FIRMS THAT AUTOMATE: Theory and evidence

14 Apr 2021

New research by Joel Kariel highlights that the fear of Artificial Intelligence (AI) and robots taking all our jobs is probably currently overstated. There is still significant potential for expansion of these technologies, and they likely to benefit certain businesses at the expense of others.

The impact on workers is also likely to be unequal: those with technological skills, working in highly productive companies, will gain, but it is important to note that automation could still harm some workers.

The author outlines the results of his study:

The rapid rise in the interest and adoption of automation technologies such as AI has led to a flurry of commentary on the potential economic, social and ethical implications. On one end of the spectrum, we have the technology optimists suggesting a world of increased prosperity and leisure time; on the other, those that fear a lack of meaningful work, and extreme inequality.

At a national level, some research suggests robots have replaced jobs. But there is limited evidence on the impact of such technologies on businesses and workers, so I investigate the case of Italy. Despite the increased rhetoric surrounding AI, I find that only around 10% of enterprises use it, and they tend to be larger, more productive, and pay higher wages.

Importantly, I find that businesses grow faster once they start using advanced technologies such as AI and robots. This surprising finding that automation allows businesses to expand, despite evidence that robots displace jobs nationally, is reconciled. I provide evidence that innovative technologies let companies grow faster, taking up a bigger share of the market, which leads to the exit of unproductive businesses.

To my knowledge, this research is the first that can explore the use and impact of AI, Big Data, the Internet of Things, robots, 3D printing and Cloud Computing. Overall, around one in three Italian companies use at least one of these technologies, but these businesses account for almost half of all employees.

But despite a significant share of companies engaging with these technologies, they spend less than 1% of investment on them. This suggests scope to embed technology deeper into the organisation, with potentially significant productivity improvements.

What happens to businesses when they start using these technologies?

This is an empirically tricky task, as my research found ‘automating’ companies were also different before they embrace these new tools. I use a method called ‘Event Studies’ to answer this question: this allows me to compare businesses before and after the year they implemented the relevant technology, and ‘control’ for other observable attributes such as company age and industry. I find that businesses expand employment by up to 10%, and see increases in turnover of up to 12%, in the years following adoption of automation technologies.

Upon automating, firms employ more workers (i.e., they expand after adopting robots), but countries adopting more robots have seen a fall in the employment rate. I reconcile these facts with a model of firms that differ in how productive they are. Enterprises can automate by paying a large up-front cost to access the technology. You can imagine that a supermarket has to incur significant fixed costs to place self-service machines in the store, including rearranging the layout and creating the software required. The long-term benefit to the store is the lower marginal cost of running these machines, compared to wages for human labour.

Due to this big initial cost of automating, only highly productive companies can currently do so. They grow due to the benefits of innovative technologies, pushing unproductive businesses out of the market. Overall this can lead to a fall in the number of people in work. Notably, I find that skilled workers are being hired in automating businesses, while less-skilled individuals don’t benefit.

What can we learn from this? My research highlights that the fear of AI and robots taking all our jobs is probably currently overstated. There is still significant potential for expansion of these technologies, and they likely to benefit certain businesses at the expense of others. The impact on workers is also likely unequal: those with technological skills, working in highly productive companies, will gain, but it is important to note that automation could still harm some workers.

Joel Kariel

University of Oxford | joel.kariel@economics.ox.ac.uk