Participation, Not Prices
How a Consumer Resale Market Absorbs Scheduled News
Information-Perishable Goods and the Missing Margin of Dynamic Pricing
Evidence from 178M observations of the FIFA World Cup 2026
Final resale order book · June 12 – July 19, 2026
Draft v9 · July 2026
The puzzle in one slide
July 15, 2026, 21:05 UTC. England–Argentina ends. The second finalist is public, worldwide, instantly.
- A Final ticket's value just jumped for Argentina's fans and collapsed for England's — for everyone, at the same instant.
- Thirty years of dynamic-pricing theory (Gallego & van Ryzin 1994 → state-dependent p(x,t,s)) predict: sellers reprice immediately.
- Sweeting (2012, JPE): even household sellers reprice along the calendar — 40%+ declines approaching the event.
Our measurement at 5-minute resolution: median continuing seller's price change at the whistle = exactly 0.0% — at every resolution, in 26 of 29 testable cells.
The asset class: information-perishable goods
Time-perishable (the canon)
- Value decays along the calendar toward a deadline
- Airline seats, hotel rooms, event tickets
- Seller's problem: when to discount
- Validated empirically (Sweeting 2012)
Information-perishable (this paper)
- Value jumps or collapses at scheduled public epochs
- Tournament tickets, playoff-conditional products
- Seller's problem: how to reprice on a resolution tree
- Measured here for the first time
The puzzle: the same households who optimize over the calendar do nothing over the state tree. Yet the market clears. How?
The benchmark: what theory says sellers should do
| Model | Policy | Prediction at a public shock |
|---|
| Gallego & van Ryzin (1994, MS) | p(x,t) | Decline toward expiry; continuous adjustment |
| Talluri & van Ryzin (2004) | target path | Shortfall ⇒ cut; scarcity ⇒ raise |
| Su (2007); Levin–McGill–Nediak | strategic customers | Flatter paths, but never zero response |
| Aviv & Pazgal (2008, MSOM) | two-period | Shock = structural break; re-solve the game |
| den Boer (2015); Feng & Xiao (2012) | p(x,t,s) | State change ⇒ immediate policy change |
One line: for thirty years, in every variant, the optimal policy responds immediately when demand-relevant information becomes public. Our sellers: 0.0%.
Contributions
- A precise zero (fact). Median reprice 0.0% at state resolutions — 26/29 cells across the entire 30-match knockout bracket; degenerate CIs. RM capability is dimension-specific.
- A measurement warning (export). Every index move decomposes into composition or a repricing minority — never the median seller. Price-index event studies read participation as repricing.
- A dose-graded participation response (mechanism). Quantities respond monotone in state resolved (R32 −0.3% → SF −12.2%); demand strikes at certainty, not surprise.
- Design & policy pricing (so-what). The measured gaps price the fixes — platform tools, organizer instruments, and a regulatory redirect (US FIFA subpoenas; China's sha-shou rules).
Data: the full resale order book
| Source | Content | Coverage |
|---|
| FIFA official resale (households) | Every seat: block/row/seat, category, USD price | 5-min, Jun 24–Jul 19; 167.0M seat-obs, 52 matches |
| TickPick (brokers/API) | Event listing count, min/mean/max | 15-min, Jun 13–Jul 19, continuous |
| Vivid, TicketNetwork | Event stats + full listings | 15-min + ~4×/day (complete) |
| Gametime; StubHub/Viagogo | Event stats / page totals | 15-min / top-grid; common ownership |
- Deduplicated: 178.2M → 178.0M rows; three panel generations unioned.
- Cross-platform: FIFA↔broker inventories ≈99% disjoint; Vivid↔TicketNetwork ~33% shared.
- One consolidated DuckDB database (19 base tables + analysis views).
The shocks: every knockout match, not a selected subset
| Round | Events | State resolved (dose) | Window |
|---|
| Round of 32 | 16 | 1 of 32 teams | Jun 28 – Jul 3 |
| Round of 16 | 8 | 1 of 16 | Jul 3 – 7 |
| Quarter-finals | 4 | 1 of 8 | Jul 9 – 12 |
| Semi-finals | 2 | 1 of 4 (finalist) | Jul 14 – 15 |
Why credible: pre-scheduled timing (no endogeneity), content unknowable in advance, globally public at one instant, 30 repetitions on one asset family.
Result 1 — The intensive margin is exactly zero
| Event (FT, UTC) | Continuing seats | Repriced | Median reprice |
|---|
| QF Morocco (Jul 9) | 680 | 20% | 0.0% |
| QF Belgium (Jul 10) | 737 | 1% | 0.0% |
| QF Norway (Jul 11) | 801 | 1% | 0.0% |
| QF Switzerland (Jul 12) | 801 | 1% | 0.0% |
| SF France (Jul 14) | 837 | 2% | 0.0% |
| SF England (Jul 15) | 870 | 13% | 0.0% |
- Same at 20 further R32/R16 resolutions: 26/29 clean cells (3 excluded: documented feed rescale).
- Full asset matrix: 27 asset×event cells (7 ticket assets × QF/SF whistles) — median 0.0% in 27/27; repriced shares 0.3–49.6% within cells.
- 95% bootstrap CI for the median is degenerate at zero.
- Not a platform lock: off-epoch hazard 0.2%/hr; the 1–20% minority moves up +42% at the two largest events.
Tier A — precisely estimated causal null
Result 2 — Dose gradient across the full bracket
| Round | Cells | Median Δ listings (TickPick) | Regime tail p |
|---|
| R32 | 8 | −0.3% | inside pool |
| R16 | 8 | −0.6% | inside pool |
| QF | 4 | −1.0% | 0.01–0.14 |
| SF | 2 | −12.2% | 0.012–0.014 |
- SF cells: −10.7%, −13.8% within 1h; ~3× the largest same-day non-whistle movement (±4.3%); day-block bootstrap p < 10⁻⁴.
- Placebo assets (3rd-place tickets at QF whistles): flat, −0.0% to −1.6%.
- Monotone in dose — and in calendar proximity (collinear; disclosed).
- No cherry-picking: every knockout match is in the table.
Tier A (timing) / B (magnitude — dose collinear with days-to-expiry)
Result 3 — Index moves misattribute the mechanism
| Event | Total Δ | Within | Exit | Entrant | Attribution |
|---|
| Morocco | +8.9% | +8.9% | 0.0 | −0.1 | repricing minority (20%) |
| Belgium | +1.1% | −0.2% | +1.1% | +0.2 | composition |
| Norway, Switz., France | 0.0% | 0.0% | 0.0% | 0.0% | — |
| England | +3.4% | +0.0% | +3.4% | 0.0 | composition (cheap exits) |
The warning: a naive index study reports "sellers raised prices" (England), "the median seller repriced" (Morocco), or "nothing happened" (France, while 10.7% of the market changed hands). Decomposition is not optional at information epochs.
Tier A — accounting decomposition (mechanical, given the data)
Result 4 — Supply fire sales, demand at certainty
Supply: wealth-graded, hours late
- Post-elimination entrants vs same-day median: Morocco −28%, Switzerland −23%, England −22%, France −13% (Norway +10%, Belgium +56% also shown)
- Entry begins ~3h after the whistle
- Exits are the cheap tail (13–49% below market)
Tier B (entry timing) / C (level) / D (wealth gradient, n=4)
Demand: certainty, not surprise
- All four QF winners were favorites — flows still wait for the whistle
- Decided match: flat in-play series for 2h, step at FT
- Contested match: −7% in-play bleed, still −22% in 45 min post-FT
- Hypothesis (pending odds): irreversible complements require p = 1
Tier D+ — hypothesis, strong circumstantial support
Result 5 — Lifecycle: consumed vs flipped (with DiC)
| Platform / Event | Flip Δ | Continuing Δ | DiC gap |
|---|
| Vivid / SF France | −1.2% | +1.6% | −2.8pp |
| Vivid / SF England | −17.2% | +1.8% | −19.0pp |
| TicketNetwork / SF England | −20.6% | +0.0% | −20.6pp |
- Whistle changes the rate of exit (40% vs ~30% baseline), not the type (whistle exits reappear more than baseline).
- No instant-arb markups (flips return 0 to −18%): pickoff is speed + continued downward discovery.
- Rent-transfer number ($1,500–2,000/ticket) not supported by flip evidence — cut, flagged for matched-counterfactual design.
Tier B (DiC: −19 to −21pp gap at England SF, within-window comparison)
Contrast panel — the calendar channel, at scale
- 30 knockout matches' own markets terminate at pre-kickoff sales cutoffs.
- Listings collapse 47–89% in the final hours — gradual (absorption, not cutoff mechanics).
- Prices drift +2.5% median (IQR −8.3 to +14.3): cheap inventory absorbed first — composition again.
One market, one margin: calendar deadlines and state resolutions are absorbed by the same margin — participation. Sweeting's discounting channel is invisible in indices in both regimes.
Tier B — timing clear; purchase-vs-delist split not measured
Interpretation: who manages the tree?
- Brokers (κ ≈ 0): approximate the state-contingent policy — reposition within minutes.
- Households (κ large): fixed policy between epochs — never reprice.
- Strategic-looking aggregates without strategic sellers: structural work estimates 5–19% strategic buyers (Li, Granados & Netessine 2014); our aggregates look strategic and decompose into a 0.0% median + composition + a ≤20% minority.
The reverse error — assuming repricing where none exists — is what we measure. No prior paper has quantified it.
Why it matters: design, priced by our measurements
| Measured fact | Agent | Instrument |
|---|
| Households never reprice | Platforms | Reprice-on-outcome tools |
| Demand strikes at p=1 | Organizer (FIFA) | State-conditional tickets (prices the tree) |
| Fire sales 3–6h late, −28% | Organizer / platform | Epoch buyback windows |
| Exit wave timed to the minute | Platform | Resolution-synchronized release; trading halts |
Playoff-conditional tickets exist in practice with zero formal analysis — the literature gap is confirmed.
Why it matters: consumer protection runs the wrong direction
United States (2025–26, live)
- NY + NJ AGs: subpoenas to FIFA (May 2026) — seat misrepresentation, artificial scarcity; covers MetLife incl. the Final
- German court injunction vs FIFA resale (Jul 2026); Texas AG vs StubHub
- NY Algorithmic Pricing Disclosure Act (Nov 2025); FTC probe; RealPage settlement
China
- "Sha shou" (big-data price discrimination): PIPL ban; SAMR+CAC rules (Jan 2026); Ctrip victory
- Ticket crackdown: mandatory real-name ticketing; 85% public-sale rule
Our redirect: both frameworks target algorithmic pricing — platforms repricing too well. The harm we measure runs the opposite direction: households can't reprice at all, and speed beats price. No current complaint names this channel.
Limitations (stated plainly)
- Certainty mechanism is a hypothesis until in-play odds are merged (Betfair planned).
- Sale vs withdrawal: exit rate is whistle-timed; exit type matches ordinary days.
- One event-correlated coverage collapse (post-England-SF); FIFA panel is a third-party aggregator with flicker.
- Dose collinear with days-to-expiry; disclosed, not fully separable.
- One tournament, one Final; asking prices, not transactions.
Conclusion
At every scheduled resolution in a complete knockout bracket, a deep consumer market absorbed the news through participation: the median seller never repriced (26/29 cells), quantities responded in proportion to the state at stake (R32 → SF dose gradient), and price indices moved only through composition or a repricing minority.
- Revenue management prices time-perishable capacity. We measure information-perishable goods — and the canonical margin is missing.
- Where sellers don't manage the state tree, the market does it for them — through who shows up.
Participation, not prices.
Key references
- Dynamic pricing: Gallego & van Ryzin (1994) MS; Talluri & van Ryzin (2004); Su (2007) MS; Aviv & Pazgal (2008) MSOM; den Boer (2015); Feng & Xiao (2012) OR.
- Empirical: Sweeting (2012) JPE; Li, Granados & Netessine (2014) MS; Hendel & Nevo (2006) Econometrica; Williams (2022) Econometrica.
- Tickets/resale: Leslie & Sorensen (2014) ReStud; Courty (2003) JLE; Courty & Li (2000) REStud.
- Indices/stickiness: Nakamura & Steinsson (2008, 2012); Bils & Klenow (2004); Cavallo (2018); Brogaard, Hendershott & Riordan (2019) JF.
- Regulatory: NY/NJ AG FIFA investigation (May 2026); NY Algorithmic Pricing Disclosure Act (2025); RealPage DOJ (2025); PIPL (China 2021); SAMR+CAC (2026).
Draft v9 · claims tiered per CLAIMS_REGISTRY.md · artifacts in analysis/