When Workers Disappear, Who Gets Paid?
The Real Numbers | Home Economics Journal
A 100-year inquiry into technology, labor, and where the money goes when machines take over
Before you ask where the money goes, ask where the time went.
In 1900, a household without help was a full-time operation. Laundry, cooking, cleaning, childcare. The work was constant, invisible, and almost entirely female. Then the machines arrived. Not all at once. One at a time, each one quietly retiring a category of human effort. The cotton gin. The sewing machine. The gas stove. The washing machine. The dishwasher. The refrigerator.
Each machine freed a block of time. That time did not disappear. It got redirected. Women entered the workforce. Domestic service collapsed as a profession. The labor moved. The wages moved with it. And the capital that built the machines went somewhere too.
That redirection, from the human doing the work to the machine replacing them, is the through line of the past 100 years. Follow it and the current market makes sense. Ignore it and the signals look like noise.
| Wave | Labor Displaced | Capital Winner |
|---|---|---|
|
Electrification
1920 to 1950
|
Physical labor: farm, craft, manual | Industrial stocks. Then gold when the system failed. |
|
Computing
1975 to 2000
|
Clerical labor: routine cognitive, administrative | Stocks overwhelmingly. Gold left behind for 20 years. |
|
AI
2020 to present
|
Knowledge work: analysis, writing, professional services | Stocks, gold, and Bitcoin simultaneously. All diverging. |
Five inflection points, 1926 to now
Mass media (1920s to 1940s). Radio and mass print industrialized information. The typesetter, the copyist, the town crier, all retired by machines that reached millions at once. Data became an asset class. Nobody called it that yet.
The postwar boom (1945 to 1965). The one period in the record where workers captured the productivity gains. Unions were strong, skilled workers were scarce, and wages moved with output. It is the exception. Everything else in this list is the rule.
The Japan and Mexico panic (1970s to 1980s). When manufacturing jobs vanished, everyone blamed Osaka and Monterrey. They were wrong. The jobs went to numerical control machines, robotic assembly, and automated welding. Capital followed the automation. Stocks began their longest bull run in history right as the panic peaked.
The internet (1990s to 2000s). The travel agent, the stockbroker, the encyclopedia salesman, the classifieds department, all gone inside a decade. Capital went entirely to stocks. Gold was irrelevant for twenty years.
The spreadsheet. 1979. The kicker nobody talks about. VisiCalc, then Lotus, then Excel. Within a decade these three applications retired an entire profession most people today do not know existed: the human computer. Large organizations employed rooms full of people, mostly women, whose job was literally to do arithmetic. They maintained ledgers. They checked each other's work. A single insurance company might employ forty of them.
VisiCalc did not make headlines. It spread quietly through the offices of the 1980s and retired the human computer so completely that the profession no longer exists in the memory of anyone under sixty. No farewell. No headline. No data point that said "human computers eliminated: this quarter."
That is the template for what AI is doing now. Not the dramatic Hollywood displacement. The quiet, office-by-office substitution that does not show up in the data until it is already complete.
What the data shows
Every prior wave of labor substitution sent capital to one place. The charts below show why that matters and why the current wave is different.
Productivity and labor share have moved in opposite directions for fifty years. The space between them is capital income: the return that goes to whoever owns the machine instead of the worker it replaced. That gap is at its widest point in the record. And for the first time, the market cannot agree on where to put it.
In Wave 2, stocks averaged 18% per year for 25 years. Gold was irrelevant. The destination was clear and capital concentrated there. In 2026, stocks are up 14%, gold is up 44%, and Bitcoin is down 25%. Three assets. Three stories. No consensus.
The market cannot yet identify which asset wins when AI replaces the knowledge worker. So it buys all three and waits for the data. Watch labor's share of GDP. Watch which companies cut headcount while growing revenue. Watch central bank gold purchases. Those signals will resolve the three-way split before any headline does.
The washing machine gave women back Tuesday. The spreadsheet gave finance departments back a floor of their building. AI is giving companies back entire departments. The question is the same as it has always been. Where does that capital go next?
Further reading
Two studies published in March 2026 put current numbers to the pattern this article describes. The first is from Harvard Business Review. The second is the underlying research from Anthropic, which built the exposure measure used in both pieces.
Sources
US labor productivity and labor share | Bureau of Labor Statistics and Bureau of Economic Analysis NIPA accounts
Historical productivity estimates | Robert Gordon, "The Rise and Fall of American Growth" (2016)
VisiCalc and spreadsheet history | Steven Levy, "A Spreadsheet Way of Knowledge," Harper's (1984)
Azpúrua, A.E. | "Research: How AI Is Changing the Labor Market," Harvard Business Review, March 4, 2026
Massenkoff, M. and McCrory, P. | "Labor market impacts of AI: A new measure and early evidence," Anthropic, March 5, 2026
Dow Jones Industrial Average 1920-1974 | Measuring Worth historical series
S&P 500 1975-2026 | Yahoo Finance and FRED
Gold spot price | OnlyGold and USAGOLD historical data
Bitcoin price | CoinMarketCap historical data
Data window | 1920 through March 27, 2026
This article is part of the Home Economics Journal published by Breadcoins. It does not constitute investment advice. The framework presented is analytical, not predictive.