Every technology wave has displaced workers. And every time, the economy found somewhere to put them.
The pattern has held for forty years. Industrial workers displaced → absorbed into services. Clerical workers automated → absorbed into knowledge work. Retail hollowed out → absorbed into logistics, e-commerce operations, delivery. Each wave was painful for the people in it. Each wave eventually found a floor. The workers who survived had a next step on the escalator.
There is no next step. This time the escalator ends in the air.
Understanding why requires understanding what was actually happening during those forty years — not just the technology waves, but the economic structure they were dismantling and what got built in its place.
The Ratchet
Start with steel.
American steel employment peaked at roughly 650,000 workers in 1953. By 1983, it was under 250,000. By 2000, under 150,000. Today, under 80,000. That's not a decline — it's a dismantling.
Manufacturing overall peaked at 19.5 million American workers in 1979.
Reagan didn't cause deindustrialization — it had been building through the 70s, driven by the combination of rising productivity, cheaper overseas labor, and a strong dollar that made American exports expensive. But the Reagan era was the tipping point: union density collapsed from 35% to 17% over his two terms. The implicit social contract that connected productivity gains to wage growth — enforced by unions, by the threat of labor action, by the postwar consensus that workers deserved a share of what they produced — was revoked.
The displaced industrial workers went somewhere. Mostly they went to service work. Retail. Food service. Healthcare support. Cleaning and maintenance. The jobs were real, but they were a step down: lower wages, no benefits, no pension, no predictable hours. The economic shock was absorbed, but the absorber was debt. Personal credit card debt began its forty-year expansion in the early 1980s. The consumer economy continued to function not because wages rose but because the ability to spend on credit expanded.
This is the first link in the chain. Remember it: when wages can't sustain consumption, debt fills the gap. And debt has a limit.
The Internet Wave and the VIG Economy
The internet of the mid-1990s did two things simultaneously that the trade press has never quite reconciled.
It created genuine wealth. Microsoft, Intel, Cisco, Oracle, and then Google, Amazon, and the rest built real products with real global value. The tech sector employed millions and paid well and generated legitimate productivity gains that spread across the economy.
And it destroyed the shopkeeper.
Big box retail — Walmart, Home Depot, Best Buy — predated the internet, but the internet accelerated the logic. The local hardware store, the local bookshop, the local record store, the local travel agent: these were not marginal businesses. They were the economic backbone of Main Street in every American town. Shopkeeping — the management, operation, and ownership of local retail — was a middle-class profession. A store owner was someone. It was a path.
That path closed. Not completely, not everywhere, but structurally. The efficiency of centralized distribution and, later, e-commerce made local retail economically indefensible in category after category. The jobs that replaced shopkeeping — warehouse work, delivery driving, customer service call centers — paid less, required less skill, and built no equity. You couldn't own a delivery route the way you owned a hardware store.
Simultaneously: clerical work disappeared. Word processors killed the typing pool. Spreadsheets killed the bookkeeper. Early enterprise software killed layers of data management and coordination that had employed millions of people in middle-level office jobs. These were not low-skill jobs — they were skilled, middle-income positions that had provided stable employment for decades. They went away and were not replaced at the same wage.
Then came offshoring. NAFTA in 1994 was the policy instrument; the underlying logic was the global wage arbitrage. American labor cost $15-20 an hour. Mexican labor cost $2. Chinese labor, after WTO admission in 2001, cost $1. The calculation was not subtle. Manufacturing that hadn't already left accelerated its departure. Then call centers. Then software development. Then accounting support. Then radiological image reading. The pattern: any work that could be digitized and transmitted could be offshored to wherever it could be done most cheaply.
What was actually happening to the American economy during this period is captured in one chart.
Worker compensation as a share of GDP has fallen from 68% in 1970 to under 57% today. Corporate profit share has more than doubled. The productivity gains from electrification, from computing, from the internet, from offshoring — they did not flow to the people doing the work. They flowed to the people owning the systems.
This is the vig economy.
A vig — for those who didn't grow up around card tables — is the cut the house takes. The percentage off the top. The economy that emerged from forty years of technology disruption and globalization is, in substantial part, an economy of vigs: financial intermediaries taking a cut of capital flows, consultants taking a cut of management decisions, real estate taking a cut of both business and household cash flows, marketing taking a cut of consumer spending. Brokers of every description, positioned between productive activity and the people who need it, extracting a percentage.
This was sustainable for two reasons. First, the actual productive work was still being done — just increasingly somewhere else, by someone cheaper, with the margin captured in New York and Chicago and San Francisco. Second, when wages couldn't sustain consumption, America had piggy banks.
The Piggy Banks
Every economy that displaces its workers faster than new sectors can absorb them faces a demand problem. The displaced workers can't buy what the economy produces. This is econ 101: you can't fire your customers. The resolution, historically, has been either redistribution (taxes and transfers that put purchasing power in displaced workers' hands) or credit expansion (borrow from the future to sustain present consumption).
America chose credit. Three times.
The first piggy bank was consumer credit — credit cards, auto loans, retail financing — which expanded massively through the 1980s and 1990s. This absorbed the shock of declining manufacturing wages by allowing workers to spend more than they earned. It worked until the debt burden became heavy enough to itself suppress spending.
The second piggy bank was housing. After the dot-com bust (2001) and the simultaneous shock of 9/11 — which arrived just as Enron was demonstrating that the corporate governance structure of the 1990s boom was substantially fraudulent — the Federal Reserve cut rates aggressively, and the financial system responded by manufacturing mortgage credit on an industrial scale. CDOs and MBS and synthetic instruments with names nobody could explain turned American housing into a global speculation vehicle. Household debt as a share of GDP went from 70% in 2000 to 98% at the peak in 2006. Nine million households lost their homes when it collapsed.
The third piggy bank was the combination of post-2009 QE, pandemic stimulus, and the brief crypto mania of 2020-2022. The Fed printed. The government transferred. Asset prices inflated. The wealth effect sustained consumption for the asset-owning class. For everyone else, the pandemic stimulus checks — one-time transfers of $1,200 and $1,400 — were a glimpse of what permanent redistribution could look like, almost immediately withdrawn.
Each piggy bank bought time. None of them solved the underlying problem: wages for most American workers had decoupled from productivity. The surplus was going somewhere — to corporate profits, to financial instruments, to asset appreciation — but not to paychecks.
The Escalator Ends Here
Now inventory what America has left.
Manufacturing: 13 million workers, down from 19.5 million, and the sector that remains is increasingly automated. The jobs are real but the headcount is contracting and will continue to contract as robotics improves.
Retail: hollowed by e-commerce and big box, employing fewer people in worse conditions than the Main Street economy it replaced. Amazon's warehouse workers are the retail workforce of the 21st century.
Financial services: a vig industry employing highly paid people who are, in aggregate, largely redistributing rather than creating wealth. The financial sector as a share of corporate profits went from 10% in 1970 to over 40% at the peak in 2001. These jobs will be among the first to be automated by AI — quantitative analysis, risk modeling, document review, client advisory services are all well within current AI capability.
Consulting and professional services: McKinsey, Deloitte, the law firms, the accounting practices, the marketing agencies. Another vig industry. Also directly in the AI crosshairs. The junior analyst producing market research reports, the first-year associate doing document review, the entry-level accountant reconciling line items — these are the jobs AI is already doing.
Middle management: the coordination layer in every large organization, the people whose job is to translate between strategic direction and operational execution. Already being compressed. Will be further compressed as AI handles the scheduling, the reporting, the escalation management, the status updates, the meeting summaries.
This is not hyperbole. This is the job description of the next wave of displacement. And unlike every prior wave, it's happening simultaneously across multiple categories. Industrial → service happened over thirty years. Clerical → knowledge happened over twenty-five years. What's coming is not sequential. It's concurrent.
The people being displaced are not the people who got displaced before. These are not workers who can retrain into the knowledge economy, because the knowledge economy is what's being automated. A 52-year-old financial analyst displaced by AI is not going to become a software engineer. A 45-year-old mid-level marketing director is not going to pivot to robotics maintenance. The retraining argument — the perennial response of economists to every technology displacement — has never worked at scale, and it is least credible when the thing people are supposed to retrain into is the thing doing the displacing.
What the Demand Side Does
Consumer spending is 70% of US GDP. The middle class and upper middle class are the core of that spending — not the ultra-wealthy, who save most of their income, and not the poor, who have no discretionary spending. The mortgage-paying, car-buying, restaurant-going, vacation-taking class that was built by postwar prosperity and sustained by forty years of debt expansion.
The companies implementing AI are cutting from exactly that class. Every financial analyst let go, every mid-level manager whose position is eliminated, every marketing department halved is a household that stops servicing a mortgage, stops eating out twice a week, stops booking the annual trip. The corporate earnings that drove the bull market are produced by reducing costs. The consumer revenues that will sustain those earnings require the people whose costs were reduced to keep spending.
You cannot do both. Not at scale. Not simultaneously.
The bull market is coming. It will be spectacular. It will be called an AI productivity miracle. It will drive S&P 500 earnings to records while the consumer economy underneath it slowly hollows out. The financial press will not connect the two because the people writing it will be among the last to be displaced, and because the lag between corporate cost-cutting and consumer demand collapse is long enough that the cause and effect are obscured.
And then one quarter — maybe 2028, maybe 2030, maybe later — the revenue line won't recover. The consumers who were supposed to buy the products didn't have the income to do it. The housing market will signal first: the same financial mechanisms that produced 2008 will produce it again, because the same conditions — wage stagnation, debt burden, displacement — will be present at larger scale.
That's when the demand side delivers its verdict.
The escalator always had a next step. Services absorbed the factory workers. Knowledge work absorbed the clerical workers. The gig economy absorbed what knowledge work couldn't. At each step, the jobs were worse, the wages lower, the security thinner. But there was a step.
This time there is no step. The escalator ends.
And the demand side doesn't forgive.
— J.P. Howlett
Next: Socialism Saves Capitalism — the structural irony at the end of the road.
Part 1: When the Levee Breaks — the leading indicators, the dam, and when it gives.
Also in the series: Transitional Tech — the gig economy was the safety net for displaced workers. It was itself transitional technology.
Sources
- BLS Manufacturing Employment Data — Bureau of Labor Statistics
- BEA National Income Accounts — Labor and Profit Shares — Bureau of Economic Analysis
- Federal Reserve Household Debt Statistics — Federal Reserve Z.1 Financial Accounts
- Automation and the Future of Work — McKinsey Global Institute
- The Declining Labor Share of Income — IMF World Economic Outlook, 2017