Second-Order Jobs

When a steel mill closed in the last century, the people who lost their jobs were the most visible cost. What people noticed later, sometimes years later, was everything else that closed with it. The diner across the street. The hardware store. The second-hand car lot. The bar. The house prices. The school enrollment. The pediatrician who could not fill her appointment book anymore because the parents had left.

Most of the economic damage from a lost factory was not in the factory. It was in the second-order jobs - the ones that existed because the factory workers existed.

We are about to relearn this lesson in an unfamiliar setting.

The current wave of AI-driven restructuring is not hitting factories. It is hitting knowledge work concentrated in specific city centers and specific neighborhoods. The people whose roles are being quietly eliminated - mid-level copywriters, analysts, coordinators, project managers - are the same people who, a year ago, were buying twelve-dollar coffees, eating at the restaurant around the corner from the office, paying rent in the denser parts of town, and spending disposable income in dense consumer economies.

When a cohort of those people loses income, the businesses that depended on their income lose revenue. The barista does not notice that any one customer is gone. She notices that the lunch rush is lighter than it was last year. The landlord notices it as longer vacancies. The restaurant notices it as empty tables at eight. The city notices it in sales-tax receipts.

This is the second-order effect. It is not dramatic. It is not photogenic. It is more important than the primary effect.

Part of what makes this hard to see is that the cohort losing income is not an economically homogeneous group in a homogeneous location. Unlike the steel mill, which concentrated its workers in one town, the displaced white-collar workers are scattered across dozens of cities and hundreds of neighborhoods. The signal gets diluted. The barista in Austin and the barista in Prague both see a slightly lighter lunch rush, but neither connects the dots to a global shift in knowledge work. They each file it under "this year was soft."

Part of what makes this expensive is that second-order jobs are the ones that employ the people who were not going to be white-collar workers in the first place. The person behind the counter at the coffee shop, the stylist cutting hair, the mechanic fixing the car, the daycare worker watching the kids. These are jobs that do not compete with AI directly. They do not need to. They compete for customers who used to have spending power, and that spending power is thinning.

The thinning is uneven. The neighborhoods that relied most on white-collar discretionary spending are going to feel it most. The neighborhoods that served lower-income workers are going to feel it differently, mostly as a drop in tips and a drop in demand for the small luxuries that used to fill the middle of the week. The retail that catered to the office lunch crowd is already reporting it, quietly. The commercial landlords in those blocks will see it by next year.

There is a specific irony in who pays. The AI tools are produced by companies concentrated in a handful of places. The revenue flows to shareholders, many of whom are not in the cities where the displacement is happening. The second-order economic damage, meanwhile, stays local. It lands on the small businesses physically adjacent to the displaced workers. That asymmetry - global gains, local losses - is not new. It is how disruption tends to work. But the scale this time is larger than any prior wave of this kind, because the primary cohort (knowledge workers) is larger than the primary cohort in any prior restructuring.

Policy has not caught up to this, and probably will not in time. Unemployment benefits cover the worker. They do not cover the barista. Reskilling programs cover the worker. They do not cover the hardware store. The conventional instruments of labor policy assume the damage radius is a person. It is not. It is a zip code.

If you are running a small business in one of the affected zip codes, the practical moves are the ones every downturn-adjacent business has had to make: get closer to your actual regulars, reduce your fixed costs, and do not assume the old pattern returns on the old schedule. Your customers are not being laid off all at once. They are being thinned at maybe one or two percent a month. The downturn is slow and will not announce itself.

If you are running a city, the practical moves are harder. You cannot subsidize your way out of a permanent change in the income base of your downtown. You can, however, notice it earlier. Which is to say, not wait for sales-tax receipts to crater before asking where the lunch crowd went.

The old lesson of industrial decline was that the factory closing was a single event, but the damage was distributed over decades. The new lesson is similar, with one difference: there is no single event. There is only the gradual, distributed, invisible thinning of a class of jobs, and the gradual, distributed, invisible thinning of every business that depended on that class having money to spend.

Automation will be narrated as a story about technology. It is also, maybe more importantly, a story about neighborhoods.