Robert Seamans On The Economic Effects Of AI, Serving As An Economic Advisor to President Obama, & Universal Basic Income

Introduction

Professor Robert Seamans teaches courses in game theory and strategy at the NYU Stern School of Business. His research focuses on how firms use technology in their strategic interactions with one another and on the economic consequences of artificial intelligence (AI), robotics, and other advanced technologies.

He previously served as Senior Economist for Technology and Innovation on President Obama’s White House Council of Economic Advisers.

In light of recent developments in generative AI technology and questions about its potential economic and social consequences, Professor Seamans and I sat down to discuss the economic implications of AI, possible solutions to the challenges it presents, and his experience advising President Obama on technology policy.

- - - - - - -

The Economic Effects of AI & Robotics

Aiden Singh: Artificial intelligence has become a dominant topic in public discourse. It features prominently in the news, and indeed, I’ve also written about it myself. 

Many observers are concerned that AI could negatively affect the labor market, particularly through widespread job loss.

However, your research presents a somewhat different perspective, emphasizing the risk of widening wage inequality between software-literate and manual workers. Could you elaborate on your research and findings in this area?

Robert Seamans: I’ll start at a broader level. Artificial intelligence, or AI, is what we call a general-purpose technology. If we look at the history of such technologies, they have ultimately led to job growth and, arguably, a better quality of life overall. In fact, the Nobel Prize in Economics awarded just a few days ago recognized work largely related to this topic.

Historically, technologies tend to boost economic growth, increase employment, and often make work more enjoyable. They do not necessarily lead to people working more hours. Sometimes they result in less work. However, they often create more leisure time and higher incomes, allowing people to spend that free time in more fulfilling ways.

In the case of AI, I think we can use the past as a guide for what the future might look like. I expect it will largely follow the same pattern we’ve seen with other general-purpose technologies. 

When thinking about the impact on employment, I find it useful to divide jobs into three categories. First, there are jobs that will likely disappear or become redundant. Second, there are entirely new jobs that will be created - roles that do not exist today. Third, and most importantly, there are the jobs we already have, which will undergo significant changes. 

I think there has been too much focus on the first category - jobs that might be lost - and not enough attention on the other two, particularly the third, which represents the largest share of the labor market. The second category of new jobs is difficult to study because it’s hard to envision roles that have not yet been created. Though, the third category, the transformation of existing work, is where I believe we need to focus more attention. 

My research, in collaboration with other scholars, aims to systematically identify and measure which jobs are most likely to change due to AI. To do this, we have been developing indices or impact scores that estimate how strongly AI will affect different occupations.

Aiden Singh: Could you elaborate on how these indices were developed and what insights they provide?

Robert Seamans: Certainly. The main initiative we undertook was the creation of what we call the AI Occupational Exposure Index (AIOE). This project actually began in 2018, when we first identified ten key areas in which artificial intelligence was showing significant progress at that time. We then linked those areas to specific human abilities, and in turn, connected those abilities to actual occupations.

Some of this mapping had already been conducted by the Bureau of Labor Statistics through a resource called O*NET, while other parts were developed by our team using methods such as crowdsourcing. Our initial findings were published around 2018.

Fast forward to late 2022 and early 2023, when we began to see dramatic improvements in image generation and language modeling. Coincidentally, these were two of the ten areas we had originally highlighted in 2018. As a result, we updated our models to reflect these technological developments and produced new indices focusing on occupations most likely to be affected by these two technologies.

Interestingly, when examining the top twenty occupations most exposed to language modeling, we found that college or university professors, people like myself, were among them. In my view, this does not imply that such jobs will disappear. Instead, our indices identify the occupations that are likely to change the most due to these technologies.

Education, and particularly higher education, is one of the sectors where we can expect to see dramatic transformations driven by advances in AI.

When we look at image generation, the top twenty affected occupations fall into two broad categories. The first, unsurprisingly, includes jobs that rely heavily on images and their creation, such as artists and graphic designers. The second category are roles like chemical engineers, air traffic controllers, and mathematicians. These are professions that often require complex visualization in three-dimensional space.

It makes sense that such roles would be significantly impacted by improvements in image generation technologies. Again, the key point is not that these jobs will be replaced. Rather, these new tools will likely emerge to assist professionals in these fields, potentially enhancing their ability to make discoveries or advance innovation in areas such as mathematics and chemical engineering.

Aiden Singh: You have briefly referred to the different categories of jobs, including those that may be lost and those that are likely to change. In your view, do AI and robotics primarily augment workers’ capabilities, or do they serve as substitutes for human labor?

Robert Seamans: That is a great question, and one that many people are eager to understand. The reality is that AI and robotics can do both. It is overly simplistic to say they solely replace workers or solely augment them. In this case, it is important to consider the nuance.

In certain industries and occupations, these tools can be used to replace human labor. This is not universally the case, but there are some areas where substitution is possible. At the same time, these technologies can enhance human work. For instance, robots in manufacturing often supplement rather than fully replace the tasks performed by humans.

What excites me most about these technologies is their potential to augment human work. Focusing solely on substitution emphasizes cost reduction, performing the same tasks slightly more cheaply, which is unlikely to generate the significant economic growth these technologies are capable of enabling.

The real transformative potential lies in using AI and robotics to create new products and innovations that can dramatically reshape the economy. To achieve this, you want to have these technologies working alongside humans, enabling us to develop new solutions and possibilities. It is my hope that we direct our efforts toward augmentation rather than substitution to fully realize the benefits of these tools.

Aiden Singh: You mentioned a list of the top twenty industries most exposed to language modeling. What are some of the other occupations you believe will face the greatest challenges from AI adoption, and what policies, if any, do you think would be most appropriate to address those challenges?

Robert Seamans:  When I think about it, there are three industries that really come to mind when considering the major changes driven by AI. Education is one, healthcare is another, and the third would be finance and insurance. Some people also like to include real estate in that category.

I think we are going to see very different things play out across those three broad categories. 

What I am hopeful for with education is that AI becomes a tool to genuinely enhance the way we teach and engage students. I imagine a future where learning is highly personalised and where AI enhances the way each student engages with material.

With healthcare, I hope for something similar, with greater personalization in the care people receive. On the one hand, that might mean more individualised treatment plans, and on the other, if including things like drug discovery under healthcare, there exists potential for tremendous breakthroughs. That is the big promise a lot of people talk about: the idea that AI could help us cure cancer or Alzheimer’s, for instance.

I will be honest, I have no idea how realistic that is, and it does not seem like major progress has occurred there yet, but hope remains. One of the main roadblocks in healthcare is regulation. In many cases, those regulations exist for good reason, since caution is necessary with patients who are sick or injured and care must be taken not to inadvertently make things worse.

Simultaneously, however, concern exists that those same regulations probably slow down a lot of innovation in the healthcare space. They likely slow the adoption of new technologies in this field.

Finance, insurance, and real estate also deserve attention. Earlier questions referenced substitution versus augmentation, and quite a bit of substitution is likely in that sector. Some early evidence already exists that this is happening. For entry-level positions, for instance, firms may rely on fewer analysts than in the past, and that could become a broader trend in the industry.

You asked what policies I would advocate, if any, to address the challenges posed by AI. I lean toward the “if any” side. We must remain cautious to avoid overregulation and the slowdown of innovation in this space. If I had to choose between regulating or not, I would choose not to regulate.

Aiden Singh: We have been talking so far about AI at the industry level, looking at how it affects different sectors. On a broader level, do you have any thoughts on the long-term macroeconomic effects that AI might have?

Robert Seamans: My general take is that, as I mentioned before, AI is a general-purpose technology. As such, I believe it will boost economic growth. I am not enough of a macroeconomist to have my own specific figures, but a reasonable benchmark might be around a 10% increase in economic growth. I would consider 10% the floor.

This estimation aligns with what we have seen from previous general purpose technologies. Some organisations have published pieces suggesting that AI could double economic growth, essentially a 100% increase, but I think that is wildly optimistic. If that happens, great, but I certainly would not count on it.

The second point is that we should not expect this growth to happen immediately. From the history of general purpose technologies, we know that while they ultimately drive growth, it takes time before those effects appear. At the firm level, companies need to make significant adjustments to fully take advantage of new technologies, and those changes can take three to five years. When you aggregate that across the entire economy, it makes sense that it could take five to ten years before we begin to see the broader growth we are hoping for.

Aiden Singh: Thinkers like Steven Pinker or Carl Benedikt Frey have these grand, overarching theories about how things work. They think about human progress and how technology shapes it. Since you work at the intersection of economics, technology, and policy, have you developed a kind of holistic, big-picture view of how these three areas interact?

Robert Seamans: Great question. The short answer is no. I wish I had some sort of grand unifying theory, but I do not. Part of the reason is that the three categories you mentioned — economics, technology, and policy — tend to operate at fairly different speeds, which makes it hard to connect them in a single framework.

For example, when I talk to technologists, especially those in Silicon Valley or at some of the AI labs, there is a strong focus on how fast things are changing. The sense is always that big transformations are just around the corner. I remember back in 2019 when Elon Musk predicted that by the following year we would have fleets of driverless Teslas on the roads. Of course, that has not happened.

Technologists often think at this very rapid pace, partly because they see how quickly technology advances in controlled lab settings. Though, once you move those technologies into the real world, you encounter all kinds of frictions and challenges. 

That is where the perspective of economists starts to come in. I think economists, or at least some economists, have done a good job of thinking about the frictions that exist, the factors that, for better or worse, cause fast-moving developments to slow down a bit. 

Then you get to policymakers, and it is definitely true that things move quite slowly once you are in the policy realm. In some cases, that is actually a good thing. You would not necessarily want policy-makers to respond too quickly to every change happening in the world. There are situations where that might be useful, but as a general principle, you do not want policy constantly flipping back and forth rapidly. If it did, it would become difficult for businesses or households to make sound economic decisions.

So even though slow policy processes might not sound ideal, there are some real benefits to that pace.

So, getting back to your question, do I have a unifying theory? The short answer is no. I am mostly struck by how different the speeds are across these three domains, and I have not yet been able to bring them all together into one coherent framework.

- - - - - - -

Robotics

Aiden Singh: How has the commercialization of robots developed in recent years, and how does the United States compare to other countries in integrating them into its production processes?

Robert Seamans: I first started focusing on robots in 2015. What was interesting is that when you look worldwide at robot adoption, roughly from 2000 to 2010, there were about 100,000 robotic arms shipped every year. There was a little bit of a dip in 2009.

Then starting in 2010, there was a dramatic increase worldwide. By 2020, somewhere between 350,000 and 400,000 robots were being shipped annually. By this past year, 2024, the most recent data I have seen, it is roughly 550,000. So we have gone from a fairly stable 100,000 to this dramatic increase of over 500,000. It appears to be stabilizing around that level. Here, I am focusing primarily on robots in manufacturing settings. I am not even addressing developments in the service sector.

Those are the worldwide trends. The United States has more or less followed them, but the most dramatic increase has been in China. Of the approximately 550,000 robots in 2024, about half were sold to China. It is a dramatic increase in the number of robots being used in production settings there.

Looking at the United States and other developed or OECD countries, robot adoption and use is more or less the same. For example, looking at Germany compared to the United States, I might get the exact numbers slightly wrong, but across the economy in the United States, about one to two percent of firms have adopted robots. It is roughly the same in Germany.

When looking at the sector with the most adoption, it is the motor vehicle sector. A lot of that adoption, by the way, is upstream from the manufacturers. In other words, these are the parts producers, the companies that make doors, brakes, windshields, and similar components. As of two years ago in the United States, roughly 17 to 18 percent of firms in that sector had adopted robots, and it is about the same in Germany. So we are seeing similar adoption patterns. 

One open question concerns usage, which relates to your earlier question on substitution versus augmentation. It is very hard to tell from the type of data economists have access to whether robots are being used to augment or substitute for human labor. This is one area I would like to explore further. There is some evidence that Germany may be pursuing more AI-driven augmentation, while the United States may show a slightly stronger trend toward using robots to substitute for human work. I do not think it is a dramatic difference, but there is a small emerging pattern.

The literature on robot adoption, and its link to both employment and firm-level performance, is interesting. There are about half a dozen papers in this area so far, and they all show that firms that adopt robots perform better and increase employment relative to firms that do not adopt robots. A big open question is why more firms are not adopting robots. 

For example, in the United States, even in the sector that is the heaviest adopter, only about 20% of firms have adopted robots. Yet these are the firms that are increasing performance, whether measured by productivity or revenue, and also increasing employment. It appears that these gains are coming at the expense of the non-adopters. Understanding why adoption is not more widespread is one of the big open questions in this field.

Aiden Singh: In 2023, you co-authored a study called “Robot Hubs: The Skewed Distribution of Robots in U.S. Manufacturing”. You found that robotics adoption in the United States was concentrated geographically and was connected to union membership. Can you unpack those findings and discuss their implications?

Robert Seamans: Yes, I will unpack that a little bit. I think union membership is less important in this case. Really, there are two other things in that paper that I think are a big deal. The first is that the findings are conditional on industry, firm size, and similar factors. Robot adoption is not uniform. I will repeat that: robot adoption is not uniform. 

That might not sound like a big deal, because one might expect more robot adoption in the suburbs of Cleveland, where there is a lot of motor vehicle manufacturing, compared to the suburbs of New York, where there is not. That is not what the paper is saying. 

Compare two cities that look very similar. For example, Cleveland A and Cleveland B, with the same mix of employment and industries. One might expect robot adoption to be very similar across the two cities, given the similar industry mix. What the paper finds is that there are dramatic differences. In Cleveland A, there is a lot of robot adoption, while in Cleveland B, which looks very similar, there is hardly any robot adoption. 

This is interesting for a couple of reasons. One is that the prevailing assumption, prior to our paper, was that robot adoption was more or less uniform across areas, conditional on industry mix and similar factors. In fact, a number of influential academic papers had been based on that assumption. We are, if you will, disproving that assumption. 

The second question is why do we see these dramatic differences across areas that look very similar? You mentioned union membership, and that is one factor. In areas with higher union membership, it appears there is more robot adoption.

What is more interesting, from our perspective, is the presence of what are called robot integrators. Areas with these integrators see much higher adoption. We cannot fully determine how endogenous or exogenous the location of these integrators are, but they are clearly very important.

So what are robot integrators? These are firms that can be thought of as engineering consulting firms. They help manufacturers determine what types of robots and associated equipment are needed on their production lines. I have visited a few of these integrators in places like rural Michigan and Ohio. 

Imagine a large warehouse-like space. They mock up a production line for a manufacturer, figure out how to integrate the robot into that line, and determine how to rearrange the production line to maximize what the robot can do well while minimizing what it cannot do efficiently. They also identify any additional equipment that may be needed to fully integrate the robot into the production process. Once they have everything running well, they break it all down and ship it or deliver it to the manufacturer. Then they reset it up there. 

These integrators play a really, really important role in bridging the knowledge gap between the manufacturer, who has done a great job of optimising their production line over the past couple of decades, and the knowledge of, if you will, the frontier of where robotics are, including what robotics can do well, what it cannot do well, and the other equipment that needs to be invested in.

This very important intermediate firm turns out to play a really big role in our story. The robot hubs, like Cleveland A, that adopt many robots, just happen to be the places where these robotic integrators are located. 

I think this point about the robotic integrators is a crucial one. It is also useful for thinking a little more deeply about our conversation around general-purpose technologies and integrating new technologies into the economy. 

It is not the case that you just sprinkle general purpose technology onto the economy and magically get growth. You actually have to work for it. You have to integrate the technology, whether it is robots, AI, computers and the internet, or electricity, into economic processes in creative, clever ways. Only then do you see the benefits of the technology.

That is the part of the Robot Hub story that I get really excited about. I am glad that you brought it up.

Aiden Singh: Interesting. One of the questions that’s often come to my mind when thinking about how factories work is that there are always so many moving parts and they’re so complex. I’ve always wondered how they even discover that a part exists and that it could be used in a manufacturing process.

Robert Seamans: That is a great point. A lot of the parts you are talking about are bespoke. They were created specifically for the product being manufactured on that production line. It is not that the manager of the factory could go down to a local store and buy the part.

These parts were created specifically for that manufacturing process. Earlier, I pointed out that in the most robot-intensive sectors of the economy, only about 20 percent of firms have adopted robots, which is relatively low. Part of the reason for this, and at least part of the answer to why adoption is not higher, is that it is not easy to adopt and use these new technologies.

Firms have to make a variety of additional investments, which I call complementary investments, in order to take full advantage of the new technology. It is not just that these investments are required; it is that firms do not necessarily know what they are ex ante. This involves a lot of experimentation, many dead ends, and a fair amount of frustration while figuring everything out.

Aiden Singh: You mentioned rural Michigan and Cleveland. Do you think automation is capable of reviving old manufacturing centers such those in the Rust Belt?

Robert Seamans: I think it is a good question. It is something that a lot of people hope for, and I think it is possible. However, I am not sure that there is a general answer to this.

Historically, these clusters of manufacturing expertise developed in different areas. For example, parts of North Carolina were known for furniture, fabric, and similar industries. I do not think it is the case that any type of manufacturing could simply be inserted into that setting. The human capital built up in those areas and that still exists today has expertise in those specific clusters of manufacturing knowledge.

I would think that what you would want is to introduce manufacturing that can take advantage of some of that local knowledge. To the extent that local knowledge has dispersed over time, you definitely need some specialized local source of human capital. Without that, it will be difficult to simply hope for new technologies to revitalize an area.

- - - - - - -

Council of Economic Advisors

Aiden Singh: In addition to your time in academia, you’ve also served on the White House Council of Economic Advisers for President Obama, focusing on technology, innovation, and competition policy. What were the major tech policy issues the administration was facing at the time?

Robert Seamans: The technology policy issues we focused on will probably not come as a surprise. One area was robots, which is part of why I became interested in them. Every year, the Council of Economic Advisers publishes the Economic Report of the President (ERP). In our report, we had a chapter focused on productivity growth, and the role of innovation and new technologies in driving it. Robots were a central focus of that chapter, and we also touched on AI.

More broadly, the administration focused on a range of technologies, including broadband internet, patents, robots, AI, autonomous vehicles, and drones, among others. It was a suite of technology-related issues.

Aiden Singh: Can you provide some insight into how the administration approached these policy issues at the time?

Robert Seamans: One useful thing to note is the role of the Council of Economic Advisers, at least traditionally, and the way it operated when President Obama was at the White House. It really is an advisory unit. The role of the Council, and of an advisor like myself, is to weigh in on the economic trade-offs of any particular policy. It is never the case that a policy is just universally good with no trade-offs. Every policy has a variety of trade-offs.

Most of my time, for any one of the policies or technologies we were considering, was spent thinking about all the various trade-offs associated with it. Historically, and presumably still today, the Council’s role has been to lay out those trade-offs and present them to the politicians in charge, who then make decisions while taking into account various other factors.

Aiden Singh: Considering recent developments in AI, if you were on the Council of Economic Advisers today, what are some of the things you might be advising the current administration?

Robert Seamans: I think there are two buckets of things that I would think a lot about. 

Bucket number one is regulation. I said earlier that, if I were forced to choose, I would focus more on making sure I was not overregulating. An important consideration is that even though the United States might choose not to regulate heavily in this area, other countries, particularly in the European Union, have a lot of regulation. I would hope that American companies have strong incentives and a desire to commercialize their products not just for the United States, but also for Europe and other places around the world. So it is the case that American companies, even if not heavily regulated at home, are thinking a lot about regulation in Europe and elsewhere.

While I would not want to overregulate, I would want to think about the extent to which we allow other countries to lead on this and how our homegrown American companies have to respond. Furthermore, under certain circumstances, we must think if it makes sense for the United States to get out ahead on regulation to set standards that are friendly to American companies. 

The second bucket is labor. One question on everybody’s mind is what the effect of this technology will be on labor. For sure, there will be some substitution and there will be some augmentation. I would hope there is more augmentation than substitution, but both will certainly occur. 

What I would advise this administration to focus on is tracking the effects of AI on labor as much as possible. That can start with things as simple as the Census Bureau tracking, at the firm level, which firms are adopting or planning to adopt AI, which it is already doing. I would push that further, since firms often have many different plants or establishments. I would like to know, at the establishment level, where AI is being adopted and where it is not, and how it is being used. 

From there, we could track the effects on labor. I would urge the administration to spend more resources on this area to monitor what is happening. It is only by tracking these developments that we can get a sense of what types of policies may or may not be warranted.

- - - - - - -

Universal Basic Income

Aiden Singh: We’ve discussed concerns held by some observers of widespread unemployment as a result of AI. One policy idea that is sometimes floated as a potential solution is universal basic income (UBI). You have argued that while it may have some benefits, it is probably not a great policy for stimulating innovation and entrepreneurship. Why do you feel that way and what other policies might be better?

Robert Seamans: I think there are three parts to your question. I am not a fan of the idea of a universal basic income. The main reason is that I do not think it is realistic.

The first point is that if what we are talking about is something like $1,000 a month to every adult, regardless of circumstance, the cost would be around $3 trillion a year. When we look at major legislation that has recently passed, such as the Trump administration’s Big Beautiful Bill or the Biden administration’s Inflation Reduction Act, those were each roughly $1 trillion spread over a ten-year period. 

By comparison, a universal basic income would cost around $3 trillion every single year, perhaps a little more or a little less. So we are talking about something an order of magnitude greater than any of the policies currently on the table. Given how divisive current policies and proposals already are, I think a universal basic income is politically a complete nonstarter.

The second point is that some people argue a universal basic income could help stimulate entrepreneurship and innovation. However, I think the burden of proof lies with those making that claim, and I have yet to see any convincing evidence.

For instance, consider Norway, which distributes payments to citizens through its oil fund, or the state of Alaska, which pays residents through the Permanent Fund (PIF). One does not really think of those as places known for high levels of entrepreneurship or innovation. Maybe there is a small effect at the margins, but there is certainly no clear indication that these kinds of payments spur broader innovation. If that were the case, I would like to see research demonstrating it, but I have not. I have not seen any evidence that a universal basic income would meaningfully increase entrepreneurship or innovation.

The third point in your question is what the alternatives are. I think a good place to start is with the policies we already have and consider how to improve them. For example, if one argument for a universal basic income is that it could help people who lose their jobs and health care, then maybe expanding existing health care programmes would be a better approach. Or, if the goal is to encourage people to enter or stay in the workforce, we might look at strengthening the Earned Income Tax Credit or similar incentives.

More broadly, we already have a range of established policies and policy experiments that provide valuable lessons. The question, then, is how we can make targeted adjustments to these systems. It might not sound as bold or exciting as proposing a universal basic income, but practical improvements to existing frameworks are often where real progress happens.

- - - - - - -

Education Policy

Aiden Singh: You mentioned higher education earlier and how AI might change it. As a professor at NYU Stern, do you think our approach to higher education needs updating for the digital age? Is there still value in a humanities education or should universities focus more on preparing students for technology-driven careers?

Robert Seamans: As we discussed earlier, education is a sector of the economy that is going to be dramatically transformed by AI, and I believe for the better. It will enable much more personalization. It will also require changes not only in how we teach but also in how we evaluate students.

On the evaluation side, one challenge that comes up often is that we cannot teach or assess students the same way we did before. For example, we cannot just assign essays, knowing students might rely on a large language model to help write them. If the goal is to evaluate something about writing, then we need to change how we do that. 

Perhaps more importantly, we need to rethink what we are trying to evaluate. It is not necessarily a student’s ability to write an essay that matters; that is a marker of something else. What we are really trying to assess is something deeper, such as critical thinking. So, are there other ways to measure a student’s progress in critical thinking? There must be, although I am not entirely sure what all of those methods look like yet.

I am experimenting in my own classes with that. I think that is the direction we will go in for evaluation. New ways will emerge to assess what students do. Within the classroom, there will be a variety of approaches in how we teach our students.

On the subjects themselves, you asked about humanities or a trade school–style approach. I think there will continue to be a variety of educational institutions. There should be for several reasons. One reason is that different types of students are very different learners at different stages of their careers. We need to be adaptable to meet students where they are.

Some things can be learned in one week, four weeks, or two months. Other things require multiple years of learning. AI may shorten that process slightly, but not by much. We will still need physicists, neuroscientists, and other specialists at the edge of the scientific frontier. AI may help people reach that frontier more efficiently, but it will not reduce years of deep training to a matter of weeks. I think we will need a variety of educational institutions to meet the needs of students and the broader demands of society. 

Will humanities play a role in that? As an English literature major in college with a strong appreciation for the humanities, I certainly hope so. I think we will see it. From what I have read, those types of degrees seem less appreciated in the marketplace than in the past. That does not mean there is zero appreciation for them. At the end of the day, skills such as critical thinking and independent thinking are what we want students to develop. Those skills can be gained from a degree in the humanities just as they can be gained from a math or science degree.

- - - - - - -

Academia and Policy-Making

Aiden Singh: You have been both an academic economist and an economic advisor. How directly do the academic theories developed in universities translate into policy formation? Is there a gap between the academic world and the policy world?

Robert Seamans: This is a good question. There definitely is a gap, not just between research and policy, but also between research and practice. The work done in an academic setting does not always align with what matters for policy in practice. Perhaps there always has been a gap, and it is unclear whether it is increasing or decreasing.

There definitely is a gap, and I would like to see it smaller. The two worlds do not need to be completely aligned. In academia, the time frame or level of abstraction sometimes differs from that of someone running a company or, returning to manufacturing and robots, a plant manager. It makes sense that there is some gap, and shrinking it to zero is not necessary.

The gap probably is wider than it should be. One hope, particularly for economics and business school professors, is more emphasis on encouraging faculty to move between academia, policy, and practice. Faculty should spend time not just in academia but also applying their work in real-world settings and then returning to research.

We see that a lot in energy, at least historically. It is also common in fields like engineering and computer science. A faculty member might have an idea, perhaps sparked by a student, and leave academia to start a company. After several years, they might return to academia. 

This type of back-and-forth is more common and appreciated in engineering and computer science. Economics and business schools show some of this, but it is less frequent. There is room to encourage more of this movement.

——————

Editing by Harpreet Chohan.