Industrial district persistence

Why does a region keep making one product for a century after the resource that started it ran out?

Sheffield made knives because the local hills gave it iron, hard grit for grinding wheels, and fast streams to turn them. All three advantages were gone or irrelevant within a century — steam replaced the streams, better ore came by rail, grindstones could be shipped from anywhere. And Sheffield went on making cutlery, then the world's specialty steels. What holds an industry in a place once the ground has stopped helping?

Try the simplest explanation first: inertia in the form of buildings and machines, sunk in place and not worth moving. Ask why that cannot be right. Plant wears out on a twenty- or thirty-year horizon, and every building and machine was replaced several times over during the century in question. Whatever persists is being rebuilt continually rather than merely lasting — so we are not looking for a stock that survives but for a process that reproduces itself.

What could reproduce? Think about what a firm in a specialised trade needs and cannot easily buy at a distance. Three things suggest themselves, each with the same peculiar property: it is more available to a firm the more other firms of its kind are nearby.

A worker with a rare skill is risky to employ, and risky to be, if there is only one employer in the district. Put fifty employers together and the risk falls on both sides at once: the worker can move if the firm fails, and the firm can hire without waiting years. Neither party paid for that; it exists only because the others are present.

The same threshold logic governs suppliers. A machine that does nothing but grind one kind of blade, or a heat-treatment shop, needs a certain volume of demand to exist at all. Once the volume is there the specialist appears — and the cost of starting the next firm falls, because what had to be built in-house can now be bought down the road.

The knowledge is the awkward one, because most of it cannot be written down: how a joint should sound when it is right, how long to leave a blank in the fire. That kind of understanding transmits by standing next to someone — and, as learning-by-standing-there.md argues from the learner's side, that constrains where it can go.

Now the step that explains persistence rather than merely advantage. How does a district get more firms? Overwhelmingly by descent. The dominant empirical finding in cluster studies is that new firms in an established trade are founded disproportionately by people who worked at existing firms in that trade, setting up near where the founder already lives. Each successful firm seeds the next generation locally, inheriting its capabilities: the cluster reproduces by budding, not by attracting outsiders.

Follow that and the original resource drops out of the story. It explains where the first firm was; everything after generation one is explained by descent from it, plus the three shared conditions that thicken as the count grows. Which is why the resource can vanish without consequence — and why districts exist with no founding resource at all. The carpet industry of Dalton, Georgia grew from one woman's revival of hand-tufted bedspreads in the 1890s, a date I am recalling rather than deriving; there was never a resource, only a lineage.

That gives a falsification test with teeth. If new firms in these trades located at random with respect to their founders' previous employers, the descent account collapses. And if districts whose founding resource remained available behaved no differently from those whose resource had vanished, the resource genuinely never mattered after the beginning. Both are checkable in company registers.

The three shared conditions — a labour pool, specialist suppliers, informal knowledge — are not services anyone provides. They are by-products of everyone else's presence, which is why no firm can create them and none can take them away. That is what makes them feed back: each arrival makes the place slightly better for the next. Formal institutions often lock the pattern in afterwards — a technical school, an assay office, a protected regional mark — and these usually arrive once the district exists, so they are consequences that become causes.

But persistence is not immortality. Lancashire cotton, the Ruhr's coal and steel, Detroit's assembly plants all had the same reinforcing structure, and all collapsed. The honest statement is that the mechanism produces inertia, and inertia cuts both ways: the density that spreads a good technique spreads a shared assumption too, so a whole district can be committed to one answer when the question changes, and fail together rather than firm by firm. A model predicting only survival would be refuted by those cases; a model of inertia predicts both the long persistence and the sudden simultaneous death.

Two neighbours differ at clear fixed points. fall-line-cities.md explains why the first activity settled at a particular spot — the founding condition this file takes as given and then discards. retail-clustering.md also has firms crowding together, but for the opposite reason: shops cluster to be found by buyers who want to compare, a demand-side effect that would work between strangers, whereas everything here runs through production, hiring and descent.

• Why some regions switch trades successfully — Sheffield from cutlery to specialty steels — while others hold on to a dying one.
• What policy can do when it tries to start a cluster from nothing rather than thicken an existing one.

Every term the collection defines is gathered in the glossary.