A two session domain expert experiment tests congestion based pricing for rail capacity and finds the pricing mechanism triggered as specified while strategic dominance by large operators held despite surcharge accumulation.
When a group of domain experts sat down to play the role of railway operators bidding for capacity on a shared network, they faced a rule with a clear bite: every round, the agent requesting the most slots paid a surcharge, and the agent requesting the fewest earned a rebate. The mechanism was built to make large operators flinch. Across two sessions of the experiment, they did not flinch.
That finding sits at the heart of a new arXiv preprint, "Congestion-Based Slot Pricing in a Railway Auction Game," which turns the abstract problem of rail slot allocation under deregulation into a real-time, browser-based laboratory. Multiple operator-agents, sized to mimic the asymmetry between incumbents and challengers, interact through a repeated auction under time pressure, watching live marginal-cost signals and competitor moves. The congestion-based base price rises with aggregate demand, and a second, asymmetric adjustment hits the high-requesting agent hardest while crediting the low-requesting one (preprint).
The mechanism functioned as its designers intended. The base price tracked demand, and the corrective adjustment triggered the way the rules specified. What did not function was the policy promise riding on top of it. Agents playing the role of large operators stuck with high-request strategies even as the surcharge accumulated, evidence that corrective pricing is necessary but not sufficient to neutralize strategic dominance in a multi-agent setting where large players treat sustained presence as non-negotiable (preprint).
This matters well beyond the lab. Across the EU, UK, and US, regulators are wrestling with how to open up track and airport slots that incumbents guard ferociously. The standard reformist instinct is to add a price signal: make congestion more expensive, write asymmetric penalties into the rulebook, and trust that the largest operators will absorb the lesson. The new preprint's exploratory data, drawn from two structured sessions with a small number of domain-expert participants, argues that this instinct underestimates how deeply market presence is wired into large operators' strategic logic (preprint).
The mechanism design problem the paper points to is a familiar one with a fresh diagnosis. Pricing rules can change what it costs to behave in a particular way. They cannot, on their own, change which behaviors an operator considers acceptable to perform. After-session debriefs indicated that participants' decisions tracked their assigned role rather than their personal disposition, and qualitative responses surfaced motives such as preserving market position and raising rivals' costs, operating alongside ordinary short-term profit seeking (preprint).
For regulators drafting the next round of slot rules, the implication is more about sequencing than about choosing between price and non-price tools. A congestion penalty that fires is a prerequisite; it is not the constraint. The constraint is what large operators believe they must defend, surcharge or no surcharge. That belief is a structural fact about contested infrastructure markets, and the paper frames it as the next thing a serious mechanism design has to absorb (preprint).
The authors are careful about what their experiment can support. The findings rest on exploratory observations from two structured sessions with domain experts in a real-time web environment where humans controlled individual agents; they are not a portrait of how any specific rail authority operates today. Generalization to real railway deregulation awaits the analytical validation and larger-scale multi-agent experiments the paper itself calls for, and behavioral claims come from self-reported, role-conditioned debriefs rather than independent measurement. None of that softens the headline finding: the price signal landed, and the strategic posture held anyway.