Has technology caused unemployment? - Economics Observatory

There are several historical precedents for large-scale and sustained unemployment following technological change. Understanding these past experiences will help researchers and policy-makers to have a clear view of the potential effects of artificial intelligence on the labour market.

Discussions about the effects of artificial intelligence (AI) on society have focused on the potential for mass unemployment. A series of technology leaders, perhaps most notably Anthropic’s chief executive Dario Amodei (Axios, 2025), have claimed that AI will replace huge swathes of jobs.

Writers who contest these worrying projections frequently claim that a historical view of technology’s impact on work shows that such concerns are unwarranted – a recent article in The Economist is one such example. These commentators argue that the past indicates that we have no cause to fear large-scale job loss from new technologies. Indeed, they claim, there was no ‘technological unemployment’ in the most famous of such episodes: the Industrial Revolution.

Similarly optimistic views also appear in academic discourse. The recent economics Nobel laureate Joel Mokyr wrote in his 2002 book Gifts of Athena that technological unemployment did not appear in various historical episodes of disruptive innovation, and he reiterated this view in a 2015 article.

If there is no precedent for substantial technological unemployment and history is a useful guide to the future, we can be reassured: there is no need to worry about a future without work, or even a future with less work.

What can history tell us about technological unemployment?

Technological unemployment has been the subject of periodic and bitter debate, at least since the Industrial Revolution itself. The Economist claims to resolve this debate by presenting data that show steady growth in men’s employment across the 18^th and 19^th centuries. The article also states that there were fewer changes in the composition of male employment by industry in the Industrial Revolution than in the 20^th century.

While these data are easily accessible, they do not provide data about unemployment. The British population grew substantially during the late 1700s and 1800s, so a rising number of employed men does not tell us whether there was any technological unemployment. Data on changes in the composition of men’s employment only show shifts in the shares of workers employed in different industries.

Concerns about technological unemployment have also been downplayed – for example, by Joel Mokyr – because some episodes of dramatic technological change do not have a documented record of ensuing unemployment.

While it is true that mass unemployment did not follow all transformative technologies, substantial and long-term unemployment has followed several past waves of technological disruption.

Public discussions of the history of technology and employment usually focus on male employment. Women have always worked, and neglecting their labour market experiences ignores a large part of the employment consequences of technological change (Agren, 2016; Whittle et al, 2025).

Recent research documents a dramatic example of the pitfalls of ignoring women’s employment and unemployment. The textile industry was probably the second-largest sector of the economy before industrialisation, and spinning yarn was a major source of employment, predominantly for women and children. This industry was mechanised early in the Industrial Revolution, with factory technologies eliminating work that had employed 8% of the British population in 1770 (Schneider, 2026).

There are no national unemployment statistics for Britain in the Industrial Revolution, so this research uses the writings of 18^th century social commentators and a one-off survey by the Poor Law Commissioners in the 1830s to measure the availability of work for women.

These materials are different from current-day unemployment surveys that provide longitudinal data on the share of the labour force who are looking for work, so the historical sources do not provide an unemployment rate. Instead, they document that unemployment and underemployment remained a large-scale social problem many decades after the introduction of industrial machinery.

A better-known example of technological unemployment during Britain’s Industrial Revolution occurred when handloom weavers were replaced by the powerloom in the 19^th century. One study suggests that there were between 200,000 and 250,000 handloom weavers around 1830 (representing 1.25-1.5% of the population), and they were replaced by powerlooms in the following decades (Bythell, 2008).

The spread of the powerloom led to loss of work and poverty for the handloom weavers. There is little evidence to track individual weavers from unemployment into new jobs, so while some of them subsequently found better-paid work, we do not know if this was the most common outcome (Bythell, 2008).

What’s more, the handloom weavers were almost entirely men, which is likely to be the main reason that their travails have been discussed extensively, while the hand spinners, who were predominantly women, have been neglected.

What can we learn from American experiences?

Britain’s Industrial Revolution is not the only example of large-scale labour market disruption as a result of new technologies. More recent research examines the automation of telephone switching, which replaced tens of thousands of American women who worked as telephone operators in the first half of the 20^th century (Feigenbaum and Gross, 2024).

Tracking operators at ten-year intervals using the US Census reveals that 50-80% of women who were in locations where telephone switching was automated lost work as operators. Roughly half of the workers who were over the age of 26 became unemployed or exited the work force.

During the same mid-20^th century period, millions of cotton pickers in the United States were displaced by mechanisation (Grove and Heinicke, 2003). Historians have yet to investigate for how long these individuals were unemployed. But this episode was followed by the large-scale migration of black Americans from the South to Northern cities, so they probably had a meaningful period without work.

The post-Second World War period has the best data for examining the individual consequences of job-replacing technology. While this era is commonly looked on as prosperous because it was characterised by low aggregate unemployment and high growth rates, contemporary reports from the US Bureau of Labor Statistics (BLS) show that many workers replaced by technology suffered long spells of unemployment (Jayes, 2025).

The BLS provides more granular detail on the duration of unemployment than is available using the Census, and its researchers found that unemployment spells of more than three months were common for workers who lost their jobs to new technology in the 1950s and 1960s.

A smaller share of workers experienced technological unemployment in the post-war period than in the Industrial Revolution. But unions were relatively strong at that time, and they ensured that many workers were reassigned within firms.

Even in this period of near-full employment and rapid growth, the fact that workers who were not placed directly into new jobs experienced extended periods of unemployment suggests a pessimistic picture for other periods for which we have less detailed information on individual workers’ experiences.

What do we know about how technological unemployment has varied across countries?

Historians have not investigated all job-replacing technologies, and many widely used innovations have only been analysed in one or a few countries. In the absence of a systematic picture, public debates commonly focus on developments in Britain during the 18^th and 19^th centuries as this is the best-known and most evocative example.

The industrialisation of spinning provides one opportunity to compare employment outcomes when similar technologies affected employment in different countries, as it has been more carefully studied across settings than other innovations.

This technological wave produced unemployment in India, Ireland and Switzerland because machine-spun British textiles outcompeted hand-produced local goods or because there was local adoption of industrial machinery. This local industrialisation also caused large-scale displacement of workers in Japan.

In some countries, technological unemployment was sustained; in others (including Japan), many women who were displaced from hand-spinning employment rapidly found new work in textile factories (van Nederveen Meerkerk, forthcoming). This variation between countries shows that the effects of technology differ depending on the setting in which it is introduced, and different technologies also, obviously, have different impacts.

Highly disruptive episodes tend to feature technologies that offer vast gains in productivity. In spinning, for example, output per worker rose about 100 times between 1760 and the 1830s through a combination of new technologies (Maw et al, 2022). There was a large share of the workforce employed in the occupations replaced in several countries, which made it more difficult for these workers to find new employment.

Workers with fewer occupational options in general, perhaps because of their previous skills or gender norms about acceptable employment, were likely to be in a particularly disadvantaged position. For example, social expectations about women’s employment and caring responsibilities limited the job options of displaced hand spinners. At the same time, if new technologies were deployed to meet large and growing demand (as in Japanese textiles), the disruption produced by innovation was short-lived.

The scale of productivity gains, the size of the affected workforce and the extent of new job options are important factors that determine which technologies are most disruptive, and how long their effects ripple through the economy. Many technologies that produce smaller gains in productivity have no effect on employment, and instead they allow existing workers to be more productive.

Conclusion

Historical research shows that there are precedents for large-scale and sustained unemployment following technological change. While prior occurrence does not guarantee repetition, only a clear-eyed view of the history of technology and employment allows us to understand the range of possible future outcomes, whether worrying or hopeful.