What Are Tariffs and How Do They Impact Us? Another Conversation with Andrew Lo

Summary

Harvard Data Science Review Podcast Summary
Episode: What Are Tariffs and How Do They Impact Us? Another Conversation with Andrew Lo
Release Date: May 30, 2025

---

Introduction

In this insightful episode of the Harvard Data Science Review Podcast, hosts Liberty Vidert and Shali Meng engage in a comprehensive discussion with Professor Andrew Lo, the Charles and Susan Harris Professor of Finance at MIT and Director of the Laboratory for Financial Engineering. The conversation delves deep into the multifaceted world of tariffs, exploring their economic implications, historical contexts, and the role of data science in shaping effective tariff policies.

Understanding Tariffs

Defining Tariffs and Their Purpose

The episode begins with a foundational understanding of tariffs. Shali Meng opens the discussion by posing a fundamental question about tariffs:

[00:56] Shali Meng: "So let's just dive into the right of way, as you know. Well, tariffs are one of those economic tools that spark huge debates. So we wanted to start off giving listeners a broad understanding of their use. What is a tariff and how does it meant to work?"

Andrew Lo responds by defining tariffs and their intended functions:

[01:13] Andrew Lo: "Basically tariffs are taxes on imported goods and they're designed to protect domestic industries from foreign competition. They also raise government revenues and they are also meant to try to respond to trade imbalances."

He further explains the economic rationale behind tariffs, referencing Adam Smith's concept of the "invisible hand":

[01:44] Andrew Lo: "You're using Adam Smith's invisible hand of changing prices artificially by imposing these taxes so that consumers will, instead of buying foreign goods, they'll buy more domestic goods."

Economic Impact of Tariffs

Who Bears the Cost of Tariffs?

Shali Meng raises a critical question about the true burden of tariffs:

[01:53] Shali Meng: "But one question for lots of people, myself included, is who actually pays this tariff in the end?"

Andrew Lo elucidates the distribution of tariff costs over time:

[01:53] Andrew Lo: "Initially, the tariffs are going to be paid by the consumers that want to buy more expensive goods. And eventually what you're going to see is society, the country itself ends up paying for those tariffs... both countries end up paying higher costs for the kinds of goods that they were previously trading with each other."

Tariffs as a Double-Edged Sword

The discussion highlights the paradox of tariffs turning potential win-win scenarios into lose-lose situations due to retaliatory measures and increased costs for consumers.

Historical Context and Economic Theory

Comparative Advantage and Global Trade

Andrew Lo provides historical insights into international trade theories, referencing David Ricardo's Principles of Political Economy and Taxation:

[03:06] Andrew Lo: "David Ricardo... came up with the idea of what is now called by economists the theory of comparative advantage."

He explains how comparative advantage fosters globalization and mutual benefits through specialization:

[05:06] Andrew Lo: "Even if the first country is better at both than the second country, it's actually still sensible for... the first country to specialize in cloth and for the second country to specialize in wine and then for them to trade."

Political Realities and Tariffs

The conversation transitions to the intersection of economics and politics, emphasizing how political pressures from affected industries often drive the imposition of tariffs:

[07:05] Andrew Lo: "Tariffs come from the population of winemakers or cloth manufacturers that don't want to compete with other countries."

Common Misconceptions About Tariffs

Shali Meng probes into prevalent misunderstandings regarding tariffs:

[07:14] Shali Meng: "What would you say are the biggest misconception people have about economic impact of the tariffs?"

Andrew Lo identifies several key misconceptions:

1. Protection of Specific Industries:

   [07:14] Andrew Lo: "Tariffs will actually help the particular industry that's being protected... eventually... the industry you're trying to protect will actually be hurt by the tariff."

2. Job Preservation:

   [07:14] Andrew Lo: "Tariffs always protect jobs... in downstream industries, those jobs may actually be lost."

3. The Rarity of Tariffs:

   [07:14] Andrew Lo: "Tariffs are used from time to time in the current world because people are concerned about the political aspects of international trade."

Data Science and Real-Time Economic Analysis

Leveraging Data for Dynamic Tariff Policies

The conversation shifts to the role of data science in shaping tariff policies. Shali Meng questions the feasibility of dynamically adjusting tariffs based on real-time economic data:

[09:12] Shali Meng: "Do we have enough real time economic data? Can we actually learn from what's going on and adjust them dynamically?"

Andrew Lo affirms the potential, citing the Billion Price Project as an example:

[09:16] Andrew Lo: "...Roberto Rigabone and his former student Alberto Cavallo... measured inflation on a minute to minute basis by scraping online prices."

However, he underscores the current limitations in economic models and funding to implement such dynamic policies effectively:

[09:16] Andrew Lo: "We don't yet have the economic models and the funding to be able to do that in real time."

Economic Models and Their Limitations

Traditional vs. Modern Economic Models

Andrew Lo delves into the tools economists use to analyze tariffs:

[11:41] Andrew Lo: "The traditional tools of economists are, not surprisingly, supply and demand... computable general equilibrium models... input-output analysis."

He critiques these models for their inability to incorporate risk and uncertainty, particularly relevant in global trade dynamics.

Emergence of Advanced Models

Lo discusses newer models that attempt to integrate risk and uncertainty:

[11:41] Andrew Lo: "Stochastic general equilibrium models as well as agent-based analysis... become computationally intractable."

He anticipates a future advancement in economic modeling that can better handle the complexities introduced by tariffs and global trade uncertainties.

Behavioral Economics and Market Dynamics

Adaptive Markets Over Rational Models

Shali Meng connects traditional economic theories with behavioral insights:

[15:13] Shali Meng: "Am I correct that these traditional models are probably a lot more based on everybody being very rational?"

Andrew Lo agrees, emphasizing the limitations of rationality in current market analyses:

[15:13] Andrew Lo: "Markets are not rational... they're adaptive... emotionally and politically driven."

Agent-Based Modeling

Lo introduces agent-based models as a means to simulate complex economic interactions:

[17:45] Andrew Lo: "You begin by assuming that agents have certain kinds of behaviors... let them interact and you see what happens to the economy."

He illustrates how these models can predict outcomes like trade wars, ultimately leading to mutual economic losses and political repercussions.

Historical Lessons: The Smoot-Hawley Tariff

Case Study of Economic Impact

Reflecting on history, Lo references the Smoot-Hawley Tariff of 1930:

[19:54] Andrew Lo: "They imposed widespread tariffs on all foreign goods. And what happened was that that drove the economy into a much, much deeper depression."

He draws parallels to contemporary tariff policies, cautioning against similar economic downturns:

[19:54] Andrew Lo: "If we were to play that out, we would see something along those lines."

Alternative Economic Tools

Beyond Tariffs: Quotas, Subsidies, and More

Shali Meng inquires about other economic instruments available to policymakers:

[21:37] Shali Meng: "Assuming politicians really learn from the data, but they still want to protect domestic industry, what are the alternative economic tools?"

Andrew Lo outlines various tools, differentiating between short-term and long-term strategies:

• Short-Term Tools:
  Quotas, subsidies, export controls, tariffs

  [22:29] Andrew Lo: "In the short run, it can look like some of those tools are working... but there are unintended consequences."

• Long-Term Tools:
  Investment tax credits, industrial incentives

  [22:29] Andrew Lo: "Such as South Korea and Taiwan... using a combination of these tools... to achieve long-term success."

Political Systems and Tariff Policies

Influence of Governance Structures

The discussion explores how different political systems impact tariff policies:

[24:51] Andrew Lo: "Democratic governments have finite terms... it's very difficult to engage in truly long-term policy... whereas in more centralized governments, leaders can afford to engage in strategies that may cost more in the short run but benefit long-term."

Lo emphasizes the need to consider the type of government when analyzing tariff strategies:

[24:51] Andrew Lo: "We have to be very aware of who we're competing against and the kind of governments that they have at their disposal."

Advice for Individual Investors

Navigating Investment Strategies Amid Tariff Uncertainties

As the conversation shifts towards personal finance, Shali Meng seeks investment advice for listeners:

[27:05] Shali Meng: "...you know, we have a retirement account... What other general advice do you have for our listeners?"

Andrew Lo provides general guidelines, emphasizing self-awareness and professional guidance:

[27:05] Andrew Lo: "Understand yourself... know your goals, your flexibility in bearing losses, and the time frame for your investments."

He advises diversification and risk management, particularly in volatile economic environments influenced by tariffs:

[30:27] Andrew Lo: "Diversify across various different assets and asset classes and be able to manage your risk... pay attention to what's going on in the world and manage your risk carefully and dynamically."

The Role of AI and Large Language Models in Financial Advice

Cautionary Insights on AI-Driven Financial Guidance

Shali Meng transitions to discussing the integration of AI tools like ChatGPT in financial planning:

[31:33] Shali Meng: "...you wrote an article about Can ChatGPT to help your financial... What your advice there for lots of listeners?"

Andrew Lo advises caution when using Large Language Models (LLMs) for financial advice:

[31:33] Andrew Lo: "Proceed with caution... LLMs are good at gathering information and summarizing... but you need to verify the information and consult certified financial advisors for implementation."

He envisions a future where LLMs could provide trusted financial advice, pending advancements and regulatory standards:

[34:47] Andrew Lo: "We can create LLMs that can provide trusted financial advice... we're within striking distance of that reality."

Recent Developments in AI

Exponential Growth and Future Implications

Reflecting on AI advancements since his last appearance, Andrew Lo highlights the rapid evolution and increasing computational capabilities:

[35:17] Andrew Lo: "The speed with which AI is evolving is accelerating... innovation happening at a much more rapid pace than we fully appreciate."

He warns of potential disruptions and emphasizes the importance of addressing AI's downsides proactively:

[35:17] Andrew Lo: "We have to pay attention to the negatives and prepare for it... bad actors using AI for nefarious goals."

Designing a Perfect Data-Driven Tariff Policy

Magic Wand Scenario

In the concluding segment, Shali Meng poses a hypothetical scenario to envision an ideal tariff policy:

[37:36] Shali Meng: "If you can wave the magical wand, design a perfect data driven tariff policy, what would be the key features and would it be."

Andrew Lo outlines his vision for a balanced and strategic tariff policy:

1. Identification of Critical Industries:

   [37:36] Andrew Lo: "Identify industries critical for our benefit and long-term survival."

2. Temporary Protection Through Tariffs:

   [37:36] Andrew Lo: "Impose tariffs carefully to provide temporary protection, not to be punitive."

3. Incentives and Industrial Policy:

   [37:36] Andrew Lo: "Couple tariffs with incentives to help industries reinvent themselves and become globally competitive."

He emphasizes the integration of data science to meticulously adjust these policies for optimal outcomes:

[37:36] Andrew Lo: "Using data science to collect and change these policies carefully to produce the best outcome."

Conclusion

The episode wraps up with a reflection on the complexities of tariff policies and the indispensable role of data science and adaptive economic models in navigating the global trade landscape. Professor Andrew Lo provides a nuanced perspective that bridges traditional economic theories with contemporary challenges, offering listeners a comprehensive understanding of how tariffs impact economies and individual financial strategies.

Notable Quotes:

• Andrew Lo on Tariffs' Long-Term Impact:

  "[07:05] "You get into this kind of a arms race where people are going to be ratcheting their tariffs... both countries end up paying higher costs."

• Andrew Lo on AI and Financial Advice:

  "[31:33] "Proceed with caution... never take large language model advice at face value."

• Andrew Lo on Designing Tariff Policies:

  "[37:36] "Impose tariffs carefully to provide temporary protection, coupled with incentives to make sure that protected industries reinvent themselves."

---

This episode serves as an essential guide for listeners seeking to understand the intricate dynamics of tariffs, their economic and political ramifications, and the future pathways enabled by data science and artificial intelligence.

Transcript

[Liberty Vidert]

Are tariffs going to make us rich, poor, or somewhere in between? Today we are going to find out on this Harvard Data Science Review podcast. I'm Liberty Vidert, feature editor of hdsr, along with my co host and our editor in chief, Shali Meng. We take a deep dive into the world of economic policy. And we are joined by one of our most wildly popular guests, Professor Andrew Lowe. Andrew is the Charles and Susan Harris professor of Finance at MIT and the Director of the Laboratory for Financial Engineering. So let's dive right in to learn the real truth of tariffs and how it will affect your pocket at home.

[Shali Meng]

Well, Andrew, thank you so much for coming on as a guest again.

[Andrew Lowe]

Thank you. Honored to be here.

[Shali Meng]

So let's just dive into the right of way, as you know. Well, tariffs are one of those economic tools that spark huge debates. So we wanted to start off giving listeners a broad understanding of their use. What is a tariff and how does it meant to work?

[Andrew Lowe]

Sure. Well, basically tariffs are taxes on imported goods and they're designed to protect domestic industries from foreign competition. They also raise government revenues and they are also meant to try to respond to trade imbalances. The basic idea is you're using Adam Smith's invisible hand of changing prices artificially by imposing these taxes so that consumers will, instead of buying foreign goods, they'll buy more domestic goods. That's the basic idea.

[Shali Meng]

But one question for lots of people, myself included, is who actually pays this tariff in the end?

[Andrew Lowe]

That's a great question. It depends on what time frame you're looking at. Initially, the tariffs are going to be paid by the consumers that want to buy more expensive goods. And eventually what you're going to see is society, the country itself ends up paying for those tariffs. Because undoubtedly when you launch a tariff, other countries are going to retaliate and they're going to impose tariffs on your goods that are being sold to them. So you're going to get into this kind of a arms race where people are going to be ratcheting their tariffs. And as happened in many times in history, you get a trade war where ultimately both countries end up paying higher costs for the kinds of goods that they were previously trading with each other. Now they're going to have to be manufacturing them domestically.

[Shali Meng]

In that sense, probably in the end, it's instead of a win win situation that become like a lose lose situation. Is it because everybody will be paying higher price? So how do they actually work? Do they really achieve the original goal and what do historical data tell us?

[Andrew Lowe]

Yeah, well, in order to Understand that it's maybe helpful to try to go back in time and ask the question, what exactly is the whole point of. Of international trade? I mean, why do we engage in globalization in the first place? And we gotta go back to 1817 and focus on the work that the British economist David Ricardo came up with. The book that he wrote was on the Principles of Political Economy and Taxation. And in this amazing book, he came up with the idea of what is now called by economists the theory of. Of comparative advantage. And let me give you a very short summary, but I'm not going to do it justice because it is subtle and a bit more complicated than meets the eye. But it's a more sophisticated version of Adam Smith's pin factory. If you recall, Adam Smith pointed out in his wealth of nations that manufacturing really gets much more efficient if you engage in specialization. And he gave the example of making a pin. So what does a pin have? It's got a shaft, it's got a head, you have to sharpen the point, and so on and so forth. And he showed that instead of having one person making a pin from beginning to end, if you have an assembly line and you have one person making the head, another person making the shaft, a third person sharpening the shaft, a fourth person assembling all of this, that having this specialization allows you to make a lot more pins per hour than the traditional approach of having a pin maker do it all from scratch. So this idea actually applies to countries. And here's an interesting example. Imagine having two countries, one that's really good at making cloth and just okay at making wine, your favorite activity.

[Shali Meng]

Yes.

[Andrew Lowe]

And the second country is okay at making cloth, but happens to be better at making wine. So the first country is more efficient at cloth, the second country is more efficient at wine. But the thing is that even if the first country is better at both than the second country, it's actually still sensible for. For the first country to specialize in cloth and for the second country to specialize in wine and then for them to trade. You can show mathematically that if you allow for specialization across countries, that both countries can win, because the total product of wine and cloth will be greater if they specialize and trade. You know, a more detailed example that all of us are used to is specialization at work. You know, a surgeon who happens to be a fantastic typist will still want to hire a secretary to help her transcribe handwritten notes about her patients into the patient's electronic medical records, even if the surgeon is a better typist than the secretary because the surgeon's time can be better spent doing surgeries. This applies to countries and that's the reason why we engage in trade. Okay, now that's just the economic analysis. If you layer on top of that the political realities. Some countries have a very active wine industry that they don't want to shut down in favor of those countries that have that comparative advantage. Now you're getting into political issues, and historically that's where tariffs come from. They come from the population of winemakers or cloth manufacturers that don't want to compete with other countries because they can't compete and yet they still want to make a living. So the political reality is in those cases, countries impose tariffs.

[Shali Meng]

So when we look at this modern trading environment, what would you say are the biggest misconception people have about economic impact of the tariffs?

[Andrew Lowe]

Well, there are several misconceptions. I guess the first misconception is that tariffs will actually help the particular industry that's being protected. It turns out that in the short run, that might be the case because you're making foreign goods that are competitive more expensive, so consumers will end up spending more money on the domestic goods. The problem is that things don't stay that way for very long because as I said, other countries will be imposing tariffs of their own. And so in the end, it may end up being that the industry you're trying to protect will actually be hurt by the tariff. Second misconception is that tariffs always protect jobs. It turns out that in some cases jobs are saved, but in others, particularly in downstream industries, the those jobs may actually be lost. So it's really hard to figure out the entire supply chain effects of tariffs. And finally, the last misconception is that, oh, tariffs, nobody does them. In fact, they are used from time to time in the current world because people are concerned about the political aspects of international trade and production. So there are very complicated motivations for using tariffs, including geopolitical considerations that we have to factor into these analyses.

[Shali Meng]

Right. Well, speaking of which. Right. We have seen as ordinary consumers, we have seen these tariffs like, you know, we heard the 10%, 145%. So it seems like changing very rapidly. And I assume these are obviously mostly driven by political considerations. But do we have enough real time economic data? Can we actually learn from what's going on and adjust them dynamically? Is that possible?

[Andrew Lowe]

Is it possible? Absolutely. Is it being done? Unlikely.

[Shali Meng]

I see.

[Andrew Lowe]

So let me explain the first part. Is it possible given where data science is today, and you know better than I do. Given all the contributions that you made to the literature, given where data science is today, given where data availability is today, it is absolutely possible. One good example is the research of my colleagues, Roberto Rigabone and his former student Alberto Cavallo. They wrote a paper a number of years ago about the Billion Price Project. It was an effort that they launched to try to measure inflation. But rather than measuring inflation on a quarterly basis, they did something really clever. They looked on the Internet and they searched for the prices of things like eggs, milk and butter. They searched online because all of these supermarkets, they post their products online and they post the prices. And so they used a search engine to be able to calculate prices not only on a daily basis, but on a minute to minute basis. They were able to calculate the rate of inflation of all of these commodities. And, and they claimed that they were basically sampling a billion prices. It was literally an amazing effort that transformed the way we think about measuring things like inflation and productivity. That's the age that we live in. Today we are able to measure extraordinarily detailed changes in various parts of our economy. The challenge of course, is being able to manage that process and processing it in ways that are gonna be useful for decision makers like policymakers and legislators. We obviously can do that. We can actually adjust tariffs dynamically as a function of the global economy if we were able to collect that data and process it. But unfortunately we don't yet have the economic models and the funding to be able to do that in real time so that decision makers can actually make use of the wonderful data that we have.

[Shali Meng]

Well, that sounds really very encouraging. And can you speak a little bit to what are the typical economical models are being used? Because we talk about data, and many of the data, they don't necessarily come with some kind of theoretical models, and we use machine learning others as a pattern seeking device. But I'm sure there are great economics models there, and if you can speak.

[Andrew Lowe]

A little bit to that, I'd be happy to. It's a bit of a mixed bag unfortunately, because up until recently, tariffs were not a really big aspect of economic analysis, since we didn't have too many of them. But let me first talk about the traditional tools. The traditional tools of economists are, not surprisingly, supply and demand. We know how to estimate supply curves, we know how to estimate demand curves. And it's the intersection of supply and demand that yields the equilibrium price and quantity in any given market. So centuries ago, the idea of generating supply and demand curves across all economies and calculating the general equilibrium of these various different economies was proposed. And there's been a lot of research done on implementing these ideas, so called computable general equilibrium models. So that's something that's been done and a close relative of it. An input output analysis that basically traces the entire supply chain of the economy. Those are tools that have been around for a very long time. Now. The problem with those models is they don't incorporate risk and uncertainty into their analysis. And when it comes to global trade, risk and uncertainty are really the key factors in thinking about how to impose tariffs, when to take them off, what their consequences are, who benefits and who loses. In fact, right now we're in a situation where because the tariffs have been proposed, but there's very little visibility into exactly what they are. The, the amount of risk and uncertainty in the economy is very high and that's really been a damper on innovation and investment. So hopefully we'll be able to get more clarity once the 90 days are up. But that's the traditional models. The new generation of models incorporate risk and uncertainty in a more sophisticated way. So called stochastic general equilibrium models as well as agent based analysis. The issue with those models is that they very quickly become computationally intractable. And while certain simplifications can be solved numerically, if you really want to simulate the entire economy and think about the implications of tariffs across all of the goods and services that we provide, that's really beyond our abilities right now. Now, I think this over this last few weeks, a number of economists have become interested again in how to deal with these kinds of issues. So I'm hoping that the next generation of economic models will be able to manage these kinds of problems. But right now we don't have very good tools to be able to tell us in any kind of certain terms what the impact of these tariffs will be on consumer benefits and costs over the course of the next few years, Even weeks, never mind months or years.

[Shali Meng]

Right. Well, this certainly reminds me. You are great work. Right? You are the one talking about all these market instead of just based on the traditional rationale assumption. Am I correct that these traditional models are probably a lot more based on everybody being very rational? Behaviors can be estimated. Well, and in this environment, probably many people would agree rational probably is not the right way of thinking about. And so how does your work, your theory, your framework can help in this truly volatile time in many sense?

[Andrew Lowe]

Well, exactly right. I think that rationality is going to give you limited success in trying to understand market dynamics right now. So my own Perspective is that markets are not rational. They're not always irrational either. Rather, they're adaptive, meaning that various different participants in the market economy are going to change their behavior in response to market conditions changing. And, you know, that's certainly rational in certain circumstances, but more often than not, they are behaviorally driven, meaning that we react not just with our rational, logical deliberation, but also emotionally and politically. So this is where I mentioned before that tariffs are a political instrument. For many years, the United States used its wealth and power very strategically and purposely decided not to impose tariffs and to suffer trade imbalances to benefit other countries because they were getting something in return. They were getting political influence to be able to maintain a certain world order. And so the tariffs that we imposed in some cases now are basically going to make it difficult for us to have that same level of political influence. So this is where the behavior comes in. Instead of looking at this as a purely economic problem, because it's not a purely economic problem, we have to understand the politics, the behavioral elements, the dynamics of the various different constituencies that are going to be affected. So by looking at the entire economy, the world economy as a global ecosystem, where you've got predator and prey and various kinds of keystone species that are all vying for survival, you'll get a better handle of how this is likely to play out.

[Shali Meng]

Right. And I was just thinking about how do you model things like what the current the government does, right? You know, talking about behavior, there's the behavior of the government and lots of things that we do not anticipate. So broadly speaking, how do you take that into account? Just like into the mathematical modeling? I'm curious, from a data science perspective, you're adding another shocking term or changing some aging behaviors. What would be the general way to think about this?

[Andrew Lowe]

Well, let me give you an example. One way to model the dynamics of the economy is to use an agent based model, sometimes now referred to as agentic AI.

[Shali Meng]

Agentic.

[Andrew Lowe]

You begin by assuming that agents have certain kinds of behaviors, and you model those behaviors mathematically in the same way that we model rational behavior. By assuming that agents maximize expected utility subject to a budget constraint, you can have a bunch of behavioral rules that will govern an agent's behavior. Okay? Now, once you've got all of these behavioral rules programmed into your agents, you let them interact and you see what happens to the economy. And so that's really where you can capture some of these irrational, but nonetheless very easily justifiable behavioral assumptions and watch as the various different agents interact with each other. So in the case of a trade war, it's actually pretty simple. Suppose that we impose a 20% tariff on foreign goods. The country that's going to be affected will likely impose a 20% tariff on the goods that we import to them. So you can think of that as a tit for tat strategy. Those tit for tat strategies have been simulated in many cases with artificial intelligence agents interacting, and you can actually calculate equilibria of those kinds of interactions. So that's one way to try to model the outcomes of these various different trade considerations. And my guess is that when you do that, I haven't done the analysis. But if you did that for this current trade war, it would end with all countries suffering dramatically and ultimately certain politicians being voted out of office and then a new economic regime being established once people understood that trade wars are in general not good for prosperity.

[Shali Meng]

I see. So that is, you're saying that based on both the economic theory as well as the historical data, things of this nature may achieve some short term goals, but in the long term it will not be achieved a good. I don't know, whatever they're trying to achieve. But you said whatever is being achieved probably is not what we wanted.

[Andrew Lowe]

That's right. A good example of that is the last time we proposed widespread tariffs, which was in 1930, the smooth Hawley tariff. Now, that event did not happen in a vacuum. The reason that we passed a law imposing tariffs on all foreign goods was because of the 1929 stock market crash and the Great Depression. Politicians thought that it was important to protect U.S. farmers and U.S. industries. And so they imposed widespread tariffs on all foreign goods. And what happened was that that drove the economy into a much, much deeper depression. And by 1932, both Smoot and Hawley, senators from various parts of the country, they were voted out of office. And that actually ushered in an entirely new regime and brought in FDR and the democratic approach to dealing with the depression. So that kind of dynamic you can easily capture in an agent based model. And I suspect that if we were to play that out, we would see something along those lines. But of course, you now have to capture the kind of decision rules that would be a good approximation for President Trump's particular policies. And as far as I've heard, President Trump has kept his particular policies fairly close to his chest, deciding on a daily basis how he's going to act based upon what he sees as the best course of action.

[Shali Meng]

So now, assuming politicians really learn from the data, but they still want to protect domestic industry, let's say, for really noble reasons. What are the alternative economic tools? Do they have?

[Andrew Lowe]

Well, there are a lot of tools depending on what your ultimate objectives are. And I think this is where it gets really interesting because if you're looking for short term gains, then one set of tools will be appropriate. But if you have a 10, 20, 30, 50 year horizon, then all sorts of other tools come into play. So let's focus on short term. The short term kinds of tools that people have used are things like quotas, subsidies, export controls and tariffs. And it turns out that in the short run it can look like some of those tools are working. For example, if you're worried about exporting very sensitive technologies like AI chips to foreign countries, you can simply say, no, you're not allowed to do that. Export control. And that's what we've done with Nvidia.

[Shali Meng]

Right.

[Andrew Lowe]

And Jensen Huang earlier this week just made an announcement in a conference in technology that Nvidia is going to be losing something on the order of $5 billion because chips that were designated for China now no longer can be shipped there. And it's not clear how they're going to be able to sell them to anybody else because they were manufactured specifically for those Chinese uses. So he's writing off something like $5 billion. Another example of that difficulty is that when you impose a tariff in the short run, that drives up foreign prices and it looks like you're being competitive, but eventually you're going to have to be more competitive than the foreign countries, otherwise you're not going to be able to maintain your market share. They'll simply sell their goods elsewhere and you will be left stranded. Same thing with quotas, which is much more inefficient because you're really capping quantity, not price. And this then creates things like artificial shortages and then black markets emerge. And so there are all sorts of unintended consequences that one has to worry about. So in the long run, the kind of tools that we are now thinking about are obviously investment tax credits. To be able to promote the buildup of certain industries incentives. The carrot often works better than the stick. And so by using a combination of these tools, but using them carefully and deliberately and sensitively to the various different changes in market conditions, it is possible to, to be able to achieve long term success. And there are examples of that like South Korea and Taiwan, but one has to be very, very adept at maintaining these kinds of controls.

[Shali Meng]

Speaking of other countries, I wanted to ask you about political systems. Right. I can See, in environment where politicians lifespan is short because they only have this many years in the office, if they want to achieve the short term things just for getting voted, they might behave one way. Some other countries you have leadership stays for very long. How do you as an economist factor into those kind of issues like the political system itself?

[Andrew Lowe]

You're absolutely right that these different forms of governments have different flexibilities, different constraints, different advantages and disadvantages. And that's going to affect trade policy, among other things. I think that Winston Churchill was the one who said that democracy is the worst possible form of government in the world, except for all the other possible forms of government out there. And I think that one he was referring to was the fact that in a democratic government you have a finite term for the leaders and they will play to that term because if they don't, they won't get reelected. And so it's very difficult to engage in truly long term policy beyond the term of your office. Whereas in other countries that have more centralized control, where the leaders have longer terms, they can afford to engage in a very different set of strategies that may involve tariffs that will cost them more in the short run. But nonetheless they know that they will be around as leaders in the long run to be able to enjoy the success after they are able to support an industry and then grow it that would otherwise have been squashed by foreign competitors. So I think we have to be very aware of who we're competing against and the kind of governments that they have at their disposal, because that really changes their strategy space. And so we want to make sure that, you know, we don't show up to a gunfight with a knife, as they say.

[Shali Meng]

Well, speaking of strategies, I want to ask you real advice here, both for me and for our listeners. You probably know what I'm going to ask is like in an environment of such, like how, how do individual investors, you know, to deal with these issues, right? I mean, you know, we are all investors, even we don't actively. We have a retirement account. I got contacts by tia, you know, ask me, talk to them. I said, well, I don't really know much, but I have a great friend here who can give us the advice. And I know you invest a lot. What other general advice you have for our listeners?

[Andrew Lowe]

Well, first of all, I don't think I should be dispensing advice given that I'm not licensed. So the first advice is for all of our listeners to seek professional advice from a financial planner. But I have an idea about what those Financial planners would say, because I teach many of them in my classes, I think that the first step is to understand yourself. Know thyself. I think Socrates once told us we have to understand what our goals are, what our opportunity set is, what our flexibility is in bearing losses, and the time frame that we're investing. The longer the time frame, the simpler the advice any financial planner can give. For example, if you're in your 20s and you're going to be working for another 40, 50 years, the answer is, put all your money in the stock market, the global stock market, and you'll be just fine. That's pretty simple advice that all of us can follow. If you're in your 70s now, this gets to be a lot more complicated because you have a shorter window with which to be able to get a certain rate of return. And you have to think more carefully about how you're going to draw down your assets and be able to pay for your expenses during your retirement. So that's why I think the first thing you have to understand is what are your resources, what are the opportunities that you can invest in, and what kind of constraints do you have in terms of expenses, medical expenses, retirement, all the things that you need to be able to live on. Once you understand yourself and the various different aspects of your goals, then I would suggest seeking financial planning advice from a certified financial advisor. But in lieu of that, the idea is to diversify across various different assets and asset classes and to be able to manage your risk, to understand whether or not you can withstand the typical losses of the things that you invest in. If you're going to be investing in a relatively short horizon way because you're older, you want to be much more careful about the potential downturns in risky assets. For example, if you're in your 70s, most likely the MA, the vast majority of your portfolio should be in bonds and other fixed income assets. In this current environment, you should be very, very sensitive to inflation, because if we do launch a trade war, it is very likely that you're going to see inflation ticking up over the course of the next few years. So things like tips which are protected from inflation would be very useful. And over the course of the next few weeks and months, I suspect that the US Stock market is going to be extremely volatile. So that's a time when you have to be extraordinarily vigilant about the kind of risks that you're taking. And if you're not able to take those risks, then it may be worthwhile to take some of your risky assets and turn them into less risky fixed income investments. So basically paying attention to what's going on in the world and being able to manage your risk carefully and dynamically so that you're never in a situation where you're facing losses that you can't recover from or that you can't live with.

[Shali Meng]

Well, thank you very much for that great advice. I can testify to what you said when you're young. You know, I think the advice here just, just forget it. Don't pay much attention. Putting everything I listen to someone just told me when I was just started put everything in the stock and I did it. I forgot it actually for a long time and turned out to be. They say that's a good strategy to have you. You don't even do anything. But now for people who are getting older like me now and I think your advice is to get, you know, get serious professional advice. But as you know, well, these days many people probably will be asking ChatGPT and you wrote an article about Can ChatGPT to help your financial, you know, investment and it's terrific work. Thanks for contributing to Harvard Data Science Review. Can you say a little bit about that? Just really this topic of if you using this ChatGPT or any of these generative AIs and how trustworthy they are and what's your advice there for lots of listeners?

[Andrew Lowe]

Sure, I'd be happy to. So with respect to getting financial advice from any of these large language models, I would say proceed with caution. And if I could put a big red sign flashing that statement, I would do that right now. It's because large language models are actually quite good at being able to gather large amounts of information and summarize them for the user. So for example, if you're in your 50s and you're thinking about retirement, but you're still relatively far away from it, it might be useful to use a large language model to say I'm in my 50s, I'm married, three kids, I've got following expenses and my life goals are such and such. If you enter that into a typical large language model and ask what should I be doing with my savings and my financial decisions? My guess is you'll get a pretty interesting set of answers from a typical LLM. But from that point forward, you need to take that information and now really, really do some research and make sure that you're not getting some hallucinations and false information. LLMs are very good at giving you suggestions for ideas that you might want to pursue. You might want to look into. But I would never, never take large language model advice right now, particularly financial advice at face value. I would spend some time betting it. And in some cases I think it would be helpful to talk to a financial advisor just to bounce ideas off of them because they will have the actual expertise to be able to make recommendations. But there's nothing wrong with getting large language models to provide a kind of a survey for all the things that are out there and for you to learn more about these different things. For example, investors might be interested in a tax advantage strategy where they end up being able to invest in certain mutual funds or other financial securities that will give them a better after tax return. However, although you can learn a lot about these tax advantage strategies from LLMs, when you implement one, you'd want to go to a certified financial planner or a certified financial advisor to actually do the implementation because they're the ones that are going to check to make sure that you're actually getting what you think you're getting from, from that strategy. So I'm a big proponent of LLMs in financial advice and as you pointed out, this article that I published in hdsr, that was really a call to arms for all of our colleagues in finance and computer science and other areas of research to use LLMs in ways that can benefit broad parts of the population that currently aren't getting access to the financial advice that they need. I think that we can actually create, in a matter of a few years now, we can create LLMs that can provide trusted financial advice that might even pass the standards of regulators. We're not there yet, but I believe that that's a goal that we should all be striving for. And for once in the history of AI, I believe we're within striking distance of that reality.

[Shali Meng]

Since the publication of your article and I think the last time you were on this podcast, you also talk about a little bit about what you understand. I think you have been always very optimistic. Since then, the whole AI space has evolved so fast. Anything kind of jump out that you feel a lot more confident or you think the research have done itself? There are more updates that needs to be done. Is there any kind of update, things we talked last time on the podcast?

[Andrew Lowe]

Oh yeah, there are a number of updates. For one thing, the speed with which AI is evolving is accelerating and I guess it's not that surprising. But you know, one of the things about human behavior that I and other authors have written about is that we are particularly bad at intuition about exponential growth. You know, humans tend to think linearly, not exponentially. And so AI may have taken a relatively slow period of growth where not a lot was happening. But then we started having things like machine learning and deep learning and now generative AI. We are now at the point where the hockey stick part of the growth curve is kicking in. And I don't think we fully appreciate just how quickly. I mean, just take a look at the difference between ChatGPT 3.5, 4, 4.0 and now 4.0 deep research and all of the various different variations. And then deep seq, which is yet another revolution in terms of a very slimmed down version of a large language model that seems to work just as well, if not better in many cases. So that tells me that we see innovation happening at a much more rapid pace than we fully appreciate. And the disruptions that can be caused by that innovation is something that we have to take into account. Now, I'm very bullish on AI and I think that the wonderful things that will come out of it will largely offset the negatives. But that's only if we pay attention to the negatives and we prepare for it. I can easily see a situation which if we ignore the downside of AI, then that balance is going to be very different. And we can see some bad actors using AI for some really nefarious goals and they will ultimately succeed if we're not careful in addressing those upfront.

[Shali Meng]

So we always wrap up with magical wand questions. And so for this podcast, the magic question bring back we're talking about tariff is that if you can wave the magical wand, design a perfect data driven tariff policy, what would be the key features and would it be.

[Andrew Lowe]

Wow, that's a tough question. But let me give it a try. I guess if I had a magic wand, I would take a look at all of the industries that are currently sensitive and challenged by foreign competition and ask all of the various different stakeholders in the economy whether or not those are industries that are critical for our benefit and long term survival. And if so, I would identify those industries. And in addition to imposing certain tariffs that are formulated in a very careful manner, not to be punitive, but to provide temporary protection for those industries. Coupled with that, I would impose various kinds of incentives and industrial policy to make sure that those protected industries are playing their role in basically reinventing themselves to become more competitive globally, so that the temporary protection, it doesn't turn into a permanent protection and doesn't induce a kind of expectation that it'll always be there. And so coupling longer term industrial policy with shorter term tariff policies to protect industries temporarily is something that I think we could do and have the ability, using data science to be able to collect the data and change these policies exquisitely carefully so as to produce the best outcome. That's what I would do.

[Liberty Vidert]

Thank you for listening to the Harvard Data Science Review Podcast. To stay updated with all things HDSR, you can visit our website at HDSR MIT or follow us on X and Instagram at the HDSR. A special thanks to our executive producer Rebecca McLeod, producers Tina Toby Mack, Arianwin Frank, Gavin Yang and Bell Riley. If you liked this episode, please leave us a review on Spotify, Apple, or wherever you get your podcasts. This has been the Harvard Data Science Review. Everything Data Science and Data Science for everyone.