In 1983, Wassily Leontief, a Nobel Prize-winning economist, concluded that human labour would go the way of the horse after the automobile arrived – “first diminished and then eliminated.”
Today, a new wave of doomsaying has emerged surrounding “technological unemployment” as AI, with its promise of business innovation, has marched to the forefront of economic debate. Pundits argue we’re now drifting towards a “world without work” where machines produce all things and algorithms provide all services. Outclassed humans serve no role in this macroeconomic dystopia.
As we argue in Shocks, Crises, and False Alarms, predictions of technology-driven job destruction have a long history and correspondingly, a long record of failure. Angst about worker obsolescence ebbs and flows with each new generation of technology. The threat from automation was once deemed so dire that Bill Gates proposed a “robot tax” on companies choosing robots over human labour to soften the blow of its impact on workers.
Yet, amid relentless technological progress over the past 80 years, during which the labour market has been remade – and then remade again – the US economy has added 120 million jobs. In 2024, as the unemployment rate remains historically low and real wages are growing, few remember Gates’ robot tax idea.
The deflationary nature of technology
Though it remains a popular narrative today, the failed predictions of technological unemployment have not been lost on all economists. More upbeat assessments contend that AI will augment workers – perhaps the least-skilled ones – rather than replace them. Others argue that replacing workers is harder than it seems because jobs are collections of tasks and AI may not be able to do all of them seamlessly.
Such arguments are important and rooted in the microeconomics of work. However, macroeconomics offers the most persuasive arguments against mass unemployment in the face of AI.
What is routinely overlooked is that technology is an inherently deflationary force. When it is broadly impactful, technology drives down costs and prices, pushing up consumers’ real incomes and demand for new goods and services – and thus, new employment. A matter of logic more than luck, such rejuvenation of the labour market has occurred repeatedly.
Perhaps the most powerful demonstration of technology’s deflationary benefits has occurred with food. In the late 19th century, nearly half of all Americans worked on a farm and spent more than 40% of their disposable income on food. Over the subsequent 150 years, successive waves of innovation have left but roughly 1% of Americans working on farms. Food budgets, meanwhile, have fallen to roughly 12% of income.
Lower prices deliver what economists call real income gains – consumers spend less and use the extra room in their budgets to increase consumption of, often new, goods and services, ultimately leading to new employment. Yes, the process has always come with sector-specific job losses and thus microeconomic pain but it has also served as a reliable route to new jobs and macroeconomic gain.
With such a record of failure, it may seem surprising that doomsayers cling to dystopian narratives of technological unemployment. But it is easier to spot technology-driven job losses, which are often concentrated and can happen fast than it is to discern new job creation, which is usually dispersed and happens over time.
“When it is broadly impactful, technology drives down costs and prices, pushing up consumers’ real incomes and demand for new goods and services – and thus, new employment.
”
— Philipp Carlsson-Szlezak, Global Chief Economist, Boston Consulting Group | Paul Swartz, Executive Director, Senior Economist, Boston Consulting Group
Will AI buck the trend?
If we scrutinize AI’s promise along each step of the well-trodden path from new technology to new employment, it seems unlikely that AI will end a history of labour market rejuvenation and adjustment.
- Cost reduction. AI is likely to succeed at replacing labour, particularly in services where digital technologies have struggled to do so in the past. Though the speed and size of this impact are likely to be more modest than many expect today, gradual and cumulatively significant cost reduction (and productivity growth) is a safe bet.
- Falling prices. Unless labour-saving technology can be monopolized, it will continue to drive cost competition and deflation. Far from being specific to food production, the effect has been seen in manufactured goods and now looks set to unfold in services. Policymakers may need to watch for and ensure competitive market structures to ensure that AI becomes a deflationary force in the economy.
- New demand. Real income gains often drive demand for goods and services that were barely known when a new technological wave began (few would have foreseen the arrival of social media marketers a few decades ago). This demand effect could fall short if consumers retreat and save their income gains. However, the resulting savings glut seems unlikely and directly contradicts the historical record.
- New employment. What if a machine or an algorithm can always meet new demand? That humans lose all comparative advantage seems remarkably unlikely. AI’s labour-enhancing properties are as credible as its labour-eliminating ones. But even if Leontief’s words turned out prescient and humans went “the way of the horse,” that would hardly amount to a macroeconomic dystopia. Instead, the overwhelming deflationary impulse would lay the ground for unparalleled prosperity.
The impact of AI will be seen, experienced, and reported primarily through a microeconomic lens that magnifies the disruptions and gyrations that will come. That should not be conflated with the macroeconomic promise that AI holds. Mass technological unemployment remains an ahistorical and unlikely proposition. AI does come with important risks but pervasive joblessness should not top our list of concerns.
A far more likely outcome is a gradual increase in productivity and wealth, punctuated by the microeconomic pain of any economic transformation. The world will not be without work but it will work differently.