This is the final installment of our three-part series exploring how robots and automation are transforming employment, featuring groundbreaking research from renowned economist and MIT Institute Professor Daron Acemoglu.
Today's technological landscape impacts workers in dramatically different ways. While professionals in certain white-collar fields—such as designers and engineers—find themselves empowered by sophisticated software that boosts their productivity, other workers face a different reality. Various forms of automation, from manufacturing robots to customer service systems, have completely eliminated positions previously held by factory workers, receptionists, and countless other employees.
A groundbreaking new study co-authored by an MIT economist reveals that automation's influence on the labor market and income inequality is far more significant than previously believed. The research identifies 1987 as a crucial turning point when jobs lost to automation were no longer being replaced by an equivalent number of similar employment opportunities.
"Understanding automation's role is essential to grasping the dynamics of inequality," explains Daron Acemoglu, the MIT economist who co-authored the recently published paper detailing these findings.
The research demonstrates that within industries embracing automation, the average "displacement" (job loss) from 1947-1987 was 17% of jobs, while "reinstatement" (new opportunities) averaged 19%. However, from 1987-2016, displacement remained at 16%, but reinstatement dropped to just 10%. The message is clear: those factory positions and phone-answering jobs are not returning.
"Many of the new job opportunities that technology generated between the 1960s and 1980s benefited low-skill workers," Acemoglu notes. "But beginning in the 1980s, and particularly during the 1990s and 2000s, low-skill workers faced a double disadvantage: They suffered from displacement, while the new tasks that emerged were created more slowly and primarily benefited high-skill workers."
The new paper, titled "Unpacking Skill Bias: Automation and New Tasks," is scheduled for publication in the May issue of the American Economic Association: Papers and Proceedings. The authors are Acemoglu, an Institute Professor at MIT, and Pascual Restrepo PhD '16, an assistant professor of economics at Boston University.
The Declining Fortunes of Low-Skill Workers
This latest paper is among several recent studies by Acemoglu and Restrepo investigating the workplace impact of robots and automation. In another recently published paper, they determined that across the United States from 1993 to 2007, each new robot eliminated approximately 3.3 jobs.
In yet another new study focusing on French industry from 2010 to 2015, they discovered that companies rapidly adopting robotics became more productive and hired additional workers, while their competitors lagged behind and reduced their workforce—resulting in an overall net loss of jobs.
For the current study, Acemoglu and Restrepo developed a model of technology's effects on the labor market, testing its validity using empirical data from 44 relevant industries. (The research utilizes U.S. Census statistics on employment and wages, along with economic data from the Bureau of Economic Analysis and the Bureau of Labor Studies, among other sources.)
The result offers an alternative to standard economic modeling in the field, which has traditionally emphasized "skill-biased" technological change—the notion that technology primarily benefits high-skilled workers more than their low-skilled counterparts, boosting the wages of the former while the latter's wages stagnate. Consider again highly trained engineers leveraging new software to complete more projects faster: They become more productive and valuable, while workers lacking technological synergy become comparatively less valued.
However, Acemoglu and Restrepo believe even this scenario, with its prosperity gap, is too optimistic. Where automation occurs, lower-skill workers aren't merely failing to advance—they're being actively pushed backward economically. Furthermore, the standard model of skill-biased change doesn't fully account for this dynamic; it estimates that productivity gains and real (inflation-adjusted) wages should be higher than they actually are.
More specifically, the standard model suggests approximately 2% annual productivity growth since 1963, whereas actual productivity gains have averaged only about 1.2%; it also estimates wage growth for low-skill workers of about 1% annually, while real wages for these workers have actually declined since the 1970s.
"Productivity growth has been lackluster, and real wages have fallen," Acemoglu states. "Automation explains both phenomena." Additionally, he notes, "Demand for skills has decreased almost exclusively in industries that have experienced significant automation."
The Problem with 'So-So Technologies'
Indeed, Acemoglu emphasizes that automation represents a special case within the broader spectrum of technological workplace changes. As he describes it, automation "differs from ordinary skill-biased technological change" because it can replace jobs without adding substantial productivity to the economy.
Consider the self-checkout systems in supermarkets or pharmacies: They reduce labor costs without making the task more efficient. The difference is that you, the customer, now perform the work instead of paid employees. These systems exemplify what Acemoglu and Restrepo call "so-so technologies" due to the minimal value they provide.
"So-so technologies don't perform exceptionally well—nobody enjoys scanning their items one-by-one at checkout, and nobody appreciates being routed through automated phone menus by their airline," Acemoglu explains. "So-so technologies are essentially cost-saving devices for companies that slightly reduce expenses but don't significantly boost productivity. They create the usual displacement effect without substantially benefiting other workers, giving companies no incentive to hire more employees or increase wages."
Certainly, not all automation resembles self-checkout systems, which didn't exist in 1987. Automation during that era primarily involved converting printed office records into digital databases or adding machinery to sectors like textiles and furniture manufacturing. Robots became more common in heavy industrial manufacturing during the 1990s. Today, automation continues to evolve through software and AI technologies that inherently displace workers.
"Displacement is truly central to our theory," Acemoglu asserts. "And it has more concerning implications because wage inequality correlates with disruptive changes for workers. It's a much more Luddite explanation."
After all, the Luddites—British textile workers who destroyed machinery in the 1810s—may be synonymous with technophobia today, but their actions stemmed from economic concerns; they understood that machines were replacing their jobs. That same displacement continues today, though, as Acemoglu argues, technology's net negative impact on jobs isn't inevitable. We could potentially develop more job-enhancing technologies rather than job-replacing innovations.
"It's not all doom and gloom," Acemoglu concludes. "Nothing dictates that technology must be detrimental to workers. The critical factor is our choice about which direction to develop technology."
