ORD Yields: American vs United

The Question

In 2025 Chicago O'Hare rebalanced its gate leases: United gained 5 preferential positions (to 95), American lost 4 (to 59). A year later American clawed back 3 and United fell to 91. The dispute has grown loud enough that the FAA has capped ORD movements twice.

United CEO Scott Kirby has publicly argued that American is losing roughly $800 million to $1.1 billion per year at ORD and cannot sustain its schedule. American disputes this and its CFO says every hub contributes positively to system profit. Neither side publishes methodology, because no airline reports P&L at the hub level.

The narrow question this report answers: using only public data, what can we say about American's and United's ticket-revenue economics at O'Hare?

How ORD Allocates Gates

Everything flows from the 2018 Airline Use and Lease Agreement (AULA), which replaced exclusive-use leases. The mechanics:

• Two gate classes. Preferential Use gates are assigned to a long-term signatory who controls scheduling but must yield when idle. Common Use gates (~24) are CDA-controlled, mostly in Concourse M, and serve new entrants, international carriers, and construction swing space.
• The unit is linear frontage, not gates. The airport measures each carrier's terminal frontage in feet (measured on a line 100 ft off the terminal face) because widebodies need more space. You can reconfigure gates within your frontage freely.
• The cycle. A rebalance triggers on Feb 1 each year if a signatory requests more frontage or CDA needs more common-use. CDA computes prior-year departure share, multiplies by available preferential frontage, and reassigns the delta. Determination lands in spring; gates change hands October 1.
• What actually counts. Scheduled departures during the prior calendar year — regardless of which gate they used. The AULA also lists seats, departures per foot, and projected common-use demand, but departures drive the arithmetic.

Why it oscillates

• Hysteresis. Gates transfer October 1, but the next round measures the entire prior calendar year, so a winner operates 9 of 12 measured months without the gates it won — and can lose them back.
• Denominator effect. It's share, not volume. American's post-COVID retrenchment raised United's share without United adding flights.
• Overscheduling incentive. Defending frontage means flying frequency you may not want. The FAA capped ORD at 2,708 daily operations for summer 2026 (extended through Oct 2027) explicitly citing the two hubs expanding to retain or gain gates.

Current state: 91 preferential gates for United, 66 for American, ~24 common-use. Concourse D (19 gates, ~2028) and Concourse E (14 gates in 2030) are under construction; net capacity growth is +28 narrowbody-equivalent gates by mid-2030s.

The Data

Airline financial data at the airport level does not exist publicly. Form 41 financial schedules stop at the entity or region level. The closest public instruments are:

DB1B Market is the right instrument for a yield question. Yield = revenue per revenue passenger-mile — the standard airline industry metric for pricing power.

Techniques Used

1. Stream-filter the raw file

The full DB1B Market CSV is 2.1 GB. Loading it whole is unnecessary — I streamed the ZIP through pandas.read_csv in 500,000-row chunks, keeping only rows where Origin == 'ORD'. Total wall time ~15 seconds, memory footprint under 200 MB. Output: a 14 MB extract used for all subsequent analysis.

2. Standard DB1B cleaning

• BulkFare = 0 — exclude tour operator and bulk-contract tickets.
• $20 ≤ MktFare ≤ $5,000 — trim voucher artifacts and coding errors.
• Passengers > 0, MktMilesFlown > 0.
• MktGeoType == 2 — domestic only, since DB1B has spotty international coverage.

Cleaning removed ~6.6% of ORD-origin records, mostly the international ones.

3. Carrier attribution — the trap

DB1B has three carrier fields: RPCarrier (reporting), TkCarrier (ticketing/marketing), and OpCarrier (operating). At ORD, where regional-jet flying is heavy, the choice matters. Envoy (MQ), Republic (YX), SkyWest (OO), Air Wisconsin (ZW), Mesa (YV), and GoJet (G7) operate under their own two-letter codes — but revenue accrues to AA or UA depending on who sold the ticket.

This report uses ticketing carrier (TkCarrier) — so a Republic-operated regional flight sold as UA 4321 is attributed to United, which is what actually reflects the airline's revenue.

4. Yield computation

For each carrier subset:

yield (¢/RPM) = 100 × Σ(MktFare × Passengers) / Σ(MktMilesFlown × Passengers)
avg fare ($/pax) = Σ(MktFare × Passengers) / Σ(Passengers)
avg distance (mi) = Σ(MktMilesFlown × Passengers) / Σ(Passengers)

Passenger counts in DB1B are already sample counts; the 10× population multiplier cancels out in yield and average-fare ratios but is applied when estimating total revenue.

Findings

Headline: yield by carrier

Figure 1. Yield (¢ per revenue passenger-mile), ORD-origin domestic, 2025 Q2.

Two things stand out. First, the legacies (DL, UA, AA) sit clearly above the LCC/ULCC cluster (B6, WN, F9, NK). Second, within the Big Three at ORD, United yields 10.5% more than American at nearly identical average stage length (953 vs 969 miles). Delta yields more still, but Delta at ORD is a spoke operation, not a hub — smaller, thinner network, better-priced.

Same distance, same route: head-to-head

The averages above could be mix effects. To test whether UA actually charges more on the same route, we compare AA vs UA nonstop yields on the top 18 O&D markets from ORD.

Figure 2. AA vs UA yield on the top 18 nonstop O&D markets from ORD, sorted by combined passenger volume.

UA out-yields AA on 16 of 18. The only markets AA wins are MIA and CLT — American hub feeders where UA sells thin. On UA's own hub-to-hub routes (EWR, DEN, IAH, SFO) the premium is 25–33%. Flat routes (DCA, MSP) are business shuttle markets where fare compression from competition levels both carriers.

Yield vs stage length

Yield falls with distance — that's the physics of airline pricing. But UA's yield is above AA's in every distance band.

Figure 3. Yield (¢/RPM) by great-circle distance band, ORD-origin domestic, 2025 Q2.

Passenger share at ORD

Nonstop vs connecting

For UA, connecting itineraries yield noticeably less than nonstops (23.55 vs 26.73). For AA, the gap is essentially zero. But the sample of connecting itineraries in DB1B is much thinner (fewer than 10% of records), because Chicago is overwhelmingly an origin-and-destination city as well as a connect hub.

Figure 4. AA and UA yield on ORD-origin nonstops vs connections.

Rough revenue sizing (ORD-origin domestic only)

Scaling the 10% sample to population and annualizing Q2:

Order-of-magnitude only. This is ORD-origin ticket revenue for domestic tickets — one direction, no ancillaries, no loyalty, no cargo, no international.

Fleet Mix at ORD

Yield tells us who charges more. To ask who makes money, we need the cost side. That starts with what airplanes each carrier actually flies at ORD.

Operator mix (measured from DB1B)

DB1B records the operating carrier for each ticket. Filtering to nonstop ORD-origin itineraries with AA or UA as the ticketing carrier gives us the actual operator split:

Figure 5. Operating carrier as % of ticketing-carrier passengers, nonstop ORD-origin, 2025 Q2.

United runs a materially more mainline-heavy operation at ORD than American — 86% of UA passengers vs 76% of AA passengers fly on a mainline metal. AA leans on Envoy (its wholly-owned subsidiary), Republic, and SkyWest for the remainder. UA leans on SkyWest, Republic, and GoJet.

Aircraft mix within each operator (modeled from public schedules)

DB1B does not include aircraft type. Route-level fleet data lives in BTS T-100 Segment, which BTS gates behind an interactive form and does not publish for programmatic download. Rather than skip the analysis, we model each operator's aircraft mix from published fleet composition and carrier schedule releases. These are best-effort share estimates, not measurements:

Derived departure mix

Converting passenger share to departure share (regional jets carry fewer seats, so a given passenger share implies a higher departure share):

Figure 6. Estimated share of daily ORD departures by mainline vs regional, using measured passenger share and modeled seat capacity per operator.

American's ORD departure schedule is roughly 58% mainline / 42% regional. United's is 69% mainline / 31% regional. That 11-point gap in mainline share is the dominant cost driver in the analysis below.

Cost Estimates

Aircraft operating expenses per available seat-mile (CASM) come from BTS Form 41 Schedule P-5.2 (Aircraft Operating Expenses), which reports fuel, crew, maintenance, and ownership by aircraft type and carrier — but at the entity level, not the airport level. There is no way to compute a hub-specific CASM directly. We combine type-average CASM with the type mix above to build a weighted ORD CASM for each carrier.

CASM by aircraft type

Figure 7. All-in operating cost per available seat-mile by aircraft type (cents). Form 41 P-5.2 published averages, calendar 2024. Regional CASM is higher because seats are fewer and turn/handling costs are similar.

Weighted CASM at ORD

Putting revenue and cost together

Assuming an 84% load factor (industry standard for hub carriers; both AA and UA have reported very close to this systemwide in 2025), we can convert yield to RASM (revenue per available seat-mile) and compare:

Figure 8. Implied RASM (yield × load factor) vs weighted CASM, cents per ASM.

Does the model support Kirby's $800M–$1.1B loss claim for AA?

Not from ticket economics alone. On the assumptions above, AA's ORD network throws off roughly $1.8B of gross variable margin annually (ticket revenue minus direct aircraft operating cost). To turn that into a $1B loss you would need to allocate ~$2.8B of overhead to ORD — plausible only if you include a substantial share of AA's corporate, loyalty program, and fleet capital costs proportional to ORD's ASM share of the system (~7–8%). That kind of full absorption accounting is exactly what airlines don't publish, which is why Kirby's number is unverifiable in either direction.

What the model does clearly show: UA's margin per ASM is 31% higher than AA's, driven roughly two-thirds by higher revenue (yield) and one-third by lower cost (more mainline). Even if both carriers are profitable at ORD in absolute terms, UA's per-unit economics are meaningfully better — which is the direction of Kirby's argument if not the exact magnitude.

Consumer Impact: Who Actually Benefits from Gate Reallocation?

The airline yield story is one question. A different question — one the AULA formula ignores completely — is: what do Chicago travelers get out of AA or UA winning gates?

The FAA cap changes the game

The April 2026 FAA order caps ORD at 2,708 daily operations. Under a hard cap, more gates does not equal more flights. Gate reallocation becomes a mix change, not a volume change — a carrier that gains gates can upgauge (bigger aircraft, more premium seats, more international) but cannot add total departures. So "consumer benefit" isn't about new flying; it's about whose flying mix replaces whose.

How competitive is ORD already?

Figure 9. Number of carriers with meaningful nonstop service (≥1,500 pax/quarter) at each of the top 50 ORD destinations, 2025 Q2.

Only 4 of 50 top ORD nonstop destinations are true monopolies (IAD-UA, DAL-WN, SMF-UA, SAV-UA). 20 are duopolies, 26 have 3+ carriers. Most ORD travelers already have multiple nonstop options. That limits how much "add a competitor" gains are still available.

What consumers actually pay when competitors leave

Figure 10. Average legacy fare per passenger under different market structures. Pax-weighted, 2025 Q2.

Two clean signals from the data:

• When UA is the only major legacy on a route (AA absent), UA charges $32 more per passenger (~14%) than when both fly. Distance-controlled. Small sample (3 routes) but consistent direction.
• In the 1000–1500 mi band, adding an ULCC (Spirit/Frontier) cuts the average legacy fare by $40 per passenger — a 14% reduction. This is the classic "Southwest effect" applied to Spirit/Frontier at ORD. It's largest in the 1,000–1,500 mile band where ULCC leisure demand is strongest.

Who benefits from which outcome

The clean framing

The AULA formula measures departures. It measures nothing about consumer surplus. A gate given to Envoy for another daily E-175 to a small Midwest city generates less measured revenue than the same gate given to UA for a 787 to Rome — but likely generates more consumer surplus, because it's the difference between having ORD service and not having it. The formula, by construction, can't see this.

If Chicago wanted to optimize gate allocation for its residents rather than for signatory contract math, it would look at three things the AULA doesn't:

1. Route uniqueness. How many gates protect the only nonstop from ORD to a given city?
2. Effective competition. Where does the "add a carrier" fare effect still have room to run? (Answer from Figure 10: 1,000–1,500 mi routes with no ULCC — a small remaining list.)
3. International connectivity. ORD's role as a global gateway concentrates in UA's widebody fleet. Chicago economic-development interests probably weight this higher than pure fare-cutting.

None of that is in the AULA. Which is why the fight over gates is fought purely on departure math, and consumer welfare is a byproduct.

Weaknesses of the Data

1. No ancillaries. Bag fees, seat selection, change fees, and priority boarding are not in fares. American's basic-economy bag policy monetizes differently than United's — none of that shows up here.
2. No loyalty/co-brand revenue. The Citi AAdvantage co-brand contract is one of the largest single revenue lines American has, and part of why AA values ORD. It is booked as marketing revenue, not ticket revenue. Same for UA's Chase relationship. This is where the AA/UA argument actually lives, and it's completely invisible in DB1B.
3. No cargo. UA's international ORD widebody flying carries meaningful cargo — a full revenue stream not in DB1B.
4. Domestic only. DB1B has partial international coverage. UA's ORD international network is substantially larger than AA's; this analysis excludes it, which understates UA's revenue advantage.
5. No cost data. Yield is a revenue metric, not a profit metric. AA could out-yield UA and lose money (higher CASM), or vice versa. The Form 41 financial schedules stop at entity/region, never at hub.
6. 10% sample noise. Aggregate yields are precise, but individual O&D estimates on smaller markets carry sampling noise. Nothing in the top 18 markets has less than ~1,300 sample passengers, so head-to-head comparisons above are reasonably tight, but any single small-market number is not.
7. Origin-directional only. This is ORD-origin. Return-direction revenue is booked to the same ticket but appears in that quarter's DB1B under a different origin. For a round-trip yield analysis you'd need to pair MktID pairs — a modest extension.
8. Timing. Q2 2025 is the last quarter before the October 2025 gate reallocation. It reflects the pre-reallocation network, when AA still had 63 gates. Post-October the AA schedule shape (and thus yields) will shift.
9. Bulk fare exclusion. Removing BulkFare == 1 drops tour operator tickets, which is standard but excludes a small revenue slice for vacation-heavy markets (MCO, LAS, FLL).
10. Reporting integrity. DB1B is self-reported by carriers under Form 41 rules. There is no audit of individual tickets. Systemic biases are believed small; individual quarter noise is not zero.

Additional weaknesses of the cost model

11. Aircraft mix is modeled, not measured. BTS T-100 Segment contains actual aircraft-by-route data but is only available through an interactive form download, not programmatic API. The per-operator aircraft mix in this report is estimated from published fleet composition, wikipedia carrier fleets, and airline schedule announcements. Individual aircraft share numbers are best-effort. However, the two headline modeled quantities — mainline vs regional split — are anchored to measured DB1B operator share, so the top-line CASM estimate is reasonably robust.
12. CASM figures are averages, not marginal costs. Form 41 P-5.2 gives fleet-wide average operating cost per aircraft type. Actual ORD marginal cost may differ: fuel prices, crew basing, maintenance contracts, and airport-specific charges are hub-specific. UA is based in Chicago; AA is not, and pays for crew basing accordingly.
13. Load factor assumed identical. We use 84% for both; T-100 (which we don't have) would give the actual load factors, and they likely differ by 1–3 points. UA's more mainline-heavy operation may actually run slightly higher LF due to network scheduling flexibility.
14. Departure counts are announced peak numbers. Real departures fluctuate seasonally and were affected by the April 2026 FAA operations cap. Annualizing peak-summer departures overstates full-year ASMs by ~10%.
15. Cost excludes non-aircraft-operating items. P-5.2 covers flying operations (fuel, crew, maintenance, ownership) but not passenger service, general/administrative, sales/marketing, aircraft/traffic servicing, or transport-related expenses. Fully-loaded CASM including SG&A typically runs 3–5¢ higher than P-5.2 alone.
16. Revenue is one-directional. ORD-origin only. To get a proper round-trip revenue figure you'd need to double the DB1B pass (once for ORD-origin, once for ORD-destination) and reconcile.

Conclusion

The public data supports the direction of United's argument but not the magnitude of Kirby's dollar figures.

What we can say: On the same routes, at the same distances, in the same quarter, United extracts materially more ticket revenue per seat-mile than American at O'Hare. The gap is 10.5% overall, 25–33% on United's premier business routes. This is not a mix effect and it is not sampling noise. It is a real pricing-power gap on the shared network.

What we cannot say: Nothing definitive about profit. American's counter-argument — that loyalty and co-brand economics, network contribution to DFW and CLT, and cost discipline offset lower ORD yields — is unfalsifiable from public data because the offsetting revenue streams are exactly the ones DB1B excludes. Kirby's specific numbers ($800M to $1.1B in annual losses) are unverifiable by construction. Both sides can be arithmetically correct.

What the cost model adds: A partial view of the cost side. On aircraft operating economics alone, UA's advantage compounds — it not only earns more revenue per seat-mile (+10.5%) but also spends less per seat-mile (−4.6%), because a higher share of its ORD flying happens on lower-CASM mainline metal. The combined effect is a 31% margin advantage per ASM. However, aircraft operating cost is only part of the cost stack; when SG&A, station costs, and gate rent are layered in, absolute margins compress and the Kirby loss claim becomes possible for AA if — but only if — overhead is fully absorbed to ORD in proportion to its ASM share.

The one clean conclusion is meta-level: the AULA gate reallocation formula is measuring departures, not economics. A carrier that yields 25% less on a given route still counts the same departure toward its next year's frontage entitlement. That is the structural defect the FAA capped its way out of in April 2026, and no amount of yield analysis will resolve it — only a change to what the formula measures would.