment prints (cross-checked against AAR employment tables) and AAR rail performance measures for terminal dwell and train speed, the industry composite shows a clear sequence. Dwell jumped hard in the 2017–2019 PSR adoption wave [STB RRM]. Headcount kept grinding lower through 2022. Service metrics partially healed afterward; the workforce did not snap back to the mid-decade peak.

The dashboard above walks the dual tape, the inflection scatter, carrier cut ladders, and craft-mix stacks. The rest of this post is the narrative behind those panels.

Scoreboard: headcount vs dwell inflection

The table’s punchline is sequence, not just magnitude. A desk that only reads the 2022 employment trough will miss the earlier service stress that shippers felt in yards and terminals.

What precision scheduling actually cut

PSR is not a single regulation or a single year. It is an operating doctrine that spread carrier-by-carrier after Hunter Harrison’s CSX campaign in 2017 and similar playbooks at peers: reduce crew starts, close or idle humps, lengthen trains, schedule more rigidly, and treat locomotives and cars as scarce assets rather than buffers.

That doctrine shows up in the employment series as a multi-year downshift. From 2015 to 2019, Class I headcount fell from 163.4k to 132.4k (−19%) even before the COVID demand collapse [STB Emp]. From 2019 to 2022 it fell further to 115.8k. Mid-2026 prints recover only to about 121k — still roughly 26% below the 2015 peak.

The craft mix shifted with the model. Yard and terminal shares slipped about 2 percentage points from 2015 to 2022 as humps closed and local switching thinned. Professional and “other” categories gained share as planning, technology, and centralized operations absorbed a larger slice of a smaller payroll. Train-and-engine remained the largest craft block, but the industry’s operational buffer — people who absorb disruption in yards — got thinner.

Did service degrade first?

Yes, on the industry composite that matters for the headline question.

Average terminal dwell rose from 22.1 hours in 2015 to 26.8 hours in 2019 [STB RRM]. Train speed fell from about 26.4 mph to roughly 23.9 mph in the same PSR-wave window. Those are not shipper invoice audits, and they are not carrier-specific STB Form C line items — they are AAR performance composites — but they are the public industry meters desks actually watch.

The inflection scatter in the dashboard makes the order visible: 2017–2019 points sit in the “employment down, dwell up” quadrant. The first large dwell jump lands in 2017, three years before the employment trough. Pre-COVID, annual employment and dwell are tightly negatively correlated (about −0.91 on the composite). After COVID, that relationship weakens (about −0.18) because 2020 collapsed volumes and temporarily cleaned yards even as payrolls kept falling.

So the cleanest reading of the public tape is:

1. PSR wave: dwell rises, speed falls, employment falls.
2. COVID shock: volumes crash; dwell improves briefly; employment keeps falling.
3. Stabilize: employment finds a lower plateau; dwell partially recovers but stays above the 2015 baseline.

Service stress led the deep employment trough. It did not wait for the trough to arrive.

Carrier cuts were uneven

Not every Class I cut the same way. Analytical carrier composites for 2015→2022 employment change put CSX near −34% and Norfolk Southern near −31%, with dwell stress in the PSR window also among the largest. Union Pacific lands around −28%. BNSF's cut is milder on this composite (about −18%) with a smaller dwell delta — consistent with a later, less abrupt operating shift than the eastern PSR flashpoints [STB Emp] [RRB Table D1].

Share of remaining mid-2026 headcount still concentrates in the western franchises: BNSF and UP together approach half the Class I employment composite on this cut. That matters for interpreting industry averages. A national dwell or employment print is weighted by networks that adopted PSR at different speeds and with different yard footprints.

Treat carrier deltas as desk composites, not audited 10-K footnotes. Merger accounting (CPKC), U.S.-only slices of CN, and reporting discontinuities can move a point or two. The rank order — eastern PSR pioneers cut harder and stressed dwell more — is the robust signal.

Cars on line and the productivity story

PSR defenders argue that lower headcount is the point: more freight with fewer people. The cars-on-line index in the annual tape complicates a simple productivity celebration. Indexed to 2015 = 100, cars on line drifted around parity through the late 2010s, dipped hard in 2020, then spiked above 100 in 2021–2022 even as employment hit its trough. Congestion returned when demand recovered into a thinner crew and yard base.

That pattern matches the shipper complaint cycle of 2021–2022: not that railroads had “too many workers,” but that the network had less slack when volumes rebounded. Train speed in 2022 printed near the PSR-wave trough again (~23.7 mph on this composite) before recovering toward 25 mph by mid-2026. Employment recovered only a few thousand heads.

Productivity can rise on a carload-per-employee meter while service reliability falls on a dwell-or-OTD meter. Both can be true in the same year. The public series do not resolve the shipper–carrier argument; they show that headcount cuts and service stress were not simultaneous accidents.

What the mid-2020s plateau implies

By 2024–2026 the composite looks like a new normal rather than a bounce-back:

• Employment: ~120–121k, far below 2015.
• Dwell: ~23–24 hours, better than 2019’s 26.8h peak, still above the 2015 baseline.
• Speed: ~25 mph, improved from troughs but not restored to mid-decade cruise.

If PSR’s promise was “fewer people, better service,” the public industry meters deliver only half the slogan. The people side stuck. The service side improved from the worst PSR-wave prints but did not fully erase the dwell premium versus the pre-PSR baseline.

Labor markets also matter. Retirements, hiring freezes, and post-COVID hiring competition in transportation trades made rapid restaffing expensive even when carriers wanted more starts. The employment series therefore mixes operating doctrine and labor supply. The dwell-first sequence still holds: yards felt stress before the payroll trough, not only because of later hiring difficulty.

Caveats and what this is not

This post uses industry composites. It is not a substitute for:

• carrier-level STB Form C employment detail,
• shipper invoice or OTIF audits,
• FRA safety event series,
• or union agreement headcount by craft and terminal.

STB monthly employment prints can bounce with seasonal and reporting timing; we aggregate to annual means for the dual tape. AAR rail performance measures are national composites that can mask corridor-level meltdowns or recoveries. Mid-2025 and 2026 values are estimated/carried provisional prints for desk continuity. Craft shares are analytical reconstructions from public employment structure tables, not a full STB craft census.

COVID is a confounder: 2020 improved some fluidity metrics while destroying volumes and accelerating furloughs. Reading 2020 as a PSR success year would be a category error. The cleanest PSR window for the sequence test remains 2017–2019.

Bottom line for desks

Class I railroading under precision scheduling became a smaller-payroll industry first measured in service stress, then in the employment trough. Roughly three years separate the first large dwell jump (2017) from the employment floor (2022). Mid-2026 employment is still about 26% below the 2015 peak even after a modest restaffing bounce [STB Emp]. Dwell has cooled from its 2019 high but remains above the pre-PSR baseline [STB RRM].

For freight desks, the operational reading is blunt: headcount reductions under PSR were not a painless efficiency glide. The public meters say yards tightened first. Payrolls followed — and stayed down.