How Advancing Automation May Affect Occupations

This 10-minute podcast was released Sept. 5, 2019.

“If a worker task is more routine-based, rather than it involves a more creative solution, then there’s a higher chance for these jobs to be replaced by the automation,” says Sungki Hong, an economist at the Federal Reserve Bank of St. Louis. He talks with Christine Smith, a digital content editor at the St. Louis Fed. They discuss what may motivate firms to automate jobs and the many forms automation may take – from machines and programs to self-driving cars.

Transcript

Christine Smith: Welcome to Timely Topics, a podcast series from the Federal Reserve Bank of St. Louis. I’m Christine Smith, your host for this episode. And joining us today in studio is Sungki Hong, who is an economist from our Research Division. Thanks for being here, Sungki.

Sungki Hong: Hi, Christine. Thanks a lot for inviting me.

Smith: You recently studied the impact of automation on income inequality: How an advancing wave of automation might affect certain occupations more than others.

So, before we dig into that subject, how did you define automation for purposes of your analysis?

Hong: So we actually used the definition that is used by Frey and Osborne. So Frey and Osborne, those are two researchers at Oxford University. So they basically define each occupation or work based on three categories.

So they are based on, first is the perception and manipulation. Second is based on the creative intelligence. And the last one is the social intelligence.

So, when we talk about automation, it’s based on how easily that a robot can replace a worker based on these three categories.

Smith: What would be like an example of a task that would involve perception and manipulation?

Hong: Oh, sure. I can give you examples for these three different tasks. So, for the perception and manipulation, you can think about it as manufacturing. Manufacturing would be one of them. And for the creative intelligence, you can think about research and development. And the last one, the social intelligence, you can think about management.

Smith: OK, management roles?

Hong: Yeah, exactly.

Smith: Interesting. So, what characteristics of certain jobs might make them more likely to be automated, then?

Hong: If a worker task is like more routine-based, rather than like it involves a more creative solution, then there’s a higher chance for these jobs to be replaced by the automation.

Smith: Gotcha. Are there things then that might motivate firms to automate jobs? Like, for example, they see an opportunity to save on human labor costs?

Hong: Yeah, exactly. One example is office administrative support and also food preparation and factories.

Smith: So let’s dig a little bit into your analysis. You wrote an essay and an Economic Synopses, which is available on our website at research.stlouisfed.org. And it was called “The Impact of Automation on Inequality.” So you took what Frey and Osborne did, which was assessing the probability of certain jobs being automated, and then you combined that with the annual income of those occupations as well as the size of employment in those occupations. So what data did you analyze?

Hong: Yes. So we have merged the probability of automation, which is estimated by Frey and Osborne using the machine learning technique. And we merged that dataset with the annual income, also the employment size of different occupations from the Bureau of Labor Statistics in 2017.

And what we find in the data is that occupations with lower income, they face a higher probability of being automated. And also in occupations with larger size of employment, they also face a higher probability of being automated.

Smith: OK, so occupations with a larger size of employment and have lower income have greater probability of being automated.

Hong: That’s right.

Smith: So what are some examples of those types of occupations?

Hong: So, for example, the occupation with small employment size and high annual income, management, also computer and mathematical related occupations. And, on the other hand, the occupations with lower income and larger employment size, for example, the office and administrative support and also production, food preparation and serving related.

Smith: OK, so those on the other end that you said—those latter examples—are the ones that are probably more likely to be automated.

Hong: That’s right.

Smith: So, when you looked at this data then, as you said, lower-paid occupations are most likely to be automated, and they also have larger-size employment. Let’s talk about how that affects income inequality.

Hong: So we have done three hypothetical scenarios for the automation.

The first case is what would happen to the income inequality if we assumed that all the affected workers become unemployed?

In the second case, what would happen if we assume these affected workers earn the minimum wage?

And the last case is, what if these affected workers take a 20% pay cut on their earnings?

And we find that the first case—the case when all the affected workers become unemployed—it increases the income inequality the most. While the last case, when the workers take a 20% pay cut, it also increases the income inequality, but much less than the first two cases that I have mentioned.

Smith: OK. And then in your analysis, what did you determine was probably the most likely or closest to reality?

Hong: So, me and my research assistant, we think that the last case, the case when the workers take the 20% pay cut, is the most likely outcome of the automation.

Smith: OK, and why is that?

Hong: Because when we talk about automation, or the robots replacing the workers, we actually are not quite sure how this automation process would take place. It’s likely that the robots could completely replace a worker.

But we think that a more likely scenario is that a human worker and a robot, they would become complementary to each other. So that’s why we think like a 20% pay cut, or maybe like 10% or 30% pay cut, is more likely than all these workers become unemployed or they just earn minimum wage.

Smith: You mentioned in your research, too, that there are some caveats to consider when analyzing not just the probability of, whose jobs are more likely to get automated, but then what the income impacts could be from that. And one of them that you mentioned is just the cost of R&D expenditures. So, if costs of automation are high enough, you know, maybe it’s more profitable to keep using human labor.

But talk to me a little bit more also what you just mentioned, about whether the labor market could adapt or adjust to automation.

Hong: So the automation probability that Frey and Osborne, they have calculated, they did not take into account like how the labor market would respond to the automation in the long run.

So, in the long run, maybe workers, they would invest in more human capital that is more related to the social intelligence or the creative intelligence that is harder [to be] replaced by the robots. So that’s the reason why we have to take into account how the humans, us, would respond to the revolution of the technologies and how their interaction would play out in the future.

Smith: So, Sungki, when we’re talking about automation, I mean, people probably have different ideas in their minds of what automation equals. Is it like physical machines that might be replacing human labor? Is it software programs or algorithms or computers that might be replacing the work that a human worker once did, like at the grocery store checkout or a kiosk at the airport?

Hong: Yeah, definitely. So the automation, it could take in terms of many different forms. So one example that you probably have in mind is autopilot cars. So that’s where you’re talking about both the software and also hardware of the cars.

But you can also think about the self-checkout at the supermarket. It’s hard to define one specific form how the automation would play out in the future.

Smith: There have been a lot of media stories about the new wave of automation. And I’m just curious to know what inspired you to study this.

Hong: So it’s actually deeply related to my personal experience. I was actually an MIT undergrad. And I was the only one who was studying social science, which is economics, among my friends.

So, when I was walking on campus and looking at my friends, I’d always have this question that, oh, all these brilliant engineers, they could come up with better technology and better ideas to advance the technology.

But, as an economist myself, the question I’m really after is, what is the interaction going to be? And what is the interaction between the research and also the society, how that will play out? Like is the advancement in the technology necessarily a good thing or a bad thing to the society? So that’s why I’m interested in this question.

Smith: Are you emailing your former peers at MIT some of the research that you’ve been doing, then?

Hong: Not yet, not yet. But maybe I will in the future.

Smith: Those engineers probably need to know.

Hong: Mm-hmm.

Smith: Do you think workers might find this topic of particular interest?

Hong: Yes. Like I have mentioned before, workers, they would actually find this very important, because it would determine what kind of knowledge that they should acquire when they’re on the job. And it would also have an important implication for future workers, the students that we are talking about on campus.

So, if they know that the automation would occur in the future, that would also determine what courses they would want to take in college. They would probably want to take the courses that are less likely to be replaced by robots.

Smith: Okay. Well, Sungki, thank you so much for sharing your insights on this timely topic of automation and jobs. Our listeners can find more of Sungki’s analysis on this subject as well as many others at research.stlouisfed.org.

And for more podcasts, you can visit stlouisfed.org/timelytopics. You can also stream this series on Apple Podcasts, Spotify and Stitcher, or ask your Amazon device, “Alexa, play Timely Topics from TuneIn.” Thanks so much for listening.