Why Won’t LAB Fix Flawed Report That Says Chicago Bike Plan Is Inequitable?

IMG_1934
Hearing attendees discuss the Streets for Cycling Plan 2020 at the Carter Woodson Library. Photo: John Greenfield

In 2012, the city of Chicago, the Active Transportation Alliance, and hundred of residents made a major effort to ensure the Streets for Cycling Plan 2020 would create a bike network with convenient access for all Chicagoans, regardless of who they are or where they live.

The Chicago Department of Transportation held eight public meetings, all over the city, to collect input for the 645-mile planned network. Nine community advisory groups, most of them led by African-American or Latino residents, were established to help make sure the network would be useful and equitable.

Last week, the League of American Bicyclists released the study “Equity of Access to Bicycle Infrastructure” by Rachel Prelog, a Colorado-based urban planning grad student. The report establishes a “Bike Equity Index” method of using Census data to explore how well a bike network provides access to underserved communities, including neighborhoods of color. Prelog uses Chicago as the case study.

The report suggests that the effort to plan a Chicago bike network with equal benefit for people of all races and ethnicities was a failure. Prelog applies her method to a map she describes as Chicago’s “planned network.” Her analysis finds that African Americans would “account for a large proportion of the residents who would not benefit from the expanded system.”

Worse, Prelog writes, “The full build bicycle network… would do little to improve access for Chicago’s Hispanic/Latino community.” She states that the network would only provide one percent more of Chicago’s Latino population with quarter-mile access to bicycle paths and lanes.

However, Prelog’s “planned network” map is not actually a map of Chicago’s planned network. “[It doesn’t] reflect the proposed routes as identified in the Streets for Cycling Plan 2020,” said CDOT spokesman Mike Claffey. He added that the department wasn’t notified about the report before it was published.

Screen Shot 2015-09-03 at 5.23.41 PM
Prelog’s map of the city’s “planned network” doesn’t correspond with CDOT’s 2020 Plan map.

Instead, Prelog’s map is based on existing bikeways and “recommended routes” shown on the city’s Chicago Bike Map, distributed for free at events and shops. The recommended routes aren’t proposed locations for bike lanes or paths, but simply lower-traffic streets that are suitable for cycling. Many of these routes don’t appear on the 2020 Plan map, and many of the 2020 Plan routes don’t appear on the Chicago Bike Map.

Because Prelog’s analysis uses the wrong map, her claim that the planned network would be inequitable is based on bad data. That’s harmful for a couple of reasons. It does a disservice to the CDOT and Active Trans staffers, and hundreds of residents, who tried to ensure that the 2020 Plan provides good access for all Chicagoans.

Moreover, it undermines support for building the network. Why would anyone, especially African-American and Latino residents, want to get behind a plan that has been statistically proven to shortchange communities of color?

The League didn’t provide a comment on the map problem before I published a post about it last Thursday. The next day, spokeswoman Elizabeth Murphy sent a statement, which she said would be the only comment the group would make on the issue.

“The report has been updated to reflect that the data being presented in the Chicago case study was pulled from the city’s public data portal in January and May of 2015,” she wrote. “As such, the conclusions made are reflective of that data only.”

Screen Shot 2015-09-03 at 6.02.39 PM
Prelog’s “planned network” map (top) shows “recommended routes” from the Chicago Bike Map (middle), rather than the actual planned network from the 2020 Plan (bottom).

Essentially, the League is acknowledging that the “planned network” maps in the study are currently wrong, but a few disclaimers been added to the report to say the maps reflect “the data available at the time of this report.” But that still isn’t true.

Before I wrote Thursday’s post, I emailed Prelog to let her know that her “planned network” map shows Chicago’s recommended routes, not the planned network. “It may be true that [the city’s] data has not been updated to reflect changes to how they want to classify their proposed lanes.”

However, that wasn’t the problem. The bike routes in the city’s data portal are labeled “existing” or “recommended.” Prelogue apparently mistook the recommended route network for the city’s proposed bike lane and path network. That’s a completely understandable error.

However, at no point in time did the city’s data portal classify the recommended routes shown on Prelog’s map as the “planned network.” That’s why the updated report still doesn’t show the planned network “according to the data available at the time.”

“This case study was created to serve as an example of the power of these mapping tools, not an exhaustive analysis or rebuke of Chicago’s planned bike network,” Murphy concluded. “The report was created to facilitate a larger discussion about bicycle equity and serve as a resource for others to investigate their own understanding of what equitable infrastructure means. To look at it this report in any other way is to cast off integral context and the larger point of the report.”

Murphy is suggesting that the fact that the report still contains faulty Chicago maps, and statements about the 2020 Plan that may not be true, is irrelevant to the greater purpose of the study. As a reporter who witnessed the some of the work done by hundreds of Chicagoans to ensure that the bike plan is equitable, I disagree.

It’s unfortunate that the League is digging in its heels on this issue, because Prelog’s model of analyzing bike equity seems to be a valuable one. But the way the organization is handling the map problem makes their report less credible, and it’s already hurting the study’s reputation.

For example, the national advocacy group People for Bikes ran a favorable blog post about the report last Wednesday. But in the wake of Thursday’s Streetsblog article, PFB removed their discussion of Perlog’s Chicago findings, with a note that the case study “is apparently built on inaccurate data.”

It’s puzzling why the League won’t simply acknowledge that Prelog made an honest mistake and ask her to plug an accurate map of the 2020 Plan network into the Bike Equity Index model. Then we could actually get some useful insight into whether Chicago’s bike plan is equitable.

Wouldn’t that be more in line with the League’s equity goals than their current strategy of refusing to admit the report needs to be overhauled?

  • Anne A

    It seems odd that LAB would issue such a flawed report and refuse to correct it. I was part of the initial S4C 2020 outreach process and was part of the public meeting shown in the photo. Many of us who were involved in the outreach then are still involved, now having meetings and continuing outreach to prioritize where bikeways get added in the next few years.

  • Can you tell us more about the meetings and outreach you’re talking about?

  • Jon C

    John, Can you please provide a link to the GIS data for the “correct map”? The current publicly available data on “Bike Routes” from the city (http://www.cityofchicago.org/city/en/depts/doit/supp_info/gis_data.html) is not the same as the map in your story from the 2020 Plan. And this is the only publicly available data from the city that I can find. Thanks.

  • Right, the GIS layer for the 2020 Plan map isn’t currently available from the city’s data portal, and it should be. However, as I explained, Prelog took her data from the “bike routes” layer, which consists of “existing” and “recommended” routes.

    It’s easy to understand how someone who isn’t familiar with the “recommended” routes on the Chicago Bike Map would be confused by this. However, the mistake would have been caught if the League had bothered to run the report by CDOT or Active Trans (an LAB member organization) before publication.

  • Jon C

    Hmm. So the data that you think should have been used in the analysis is not actually available? Is that correct?
    I had to take a look at the Leage report again, and the never actually say they are analyzing the 2020 plan. Do you know who has that data?

  • My understanding is that the GIS layer for the 2020 Plan is not currently available from the city’s data portal. However, it’s likely that CDOT would have provided the right GIS data if they’d been asked.

    No, Prelog never refers to the 2020 Plan in the report. It appears that she wasn’t even aware of its existence, since her study quotes Chicago’s outdated Bike 2015 Plan. However, she says that her maps represent “the planned network,” and that has never been the case.

  • tooch

    I’m guessing LAB paid Prelog to conduct the study? If so, they’d probably have to pay her more to update the study with new data…my guess is that the money for the study ran out.

  • That may be a factor.

  • grifter1910

    But presumably most of the intellectual heavy lifting, i.e. developing a sound methodology, has already been done. With that in mind, so long as the paper describes that methodology in sufficient detail, what’s to prevent someone (anyone) from applying it to good data, assuming it’s available?

  • cjlane

    “This case study was created to serve as an example of the power of these mapping tools, not an exhaustive analysis or rebuke of Chicago’s planned bike network,”

    If that was the purpose, then they should have removed all of the criticism of Chicago’s “plan”, because that all tracks a a “rebuke” of the “plan”.

  • Anne A

    I attended a meeting last week of folks from CDOT, Active Trans and representatives from communities across the city, many of whom were community reps for the initial S4C outreach process. This is part of an ongoing series of meetings where we worked on maps from that process.

    The continuing outreach I refer to is various members of the above-mentioned group asking for input via contact with community groups and individuals and forums like Chainlink. It’s a smaller scale project using data from the earlier round.

  • LowSeason

    Here’s Strava’s global heatmap for Chicago, which is pretty interesting. Like it or not, it speaks for itself.

    http://labs.strava.com/heatmap/#11/-87.64861/41.86794/blue/bike

  • Interesting map. But, obviously, Strava users represent a *very* narrow subset of Chicago bicyclists.

  • dr

    Yes, it speaks for itself, and it says: “strava users live on the north side.”

  • LowSeason

    I agree that it suffers from selection bias based on the use of Strava, but still, the difference here is pretty shocking to the point where you can’t just dismiss it.

  • LowSeason

    People see things how they want to see them. This is a very large dataset about cycling habits around Chicago, so to outright dismiss it seems silly and evidence of prejudice.

    Obviously, this isn’t the data that you other others want to see, but it’s there.

  • From their website: “Strava lets you track your rides and runs via your iPhone, Android or dedicated GPS device and helps you analyze and quantify your performance.” Lots of people of people who ride bikes don’t even own smartphones, let alone use an app to track training rides. Even on the North Side, people using Strava represent a tiny fraction of bike ridership.

  • LowSeason

    Like I said, you will see what you want to see. It’s a large, albeit imperfect dataset. It has relevance whether you like it or not.

  • Come on, do you really believe Strava users are representative of everybody who rides a bicycle in Chicago? For example, that map suggests that no one is riding bikes in Little Village. Hang on out by the 26th Street bike lanes on a nice day and you’ll see that’s not the case.

  • LowSeason

    No, I do not believe Strava represents everyone. I repeatedly said this was imperfect data if you would like to re-read my posts. But it is interesting nonetheless, and the disparity appears strong enough that perhaps there really is truth behind it.

    Furthermore, your example is wrong because the heatmap calculates amount of RELATIVE usage. Neither the map nor I are suggesting NO cycling in Little Village or 26th Street or anywhere else. But it does show far LESS cycling. Zoom in anywhere you want to get a better sense of activity on a micro level.

    Again, this is a large, multi-year dataset and to instantly shrug it off as so unreliable to be literally useless seems extraordinarily unreasonable to me. I have little doubt that if it painted the picture you obviously want to see you would be singing a different tune.

  • LowSeason

    It represents a tiny fraction of overall rides but it’s still a gigantic enough of a dataset to be useful. This is how statistics works, John, by sampling. Is Strava a perfect sample? No. For example I wouldn’t be surprised if fewer people owned smartphones (a requirement for Strava) on the south side. But that doesn’t render it worthless.

  • I think we can agree that this map is very useful for seeing what routes are popular with people who like to track their training rides using a smartphone app.

  • LowSeason

    Are you suggesting that no one uses Strava to casually track rides unrelated to triathlon training? The app is free, mind you, and quite cool to use in general. Why are side streets so heavily ridden if this is purely about training?

    John, you make your inner workings obvious. No point in discussing this any further, you will obviously brand anything counter to your worldview as a non-starter for whatever reason is convenient.

  • Man, where’s Bike Snob NYC when you need him?

  • LowSeason

    Your diversions are kinda telling here. You might as well stick your fingers in your ears, close your eyes, and shout “nananana I cant hear you nananana!”

  • Mas

    Agree (LAB contracted with TAMU). It was probably a grad school thesis project. She’s already at another job.

  • Mas

    Page 5 of the report states: “this analysis [Chicago case study] seeks to investigate the equity of access to Chicago’s current bicycle network and identify areas that would benefit from better access.” The case study conclusion mentions only the current discrepancies. She probably would have been better off stopping the analysis at page 12 instead of getting into the “Expanded System” (and maybe pp 13-15 could simply be removed). However, IMHO, I don’t see this as being as egregious an indictment on the 2020 plan as the article makes it out to be. Nowhere does it state that the 2020 plan is a failure. It is unfortunate that the “expanded network” maps are in there because now we are talking more about that instead of evaluating the actual methodology and whether or not our current and (if applied to it correctly) our future network.

  • cjlane

    Really should have stayed out of the “planned” network, entirely, since there was plainly inadequate time invested in researching the “plan”.

    I think we can all agree that the *existing* network distribution is skewed.

  • cjlane

    So, sampling 90% wealthy(ish) white folks who live on the northside of Chicago is a representative sample of Chicago as a whole under what standard of statistical analysis?

  • Mas

    Agree w/ dr and John on this one. I think you need to take into account the bias of your “sample”. If you are looking for equitable bike infrastructure for a city made up of many types of people, you need to understand the bias you bring to your analysis.

    One of the culprits of inequitable planning is obliviousness – obliviousness to the biases of the planner and to the actual needs/desires of the community.

    Kudos to the planning process mentioned in the article for reaching out to more diverse groups. At least the effort was there.

  • Anne A

    There’s more to it than what I said in the previous comment. Related announcement coming soon from Active Trans.

  • dr

    OK, I may have came off a little snarky, but I was actually being literal. What strava shows is where strava users bike (I am a strava user, by the way). You are making the assumption that strava rides = a representative portion of all rides, which is an assumption based on zero evidence. It is incumbent on you to prove the relevance of the sample – the correct baseline assumption is that the sample is not relevant. Assuming data relevance is a dangerous precedent with the potential to undermine truly empirical conclusions.

    You’ve suggested the sample size alone indicates relevance, but this an extremely fallacious statistical assumption. Sample size alone cannot indicate relevance, regardless of the fact there is no indication of how large the Chicago sample size is, in relation to the total strava dataset, or in relation to the “world dataset” of all bicycle trips. “Large” is meaninglessly relative.

    There are many considerations when constructing a representative statistical sample. Size is in fact a lesser consideration, and in the era of “big data” more data points in closed environments are not in fact creating more solid sociological conclusions, but moving the required p-value threshold, the alpha, lower and lower. I would suggest that the self-selecting nature of strava users make it nearly impossible for strava data to be generalizable to society, regardless of sample size.

    You’ve introduced a cool data-set for forming hypotheses, but have fallen into the increasingly frequent internet-era trap of assuming data is inherently meaningful, which unfortunately for society, it rarely is.

  • Of course not. My husband uses it for his commuting — he is no kind of triathlete and doesn’t go for time trials, but he likes its functionality over RunKeeper and other apps he’s tried for bike logging.

    However, almost no Strava users live outside the white areas of the North Side, meaning if you rely on (or even are ‘strongly interested in’) the Strava dataset you’re going to do what everyone has always done in the history of Chicago infrastructure, and pretend only white northsiders exist or matter.

    There are quite a few HEAVILY used bike commute corridors that show as zero on Strava. It’s missing a lot.

  • I don’t believe less cycling. I believe less Strava usage.

  • LowSeason

    Yes, yet again, I admitted there is sampling bias repeatedly. However, the claim that there is sampling bias at all is itself purely an assumption. Your (and others’) claim that the Strava data is nearly worthless is predicated on an assumption that there are disproportionately fewer people from lower income neighborhoods that use Strava, yet you provide no factual basis for such claim. If I was so inclined, I could just as easily dismiss your own claim of sampling bias since you haven’t provided any evidence that it exists.

    But look, I’m not stupid. The idea that there is sampling bias is based on an assumption, and one that I would agree is perfectly reasonable, which is that lower income people probably have both lower propensities to own smartphones and use Strava. So I’m in agreement with you.

    But there is another reasonable and relevant assumption that goes against your argument, and that’s that people from lower incomes are less likely to regularly ride bikes in the first place. A person without a smartphone is less likely to own a car, bike, truck, etc. Like your own assumption about their willingness and ability to use Strava, this assumption seems pretty fair. And I think we all see this across the city. If you ride the LFT end-to-end regularly like I do, we all know that the traffic on it south of 31st Street is a fraction of what it is north of Fullerton. There’s no sense in denying what is easily observable to us all. Bike riding appears to simply be an activity more popular among higher income people, and I don’t think there’s anything so far that suggests otherwise, no do I think this is a controversial statement.

    I think the most likely explanation is that the Strava data’s flaw is that is AMPLIFIES a disparity which already exists in the first place. So where Strava might show that Lincoln Ave is ridden at a rate 50x vs MLK Drive (just to pick a number for illustrative purposes), in reality it might only be a rate of 20x. Again, amplified, but the main point is that disparity is still there.

    I think it’s pretty naive to think this data isn’t relevant. On a scale of 1-10, it’s probably a 5 or 6. Anyone who claims it’s scientific fact (9 or 10) or completely useless (0 or 1) obviously just has an agenda and is in denial. The notion that there are an EQUAL number of bikers on the south side who simply don’t have a smartphone app turned on is pretty far-fetched. The data isn’t proving anything but it’s definitely pointing us in the right direction.

  • LowSeason

    What “HEAVILY” used corridors is Strava missing? Please provide a basis for this claim.

  • It shows 26th lighter than northern Lincoln, which is nonsense, and Vincennes doesn’t show much either, when I know it has nearly as large a peloton as Milwaukee.

    Strava is only useful data in a positive sense: its very brightest, strongest routes are definitely heavily used. But dimness on Strava only means not many Strava users use it, not that it is not heavily cycled.

  • LowSeason

    “It shows 26th lighter than northern Lincoln, which is nonsense”.

    You’re doing it again. Please provide a basis for this claim.

  • Go sit next to 26th during rush hour and count the bikes. I don’t need to do your work for you.

    Also, every single suburban forest preserve is brighter/”more heavily used” than any of the heaviest used south-side bike commuter corridors, which is nonsense.

  • LowSeason

    Well, like I said, it’s not perfect, but it probably is pointing us in the right direction, and to scoff at it is an admission that you’re not interested in giving much thought to it, since such thought runs against your ideals.

  • LowSeason

    The other thing is, you really have no idea, so forgive me for rolling my eyes about the idea that you’ve literally sat on a sidewalk over a period of weeks and manually counted bikes on 26th Street vs Lincoln. So just stop because this is a comical statement. Does 26th Street get ZERO usage? No. Are there more riders on Lincoln? PROBABLY. That’s all…good grief.

  • dr

    I have voiced no opinions on probable bike disparities in Chicago. You’re projecting onto me a bias I don’t have.

    I have made precisely zero assumptions. I have only made the factual observation that the strava data is not relevant until proven so, and that assuming relevance is scientifically and intellectually imprudent and dangerous. You don’t seem to understand that non-relevance of data is the baseline, and not an “assumption” or “guess.” In the world of empirical evidence and scientific investigation it’s not “innocent until proven guilty,” it’s “guilty until proven innocent.” Data relevance is not determined through the discussion of competing hypotheses, related “reasonable assumptions,” or pattern-fit.

    Your entire post above demonstrates what you’re accusing everyone else of doing: assuming truthfulness based on prior convictions. Your blind acceptance of data-relevance is based on your preconceived notion that, as you say, “people from lower incomes are less likely to regularly ride bikes in the first place,” rather than any evidence of actual data-relevance. This is problematic, not only from the obvious standpoint of logic, but from the standpoint of available evidence.

    I encourage you to read the transportation department’s report on bicycling, excerpted here: https://www.transportation.gov/fastlane/who-bikes-work-america – and available in full via links from the site.

    The evidence we do have suggests that poor people commute by bicycle at the highest percentages, that young people commute by bicycle at the highest percentages, and that Hispanics commute by bicycle at the highest percentages. I have no idea how this translates to Chicago’s residents, but what it does suggest is that the basis for your priors is neither “reasonable” nor evidenced.

    In short, you’re assuming data-relevance on the basis of priors, which would be wrong, even if your priors were founded in evidence, which they are not.

  • It’s interesting — in the areas where it has solid uptake. Strong presence of Strava heatmap on a given route is probably indicative of strong cycling interest in that route (though given how heavily all the forest preserves are marked, not necessarily TRANSPORTATIONAL cycling). Absence of a line on the Strava heatmap is not, however, any evidence of absence of actual cycling.

  • “There is another reasonable and relevant assumption that goes against your argument, and that’s that people from lower incomes are less likely to regularly ride bikes.” Chicago’s predominantly Latino neighborhoods tend to be low-to-moderate income, and Latinos have the greatest rate of biking to work of any racial or ethnic group, according to the Census.

  • Interesting to hear that about Vincennes, which just got a new buffered lane on its northern stretch. Please share a photo of a large group of cyclists biking on it, if you come across one.

  • http://chi.streetsblog.org/2014/06/23/eyes-on-the-street-checking-out-new-bikeways-across-the-city/

    “I’ve got one more new buffered lane street to check out: 26th between Pulaski and Kostner… The lanes are getting good use from blue-collar workers coming home during the evening rush.”

  • LowSeason

    You’re wrong. Good day.

  • dr

    Amusingly, this is the most compelling argument you’ve made yet.

  • Let’s combine the Strava data with the Divvy data and see what happens.

  • That would be awesome. I wonder if someone could hack together a Strava-like heatmap (using the same assumptions about routing between stations as the other visualizers have made, probably) to aggregate a year of Divvy data?

ALSO ON STREETSBLOG

Chicago’s Bike Plan Is Inequitable, Says Report Based on Wrong Map

|
There’s great potential to use statistics and mapping technology to help ensure that bicycle resources are distributed equitably to people of all races and income levels. For example, Streetsblog’s Steven Vance and data scientist Eric Sherman recently recently worked with Slow Roll Chicago cofounder Oboi Reed and analyzed Census data to get a sense of how well […]

What Will It Take to Build a More Equitable Chicago Bike Network?

|
Earlier this month, the League of American bicyclists released a report with a method for using Census info and geographic information system data to measure how well bicycle networks serve communities that have the greatest need for better infrastructure. Using Chicago as the case study, the author concluded that the city’s “planned network” of new […]