Podcast Transcript: Episode 275 – The Communist Leanings of Charlotte Streets

April 22, 2020

We’ve been posting transcripts to our podcast episodes after we clean them up.  Click through for a full transcript of the episode below.

This week, we’re chatting with Geoff Boeing, a professor of Urban Planning and Spatial Analysis at the Sol Price School of Public Policy at USC. Boeing talks all things data and streets, focusing first on data usage, moving on to street network design, and then to urban design.

JW: Geoff Boeing. Welcome to the Talking Headways podcast.

GB: Thanks for having me.

JW: So before we get started, can you tell us a little bit about yourself?

GB (1m 18s): I am an assistant professor at USC in the Sol Price School of Public Policy in the Department of Urban Planning and Spatial Analysis. Prior to USC, I was an assistant professor at Northeastern University, and prior to that I did my PhD and postdoc at UC Berkeley in the Department of City and Regional Planning. So most of my academic background is in the urban planning discipline, and within that most of my research and teaching focus on the sort of data science/urban analytics realm that has been growing in urban planning over the past few years. So I look a lot at different sources of data and new sources of data, bigger sources of data, how we can use use code and computer programming to work with them and get new insights from them, and what they can tell us about the history of cities and current conditions and predicting the future of cities to inform better policy making and planning.

JW (2m 21s): So what got you into urban issues initially?

GB (2m 24s): Yeah, I’ve always been interested in it. When I was a kid, it was always something I was really interested in. I was one of those map nerds when I was little but when I did undergrad I went much more for the sort of pragmatic degree. I went into computer systems initially thinking, whatever I do in the future, this will get me a good job, right? And what I found was that it allowed me to find work, but not necessarily work that I was interested in. And as the years went by, I discovered increasingly what I spent my free time doing was thinking about cities or fretting about cities or trying to understand how individual planning decisions produced these different spatial outcomes that we all have to live with or that we all celebrate. And so when I went back to school, I took that computer systems and computer programming background into these spatial questions and questions about urban planning.

JW (3m 20s): What were you fretting about cities?

GB (3m 22s): Pretty much everything. (Laughs) I’ve always been kind of a dyed-in-the-wool urbanist, so anything from the quality of the sidewalks I was walking down, to the inexplicable bus routes and headways I was dealing with, to the really obvious nature of racialized and class-based spatial segregation and social segregation and unequal outcomes resulting from it.

JW (3m 48s): What do you think about the term data science?

GB (3m 50s): I tend to be kind of skeptical of it in part because, what is science if it doesn’t use data? So science itself is explicitly an empirical endeavor, where we collect data, we try to process it into information, and we test hypotheses with it to try to build theory, test theory, better understand reality. So it’s all very much data dependent. So if all science uses data then what is data science? The typical definition would mean that data sciences focuses more on these processes of collecting data, processing it, cleaning it, wrangling it, manipulating it into something that’s easy for analysis. Which are very useful parts of this broader endeavor, and I think data science is useful in labeling them and for grounding them and highlighting their importance, but the strange thing is that all of those processes need to be embedded within some broader discipline. So for urban data science, it would be embedded in urban geography or urban planning urban studies, same thing for biostatistics and information science and biology or biomedicine. So thinking about data science as this sort of stand-alone abstract field can be problematic because too often you’ll end up with people with a computer science degree getting their hands on a little bit of data, making some naive or atheoretical claims without actually engaging with policy-making context or the political context and saying, “Well, there it is. We’ve solved cities.” I think that that naivety and data science can be troublesome when we don’t think about it as an explicitly applied field within some broader knowledge domain.

JW (5m 37s): I think I saw a tweet the other day that somebody had solved cities with data.

GB (5m 40s): Yeah. Finally, I’ve been looking forward to that. (Laughs) I guess, I guess we’re done, podcast is over and we can enjoy our good cities now once and for all.

JW (5m 50s): How much do you think the general public knows about cleaning and processing data? I mean, I think we see these outputs and we see kind of the inputs but how much do you think people know about the in-between?

GB (5m 60s): Well, I guess in my teaching experience, very little. So at USC I teach a course that’s kind of the gateway to data and evidence and communicating it for our cohorts of Urban Planning masters students, and a lot of them already come in with some idea of the different types of data visualization. A lot of them know basic descriptive statistics or what a regression model is for predicting or explaining trends. But very few of them have any idea what to do if you open a CSV file or an Excel spreadsheet or even god forbid a PDF of data…

JW: Ugh, the worst…

GB: …and try to get pieces of information out of it or turn it into something that you can do an effective data visualization on. A lot of that is because it’s not the sexy part of the job right? It’s not like when the New York Times puts up a really nice data visualization interactive and really gets to the heart of some important public policy question, that they also accompanied it with the thousand lines of code that they had to write to get it into something workable for them.

JW (7m 11s): Yeah. It’s pretty crazy. I mean, I remember going through parcel data sets for the city of Boston and things like that and having to look through and see if everything was on the right scale or everything was cleaned up, because there’s a lot of data entry issues that happen when there’s a collection like that especially in those large data sets.

GB (7m 28s): Exactly. Yeah, and I think that kind of example is a really good one, because nowadays there are so many city data portals or county data portals out there with thousands or tens of thousands of different data sets on offer. But when you actually dig into them, you see a lot of them aren’t really usable. Either they haven’t been documented, so you don’t know what units they’re in or how the different variables were measured, or the data is really dirty to the point that it’s hard to tell what you can really salvage and still make reasonable claims from it. So a big thing I do in class as I make the students go through a data portal and try to see what they can actually figure out from the data that do exist there. Because the people who put these things online, it’s not, they’re not hired to do that. They’re not trained to do that. It’s usually just an edict from on high that we have a data portal now, you should make your data public, you figure it out. (Laughs)

JW (8m 23s): Well, that’s another question. I mean, how can, you know, folks that work in those offices that have these edicts or even lawmakers become more informed about this data collection and what they’re actually legislating? It seems like they’re trying but they don’t know a whole lot about it and it might make things worse.

GB (8m 40s): Yeah, they’re trying but a lot of the groundwork isn’t already in place to do it well, and I think part of that is a training issue. These haven’t been things that we tended to train policymakers or urban planners in, historically. A lot of that is changing now with this sort of big data data science revolution over the past 10-15 years in the social sciences and in the policy world, but a lot of it is infusing all of our jobs with a greater respect for data’s power and for its limitations. Too often we treat it as a black box or a panacea, and that means that we can’t get a lot of the benefits out of it that we could if we thought about it more as just a practical problem. We need to document it, we need to carefully explain what we did and how we did it, we need to make these things broadly and widely available, and clean enough that we can actually have second order effects from other people being able to look at it. Instead we do it as a sort of ticking boxes on a checklist. Just, I did my job rather than what’s the purpose of doing this job?

JW (9m 46s): Is there too much data?

GB (9m 48s): Sometimes. It really depends on the domain and the question. So there are some things we have lots of data about but other things that we have almost no data about.

JW (9m 59s): Yeah, I think public health is one of those “not enough data”s. It seems like whenever I was looking for data on certain specific illnesses or anything along those lines coming from hospitals. And I think part of it is because of HIPAA and requirements related to privacy. That’s another question as well, privacy, and then there’s other places where we have too much data. It seems like urban planning data, there seems to be a lot of it as well from cities, especially since like you said, they’re pumping it out at a great rate because they’re required to.

GB (10m 25s): Yeah. We often tend to have a lot of low hanging fruit data widely available. So even for your public health example, we tend to know a lot about the relationship between the built environment and travel patterns and active mobility. But conversely we don’t know much about illness or disease transmission and like hospital data that you’re talking about. This kind of goes throughout all of urban planning. So we have a lot of data about home sale prices. For example, we can look it up pretty easily from county assessor offices there are a lot of third-party services that aggregate it for us. So we can get a good sense of what homes are selling for, but we know almost nothing about current rents, which is a problem, you know when all of our data are lagged, it’s really hard to understand the affordability crisis and the rental market because the data aren’t there, even if we’re swimming in data from a very similar topic.

JW (11m 21s): I mean you wrote about this in your papers, right? So thinking about these data portals, or they’re not really data portals, they’re just services that people use to rent places or buy places whether it’s Zillow or Craigslist or wherever, but you wrote about how they’re not really representative of the overall market. How is that? How can that be, if data is so perfect and representative? (Laughs)

GB (11m 41s): Yeah, go figure. (Laughs) Yeah, so for that work a lot of what I’m interested in is getting at that problem of where we’re currently lacking data and why. And so for the rental market a lot of the reason why we don’t have really good up-to-date data about current rents and current asking rents is because those market behaviors aren’t necessarily being captured and logged. They’re often informal transactions between a landlord and a renter. Either a landlord from a large building or even just someone renting out a granny flat behind their house or an upstairs room and in their home. Because of that inherent informality to a lot of the rental market, we tend to not have good data exhaust producing the sort of transaction trail that we can use to keep track of what’s happened and when and put some numbers to it. So a lot of the work that I have done in the past few years has collected rental listings from platforms like Craigslist to try to understand who is posting what and what do those acting rents look like. Now the caveat is for rental listings, they are asking rents not final consummated transacted rents. So they’re just a proxy for the ladder. And they could be biased up or down depending on market conditions and how negotiations would work in that context, but they do tell us something about current asking rents on the open market. Now the problem is that it is a biased data set. So what that means is that on Craigslist, we don’t have a representative sample of all of the rental units in a city. So in most cities we tend to see this fairly strong over-representation of whiter, wealthier and better educated neighborhoods. Those are the kinds of places where we tend to see more rental listings relative to the number of vacant rental units. And then conversely in black or Hispanic neighborhoods, especially poorer neighborhoods or neighborhoods with lower educational attainment. We have these sort of blind spots where we can’t see as much information about them because they don’t dissipate in these online exchanges. Now this is a common problem that plagues most volunteered geographic information or user-generated spatial data, where different people will self select into the platforms. Those will shape future self selection. So if there are a bunch of housing listings available in the kinds of neighborhoods that my demographic profile tends to look at, I’m more likely to continue using it than someone else is who can’t find relevant information in their neighborhood or in neighborhoods like theirs. So you get these sort of long-standing entrenched sociospatial biases being reproduced in online platforms through these various processes of discrimination, outright prejudice and even just simple self-selection.

JW (14m 48s): How does that get fixed though? I mean, it seems like this cycle continues over and over again. You have now a digital space where the biases are continually getting reprocessed and re-represented. How do you get away from that?

GB (14m 59s): Yeah, it’s tough. So a lot of what we see online today really reflects the same sort of things. We saw in the mid century where rental brokers and real estate agents would steer people of different races toward different neighborhoods, and we see that now. So if I’m white and I log into a website that other people with my demographic profile tend to use, I tend to just enter into this filter bubble of only seeing stuff that’s more tailored for me. Partly because I went after it, but partly because all of these broader social factors throughout my life have filtered my sources of information and the kinds of people who I meet and the kinds of neighborhoods where I live and the kinds of things that I know about the city. And so that gets reproduced online just like it does in all of our other social interactions. Now, there are a few things we could do, but at this point a lot of it is really speculative because this isn’t something that has seen a lot of policymaking yet. So one thing we could do is require landlords to track their leases. So when they have either openings for a rental unit, post it on a city database but also enter into a city database rental start and end dates and the transacted rent amount they’re getting from their tenants so that we have better information about it as planners and policy makers, but also so that anybody can access a single source without getting that filtered information based on your prior knowledge. A second thing that we can do is to have different housing agencies collect information from rental listings and share them online so that if I am using a housing voucher, for example, which tends to be correlated with race and class and income levels, I can look in a single source of information and see what rental listings are available in other parts of the city that might be an excellent match for my individual housing needs, might have good access to jobs, it might have the right number of bedrooms square footage for the size of my family. But importantly, it might be in a different geographical area of the city then what I’m familiar with. If we can help people break out of those geographical knowledge silos that keep them stuck in place so much of the time, you can very often find affordable units in choice communities to use a voucher to try to do that sort of community upgrading that underlie the purpose of using vouchers a lot of the time.

JW (17m 26s): It seems like this is something that housing agencies and I think in Seattle does this, where they actually assign people to potential renters and their outcomes are much better. I don’t know if you’ve seen this research or not or this process, but it’s been actually a fairly positive outcome for a lot of folks who are trying to change neighborhoods or get out of a certain situation.

GB (17m 45s): Yeah, and this is a really interesting problem for planning and policy making more generally about you know, this nature of freedom of choice. Even when freedom of choice creates all of these dis-amenities through our entrenched biases and our inability to to really get out of our own silos in life. There are different ways that we can approach it. And I think part of it is a political problem where some solutions will work in certain places and different solutions might work elsewhere, but I think it’s it is a really important problem to solve now while we are still in the early stages, you know, the first decade of these online platforms really taking over how we acquire housing and think about cities before these processes get too deeply entrenched.

JW (18m 33s): And you also have a paper about discussing the divergence between the data that you’re collecting and then federal data sources like the census and the ACS. How important is the census to all of this and data generally?

GB (18m 46s): Yeah, it’s really important. You know, the decennial census is is is very important, but I think even more so the American Community Survey is now, just because it gives us more up-to-date data as planners on an annual basis or you know for finer grain scales like the census tract on a rolling five-year basis. That’s really important, and it’s really useful. The problem with something like the American Community Survey is that when we’re looking at those finer scales to try to understand say neighborhood effects using tracts as a rough proxy for neighborhoods, they tend to have really high errors. So we have these kind of wide confidence intervals because you have small samples at that scale. And because of that it’s hard to pin down a really precise estimate of what the true value is when you’re just taking a sample. So it’s hard to know what’s really going on when you have these confidence intervals that will often include a zero value when we’re talking about something like median contract rent from the ACS at the tract level. So they’re useful, but if we can supplement them with other sources of data, we can round out the picture where there are these biases and something like Craigslist data, we can also collect a lot more information at fine scales, which we can then use try to ground truth against the census, but we can also get a different side of the story from it.

JW (20m 14s): Are you excited about the coming 2020 census?

GB (20m 17s): Yeah. It’s what a weird thing to be excited about but I am very much.

JW: That’s not weird!

GB: Yeah, what a life we live. (Laughs)

JW (20m 23s): Oh, I saw some ads for collectors here in San Francisco in front of the gym place, I think they were some posters and stuff and I was like, oh that’s interesting that might be an interesting side thing to do. I don’t know if I have time, but it might be interesting to be one of those people going from house to house and asking those questions and seeing what people is reactions are and how hard it is to collect the data and put it in, because I imagine that’s just a whole other undertaking as well and just thinking about how much positive there is but how much can be messed up along the way too.

GB (20m 52s): Yeah. I felt similar to you in the past where it would be interesting to see what actually happens as the sausage is made, from someone who always uses it as this sort of secondary data source that I’ve just been handed. It would be really fascinating to try to understand the processes involved from a collector’s perspective. And also we know from a lot of the subsequent validation work, how much of it is this sort of guesswork on the part of the people who are out there collecting data in the neighborhood and trying to impute or infer from what neighbors are telling them.

JW (21m 27s): So you have a lot of published papers and a lot of them have been on this housing subject lately, but you’ve also done street networks. What paper or what topic tells us the most about you?

GB (21m 37s): Hmm. I would say the the street network research is probably my primary individual subfield of inquiry, and that’s what my dissertation when I was doing my PhD focused on and it’s what a lot of my work outside of just regular paper publication tends to focus on too. So beyond the empirical analysis of street networks to try to understand cities and urban form better, I also have developed a software tool called OSMnx which stands for OpenStreetMap and NetworkX. It lets you download and model street networks really easily, visualize them, analyze them, simulate trips on them, all that kind of stuff. And it’s a python package that I created and maintained and that’s kind of been the bread and butter for a lot of my empirical studies that I’ve done about city streets.

JW (22m 32s): And what made you want to do that? What made you want to start the package?

GB (22m 37s): It was really just kind of a pragmatic problem. It started when I was working on my dissertation. And when I did my dissertation prospectus, I had all these questions about urban form in the structure of street networks. I was really interested in bringing in some of the new tools coming out of network science and statistical physics to try to characterize city street patterns in different ways to try to understand resilience through these kind of advanced measures of a network topology. And what I kind of took as a given when I wrote up this prospectus was that surely the data are all out there and surely the tool already exists to just process it into a nice neat model where I can then run some algorithms on it. It turned out obviously that was all completely untrue. (Laughs) And so I really had to start from scratch of trying to pipeline from a data source into a really theoretically sound modeling toolkit that we could then ask interesting transportation and urban design questions about.

JW (23m 37s): Some of the findings that you had in your histogram specifically got really popular. Were you surprised at the kind of the viral sensation of it all?

GB (23m 45s): Yeah somewhat. I mean, I, you know when I was doing and I put it online because it was I thought it was a really nice visual content, you know, it looked pretty and it told an interesting story but it’s always kind of surprising what people get excited about for those histograms. For example, they don’t really tell a clear policy story or a planning story. There’s nothing that is really actionable from it. There’s no causal model there that allows us to build better cities. They just help us understand what patterns already exist. I think what was popular about it, at least from what people were telling me, is that it compressed a lot of information about the city that we can all kind of put our thumb on from living in these different places into a really simplified artifact. So when you look at Manhattan, for example, you can trace the pattern of its grid offset from true north. When you look at a place like Detroit, you can see its two offset grids, one closer to the river and one kind of surrounding it. And I think people who had some experience in these different places were really able to kind of track their own knowledge of the place and unpack some of those urban planning histories that produce these different sort of patterns over time.

JW (24m 60s): And for folks that might not have seen the histograms or the maps or understood the research, can you tell us a little bit of like what you were looking for?

GB (25m 7s): Yeah, so I was kind of broadly interested in different patterns in street networks and different places around the world. And one of the key ones that I was interested in was this idea of entropy. So entropy tells us how disordered a system or a data set is in my case. I was interested in applying the idea of entropy to the order or disorder in howa  street network is oriented. So a grid is a really simple example of a very ordered street network. All the streets in a traditional orthogonal grid point in one of four directions. If it’s aligned to true north then they all point either north, south, east or west. In contrast if you think about the streets in a place like Rome or Venice or Sao Paulo, they tend to point in all different directions. A lot of that is the more organic evolution of those places over time without really strong centralized geometric planning laying it all out according to a single gridded spatial logic, part of it is just typography where hills will shape the ability to lay out grids in most places other than San Francisco, of course. (Laughs) And then part of it is just the history and age of the place, where you accrete these different urban fabrics over time and they tend not to be aligned with each other, and then just the wearing on them over the ages will shape and shift how they look and feel. So from an extreme example of a place like Chicago, a very ordered gridded street network to, on the other extreme, a place like Rome, a very disordered street network. And disorder here is really only in the context of how much streets are oriented. It doesn’t imply anything about social disorder or an inability to use the streets, just means the orientations. And so I came up with a set of a hundred cities that I visualized in terms of the bearings of their streets, so where the streets point. And you can really quickly pick out these patterns where one or two grids exist or in a place like Charlotte, North Carolina, where you just kind of have suburban spaghetti as the streets.

JW (27m 18s): (Laughs) I was surprised that Charlotte not Atlanta was that suburban spaghetti, but I guess maybe Atlanta is up there, too.

GB (27m 23s): Yeah. So Atlanta is a little bit more grid-like partly because it has, so I was looking at municipal city limits. If you look at the broader metro area it’s very similar to what we see in Charlotte, but in the city limits of Atlanta, although there’s some curvature of the streets, they tend to overall point north, south, east and west.

JW (27m 43s): All the peach tree streets.

GB: Right. (Laughs)

JW: What’s your favorite network? What’s your favorite design of a city streets?

GB (27m 51s): You know for me, it’s really context-dependent. I really love downtown Portland’s street network on those 200 foot by 200 foot short blocks, and that’s a nice compact orthogonal grid. But conversely, you know, I really like the sort of of wonder and joy as you’re turning corners in Venice and you have no idea which way you’re facing anymore, but there’s a canal but as you got to double back and try to route some other way. You know different street networks work well in different transportation contexts, cultural contexts, political contexts, you know, it’s really a function of the economy and the way that people live in different places. What I like is diversity, so being able to travel to different places and just walk the streets is the greatest pleasure for me really.

JW (28m 34s): Yeah, I love that too. I think I drive my family nuts sometimes though. (Laughs)

GB (28m 39s): Yeah, I definitely have taken my wife on some death marches in different cities around the world. (Laughs)

JW (28m 43s): I feel you on that. So when was it first possible to do this type of research? I mean at some point we didn’t have enough computing power or the data wasn’t there. When did it kind of coalesce into something that could be done?

GB (28m 56s): So it’s really a question of scale. So a lot of this kind of street network work has been done in some way or another for a long time, you know back in even the 60s there was a lot of thinking about early street networks and in the 70s, the mathematical representation of them, as graphs started becoming popular and people started trying to think about it in that way. But a lot of transportation modeling to this day still tends to focus on zones, which abstracts us away from a lot of the underlying fine-grained street network for the sake of simpler computational complexity. Being able to get at the heart of the matter without having to model, you know, millions or hundreds of millions of little individual elements to paint the same picture. But what’s lost is that we lose all the nuance of what individual residential streets look like or these individual planning and design patterns that show up in the urban form differently in different decades or centuries or even millennia. To get a lot of that, we needed one, more data to be able to model things at scale and two, faster computers to be able to load it all up, process it all, etc. And those have both changed really rapidly since the year 2000. A lot of what we’re doing now in network science has really percolated over into the social sciences from computer science, statistical physics, a lot of machine learning research, and now we’re able to adapt a lot of what they’ve done on the methodological side, now that our computers are fast enough and a lot of social scientists are being trained in basic data science and machine learning to understand some of the languages needed to do this work. So really for the street network stuff, in terms of urban planning questions, it really looked a lot different in the past decade than it did in the previous six decades prior to that.

JW (30m 50s): What does this mean for the future of policymaking in cities? I mean, does it mean better outcomes or is it something along the lines of just learning more? I’m curious.

GB (30m 59s): Yeah. Well, I certainly hope it means better outcomes. You know I think at this point today, a lot of the street network research has been, you know, it’s often called basic science. So asking these sort of basic scientific questions to try to understand and learn what patterns are there, trying to measure the patterns that we’ve often talked about theoretically or through case studies, but we don’t really know what the trends are like at scale across the entire country for example, that’s been really hard to analyze. So partly it is a process of hypothesis testing existing theory. Partly, it’s about trying to build new theory by looking at a bunch of observations, talking about what patterns are there, talking about individual political context or planning processes that produce them and then using that as a new explanatory framework for what’s going on and how cities look at how they function. On the policy-making side, I think we can do better in a couple of ways, and I think some of the street network work can really help us in terms of what I was just talking about. So getting at those finer-grained models of city streets. If we can no longer just privilege arterial roads as these simplified rough approximations of the urban circulation infrastructure, we can look at more interesting questions about local access. the structure and resilience of street networks in individual neighborhoods, and how they interface with each other to create boundaries between places or to create a continuous mesh of urban fabric across different neighborhoods. We can look at how those things tend to separate and divide different racial groups from each other where those boundaries do exist or in other places where the urban fabric serves to knit communities together. That might even be an important factor in the process of gentrification as the dynamics of gentrification often will spread from one neighborhood to another. How does urban circulation infrastructure shape that? What can policymakers do to either knit cities together or to try to protect vulnerable communities? And one really important thing too is thinking about transportation beyond just the single occupancy vehicle. A lot of our transportation models that focus on arterial roads tend to focus on that as the way we get around the city. If we can think more about the dense mesh of local streets in between those arterial roads, we can ask more critical questions about micromobility and walking trips that go beyond just having to move along our big arterials.

JW (33m 34s): Well you talk about streets, and we talk about transportation a fair amount. But what about urban design?

GB (33m 39s): So design is a big part of this and a lot of the work that I do tries to think about transportation planning as in large part an urban design project. So rather than just thinking about it as engineering and how you know to keep traffic flowing through intersections, I’m interested in the design aspects of it. And it’s usually designers who are at the forefront of questions about road diets more so than traffic engineers who are more interested in free-flowing capacity. I’m interested particularly here and how design produces different patterns that have different characteristics when we do urban morphology, or the study of urban form. So different kinds of street networks that are very much a design exercise will produce more or less resilient patterns. And an easy way for us to look at that is to think about how important the most important nodes in a network are or on average if we knock things out of a network like through earthquakes or fires or traffic collisions or flooding, if links in the network are failing one at a time, how many need to fail before we disconnect some randomly selected point A trying to get to point B? With those kind of measures we can start to evaluate how resilient the network is to different natural disasters, traffic jams or traffic collisions, evacuation scenarios. And we can use that to help create better design guidelines to go beyond just design as art to design as an integral part of engineering our network and transportation infrastructure itself to make them aesthetically pleasing, sure, but also to make them function in different ways than we’ve tended to optimize for in the past around cars and driving.

JW (35m 27s): Well, it feels like before it was kind of a “you know it when you see it” type of feel and I think that you’ve criticized that to a certain extent.

GB (35m 33s): Yeah, I think it’s important for us to be able to put numbers to it beyond just you know, the look and feel of, “This looks well-connected. This looks disconnected.” Which we can do pretty effectively, but there’s more nuance than that. If you think about something like LEED-ND certification where you know, we’re counting up intersections or you know four-way junctions and stuff like that trying to get at this neo-traditional design. There’s a continuum here, and on this spectrum of good or bad urban form, there’s a lot of gray area where some are politically feasible in some places and others aren’t. And if we can measure them, we can start talking really pragmatically about the trade-offs, you know, anything from Vision Zero and pedestrian safety to encouraging more cyclists’ trips because of safer streets for biking.

JW (36m 26s): You came up with some typologies as well for putting these into data.

GB (36m 32s): Yeah, so in a previous paper when I was looking at some of those street network patterns and that context of how grid-like the networks are. So how much they all point in the same direction or not. I looked at a few other variables too, so I was also measuring block lengths, intersection densities, the circuity versus straightness of streets, and the connectivity. So the proportion of four-way intersections versus dead ends or 3-way intersections. So in the end we get this basket of different variables that characterized different aspects of the street network, and then I clustered a hundred cities in the world into different subgroups based on where they fell in terms of these different variables. And there are some interesting patterns there. Most of the older grid-like US cities tend to cluster together in these kind of fine-grained orthogonal grid cities. And then we also had fairly grid-like but much coarser-grained cities that clustered together, so places like Los Angeles or Las Vegas or Phoenix tended to be more similar to each other than they were like say Chicago or Manhattan. There was also this cluster of communist cities, so cities that in one way or another were shaped by 20th century communist government. So Kabul in Afghanistan, Warsaw, Kiev, Moscow, on and on. Those cities tended to cluster together as well with a couple of American cities too, ending up in there. I think Charlotte might have ended up in that cluster with those communist cities.

JW (38m 13s): Huh, interesting. There’s so much you can look at and see what the which cities are similar and which cities are different. They all share some similarities, but they all have some interesting little factoids or pieces that separate them just because of how they grew up seems like.

GB (38m 27s): Yeah, so we get this cluster of communist cities with stuff like Pyongyang, Moscow, Kiev, etc in it, and Charlotte ended up in there rather than closer in this clustered space to the other American cities not because it’s a communist city in anyway, but because just across these measures it tends to have this urban form that looks more like those kinds of places in terms of the circuity of its streets, the very low proportion of four-way intersections and the really large grain of its blocks.

JW (39m 2s): Huh. Maybe that’s the title for the episode. “The Communist Leanings of Charlotte?” (Laughs)

GB (39m 6s): That would be would be great clickbait to get them in for that one. (Laughs)

JW (39m 11s): That would definitely pull people in. I don’t think they’d be as revved up about the answer though. (Laughs)

GB (39m 14s): Yeah, a big letdown in the end. Though we will see though. I got a lot of contacts from North Carolina people who are interested to talk about, “I knew it all along, Charlotte is the worst, who laid out these streets, finally someone’s proved it scientifically!”

JW (39m 30s): (Laughs) But you’re not saying it’s the worst though. You’re just saying it’s a certain way.

GB (39m 36s): Exactly, and I had to explain at excruciating length. I’m not placing a value judgment on it here. What I’m talking about is how things are alike or different and how we can measure them. Now certainly there are things that are that are good or bad about Charlotte. If it has really long blocks or really low intersection density, we can place a value judgment on those different characteristics effects. So if they make it harder to walk or if they make it harder to bike, if they make it harder to run effective transit lines through it, it might make us more car-dependent. And we can put value judgments on car dependence, about physical inactivity, public health, greenhouse gas emissions, particulate matter pollution in the air. We can talk about those kinds of things, but the intersection density, the block lengths, the circuity of the streets are only part of the story that could be part of a causal framework, but it is only part of it. Trying to explain that to one North Carolinian after another has been a little bit cumbersome, but I’m really glad to see how engaged they are by this finding though, by trying to place Charlotte into this broader context of what cities are like around the world.

JW (40m 46s): Yeah. I think we’re having Danny Pleasant who used to be the transportation director at the city of Charlotte on at some point in the next couple months. So I’ll ask him specifically. (Laughs)

GB: That would be great.

JW: Well, so what’s up next for you, Geoff?

GB (40m 60s): So currently the project that I’m working on kind of polishing at this point is building on a lot of this street network research. In particular it’s looking at the history and evolution of American urban form particularly through that street network lens since World War II. So using something like 1940 as a sort of breaking point after a couple of decades of rapid automobile adoption, when American planning & policy making really focused fully on the automobile-oriented suburbanization of our national landscape, using that as a cut-off point to then say, what has happened since then? So I’m looking across a big basket of variables and kind of framing it around a story about the street grid which America used to lay out its terrain for centuries prior to, by and large World War II, when we started shifting toward these more circuitous patterns, T-intersections and cul-de-sac instead of four-way intersections, the patterns that Southworth and Ben-Joseph termed “loops and lollipops” back in the 90s in our suburbs. So what I did is I have modeled the street network of every census tract in the United States. I then used a few different algorithms to try to tag the vintage of each of those census tracts, so when they were primarily built. Typically if we look at when the earliest structures were built in a place, it’s a rough proxy for when the street networks were originally laid out there. So I’m looking both at structures built data from the census, and I’m looking at assessor data and property records that come from Zillow to try to understand when things first appeared in different places. And then once I’m able to estimate the vintage of these different street networks, I’m able to look at how their typical patterns have changed over time. So we see this really clear story across a basket of like eight different variables: griddedness declined, four-way intersections declined, block lengths grew, dead-ends grew, the cars per household grew as well across 1940, ‘50, ‘60, ‘70, ‘80 to 1990, when we see the beginnings of a rebound in the other direction. So in the 2000s and 2010’s we tend to see indicators that are returning more toward their historic values. So the 90s look to be something of a local maximum or minimum where it’s the most coarse-grained, the least grid-like, the most car-dependent cul-de-sac and curving roads. And since then we’ve seen a little bit of a shift back toward more grid-like and denser street networks.

JW (43m 46s): I look forward to seeing that. Are you going to look at collisions as well to see if there’s differences in the network style and the number of collisions?

GB (43m 54s): Yeah. So that’s the next step for me. I’m starting to look at the relationships between the design of different street networks and both vehicular collisions and also pedestrian and bike safety there. That’s just in its preliminary stages, but I think that’s a critical set of questions to ask of these data.

JW (44m 15s): And now I’m starting to think of all the things you could possibly do, which you only have so much time in the day and so much resources. (Laughs)

GB (44m 21s): I know, I have the same problem. There are a million things that are so fascinating about these topics and these data, and I just wish I had a few more 40 hour weeks.

JW (44m 32s): Stop time or something along those lines.

GB: Yeah.

JW: Well, so Geoff, if people want to find out more about your work where can they do so?

GB (44m 38s): Yeah, so, I’m on Twitter at @gboeing, my website Geoffboeing.com has at least once a month I blog about this work, try to throw up some data visualizations or summarize some of the findings that I’ve been coming up with, looking at these different housing data, street network data, thinking more broadly about data science and analytics and planning. And I try to make the blog pretty user-friendly rather than having to go to what are frankly pretty boring research papers that all of us academics write. I try to dig out the nuggets in a nutshell for what I did, why it’s important, and what it might mean for the real world.

JW (45m 14s): Awesome. Well Geoff, thanks for joining us. We really appreciate it.

GB (45m 17s): Yeah. Thanks so much for having me. It’s been fun.


JW: All right, thanks for joining us. The Talking Headways podcast is a project of the Overhead Wire, on the web at theoverheadwire.com. Sign up for a free trial of the Overhead Wire Daily, our fourteen-year-old daily cities news list by clicking the link at the top right of theoverheadwire.com, and please please please support the pod by going to patreon.com/theoverheadwire. Many thanks to our current patrons for their ongoing support. And as always you can subscribe to this podcast on iTunes or stitchr or soundcloud or overcast or Spotify, or wherever you get your podcast. And you can always find its original home at USA.streetsblog.org. See you next time at Talking Headways.

Listen to the Talking Headways Podcast


…the first thing I read every morning is the newsletter to see what’s been out there. It’s great to have an aggregator that pulls everything together so nicely.

Joe Cortright, City Observatory

I think that the email newsletter that you do every morning is the best one that I get, and I get a lot of them.

Mary Newsom, The UNC Charlotte Urban Institute

Really is the best daily urban newsletter out there.

Eric Jaffe, Editorial Director Sidewalk Labs


To Receive The Overhead Wire in Your Inbox Daily

Premium Daily Subscription

The Premium Daily Subscription is our most information packed offering, chock full of over 30 pieces of news every single day. Included are popular features such as the quote of the day and the most read article from the previous day. Also included is our weekly roundup for times when you are strapped for time but need to know what’s going on.

Premium Weekly Subscription

The Premium Weekly Subscription is for professionals constantly under a time crunch. We take the most read items from the week before and share them with subscribers along with more in depth analysis of a single popular topic.

Learn More and Subscribe

Top Links

of the past 48 hours

Video of the Day

Friends of The Overhead Wire

Back To Top

Welcome to The Overhead Wire

What Can We Help You Find?

Try Our Newsletter For Free