UBS Group AG (UBSG) Earnings Call Transcript & Summary

Good morning, everyone. Thanks for joining us today. My name is Ben Nunnery, and I am on the client team for Evidence Lab. For those who may not be familiar, Evidence Lab is a group at UBS who provide data and insights to clients and decision makers. We offer more than 2,000 data sets, ranging in types from pricing to geospatial to social media, and that cut across nearly all industries and geographies. So if you're interested in a free trial or learning more, please reach out to myself, Jeremy or anyone else at UBS. Today, though, of course, we will be focusing on the industrial and transportation sectors, and showcasing some of our most popular data sets and also touching on hot topics like mobility, automation, city flight and infrastructure. You should be able to open the presentation and follow along. We'll reference the slide numbers throughout so that you're able to do that. [Operator Instructions] I'm joined today by our Global Head of Frameworks, Jeremy Brunelli, who'll be walking us through some of these data sets and highlighting trends. Jeremy, do you want to introduce yourself and take us through the agenda?

Sure. So if we move to Slide 1, I'm Jeremy Brunelli, I head up Frameworks group, which is essentially one of the main product groups for Evidence Lab that creates different types of data solutions to answer investment debates. We work with buy-side clients, corporates, PE and others and certainly across the globe and various sectors, including industrial. So for the agenda today, we're going to kind of split this up between activity we're seeing now and some data sets we think for the industrial sector that are helpful or useful to look at, to understand what's going on. And then also some themes around what's next, including looking at things like airline forward pricing, social distancing and what it could mean for factory automation. A very new topic that we're getting a lot of questions on is around city flight or de-urbanization, which is not necessarily a new theme, but certainly one maybe -- or could be accelerated given what's going on with the pandemic. And then lastly, stimulus and infrastructure, given that there's a renewed debate about using infrastructure as a means for kind of a work program coming largely from Washington in The States. So we'll kick off, and feel free to enter any questions as we go. So starting with mobility, we have several measures of mobility, and this is Slide 2. We have -- besides the 3 that I'm showing here, we have others. But the 3 that we found that has been particularly efficacious during the pandemic and understanding which countries, which economies were decelerating, which ones were kind of flattening, which ones were recovering, we've looked at auto congestion, which is a relative measure of congestion on the roads. We look at public transit, so we partner and we have an exclusive partnership with a company called Moovit to get usage of public transit globally. Moovit is the largest apps provider for public transit, so they have a very rich global data set. And then also, we leveraged Google foot traffic trends to understand how people are moving. And in these charts, we do several things to kind of make the data more harmonious so that we can kind of understand what's going on. For one thing, we adjust for key holidays, which will cause a lot of volatility in these data sets. In this case, because we're trying to understand how people are getting back to work, we've filtered out weekends. So we're only looking at the work week. And then we also do different types of smoothing so that the data kind of -- doesn't have as much volatility. Even with that said, you can see from these charts, there's a lot of volatility. And then we split the data between -- to several of the major regions. And so the next 3 slides, we're going to show you the top 5 economies minus China in each of the major regions. So what we see here in APAC and looking at the curvatures of these different mobility metrics, we see mixed results. But certainly, India, Australia, South Korea are showing kind of a steady recovery trend. Japan had notably, after a several-week and several-month slide, had kind of done a step change up, has kind of paused again, slowed down. And actually, in Indonesia, we've seen, based on containment efforts around a major holiday there, that's actually taken a step down. Shifting to Europe, and this is Slide 3 now. The big 5 economies are showing a broad recovery across all these economies that we're looking at. Germany, for example, is back to the normal base rate in terms of congestion, looking at, and the base rate in this case is looking at the 2019 average congestion of the same day. We see that France is also showing a step change in a broad recovery, even countries like Italy and Spain, which really kind of suffered some of the worst incidence rates of the pandemic are kind of moving off the bottom at a healthy pace into recovery. Okay -- even U.K., which is kind of, has been sort of lagging in terms of these mobility metrics, it has even shown bottoming trends in public transit and starting to move up as well. So kind of Europe are, at this point, showing kind of a consistent recovery trend across the major economies there. Shifting to Slide 4, we look at the Americas. And Americas, based on these metrics, clearly appears to be lagging the rest of the regions. In the U.S., for example, it's showing a steady moderate recovery in the 3 mobile metrics, but the curvature of the chart is much slower than what we're seeing, for example, in Europe and in some parts of Asia. And there's no notable kind of step change in the national data, even as we -- several states have now kind of moved to partial openings. Myself living in New York, we're moving into Phase 2. We still don't see a major kind of recovery trend in the data, just kind of a slow moderate recovery. Canada also, similar to the U.S., showing a slow, moderate recovery. Mexico, Argentina, kind of mixed trends across the different mobility metrics. And Brazil actually appears to be flattening and decelerating at this point. Shifting gears…

Jeremy, just on foot traffic, is that just anyone moving around? Or what's kind of that measuring?

Yes. So this is foot traffic to workplaces. And we don't have a clear kind of understanding of the algorithm that's going here. But myself and several members of our team kind of have worked with these types of data in the past. And there's several heuristics that are used to kind of understand place detection. And so they're looking at what's called a software development kit within your app, which carries a GPS, and it records a signal of people moving. And then the heuristics that I'm talking about, they'll see is like how long are you staying in a location, for example. So it's not enough that you kind of go to an office building or a factory or something. They'll actually see, are you staying there for 8 hours? So you're doing a workday. So it starts to create these -- this algorithm to understand where your workplace is. And so what we're showing here, again, for the use case of understanding are people getting back to work, looking at public transit, looking at auto congestion, which are 2 models to getting to work, but also looking at people's foot traffic to their typical workplaces, this is showing that case. The foot traffic data also shows other categories, like foot traffic to grocery, foot traffic to retail and some others. But given the use case here, we were focusing on workplaces.

Got it. So it's not just it's the weather is getting nicer, people are walking around more, that it's more -- they're going to specific places?

That's exactly right. Yes, yes. And Google has a very rich global data set of what's called geofenced or polygon structures. So they digitize, for example, retail stores. They digitize manufacturing plants and a lot of other things. So they have an understanding of different types of places, and then they're taking these movement data and associate into these places, using those heuristics to understand, like some of the behaviors that are people are doing. And this has been a very useful data set to understand: are people kind of still self-quarantining even as the government has said we're opening up? Or are they working from home and different things. So we've been using this in pretty much every geography alongside these other mobility metrics to kind of paint the picture around that. And so shifting gears to another measure of activity, in the early stages of the pandemic, there was not a lot of these mobility metrics. A lot of these mobility metrics, kind of based on kind of different corporations that have these type of data, they've offered these as metrics to help with the understanding the data of what's going on with COVID. We didn't actually have a lot of this in the early -- a couple of months ago. So one of the data sets that we looked at, and this is on Slide 5, is what's called the Air Quality Index. So we scraped data from roughly 90,000 air measuring stations globally. These stations are maintained by what is effectively like an environmental agency for each of the different countries that they're located. And they're trying to measure different types of particulates in the air. So this could be sulfur dioxide, it could be ozone, it could be PM2.5. They take those measures -- those particulate measures, those measures of pollution, and they transform them into an Air Quality Index, which gives you kind of a measure of health. Now when I was watching the Beijing Olympics, I actually was -- I was quite amazed because the government had instituted a blue sky law, because they wanted to make sure that when they were broadcasting this globally, they would have this nice pretty blue sky behind them. And they basically shut down significant amount of capacity in the industrial sector just to make this happen. And I started to investigate whether that could be measured because in my head, I said, well, this is interesting because this could be a data set that could help understand local industrial production. Now I found the AQI. Now what's important, though, is you can't just look at the pollution measure on a stand-alone basis because pollution is impacted by a lot of different things. There's different types of pollution, different types of polluters, what a car is polluting versus a factory, whatnot. During the winter season, there's a lot of coal-burning in certain geographies that kind of creates a seasonal effect. So in order to kind of isolate the impact of industry from pollution, you have to do certain things. So we actually have 2 data sets that we leveraged here. And working with our proprietary Evidence Lab data science team, we adjusted the measures of pollution for winter heating, for seasonal effects that we were able to isolate. We isolated -- we removed the trends in the data because the data, over the years, has been getting better as different companies are instituting more kind of environmentally friendly kind of exhaust systems. So we remove trends from the data. And then we kind of isolated around key geographies where we knew that there was kind of a pullback in activity because of COVID. So switching to Slide 6. After making these different adjustments, we're able to create an index, an adjusted pollution index, to look at key provinces in China to understand whether they were recovering and whether they'd come back. And again, at this time, we didn't have mobility metrics. So we were using this as an indicator to understand whether these different provincial regions were polluting. Again, the idea is if they're polluting and adjusting for all those different factors, this could give us an indicator of whether people are going back to work and whether industry was coming back. And you can kind of see, like we were able to see for these very key provinces, which are selected because of their industrial capacity, started to bottom about mid-February, and start to move out and recover. And this is consistent with what we found out after the fact that the mobility metrics and how the recovery was happening in China. This is a global data set that we can use for most of the major economies that we have data for. So it was a good early use case for us to kind of understand activity. Then shifting gears, another data set, this is Slide 7. In the energy sector, and it's not -- it's well-known that oil prices have really -- kind of falling off, partially due to, if not largely due to, the pandemic. We have 2 key data sets that we capture. One is for flaring. And what's flaring is, basically, we use remote sensing imagery from these weather satellites that are fitted with this infrared technology. These satellites cross the equator about 14 times a day, so we get about 2 pictures per day for every piece of land in the world. Of course, there's clouds and stuff that you have to adjust for and different things. But what's -- the use case here is that when different oil and gas operators are drilling for oil, they may release gas. And when they release the gas, they, sometimes because of the economics, sometimes for safety reasons, they burn that gas off, which is called flaring. These satellites are able to pick up these flaring events. And with that flaring, we're able to actually see how -- the activity of oil production in any part of the world. So this data set allows us to look at well-known regions like the Permian Basin in the United States, but also we can see harder-to-reach areas in Africa and the Middle East, where there's not a lot of data, so we can understand more about energy, oil production and the energy complex. And the second data set that we use, which is actually a little bit earlier stage in the oil kind of -- in the phase of extracting oil from the earth is a rig dataset. And basically, when these operators are drilling for oil or gas, they hire rig operators, these big drilling platforms, to come in and to drill. And so we capture rigs. And this data set that we have is actually the GPS signal on the rig. So we actually know where these rigs are going. And then we associate the location of those rigs near key assets of different types of energy companies. So it could be near a processing plant, that could be near -- it could be the actual land ownership of an energy company. So we're actually able to see, for those different companies' assets, who has the most rig activity, which again, is an indicator of oil activity. What you can kind of see here on Slide 7, is that both figures for the U.S. are down over 50%. I think it's even more at this point, we've updated the model because this is from last week or so. So this is a very useful data set to understanding the energy complex, which has obviously has a lot of important implications for industrials. The next slide, this is actually a new data set that we launched this past week. We partnered with Dodge Data, which is a -- it's a well-known construction data provider in the United States. The U.S. government uses Dodge Data to actually create the figures for the nonresidential construction measure that's reported on a regular basis. We're getting the data at a county level, which gives us a little -- like a very high-resolution to understand regional dynamics, and we also get different types of construction categories. So whether it's amusement or social recreational buildings, apartments, dormitories, government service buildings and so on and so forth, this allows us to kind of think about a lot of different themes that are happening. This data, for example, we could look at office. And I'm going to talk about city flight in a few slides. As different companies are considering their real estate strategy, and whether they're going to keep concentrations in center cities where pandemics can kind of be the worst for people, whether they maybe drive a hub-and-spoke model, and so we can start to track the activity of construction of office parks and different things like that. We also capture things like data centers in this data set. A quick takeaway from Slide 8. You can see here that on -- the left chart is showing a map of the U.S., and this is residential construction, a 12-month trailing figure on the year-over-year growth of that. And largely, you can see that it's kind of -- it's a mixed result across all the different states. However, if you look at the nonresidential construction, you see a lot more diversity in the year-over-year trends, with some states that are actually quite green, kind of positive and then probably a larger percentage of states that are red and some that are considerably very significant -- indicating that they're down significantly. On a total basis, the total U.S. figure for April was down 2.5% year-over-year for the 12-month trailing figure for non-res. And for housing, at this point, was up 2.5% year-over-year. Switching to Slide 9. This is a -- one of our data sets from our Nowcasting products. And Nowcasting is a form of predicting near-future data sets. You can almost think of this as it's kind of like weather forecasting. And the accuracy of the forecast in the near term, it gets better as you get more data points. What's different about Nowcasting than weather though, however, is that weather; you actually know what the weather is today. Oftentimes with different types of macro factors, you're not getting that read from the government or from a company, oftentimes well after the actual period. So what Nowcasting has developed over the years is a way to kind of increase the frequency of signals that can be used to understand what's going on. And this is an example of one of our U.S. macro Nowcasting products. We have a combination of several different weekly activity factors, that we're able to use to create a prediction model for the U.S. manufacturing industrial production. And you can kind of see here in the last forecast that we put out, we were forecasting about 8%, and the actual came in about 8.3%. So the efficacy of this model has been pretty good, and I think what also is helpful for analysts in the industrial sector, you'll see that our weekly index actually, which we publish as well, gives you a kind of a leading kind of a measure as well. And what we can see in the data is it started to flatten out and started to bottom and to move up. And so this is a data set that we put out -- we update on a regular basis, and we put out on a monthly basis onto our portal. The second Nowcasting set of data I want to show you on Slide 10, this is from our global end market Nowcasting monitor. And this is a very probably difficult chart to see, and I apologize for that, but I'll give you the kind of the takeaway here. Basically, we have about 50 different end markets that we've defined, and we used 20,000 different indicators to create a signal for each of those end markets within each of the regions. We split the end markets by major region, so there is one for the North America, one for Europe and one for Asia, and then there's different end markets that we look at, ranging from things like animal feed, to commercial vehicles, to medical supplies and so on and so forth. We then normalize these data to be a second derivative measure, so we can understand the acceleration or deceleration in that end market. These data are then used by our partners in research and our strategists, in particular, who uses this to predict different types of measures of sales and sales trends for different companies that they cover. We don't actually do that, but we help them with putting together these more macro-level figures. A broad takeaway that you can see here, and this is very consistent with what we saw in the mobility metrics and the AQI and some of these other metrics that I just showed, is that Asia, there's a good portion, about 40% of the end markets in the Asia cohort of the data that's actually showing an accelerating trend, improving trend. However, within the North America and Europe, we see about 70% have actually decelerated. So this is very consistent with the cadence of what's happened with the pandemic. The pandemic started in Asia, it spread, it started to hit Europe a bit harder as a secondary kind of region, and then it started to hit The States. And so we're seeing that the end market data is kind of consistently following that recovery trend with Asia first, Europe second, Americas third.

Great. Thanks, Jeremy, that's really interesting and helpful to kind of showing us things that are going on right now, a lot of the data that you brought us through is very high frequency. Some of it's daily, much of it is weekly or more frequently. So you can really get a sense of how quickly things are moving there. And so just kind of now pivoting to looking forward, can you take us through some of the data sets that might be kind of indicative of things that are developing versus kind of what's going on now?

Sure. And a lot of these, what we put together, is based on feedback we've gotten from our clients. And also we work in partnership with the research department. So we have hundreds of analysts globally that we talk to, not just in industrials, but across the board, that help us to focus our attention on which types of data sets are commercial, useful and relevant for different debates that are happening. So we've -- this isn't comprehensive, but kind of a best of. So I wanted to start, given this is the -- it's a transportation conference, we'll start with one of our very, very big data sets, which is on Slide 11. This is our airline yield management tracker. And basically, we get about $2.5 billion price observation covering 17 million flight departures every year, 2,000 airports. So different pricing trends by airline. We cut the -- we're able to get information about which routes, different flights are going on; whether they're one way or round trip; what type of cabin class the trip is, whether it's business, it's first-class; and various other kind of key variables that go into the pricing. And this data set can be used to kind of, in a sense, reverse engineer different strategies that are observed by pricing for the different airlines within the data set. And this view that I'm showing here for the U.S. airlines kind of cohort, is what's called a forward curve. And basically, we create different metrics to understand what is the forward pricing ahead of a trip. And why this is useful is these data can help to tell you whether or not the utilization of the different assets of the airline are kind of at-plan, above plan or below plan. If you're seeing rising prices in the forward curve, based on kind of industry understanding, this typically means that utilization is at or above plan. And so the airlines are starting to try to kind of optimize margins by raising prices. In reverse, if you're seeing the forward price curve going down, it suggests like utilization is not good. So this gives you an insight into how the airline operators are kind of seeing kind of what's going on. They obviously have a lot of information about demand and some other indicators as well. But this is kind of an independent view from the outside, where we're able to kind of understand what's going on. And what you can kind of see here, just looking at the U.S. in terms of insight on Slide 11, is the price trend is deteriorating for Delta and it's kind of weak-ish for United. Generally across the Americas -- and I would say, this is consistent with, again, the mobility metrics and the various metrics we've seen, we're seeing more kind of weaker trends in the airline forward curves in the U.S. airlines than we're seeing, for example, in Asia for sure, which is actually the forward curve has actually turned positive, and Europe, which is more neutral to positive over the next few weeks and months. The next slide, Slide 12, this is actually a view from our airlines. And you can see, for example, some of the airlines' forward curves have started to either bottom or, in some cases, move up. For example, BA price is deteriorating, but EasyJet inflected higher in recent weeks. So again, this is a very big data set. It's global. You're able to look at airlines in every major region, and it gives you insight into understanding kind of the strategy of pricing as it relates to airline travel. The next data set, this is Slide 13. This speaks to automation. And the reason we selected this is because we've gotten a lot of questions about how different companies are going to react or implement social distancing as we get back to normal. And one of the things people have asked about is around automation, and whether we are seeing anything to indicate that automation trends are potentially accelerating, maybe, orders or whatnot. So this is one of our data sets on automation. We also look at patent data, for example, which is helpful in that space as well. We work with a leading consulting firm that specializes in China industrials, and tracks basically market size and share of the industrial automation market. The vendor gets these data through channel checks along the supply chain and also different forms of market research that they apply. We take the data, we break it down into several categories, and then we're able to see -- the categories within industrial automation, then we're able to kind of see which ones are seeing kind of the most demand or kind of the most implementation. So here, we show a couple of insights, for example, low voltage inverters account for the largest part of the product segment. And just to step back once more, these data are indexed. So this is not like -- these numbers that you see here, don't assume that these are like billions of units or anything. This is an index number just to kind of normalize the data so we can look at it harmoniously on the chart. So we see low voltage inverters account for the largest part of the segment, but this sector saw a decline in 4Q and -- 4Q '19, 1Q '20. AC servos, second largest market, remaining relatively stable. Generally speaking, there's no indicator from these data at this point that we've seen a step change related to what's gone on with the pandemic. You can see across these charts, they're pretty stable at this point. Slide 14 just shows you kind of a more deeper view. So for each of the product segments of the industrial automation market, we also track share by buyer industry and share by region of the supplier. So we can see, for example, which types of companies are buying these different automation parts, and then also which countries are kind of investing in automation a bit more. Here, you see a little bit more diversity. For example, elevators manufacturers were the largest buyer in the industry for low voltage inverters. But their share declined significantly in the first quarter, while HVAC and lifting machinery buyers accounted for a larger share in 1Q. So this gives you a sense of kind of share shifts and different things like that, but also who is buying these automation, and then also which countries are kind of seeing the most increase in the automation. Now again, shifting gears a little bit. When we think about city flight and de-urbanization, so we've gotten a lot of questions about does the pandemic create a more remote workforce. I know that this is a conversation that's probably being had across many companies in terms of what is their real estate strategy once we go back to normal. And now that they've actually, in a sense, proven that people working from home can be productive, we can still kind of do work. And what does that mean? One of the data sets that we use to try to understand de-urbanization and city flight is understanding are people working at home were -- is app data. So we get apps, so apps on your mobile phone, we create different types of cohorts -- and I'm sorry, this is Slide 15, for those of you that are following. We get different app publishers for, in this case, workplace solution apps. And these are apps for things like Webex and Zoom and different things like that. And basically, we are able to see which apps are ranked the most in the App Store, both for Apple and for Android. And the idea is the higher the ranking of an app, that means the more downloads it's getting, that means the more usage potentially that app is getting. And so we create these cohorts for every single country because there's different publishers in every country, not all of them are global. And then we see how that's trending. And what we saw as the pandemic kind of continued from the start to where we are today is the relative rankings across every geography that we're showing here, both in the Americas, in Europe, in APAC, you saw the rankings getting higher and higher, meaning more and more people were downloading these apps, which at this point seems pretty intuitive. Now we're starting to use this to understand which people are actually staying home or reverting back to normal. And we're starting to see inflections in the data. China was the first to show an inflection. Again, going back to the mobility charts and some of the other charts we talked about, China was the first to be hit and the first to recover, and we're seeing that they're recovering in the workplace apps with the app rankings going down now. We also see that in South Korea, Australia, so also in APAC region. Germany kind of, I would say, flattish at this point, but maybe starting to trend to inflect. U.K., also just starting to inflect and France as well. So again, this is giving us an indicator of who might still be working at home. The next data set we look at, and this is on Slide 16, we have an exclusive relationship with a company called Ookla. And Ookla has an app that understands network strain, so how fast is your network wherever you are, whether from a mobile network or if you're checking from broadband. Now why that this is important for industrials is that if the workforce becomes more remote, and they have to work from home more, this is going to put outside strain on the broadband network. And so this could drive the need for spending to kind of upgrade these networks. And so we used this data. This is a view from looking at 9 of the global cities in the data set. We also can see these data at an operator level and whatnot, to see which carriers have kind of the worst or the best network quality statistics and potentially have the biggest need to kind of upgrade their services to remain competitive, especially in light of the outsized demand that's been put on the networks from the pandemic. The next slide, Slide 17, looks at housing. And housing in the context of city flight is also an important thing to look at. So one of the first things we saw several weeks ago, we did kind of an analysis of some of the major MSAs in the U.S. that were the hardest hit, like New York, Chicago, San Francisco, and then we indexed the data. And what we were seeing is that these housing markets were definitively under pressure more than we are seeing in some other markets that were not as hard hit. And I think it bears the question -- the questions that we're hearing is, is this just pandemic related or is this a new trend? Because there's a lot of anecdotal evidence out there that people are fleeing the cities because they don't want to be in these kind of -- in the ground zero if another pandemic happens. So we did another case study, getting a little deeper into the data. We took our U.S. housing monitor, this is Slide 18. We looked at the New York City metropolitan statistical area, which is the way that the government aggregates geographies. In this data, starting from January 1, there's about 1.2 million observations, a 100,000-plus unique properties. And then if you look at the breakout across the region, from a property perspective, there's about 63% of this is single family homes; 13% is condo or town homes; 10% is multifamily; and then we have kind of single-digit rates for condos, co-ops and townhouses. Within this region, this is primarily the New York state, and so 61% of these listings are in New York state; 38% in New Jersey; and just 1% in Pennsylvania. Further, we wanted to understand, because the debates about whether people are fleeing a city, is there any indicator that there is pricing support for non-city center geographies? So are we seeing pricing support or different types of housing trends within the suburbs, within the rural areas of this MSA? So what we did is we leveraged data or geography definitions from the U.S. Department of Education, which actually goes very -- like fine detail to qualify different school systems based on the type of community that they're in, are they a suburb school, are they a city center school and whatnot. So we leveraged that data. So we took the data and we classified it into these categories that you can see here on the right in this map, between a large city, a midsized city, midsized suburban and whatnot. And then the map kind of visualizes those colors to show you. That dark red in the middle of the map, that's obviously, that's Manhattan, that's the 5 boroughs of New York City, that's Hoboken in New Jersey. That's kind of the very center city, very urban areas. And then that darker orange around it is the large suburbs, which covers counties like Westchester, parts of Connecticut, Long Island and then Northern New Jersey. So then we looked at the data in these different things, and Slide 19 shows you a cut of the data for average home price listings and inventory. And then we did that for each of those community types, the large cities down to rural distance. And then we indexed the data about the time when the lockdowns or the quarantines really started to happen in the New York region, so right around like mid- to late March. So we indexed all the data. And what you can see is some indications of the average home price that is there is price support, or like more price increases, relatively speaking, in the more fringe areas. So for example, a town distance, which is pretty far from the large city center, you see prices are actually up 15%. A town fringe category is up 2%. A rural distance is up 1%. Now notably, we see in the suburbs that these are actually, at this most recent data point, are actually trailing the large city cohort, which is predominantly New York City, Manhattan, the Bronx and whatnot. Now another kind of notable trend here, if you look at the inventory, you see that in the most recent reading that we've captured from the millions of listings that we look at, the inventory in the large city, which again is primarily Manhattan, continues to stay relatively below where we indexed it to. However, in some of these other cohorts, we see that inventory has started to increase, and it's actually up from that -- kind of that base rate in the suburbs, in the town fringes and so on. So we're starting to see an indication that this possibly could be supported by a kind of urban flight. Now looking a bit deeper into the data, single-family homes, splitting it out between what's in New York State versus New Jersey, you see that actually the suburbs, which are like Westchester and large parts of Long Island, actually, again, have price support versus that large city center, which is, again, predominantly New York City. So the suburbs, in this case, when you look at this in even finer detail, you get a bit more local, are showing more support in pricing than what you're seeing in the city. Getting a little bit deeper. Because in the city, the mix of properties that are listed are very different than the full metro area. So if you look at this Slide 21, for example, you can see that single-family homes for the whole metropolitan areas are 63% of listings. But for New York City, the most important or the largest listing is actually condos and multifamily homes, which represent 50% or more of the mix. And what you can see once you look at this level is that condos, the average price is still under pressure in large cities, primarily New York City, in this case, and multifamily also still under pressure. Single-family home has recovered. And then inventory levels, more than any other cohort that we looked at for condos in New York City are down almost 30%. So clearly, there's a lot of listings coming off the market. So there are indications that this could be pandemic-related, but this could also be a broader de-urbanization kind of incident that's happening here. And then I'm just going to skip, so that we can get to Q&A a little bit early, to Slide 23. Another big theme that we've got a lot of questions about, and then this was fostered by the White House's, President Trump's announcement that he'd like to revisit the using infrastructure as stimulus for Jobs Act and different things like that. We've done extensive work partnering with our colleagues in the UBS Research department, collecting various types of infrastructure data. This first view on Slide 23 is a view of some of the data sets we put together around bridges and dams. So we have 70,000 -- an inventory of 70,000 bridges and dams, which are tagged as structurally deficient or functionally obsolete. That is core data that we've collected. But then what -- we've taken it a step further, we've collected locations of different industrial players. So rental companies, construction equipment dealers, gas distributors and different things. And we calculated a service area around every single branch, and then we actually created metrics to see which one of them have the most structurally deficient or functionally obsolete bridges within their service area. So we're able to create a company-level metric to actually start ranking which companies have the most opportunity if this $2 trillion of stimulus is passed and put into the economy for Jobs Act. So we've done this for bridges and tunnels. Slide 24 shows us another infrastructure category. We collected nearly 100 million different road segments in the U.S. There is different metrics that we've created with it around what's called the International Roughness Index, which is a metric used by different transportation agencies to lobby for funds to their state and federal governments because if your road is more rough, you can -- it causes more accidents, and so therefore, they use that as a metric to kind of lobby for funds. We created a measure of square footage of pavement. And then similarly to what we did for bridges and tunnels, we took different types of industrial companies, created a service area around each of them, and created metrics to see how much of these worse-off roads, the roads that had the most opportunity, were within their service area, therefore, they potentially have more opportunity to gain from this program if it goes through. And the last 2 slides I want to highlight on the infrastructure side of it is, on Slide 25, this is through different Freedom of Information Act requests. We've gotten a very broad survey of water infrastructure on data. So we can understand how many feet of pipe, types of pipe, types of facilities are needed for different water municipalities, both kind of publicly-owned or privately-owned. And again, with this data, we're able to kind of associate them to different relevant industrial companies to see who potentially has opportunity in kind of close proximity to their branch network. And then lastly, Slide 26. This shows us some of the work that we've done on the electric grid. In this case, we looked at all the public utilities, their service area, and what is the density of electric vehicles in their area. Because electric vehicles -- and we've done a lot of work on electric vehicles to understand, for example, consumer adoption, the different parts of the supply chain that goes into it, even going back to like looking using remote sensing on copper mines to understand where different copper kind of commodity inputs are coming from. But we've done work to understand the density of and adoption rate of electric vehicles and then which utilities kind of have outsized adoption happening. Some of them are pretty obvious, like the utilities in California. But we're tracking this data on a regular basis to kind of understand the grid and the opportunity for upgrading the grid over time. So that's the last part of the infrastructure. So I'll hand this back over to Ben for any questions.

Yes. Thanks so much, Jeremy. That was really helpful, and we had a ton of content there, and we only have just a couple more minutes. So if you could just, in a quick few minutes, tell us anything that's coming, to look out for in this sector, that would be helpful.

Sure. We have a very -- another very big data set coming. We've been working in kind of the shipping field with data called AIS data. So these are kind of the traffic control system of different types of boats, so tankers and bulkers and containers. And we can actually see ship movements. This data is very messy. It takes a very big processing system to kind of work with it, but we should have, in the next few weeks, a global data set that allows us to see different types of quarter usage and then we'll be enhancing that data to understand which commodities are moving. This data set will be very useful for understanding, for example, bottlenecks that have been presented because of COVID, or even things like trade war or just kind of normal kind of economic activity. We're also doing a lot of work on bankruptcy data and trying to understand. We did a case study recently in one of our webinars last week -- if anyone is interested, we could share with you, looking at how bankruptcies in Texas are leading the entire U.S. and their second in liquidations. And one might think it's all energy and gas, but it's not. It's more broad-based. And so -- and we see concentrations in retail, in construction and some other areas as well. And we're going to be doing a lot of work on bankruptcies. We've also been getting a lot of questions about, will countries start nationalizing their manufacturing? So we're thinking of ways to start collecting data, thinking about that. And then broader trends that kind of are more perennial, like we do a lot of work in ESG, thinking about automation, thinking about kind of different trends that kind of are always underlying more long-term in nature kind of in the industrial space.

Great. Thanks so much, Jeremy. That was really helpful. And anyone, please reach out with any questions, happy to follow-up any time. And I hope everyone has a great rest of the day.

Thank you.