With a fixed cost per plant, two plants make sense when freight is dear; once freight falls, one plant can serve both towns and jobs concentrate on the larger side.
The video builds the intuition; sources and limits are set out in the article. English- and Japanese-captioned versions of the vertical film are still on the to-do list.
Can better transport empty small places? Studies find exactly this: small towns where the Shinkansen stops on its way between big cities saw their share of national jobs fall by about a third over decades, and Chinese counties connected by national highways grew industrially about 26% less than those left out.
A plant’s accounts show why. Two equal towns, east and west, each need 5 units; a plant costs 4, and bigger plants are cheaper per unit (increasing returns).
- With poor roads, shipping one unit across costs 2. One plant: shipping 5 units costs 10, plus 4 for the plant, 14 in all; one plant in each town: 4 plus 4, 8. So each town keeps its own plant.
- With good roads, shipping a unit costs 0.5. One plant: freight 2.5 plus 4 is 6.5, cheaper than two plants at 8, so one closes. A plant cannot be built one and a half times, so two plants become one at once.
- Expensive freight used to protect the small town’s plant; once the road opens, the big city’s plant can sell there. One side is always slightly larger, the plant tends to stay there, workers follow, and the large side grows: the core–periphery model (Krugman 1991). The Shinkansen carries people rather than goods, but travel time and tickets are part of the cost of doing business.
There is counter-evidence: Chinese prefecture-level cities with a high-speed rail station have about 7% more employment than those without, and the bridges between Honshu and Shikoku benefited large cities such as Hiroshima and Okayama most. What the studies measure falling is mainly plants and jobs; registered population need not decline. That small places near large cities lose most is our summary of several studies, whose evidence is not fully consistent.
Japanese classroom version (16:9)
A slower Japanese 16:9 version paced for a lecture, in two parts, with the full derivation, worked numbers and review questions.
Sources: Faber (2014); Koster, Tabuchi and Thisse (2022); Lin (2017); Krugman (1991); the numerical example follows the simplified model in Krugman’s Geography and Trade, lecture 1, with illustrative numbers.