Austin, TX

Austin is one of the fastest-growing cities in the South; its population rose 39% from 2000-2014, from a combination of migration and city expansion. Its primary industries are technology and music. It’s also the state capital, and houses the main campus of the University of Texas, which enrolls over 40,000 students.

Thoughts from Austin

Map output

I wanted to get as much data analysis done on my home computer as I could before I head out for study abroad in Berlin. It’s not entirely perfect yet, but I have a workflow which generates these maps right out of Python for all of my cities, with some provision for longitudinal comparisons (limited by data availability). Code has improved enough that on that setup (Mac Pro 2013, 6-core, 12 GB) it takes about a day to generate these for 100 cities.

To do still are to deal with projection issues (they’re all in “web Mercator” because that’s what Leaflet uses), and centering issues caused by inconsistencies between folium, selenium, and PhantomJS. And to improve legends and captions. But it’s pretty cool, if you ask me.

2015

[flickr_set id=”72157683818724493″ max_num_photos=”21″]

See all maps from 2015

2010

[flickr_set id=”72157686472156785″ max_num_photos=”21″]

See all maps from 2010

Intersection density

OSMnx makes it easy to generate statistics on street networks by city extent, or by polygon shape. Here are the stats for the study cities, clipping the bike network (including bike paths, excluding freeways) to the 5-mile circle:

Austin Charlotte Columbus Minneapolis
Nodes (intersections) 10470 5417 12208 17892
Streets per node (avg) 2.98 2.82 3.16 3.13
Segment length (ft., avg) 323 394 321 299
Street segments, total 17733 9291 20976 30909
Bicycle mode share 3.6% 0.7% 2.3% 4.7%

This covers the entire circle, not just the census tracts, so the area is identical for the four cities. Once again these aggregated stats show very strong correlations, especially between bike mode share and intersection density (r=0.91). The fact that there are three times as many intersections in Minneapolis’ central city than in Charlotte’s is a somewhat remarkable finding, though it’s consistent with the experience of riding around those cities. Higher intersection density means that there are more choices for people getting around the city, which in turn means that the arterials have less traffic on average, and there are more alternate routes for bikes.

There’s an analogy to the geophysical concept of drainage density. A river network with fine drainage density (relatively small space between river channels) will drain very quickly during storm events because there are so many different ways for the water to reach a river channel. If you think about the networks below as representing the morning commute, the “coarse” network will require much larger, higher-speed roads than the “fine” network. And it’ll suck for bicycles.

Image depicting river systems with three different drainage densities (fine, medium, and coarse)

One of the problems of cycling advocacy is that there’s often not much that can be done to change this indigenous condition. Charlotte is already built out with high-volume streets and long blocks, and at this point you can’t make its street network look like Minneapolis’ (or Copenhagen’s).

Density = destiny?

One striking result from the target area analysis is the correlation between residential density and cycling rates in the target area.

Austin Charlotte Columbus Minneapolis
Target area (mi^2) 4.86 4.11 5.11 5.25
Population (persons) 147676 71312 129290 214592
Cyclists (work commute) 5380 464 3006 10040
Density (persons/mi^2) 30396 17342 25311 40850
Bicycle mode share 3.6% 0.7% 2.3% 4.7%
Source: ACS 2015 5-year estimates

Note that the square mileage in the above table is the square mileage of the census tracts fully contained within the target circle. Obviously the circles themselves have the same area, but because of differences in tract shapes, and non-land areas (lakes and rivers) within the circles, the census tract areas differ. This table reflects only the census tract areas, because it’s using census data.

For these four data points, the correlation between density and bicycle mode share is dramatic (r=0.97), which seems to speak to the importance of indigenous conditions in people’s mode choices. And there’s probably a real effect there. Unfortunately, the effect disappears when examined at the census tract level, becoming slightly negative in all four cities. My hypothesis is that there are threshold effects, where the most dense census tracts actually have less cycling because their walk mode share is higher. I didn’t have time to dig into that question, but it would be consistent with some of the work done by urban design superstar Jan Gehl.

Two things I would like to do are to automate the analysis to the point where I could examine this same question across many more cities, and, to combine the city data to see if the correlation flips back to the positive side when looking at census tracts across the four cities. I expect that both of those methods will show a strong positive correlation between density and bicycle mode share, but not near r=0.97.

Maps of network connectivity to jobs

One of the factors I’m trying to measure in valuing facilities is their usefulness; do they actually go where utility cyclists want to ride? I was unable to get BikeScore data (which looks pretty questionable anyway), so I had to build my own metric around it. One of the factors I used was access to jobs along the cycling network (with cycling network data obtained from OpenStreetMaps via Geoff Boeing’s awesome OSMnx tool). These maps show access to jobs within 1.5 miles from each intersection in the city; cycling routes through and near the dense areas here were scored higher.

One of the interesting things about these maps is how they visualize the density of the street network. The difference between Minneapolis’ densely connected street grid, and Charlotte’s broken-up and sparse network visualize really substantial differences in the indigenous conditions for cycling in the two cities.

These maps are using the same extents as the target area maps (the circle extenst rather than the census tract extents).

Maps of commute biking and value-added facilities

Combining my concepts of target area and value-added facilities, I developed maps for each city which select the circle of census tracts near the central city which have the great number of commute cyclists (by the 2015 ACS 5-year estimates). Each of these maps is equal area and uses the same scale.

One thing that’s visibly notable is the relatively weak connection of value-added facilities to cycling rates, by census tract; only in Columbus are they well-connected. Part of what I conclude is that cultural factors and the indigenous qualities of the street network are more important than value-added facilities in cycling mode choice.

One of my summer projects is to try to re-generate these in Python. I was trying to do all the GIS work in Python, but when it came to final output, I ran out of time and wound up having to export shapefiles and do the maps manually in QGIS. That’s why they’re not perfectly identical. Once it’s working in Python, I should be able to auto-generate the circles and the census tract data for any city in the U.S.

The methodology is potentially interesting for other kinds of analysis; I’m using bike mode share data, but you could just as easily optimize for any other census data, like median income, non-white population, educational attainment, etc. It could be a useful way to make urban areas more comparable for data analysis.


Target area data analysis

During the fall semester I took two relevant classes, one an introduction to GIS, and one on Active Transportation (with Professor Daniel Rodríguez, formerly of UNC-Chapel Hill, who will act as my thesis advisor). For my final project in the Active Transportation class, I used GIS tools to analyze bike mode share, bikeway mileage, and crash data for the four cities.

One problem I noticed while doing field research is that the city extents vary greatly; Austin, for example, has over six times Minneapolis’ land area, which means that Austin includes some sprawling, low-density areas with low cycling rates that are excluded from Minneapolis’ mode share numbers. Nathan Wilkes (from Austin’s Active Transportation Department) had done some work comparing mode share in the downtown area of Austin to San Francisco, to point out that the city has actually made more progress than would be evident from looking at the city-wide data.

Map of Austin bike mode share by census tract, showing much higher rates in the central city

Bike mode share by census tracts in Austin city extent (1:250,000)

Map of Minneapolis bike mode share by census tract, showing much higher rates in the central city

Bike mode share by census tracts in Minneapolis city extent (1:250,000)

I looked at creating a 4-mile radius circle to encompass the highest-cycling area of each city (by ACS mode share data), to be able to develop more comparable metrics from city to city. I think the method will prove very useful in the data analysis portion of the work, though I’ll make some tweaks to it in the final version. For this work I selected all census tracts which intersected the circle, but the resulting extents are still quite different in land area (by almost a factor of two). For the paper I’ll probably use a 5-mile radius and select only the census tracts which are completely included in the circle. I’ll also include census tracts outside of the central city if the circle includes them (which might affect Minneapolis, but not any of the other cities).

Bike mode share, facilities, and crashes within 4-mile radius circle in central Austin (1:100,000)

Bike mode share, facilities, and crashes within 4-mile radius circle in central Austin (1:100,000)

Bike mode share, facilities, and crashes within 4-mile radius circle in central Minneapolis (1:100,000)

Bike mode share, facilities, and crashes within 4-mile radius circle in central Minneapolis (1:100,000)

There are still a lot of issues with the data, but at least I’ll be comparing rotten apples to rotten apples.

Here’s the full paper if you’re interested.

Wrapping up Austin

Austin ride data. I'm in pink, Nancy's blue, April's orange.

Austin ride data. I’m in pink, Nancy’s blue, April’s orange.

We covered a lot of ground in Austin; 222km, over double what I did in Charlotte. There is some duplication in that number, but there are also some missing segments. That curving path all the way out on the right side is the Southern Walnut Creek path; you can see how it would have limited utility if you don’t live northeast of the city. The loop in the northwest is also not great for utility, too much steep up and down, but relatively pleasant residential streets for recreational riding.

It’s unfortunate that UT Austin (more or less in the center of this map) was not in session; you can see how much of the network is oriented towards the university. That’s a fairly fundamental problem for a summer research project on bike transportation–UMN and Ohio State will also be out of session when I visit Minneapolis and Columbus. I’ll have to see if I can manage to get back when the students are around.

It will be interesting to think about the analogy between southern Austin and Saint Paul. The area south of the river is way less bike-friendly than central Austin; high-speed roads with tacked-on, discontinuous bikeways. The “boardwalk” bike path along the river is pleasant but has issues for utility cyclists (difficult connections, potentially heavy pedestrian traffic). If central Austin is like Minneapolis, and southern Austin is like Saint Paul, what does that say about relative cycling rates?

Exclusions and “invisible cyclists”

“Cities have the capability of providing something for everybody, only because, and only when, they are created by everybody.”
― Jane Jacobs, The Death and Life of Great American Cities

The most interesting finding of my first field trip came as I was riding through East Austin. I was surveying facilities, just riding around and looking for things that cyclists might ride, when I came upon what look like a well-constructed bike path (the Southern Walnut Creek Trail, I would later discover). In the spirit of exploration I decided to ride it.

Southern Walnut Creek bike path

It was a strange experience. I was looking for an opportunity to turn to my left to get back into town, but there were none for miles and miles. I rode on this freeway-style path for over four miles, and in that time it had almost no connection to the street grid, only a single right turn about a mile in, despite the fact that I could see buildings on both sides.  It seemed a little weird, and definitely not a facility that makes sense for utility cycling. Because of the lack of connections, the path would be a long detour almost no matter where your destination is.

The next day, my last in town, I happened to run into Mercedes Feris and Miller Nulte (Bike Austin’s advocacy manager) at a café in East Austin. (Cue Jane Jacobs on the importance of chance encounters). They were able to give me some of the background on this facility and the controversy around it. There is a spot not far from where I joined the trail where local (mostly Latino) kids who live on the west side of the path go to school on the east side. Because of the disconnection of the street grid (not only this path), the kids have to go half an hour out of the way, or else walk along a train trellis and through Boggy Creek. Neighborhood groups had been asking for a bridge for years, but this $7M project was built through the same area and did nothing to help the people who live there. Some of the kids made a documentary about it:

More recently, Linda Guerrero of Austin’s Bond Oversight Commission rejected a draft bond for bicycle and sidewalk infrastructure, arguing that it was inequitable because “a lot of Mexicans aren’t going to be riding bikes.”

These are examples of the dismissal of the invisible cyclist. Stats show that about a third of all cyclists come from the poorest quartile of the population, but the image planners have of bike facility users tends to be mostly of affluent whites. So where does the invisibility of poor ethnic cyclists come from?

My thought is that these populations don’t identify as cyclists in the same way that affluent whites do, because they don’t experience the disprivilege of cycling in the same way that affluent whites do. They’re accustomed to being disprivileged, and they’re more concerned about the disprivilege they experience in policing, housing, jobs, and schools, so they don’t view bike facilities as addressing their needs. One of the cyclists I spoke with lives in East Austin and says that his Latino friends there call bike lanes “white lanes.”

I believe that good streets can benefit everyone. I also believe that bike advocacy often takes a narrow and hegemonic view of what good streets look like, and that much advocacy work loses sight of the core goals of street projects (increasing safety and reducing car trips) in favor of projects which symbolically reassert cyclist privilege. By doing so, bike advocacy misses opportunities to make cities better and more equitable places. Digging into this dynamic is what I view to be the most important part of my research.

I should note that Mercedes herself, a self-described first-generation American, is well aware of the issues of equity and exclusion, and was not involved with Bike Austin at the time of the Southern Walnut Creek project. She’s apologized to the neighborhood advocates and is working to get funding to connect the neighborhood.

Austin Transportation Department

A number of people had mentioned that I should meet with Laura Dierenfield, the City of Austin’s manager of “Active Transportation.” Fortunately I’d already had a meeting set up with her, and she was kind enough to invite a number of her staff members to discuss their projects. One of them, Nathan Wilkes, has a strongly quantitative approach and has done a bunch of analysis of different aspects of active transportation.

Achievable mode shares. From 2014 City of Austin Bike Plan

One of the more interesting points Nathan raises is the concept of “attainable mode share.” Based on data from Amsterdam, he looked at the percentage of trips of different lengths which can be captured by different modes. For extremely short trips, walking will be the preferred mode. For extremely long trips, car or train will be preferred. But for intermediate trips of 1-2 miles, it may be possible to capture a large percentage of the trips on bike. And because those trips are more common than extremely long trips, you can make real progress on auto trip reduction goals if you can increase the amount you capture those intermediate trips.

Another point he raised is that the city limits of Austin have expanded over time. If look at the mode share data from ACS on a census tract basis, and take an area of central Austin comparable to the area of San Francisco (~49 sq. mi.), the cycling mode share is close to San Francisco’s. But in the past 20 years Austin has annexed a lot of low-density outlying areas where cycling is uncommon. His assertion (which seems plausible) is that the real cycling mode share in Austin has been going up faster than the ACS data would indicate. I’ll have to think about that in the context of Minneapolis/St. Paul; Minneapolis’s mode share is at least double St. Paul’s, and they really could be considered a single city. What if Austin were two cities divided by the river? North Austin would have a lot more cycling than South Austin.

Anyway, the folks were all really helpful and engaged, and one of the staffers (Lizzie) followed up with GIS data indicating all of the bike projects they’ve built, tagged with location and date. That’ll be super-helpful as I try to relate it to areas where neighborhoods have changed over time.

Bike Austin

Mercedes Feris leading Bike Austin's Spring Fling ride

Mercedes Feris leading Bike Austin’s Spring Fling ride

At the Spring Fling ride I got to have a couple of good conversations with Mercedes Feris and other folks from Bike Austin. My impression is that their strategy is to normalize cycling. For many years in the U.S., biking was an exceptional activity, which required specialized equipment, special clothing, safety gear, and usually a car to drive to the place where the activity would commence. Like snowboarding or golfing, cycling was generally not viewed as something that could be integrated with daily life.

Leading a group ride on a city bike, in street clothes, and without a helmet is more than a collection of personal choices: it’s a statement about a different way to conceptualize cycling in the city. You’re a lot more likely to ride from the brew pub to the café if it’s OK to jump on any old bike in the clothes you happen to be wearing. Bike Austin’s vision statement includes the idea that, “cycling [should be] a common aspect of daily life for everyone.”

Until recently, Bike Austin was named the League of Bicycle Voters. The organization was founded to fight (successfully) an all-ages mandatory helmet law which was put on the books in Austin in 1996 (changed to be limited to children under 18 in 1997). And now I’ll see if anyone’s reading this blog, because there are few things more sure to start a comment-section flame war than a picture of a helmet-less rider.

My take on the helmet issue is this: Utility cyclists in other countries do not regularly wear helmets.  We do not see massive carnage on the roads of Italy or Denmark or Japan because utility cycling, by any reasonable measure, is a relatively low-risk activity. My personal opinion is that reinforcing the idea that cycling is a low-risk activity is the most important thing a cycling advocacy group in the U.S. can do. Many advocacy organizations use fearful language to promote their agendas, and I think that approach is counter-productive.

Bike Austin also has some language in their vision about diversity and social equity; more on that later.

Spring Fling

From the bike rental place we were headed out to Bike Austin’s Spring Fling ride. This was a slightly more organized social ride than the ones I’d done in Charlotte; a $10 entry fee got you a beer at the start, coffee at the midpoint, and beer at the finish, all from local businesses who support cycling. (Pretty good deal!) Like the social rides in Charlotte, the idea was to have a non-lycra, low speed, fun group ride with a focus on seeing different parts of the city.

Bike Austin Spring Fling ride

Bike Austin Spring Fling ride

It was quite fun, and Mercedes Feris, Bike Austin’s director, was a great ride leader. Most of the participants seemed to be casual cyclists, some appearing utility oriented, others more from the lyrcra crowd, and a few who probably didn’t ride much other than this sort of social ride. The group, about 50 strong, did about 30km, with a bit of climbing, with two stops along the way (At a bar and a cafe. It is interesting how integrated social cycling seems to be with beer and coffee. Not that I am complaining.)

Riding in a group meant that the road conditions were not as noticeable, but some of the pathway conditions became more obvious–for example, the difficulty of getting up and down from the river-level bikepaths to the Highway 35 bridge. The pathways are moderately steep and have narrow turns, a real problem with 50 bikes trying to get through at the same time.

I had some good conversations on the road and at the stops, and made some good contacts for later in the trip. On the way back to our place we rode by the weekly bike polo game, where I talked with Daud, the owner of the pedicab operation. There’s a lot of bike stuff going on in East Austin.

Protected bike lane

Third Street protected bike lane

Third Street protected bike lane

We picked up rental bikes at Austin Bike Tours, a local operation run by folks who are pretty well integrated with the bike culture in town. The operation is run out of a shipping container in the hotel district, on a street with one of Austin’s first protected bike lanes. The rides are city bikes from Fairdale, a local manufacturer with a flair for utility bikes. These are much different than the typical tourist rental bikes, cooler and more fun to ride, and also more practical; the fenders would later prove useful during a couple of summer thunderstorms. We had some good conversations with the staff about biking in Austin, and headed off to check out the streets for ourselves.

Left turn bike box

Left turn bike box

My personal opinion on this protected bike lane: meh. The traffic on this road was likely low-speed even before the installation, so it was probably fine as a natural bikeway. The paint and curb make the space feel safer, and that might be appropriate for a tourist-oriented space, but for me as a confident cyclist it’s not providing much value. And there’s a cost; because you’re separated from traffic by a curb, you can’t make a normal left turn out of the lane. To deal with that problem, they’ve set up left turn bike boxes in front of the cross-street traffic, which at least gives you a way to make the turn without going on the sidewalk. But it’s not a great trade-off at the high-volume cross streets. We were heading out towards Barton Springs, a popular tourist spot, which meant we needed to cross the Congress Street bridge. Without the protected bikeway, I would have made a left turn on a green light from a low-traffic street, and been well established on the road before I had to deal with significant traffic. The pedestrianized turn takes an extra light cycle, and requires me to start from zero, with a line of cars behind me on a two-lane, one-way street with parking and no bike lane.

Good design is contextual. In my opinion, a protected bikeway makes sense on a high-traffic street with few intersections or predominantly low-traffic cross streets. When those factors are flipped around, and the facility is on a low-traffic street with high-traffic cross streets, it can cause more problems than it solves. (My opinion).

Research assistants

The Bike Lab hard at work in Austin

Bike Lab hard at work in Austin

I was joined in Austin by two research assistants: my wife and our friend April. We headed out to East Sixth Street to discuss the logistics of doing surveys, and to conduct an evening count in the district.

One of the things I’m trying to look at is utility cycling outside of commute hours, which feels like it’s increased more than commute cycling in the past 10 years. East Sixth is the heart of Austin’s music district, so it gets a ton of traffic in the evening and night.

Watching the street activity, it was clear that most people get to the neighborhood by driving–in fact, the parking meters are enforced only from 6PM to midnight–but most of the drivers are carpoolers, people going out with friends or family. As I’d seen in the afternoon, there was light but steady bike traffic, and a handful of pedicabs.

While we were counting people we talked about procedures, methods, and logistics, and our shared impressions of transportation in the neighborhood where we were. The next morning we’d pick up rental bikes, join in a casual group ride, and start doing our own surveys.

Landing in Austin

Even the coffee roasters are big in Texas

I arrived in Austin around noon and had a few hours to kill before I could check in to my place. I decided to use the time productively by doing a count, so I had the taxi (An actual taxi. Uber/Lyft had just left Austin in a hissy fit) drop me off at a coffee place near where I was staying in East Austin. As we pulled up we saw a group of cyclists sharing a cup on the outdoor patio; a promising sign! Going in, I got to see a bewildering array of coffees, and was led through a tasting of a single type of coffee brewed three different ways. (They all tasted like coffee to me).

The place would have been right at home on an episode of Portlandia, which meant I was in the right place. East Austin has been a low-income Latino and African-American neighborhood for decades; I was staying near a federal housing project built in 1939 which was almost certainly segregated at the time. But Austin’s industrial expansion, along with the continued development of the music district along nearby East 6th Street is clearly driving changes to the neighborhood. (The amount of live music in town is really astonishing.)

Some of those changes are connected with cycling. The group I saw having coffee when I rolled up were recreational cyclists back from a Saturday ride, but during my count I saw a light but regular stream of utility cyclists, including a couple of cargo bikes. It turned out my spot was just down the street from the pedicab garage (Idea: Use the existence of a pedicab operation as an indicator of bike culture in a city) and I got to include a couple of pedicabs in my count.

Comparing cities

A question I’m trying to answer is, why are utility cycling rates higher in some U.S. cities than others? It’s tricky to investigate for a number of reasons. The first is simple: Our data sucks. The cycling mode share number commonly used as a comparator is based on commute cycling only, not utility cycling in general, and it has a number of other issues. (For example, it’s based on the city limits, which are handled differently in each city, and which change over time). There is no generally accepted standard for collecting count data. Bikeway mileage numbers bear only a loose correlation to the utility of the street network for cycling. And so on.

The second is complex: It’s just a difficult question. Utility cycling rates are driven by countless factors: the density and nature of the built environment, the character of the street network, weather, demographics, socioeconomic status, culture, and more. Each city has its own unique combination of these factors; extracting causal relationships from the paltry data is quite challenging.

Still, I have to start somewhere. My plan is to look at pairs of cities with relatively similar demographic profiles, but significantly different cycling rates as reported by ACS mode share, to see if it’s possible to get a little closer to what’s going on. Using some overview statistics and a bit of hand waving, I paired Charlotte with Austin, TX. Here are some of the reasons why:

Austin Charlotte
Demographic
Population 912,791 809,958
Pop. change since Y2K 39.0% 49.8%
Median household income $56,351 $51,034
Median resident age 32 33.3
Geographic
Land area (sq. mi.) 242.3 251.5
Density (persons/sq. mi.) 3520.2 3272.7
Housing density (units/sq. ml.) 1099.8 951.7
Ethnic
White Non-Hispanic 49.7% 42.9%
Black 7.2% 35.2%
Hispanic 34.0% 13.9%
Asian 6.1% 5.4%
Two or more 2.6% 2.1%

(source: City Data)

As cities 1,000 miles apart go, this is about as close as it gets. The Black and Hispanic ethnicity numbers are flipped, but other than that it’s pretty close.

A qualitative thing I like about this pairing is that Austin and Charlotte are different examples of the new South: Charlotte now being the second-biggest banking center after NYC, and Austin being possibly the fastest-growing tech hub in the U.S. These industries are driving population expansion (and a ton of construction) in both places, driving up median incomes, and creating neighborhood change.

In terms of biking, Charlotte’s ACS mode share is 0.2%, among the lowest in the country. Austin’s is 1.2%, about double the national average of 0.62% (2014 numbers). Counting bikes in Austin should be a lot more fun than it was in Charlotte.