Cross-Elasticity Mode Shift, What Moves From Truck to Rail
Pull-quote: “There is no such thing as the truck-to-rail diversion rate. There are thousands of flow-level answers, and the average of them is the least useful number in the set.”
Why this matters
The question arrives in every freight plan, toll study, fuel-price scenario, and emissions strategy: if trucking costs rise, how much freight moves to rail? It is usually answered with a single diversion percentage borrowed from another study, applied uniformly across a state or a corridor. That answer is wrong in a specific way: mode shift is not a property of freight in general. It is a property of each flow, and the flows differ enormously. The analytical tool that respects this is cross-elasticity, the responsiveness of demand for one mode to a price change in another, applied at the level where the decision actually happens: commodity by commodity, corridor by corridor.
Why diversion is so uneven
Three structural facts drive the unevenness. Commodity: bulk goods that tolerate transit time respond to price; time-sensitive and high-value-per-ton freight barely moves for any plausible cost change. Distance: rail’s economics improve with length of haul, so short-haul flows are largely captive to truck regardless of price. Infrastructure: a flow cannot divert to a service that does not exist, and drayage at both ends taxes short movements out of contention. A uniform diversion rate averages across all three dimensions and gets every individual corridor wrong.
| Flow profile | Response to trucking cost increase |
|---|---|
| Bulk commodity, long haul, rail-served corridor | Strongest diversion candidate |
| Bulk commodity, short haul | Little response; drayage eats the saving |
| Time-sensitive or high value per ton | Minimal response at plausible cost changes |
| Any commodity, no competing rail service | None; there is nowhere to divert to |
The simulation shape
Cost scenario: trucking cost + X% on corridor set
│
▼
Elasticity assumptions, by commodity class
and distance band (own-price and cross-price)
│
▼
Applied flow by flow across FAF5 records
(commodity, corridor, mode, constraint checks:
is there a rail alternative on this movement?)
│
▼
Diverted tonnage by corridor and commodity,
reported as ranges under elasticity uncertaintyTwo disciplines keep the output honest. Constraints before elasticities: the simulation first asks whether a competing service plausibly exists for the movement, because arithmetic without that check happily diverts freight onto tracks that are not there. And ranges, not points: elasticity estimates carry real uncertainty, so results are stated as bounds under varied assumptions. The finding that survives is usually ordinal, which corridors and commodities respond most, and ordinal findings are what infrastructure prioritization actually needs.
In practice
Cross-elasticity belongs in a wider family of simulation methods, alongside growth forecasting, gravity modeling, network disruption, Monte Carlo, nested-logit mode choice, and economic shock propagation. Run it against a unified model of the public federal freight data, the FAF5 core of roughly 5.7 million flow records, and a scenario applies to the actual flow structure of the network rather than a stylized aggregate. Where a question warrants richer behavioral structure than elasticities carry, nested-logit mode choice takes over; the two answer the same family of questions at different resolutions, and both need back-testing against published U.S. freight benchmarks before their outputs deserve trust.
Closing
When costs change, freight does not shift; particular flows shift. Cross-elasticity simulation applied at the flow level, constrained by what infrastructure exists, and reported as ranges tells a planner which corridors and commodities will actually respond, in what direction, and roughly how much. It replaces one borrowed percentage with a distribution of grounded answers. The single number was never the analysis. It was the absence of one.