New Insights about Uber’s Impacts on Air Pollution, with Luis Sarmiento

Notable Quotes:

• Ridesharing can complement low-emissions transportation: “Within a city, there are several mechanisms that can drive the effect of Uber … [one] mechanism is called complementarity with the urban transportation system: The idea is that Uber can aid commuters to reach the public transportation system. For instance, in Chicago … all of the subway stations start in the suburbs, and they coalesce toward downtown Chicago. However, the connection between the suburb stations is very, very bad. And the idea is that if you live in the suburbs, and you’re not within walking distance of a subway station, perhaps Uber can help you with a short ride to reach the subway station, and then you use the public transportation system.” (5:04)
• Uber decreases air pollution overall: “We compared the difference in the pollution values of cities with and without Uber, before and after the company started operations, in order to try to assess the effect of Uber … We found very consistent evidence that, on average, Uber improves air quality. Just for you to have an idea of how large this effect is, we find that, on average, it decreases the air quality index by around 10 units, or 7 percent of the value of the air quality index before Uber was implemented … The Environmental Protection Agency considers a day as unhealthy for the population if it is beyond 100 units. What we find is that Uber decreases the number of days beyond 100 units by 2.53 days. This means that, overall, we have 2,000 fewer days of bad air quality per year, aggregating across all of the cities. This is a very significant finding, and this is a very significant effect.” (13:28)
• The Uber effect varies across cities: “There are some caveats to this study. I think one of the main issues is that we’re talking here about the average effect of Uber across the United States. This doesn’t mean that Uber improves air quality for all of the cities in the country; this is an average effect. We can have cities where, because of their specific characteristics, Uber decreases air quality—whereas there can be cities in which the contrary happens.” (17:22)

Top of the Stack

• “The Air Quality Effects of Uber” by Luis Sarmiento and Yeong Jae Kim
• Iran: A Modern History by Abbas Amanat

The Full Transcript

Kristin Hayes: Hello, and welcome to Resources Radio, a weekly podcast from Resources for the Future. I'm your host, Kristin Hayes. My guest today is Luis Sarmiento, a postdoctoral research economist at the RFF-CMCC European Institute on Economics and the Environment, or EIEE. Luis is based in Milan, but is also affiliated with the Energy, Transportation, Environment Department of the German Institute for Economic Research, or DIW Berlin.

Luis and I are going to be discussing a paper that he and his coauthor, Yeong Jae Kim, released earlier this fall, looking at the air pollution impacts of Uber. The introduction of Uber and other ridesharing services has been getting mixed reviews in terms of environmental impact. Luis and I are going to look at his new research and discuss what the evidence shows. Stay with us.

Hi, Luis. Welcome to Resources Radio. It's really nice to talk with you.

Luis Sarmiento: Hi, Kristin. It's a pleasure to be here.

Kristin Hayes: Great. Quickly, before we talk about this new research, I would like to introduce you a bit more to our listening audience. Why don't I give you a chance to say just a bit about your background and your overall research interests?

Luis Sarmiento: Well, overall, I'm an environmental economist with a particular focus on the determinants of air pollution—and, of course, the effect that air pollution has on human welfare. And what do I mean by human welfare? Well, it's the effects of pollution on health, on labor productivity, on cognitive abilities. Besides that, I also focus on other environmental topics, such as the North American energy transition, or the effects of natural disasters on labor markets, or the transportation sector. Overall, those are my main topics.

Kristin Hayes: Okay. And how did you get interested in working on air pollution in particular?

Luis Sarmiento: Well, it is a complicated story. I used to be a banker back in the day. And at one point in time, after my master's degree, a professor came close to me and said, "Hey, do you want to do your PhD in economics?" I started doing my PhD. And while I was doing my PhD—where I was more focused on macro—I just felt really connected with environmental sciences and environmental research. I completely shifted my focus.

And within environmental research, something that—because I come from Mexico City, and Mexico City is constantly having pollution related issues—once I started doing environmental research, I was drawn to watch air pollution, since it's something I've been hearing about in my day-to-day life since I was a kid. That's the main reason why I'm doing air pollution research.

Kristin Hayes: Right. Well, that makes a lot of sense. I think issues that are close to home are often most compelling to try to study.

Let's talk about this new research that you and Dr. Kim are putting out there right now. And I want to start with a little bit of context by talking about what we do and do not know about Uber's impact on air pollution.

You noted in the paper that then–New York City mayor Bill de Blasio had argued for restricting Uber's growth in the city, based in part on concerns over air pollution increases. But you also note that, and I quote here, "The existing literature has yet to understand Uber's effect on overall air quality." There's some narrative that Uber is making things worse, but there's clearly still some confusion. Where does the sense that Uber has led to increased air pollution come from?

Luis Sarmiento: Well, it is not only Mayor de Blasio. We have evidence, well, anecdotes also from the mayor of London, the mayor of Milan, even the mayor of Mexico City, being worried about the pollution effects of Uber. This is a very easy connection and I think a very logical one. One that I, myself, when I started this paper, thought very logical.

And the basic idea is, well, cars pollute. Uber increases the number of cars, and naturally, more cars means more pollution, which is this very simple explanation. And human beings are driven to these simple explanations. However, what we want to point out in this research and in this paper is the relationship between Uber and air quality is way more complicated. And it depends on other factors beyond just the increasing number of vehicles.

Kristin Hayes: Yeah. And you often reference, in the paper, some of these challenges studying the link between Uber adoption and air pollution impacts. Can you say a little bit more about what makes that study particularly challenging?

Luis Sarmiento: The first challenge is that, of course, the effects of Uber are going to be heterogeneous by cities. And this depends on the architecture and the transportation environment of the specific city. Now, within a city, there are several mechanisms that can drive the effect of Uber.

For example, one of these mechanisms is the substitution of all cars. Once Uber enters a city, it can substitute all taxis. For example, imagine New York City: all the taxis in New York City were substituted by Uber rides. Or even it can substitute all private vehicles. If this happens, Uber will tend to decrease air pollution.

Another mechanism is called complementarity with the urban transportation system. And the idea is that Uber can aid commuters to reach the public transportation system. For instance, in Chicago, the subway network is built in an asterisk form in which all of the subway stations start in the suburbs, and they coalesce toward downtown Chicago.

However, the connection between the suburb stations is very, very bad. And the idea is that if you live in the suburbs and you're not within walking distance of a subway station, perhaps Uber can help you with a short ride to reach the subway station, and then you use the public transportation system.

However, not everything is good, of course. We also have evidence of the scale effect, which is the reason why Mayor de Blasio was saying that Uber was increasing pollution. Because indeed, Uber does potentially increase the number of cars in the streets.

Also with the transportation system, it does not only complement it, but it can also substitute it. At times, people that use the subway or use the public transportation system might decide to use an Uber. All of these effects in the transportation system make it very complicated to make a general assessment of what is the true effect.

However, it is not only that. Once you move from the complexity of the transportation system, you also have complexity in terms of atmospheric chemistry. Even if Uber were to increase the combustion of fossil fuels, and thus the pollution that is related with this combustion—for example, the emission of nitrogen dioxide—this does not necessarily translate to worse air quality. And this happens because pollutants in the lower atmosphere interact with one another.

An increase in nitrogen dioxide, for example, can lead to decreasing atmospheric ozone. And ozone is one of the pollutants that are most harmful for human beings. Even if you increase nitrogen dioxide, perhaps this will improve air quality.

As you can see—not only from the complexities of the transportation system, but also from the complexities of the chemicals in the atmosphere—trying to assess the effect of Uber is very complicated.

Kristin Hayes: Yeah. It sounds like you guys set up quite a challenge for yourselves. I always appreciate that, when researchers are willing to dive in and tackle the hard questions here. Let's talk a little bit about what you and Dr. Kim did to try to fill an information gap in that case.

I want to start with talking a little bit about the data that you were working with. And I'd love to hear a little bit more about the geographic scope of what you were looking at—where the data came from. Did it come from Uber directly? Where you pulled in the air quality data? Tell us a little bit more about that piece of the puzzle.

Luis Sarmiento: Yes. Definitely. We decided to concentrate on the United States, both because Uber suffered no restrictions in the United States (it had a very easygoing rollout), while in Europe, a lot of governments and a lot of cities restricted Uber. We decided to concentrate on the United States.

Regarding the introduction date of Uber, back in the days—I will say 2016—Uber used to have in their website their cities and the day they started operations in the cities. However, they stopped publishing this information, and they took it out of their website. I don't know what is the reason.

We needed to rely on other peer-reviewed studies that work with these data in order to get the specific introduction date of Uber to each city. But also, we went into Google, and we made manual searches. Like, well, what is the Uber introduction date in Chicago? And then, from newspapers or articles, we extracted the introduction date of Uber to each metropolitan area in the United States.

Regarding air pollution, we decided to concentrate on the air quality index. The air quality index is just an index that goes between zero and 500 units. And what it does is it transforms the concentration of the most important pollutants in the atmosphere into a single scale that is easy to understand. And we got these data at the county level from the Environmental Protection Agency.

Kristin Hayes: Okay. Is there any temporal variation that you all were thinking about as well? Is this dependent on the number of hot days versus cold days? Does that depend on geography? Chicago obviously has more cold days than a place like Miami. Did the geographic or temporal variation come into play at all?

Luis Sarmiento: Yes. In robustness scenarios in our paper, we look at specific areas of the United States. However, we find that, across the entire country (it really doesn't matter if you go to the northern part of the country, to the south), we see evidence that Uber has very similar effects. However, while trying to understand heterogeneous effects by time of the year, we find out that the effects of Uber on air pollution were higher in summer months. Yes, there's some variation. Specifically, the effects of Uber on air quality are higher in the summer.

Kristin Hayes: Okay. Well, I feel like we're teasing the findings a little bit, and I don't want to keep our listeners waiting too much. Why don't we turn to that? And if you can share a little bit about how you pulled that data together in designing and carrying out your analysis. Then, obviously, feel free to tell us what you found, as well.

Luis Sarmiento: Okay. What we did to try to identify the effect of Uber and air pollution was, a famous econometric technique to infer the causal effects. And this is called difference-in-differences. However, we needed to adapt these techniques to the particularities of our framework.

For example, Uber implementation was staggered across the United States. Uber started operations in cities at different points in time, which increased the challenge of analyzing the effects. Also, we have dynamic treatment effects. The effect of Uber is not a one time jump. It's dynamic. It evolves across time.

Another challenge, of course, is that you have competing policies and competing shocks. You can think of forest fires as a shock that is there, bringing noise to our estimates and that we need to control for. But in layman’s terms, just to make it pretty straightforward, we compared the difference in the pollution values of cities with and without Uber, before and after the company starts operations, in order to try to assess the effect of Uber. Of course, it's a bit more complicated than that.

Kristin Hayes: Certainly.

Luis Sarmiento: For that, you have to read the paper—but this is the basic intuition for it. Well, the findings were, for me, very surprising, to be honest, the first time I saw them. It was that we found very consistent evidence that, on average, Uber improves air quality. And—just for you to have an idea of how large this effect is—we find that on average, it decreases the air quality index by around 10 units, or 7 percent of the value of the air quality index before Uber was implemented. This is a 7 percent reduction.

Kristin Hayes: Wow. Okay.

Luis Sarmiento: It's significant, indeed. But to provide further intuition: The Environmental Protection Agency considers a day as unhealthy for the population if it is beyond 100 units. What we find is that Uber decreases the number of days that cities in the United States report days beyond 100 units by 2.53 days.

This means that, overall, across all of the cities in the United States and all of the cities in the sample, we have 2,000 fewer days of bad air quality per year, aggregating across all of the cities. This is a very significant finding, and this is a very significant effect. If you think about it, it even surprised me.

Now, looking at heterogeneous effects, as I said before, we find that most of this decrease in the air quality and most of the reduction in the days of bad air quality is concentrated in the summer months. And it comes from a decrease in atmospheric ozone. Although, we also find decreases for fine particulate matter—that is dust. These results are less robust to alternative specifications. However, the ozone effect is very, very robust and it holds across any number of specifications that we do in the paper.

Now, across the different specifications that we do in the paper, we try to understand if this effect holds across different census regions in the United States. We divide the country into the western states, southern states, the Midwest, et cetera. And we find that the effect is very consistent across regions in the United States. We find reductions in the air quality index for the West Coast. We find reductions for the Midwest, from the South, Texas, the Great Plains. These reduce worries that our point estimates come from specific shocks that we're not accounting for, because if this were the case, the shock will need to exist across all of these regions—right?

However, we worry that perhaps there are some other shocks that we're not accounting for. We also reduce the size of our data by excluding all of those counties that violate North American air quality standards. These air quality standards are when you have very large levels of air pollution, the Environmental Protection Agency fines you, and then you are forced to implement public policies to reduce air pollution.

We took away all of the counties that receive these fines, because there's evidence that these counties decrease their levels of air pollution after they're fined by the Environmental Protection Agency. And if this is correlated, when Uber is going into these counties, this might bias the estimates a little.

And in the end, we find that taking away these counties really doesn't change the point estimates or the results. And also, we do two additional robustness scenarios in which we take away all of those counties that have changes in their power plants fleet—so, if you close or you open a coal power plant. And also, all of those counties that have forest fires in their vicinity.

Kristin Hayes: A significant air quality impact.

Luis Sarmiento: Yes. And in the end, it doesn't matter which counties we consider, which counties we take away: The effect is very consistent. The effect stays there, which is very good news for our research. And we were very happy that it was such a stable effect across different subsamples.

Well, that's almost it in terms of results. However, there are some caveats to this study. I think one of the main issues is that we're talking here about the average effect of Uber across the United States. This doesn't mean that Uber improves air quality for all of the cities in the country. This is an average effect. We can have cities that, because of their specific characteristics, Uber decreases air quality, whereas there can be cities in which the contrary happens. This is an average effect. We couldn't go in the paper to analyze each city, because it was just too long—too much.

We decided, okay, let's just focus on the average effect this time, identify this effect and be very certain on what is the average. And leave for future research trying to understand what is the effect of each specific city. And, well, that, I think, is the main caveat.

Another thing that was missing in our study was to try to identify all of these mechanisms that I told you, at the start of the paper. We cannot really identify them, because as we're identifying an average effect, these mechanisms will be heterogeneous by urban center. For example, the substitution mechanism of all taxis might be very relevant in New York, but not so relevant in smaller urban agglomerations like Nashville.

To identify the mechanisms, future research should try to understand standards at the city level and see what mechanisms are the ones driving the effect of Uber and air quality, either to the positive or to a negative side.

Kristin Hayes: I find myself as surprised by the results, as you were. And then also, my mind quickly turns to, "Well, why would that be the case?" And so, getting to the “why” seems like the next step here.

And if I'm hearing you correctly, the idea of delving down into specific cities, where you can actually understand the interplay between the factors that you noted at the very beginning of our conversation—the structure of the transportation system, the geography, all that stuff—looking at that in more detail might allow you to understand some of these mechanisms a bit more. Is that right?

Luis Sarmiento: Yes, exactly, Kristin. But it's very challenging.

Because let's imagine a city like New York, and let's imagine that the two most relevant mechanisms in this city are the substitution of all taxis, but there's also a scale; there are more cars in the streets. These two effects work contrary to each other. Perhaps you find a null effect of Uber because these two effects are fighting each other. Each analysis of each city has to be very specific. However, for New York, we do find a very large decrease. Just for the future and for the future paper, we do find it for New York. What we want to do is make the analysis for each city with what's called a synthetic difference-in-differences. It's a new economic technique. I think it was just published last month in American Economic Review.

The idea is to use this to identify the effect on each city. And then apply machine learning to try to understand what are the specific factors that predict that a coefficient will be positive or negative. That is the idea of the next paper.

Kristin Hayes: I wanted to ask a follow-up, too, about whether it's also possible to look at the introduction of other new forms of transportation in cities. For example, the introduction of bike-share programs, or scooters, or any of the other things that certainly float around Washington, DC, on a regular basis.

Obviously, those being zero-emissions options, the air quality impacts would be, I think, clearly focused in one direction. But is there an opportunity to use similar techniques to look at the impacts of those introductions on various factors, as well? At least understand the magnitudes of the impacts that those introductions are having?

Luis Sarmiento: Yes, definitely. What you're saying about this bike revolution: The new bike lanes, rental bikes in big cities in the United States—I think this will be very interesting to try and understand. Also, it has very similar challenges as the introduction of Uber. The transportation system is a very dynamic animal, and it's quite difficult to try to pinpoint how all of the different elements of the transportation system interact with each other.

But of course, trying to understand the effect of people riding bicycles will be very interesting. There are some studies in Germany trying to analyze the effect of building regional trains between cities, and they find decreases in air pollution when you build regional trains with similar techniques. Yes, there's a lot of space to try to understand this.

I believe, on average, we tend to try to simplify the transportation system. For example, with bicycles, we say, "Okay, more bicycles will certainly decrease pollution." As with Uber, it's very complicated. A priori, we don't know if the people who are riding the bicycles were people who were riding the subway. Perhaps the effect is zero because these persons are not moving from cars to bicycles—they're moving from the subway to bicycles.

And of course, this forces us to do a lot of research and try to dive into the specificities of each transportation mode. However, well, this is kind of the idea. It's a complicated system, and the answer is not as straightforward as we believe it is.

Kristin Hayes: That's a really good point. If my intuition was that adding Uber would've increased air pollution, maybe I shouldn't make the assumption that adding bicycles would decrease air pollution.

Luis Sarmiento: Well, and not only just you, Kristin—even mine. My intuition was the same. Uber should increase air pollution. And as an environmental economist, I swear that I tried to destroy this result—I really tried to be like, "No, something is wrong here." But then, the more I was really going into the paper and tried to explain it, the more I realized that my beliefs, my a priori beliefs, were just wrong.

Kristin Hayes: That's why research is so valuable—because we do have intuitions, but checking them against the data is really valuable. Certainly for decisionmakers.

Maybe I'll close the discussion of the paper itself with a question about how you see this information being used. Again, there's a lot more to do here. We've talked about the research that you want to undertake. Next, we've talked about some of the different kinds of ways that this type of research could be implemented, looking at other introductions. But let's say we’ve got a robust enough body of information: How would you anticipate that someone in a city might look at this and make decisions?

Luis Sarmiento: Well, this is challenging to say, but if I were a politician or a public administrator …

Kristin Hayes: Or an urban planner, sure.

Luis Sarmiento: Or an urban planner. And now I get the information that ride-hailing technologies have the potential to improve air quality. This will be part of my cost-benefit analysis. For example, in Europe, where they restrict Uber because of environmental concerns, and Uber cannot enter cities—well, in the cost-benefit analysis, I will say, "Okay, Uber will not pollute more."

However, this goes hand in hand with the taxi industry. That is a very sensitive topic. And, of course, we need to take care of taxi drivers and their livelihoods and et cetera. It's also a disruptive technology by itself. I think, for a politician, it’s way more tricky, because he has to balance the effect of Uber on taxi drivers and the disruption of Uber in the transportation system with environmental claims, et cetera.

What I want to say is, the environmental part that we just found out—that on average, Uber improves air quality—in the cost-benefit analysis that is often done, it will go now to the benefit side of Uber. That is the only thing that I can provide at the moment. Not saying that Uber is perfect. Of course, the technology has a lot of drawbacks. And we will strive for—for example, I think in California, they're trying to make Uber completely electric. All of the Uber cars.

Kristin Hayes: I was just going to ask about that—if there's a change in the fleet, either the taxi fleet or the Uber fleet, I'm assuming that would further change the cost-benefit calculation.

Luis Sarmiento: Definitely, yeah. Even here in Milan, if you go to the airport, you have the option, if you order an Uber, to order an Uber Green, and you get an electric vehicle. If people start ordering more of these Uber Greens, then when Uber enters a city, if most of the Uber cars are electric cars, the improvements on air quality are going to be way larger.

And even from a company point of view, perhaps the company can say, "Okay, if we're having such a big challenge to enter the European market, for example, why not start making our fleet electric? That will allow us to have more negotiation power with mayors in Europe."

Kristin Hayes: Very interesting. Well, this makes me want to come to Milan, just to take an electric Uber.

Luis Sarmiento: Yes, definitely.

Kristin Hayes: See you in 2022, I hope.

Luis Sarmiento: Yes, come here. We go for some pasta.

Kristin Hayes: That sounds great. Well, Luis, this has been really interesting, and I really appreciate you taking the time to talk to me about the paper. Let's close with our regular feature, Top of the Stack. I wanted to ask you, What's on the top of your stack? What good content would you want to recommend, of any media format, to our listeners?

Luis Sarmiento: Well, right now, just from the top of my mind, I started a book on the history of Iran. I know it sounds strange, but a couple of months back, I realized I was very ignorant about the history of a lot of the world. I decided, okay, I'm going to start with just one random country. I bought a very large book, 700 pages.

Kristin Hayes: It's a very large history, so that makes sense.

Luis Sarmiento: Yes. I was not ready. I was really not prepared for this. I bought this book. It's called Iran: A Modern History, by Abbas Amanat. And it's incredible for me. It's like reading The Lord of the Rings. It's history that I had no idea it happened, very incredible things. Treachery, wars—it's very, very interesting. If anyone wants to know about the history of Iran, you can read this book. But also, the idea of just reading the history of random nations is extremely interesting.

Kristin Hayes: That's really cool. I’ll look forward to the next conversation, when we figure out what you're reading next. And then you'll be able to educate us about lots of countries in the world.

Luis Sarmiento: Yeah. About Iran.

Kristin Hayes: Mexico, Italy, and Iran. Yeah—that sounds great. Thanks for that very creative recommendation. That's much appreciated. Well, it was great to talk with you, and I hope you have a wonderful rest of your day.

Luis Sarmiento: Yes, it was a pleasure. Have a good morning—good night over here.

Kristin Hayes: Okay. Take care.

Luis Sarmiento: Ciao, ciao.

Kristin Hayes: You've been listening to Resources Radio. Learn how to support Resources for the Future at rff.org/support. If you have a minute, we'd really appreciate you leaving us a rating or a comment on your podcast platform of choice. Also, feel free to send us your suggestions for future episodes.

Resources Radio is a podcast from Resources for the Future. RFF is an independent, nonprofit research institution in Washington, DC. Our mission is to improve environmental, energy and natural resource decisions through impartial economic research and policy engagement.

The views expressed on this podcast are solely those of the podcast guests and may differ from those of RFF experts, its officers, or its directors. RFF does not take positions on specific legislative proposals. Resources Radio is produced by Elizabeth Wason, with music by Daniel Raimi. Join us next week for another episode.