Most Jobs Today Existed in 1880s, Swedish Data Shows

Roughly seven in ten Swedish workers today hold jobs whose core functions already existed in the late nineteenth century, according to research tracking occupational data across 140 years. The finding challenges widespread predictions of AI-driven mass unemployment and reframes what we should actually worry about.

Researchers William Skoglund, Jakob Molinder, and Kerstin Enflo traced Sweden's occupational structure from 1880 to 2019 using full-count censuses and administrative registers, as detailed in research first reported by CEPR. They harmonized occupations across the entire period to enable direct comparison.

The persistence is striking: nurses, retail workers, and other dominant occupations today perform the same economic function they did in the 1880s, even though their task content has transformed completely. The carriage driver became the lorry driver. The clerk acquired a computer. The occupation endured while individual tasks did not.

Why it matters

This research exposes a critical flaw in how we measure automation's impact. Most AI displacement studies count automatable tasks, but labor markets hire and pay whole occupations, not tasks. When some tasks within a job get automated, the typical result is not elimination but transformation. That distinction matters enormously for policy makers and business leaders trying to anticipate workforce disruption.

The methodological gap

The Swedish pattern holds elsewhere. Applying the same approach to US data, the researchers found only 9% of 2019 employment in occupations new since 1940. This contrasts sharply with other studies suggesting a majority of current work is new, but that gap stems from methodology: counting job titles versus counting actual workers.

When other researchers moved from title-based estimates to person-level census data, they found just 18.3% of workers in 2011-2023 employed in work introduced since 1970, much closer to the Swedish findings.

Where new work actually comes from

The research distinguishes two sources of new occupations. "Schumpeterian" jobs emerge directly from new technology—electricians, computer programmers. "Smithian" jobs arise from division of labor as markets grow and organizations become more complex—accountants, specialists, logistics workers.

Durable, well-paid employment accumulates overwhelmingly on the Smithian side. Specialized occupations are roughly half as likely to decline, while occupations born in recent technological waves are several times more likely to vanish. The punch-card operator is the textbook case.

The mechanism forms a chain: technology creates scale, scale creates specialization, and specialization creates durable labor demand. Technology raises productivity and enlarges markets, but the employment it underwrites appears mostly as specialized roles within enduring functions.

Warning signs in the digital sector

Computer programmers, once at least a decade younger than the average worker, have aged to roughly the workforce average—a demographic signature of an occupation that has stopped attracting young workers. The youngest workers increasingly appear not at the technological frontier but in low-specialization service work.

Earnings data reinforces this pattern. The market rewards specialization depth, not technological novelty. Twenty-first century occupations command no premium over occupations inherited from the nineteenth century.

The real AI question

The genuine concern is not whether AI will displace workers from tasks—occupations have historically absorbed that shock repeatedly. The question is whether the current technological wave will generate enough productivity and scale to spin off durable, specialized, well-paid new work the way earlier waves did.

Earlier general-purpose technologies reliably converted innovation into broad employment through division of labor. The digital wave has not visibly repeated that feat. That should lower alarm about mass technological unemployment while raising concern about a quieter problem: whether AI can sustain the historical bargain of turning productivity gains into new specialized work.

The research was published by CEPR and draws on IPUMS Swedish censuses from 1880-1910, the 1930 census, and FOB/LISA registers from 1960-2019.