Greater Greater Washington

Bicycling


What makes some CaBi stations more used than others?

Open trip data lets researchers analyze bike sharing systems in detail. They are making useful discoveries about how culture and urban spaces affect the way people use bikeshare. These conclusions can help cities refine their bikeshare systems as they grow and mature.


Expected monthly Capital Bikeshare ridership based on October 2011 usage.

My recently completed master's paper analyzes the factors behind the number of trips at different Capital Bikeshare stations. I created a regression of trips in October 2011 that began at stations in the District. After controlling for 14 variables, the analysis concludes that 5 key factors primarily determine a station's usage:

  • The population aged 20-39
  • The level of non-white population
  • The retail density, using alcohol licenses as a proxy
  • Whether Metrorail stations are nearby
  • The distance from the center of the CaBi system
I measured each variable based on what's within a ¼-mile walk of each station. With that information, I created a "suitability map," above. For any spot in DC, it projects how much ridership a station would get if DC placed one there. You can also download the KML file to view the analysis in Google Earth.

The map shows that as you get farther from the main activity centers in central DC, there's a dramatic drop-off in station demand. Approximately 13% of Capital Bikeshare stations, as of March 2012, are located in areas where we would expect fewer than 18 trips a day. The actual usage data shows that a significant number of these stations at the edge of the system have even fewer trips.

There are equity reasons to place stations outside the core; policymakers want to make sure that money spent on Capital Bikeshare benefits more than just those who live and work in central areas, and it builds political support from councilmembers representing wards farther away. However, there are multiple areas around the District that are under-served by bikeshare today, yet highly suitable under the analysis.

Planners and policymakers should consider these areas as they build out and tweak the system in the coming years. The figure below shows the coverage gaps by overlaying the existing bikeshare stations and the suitability map.


Suitability gaps in the Capital Bikeshare system.

What can we conclude from this? The Washington region and other cities should consider the following issues when they plan and expand their systems:

Distance from the center matters. This variable accounts for 60% of the variation among station usage, by far the most of any factor. This matches a principle known as the "gravity model" in transportation planning, which predicts more trips between closer locations. Capital Bikeshare's pricing structure also encourages shorter trips by charging for using a bike over 30 minutes, which strengthens this factor.

Carefully weigh goals of equity and coverage against ridership. It's very important to provide active, multi-modal transportation options to low-income and minority communities, and this study does not dispute that. That being said, it is important to carefully assess the tradeoffs among various objectives, especially in light of the relative costs of providing other mobility options for individuals of lower socioeconomic status.

Suburbanization of bikeshare has opportunities and pitfalls. The prospect of a region-wide bicycle sharing system in the nation's capital is an alluring one to advocates. It is easy to imagine a robust, polycentric system around dense nodes like Alexandria, Arlington, Bethesda, College Park, and Silver Spring.

However, some facts could temper that enthusiasm. Even some relatively close-in stations in the District have very low usage. Nearly 40 of the 97 stations in operation during October 2011 experienced 15 or fewer trips a day. Similarly, the densest parts of Arlington, with 18 stations during the same period, had 15% of stations system-wide but just 5% of trips.

To successfully expand bikeshare into the suburbs, planners need to choose station locations wisely, and elected officials need to invest enough to create a critical mass of stations early on. If we rush to build an inadequate suburban system, then it will likley not meet expectations and could act to blunt public support for the program, precluding a more economically sustainable system later on.

Stations can easily move as we learn more. Within a matter of hours, bikeshare operators can load stations on a truck and redistribute them to more suitable locations. While Capital Bikeshare operates year round, colder cities like Montreal and Boston take their stations away each winter. Planners there use the spring launch of the system to refine the location of their stations based on station performance the previous year. Capital Bikeshare should schedule an annual station redistribution.

Promote open bicycle sharing data. Having this data available to graduate students and anyone else promotes transparency, scholarship, and innovation. Bicycle sharing systems are proliferating rapidly, which is very encouraging, but few systems nationwide release trip data.

For instance, despite $4.5 million in grants from public sources ($3 million from the Federal Transit Administration), data from Boston's Hubway remains proprietary because of a private sponsorship agreement with New Balance. New York also hopes to fully fund its system with private dollars, which creates a danger that the same may happen there.

Like other North American cities, DC relied on international practices to plan its original system. Now, with an ample stream of data and more than $13 million in public funding committed to the regional system, it is time to strategically reassess station locations to ensure that bike sharing remains viable for the long term, as a true transportation investment.

David Daddio is a master's student in the Department of City and Regional Planning at the University of North Carolina, Chapel Hill. Originally from Columbia, Maryland, David founded Rethink College Park while an undergraduate at the University of Maryland. He is currently the Second-Year Editor of Carolina Planning, the oldest student-run planning publication in the country. 

Comments

Add a comment »

I'm guessing there was some problem in obtaining or analyzing data from the Arlington portion of the system?

by Michael Perkins on Apr 10, 2012 12:23 pm • linkreport

What were the correlation coefficients for each of the factors and their significance?

I would be very interested in the sign of the coefficient for non-white population.

by Michael Perkins on Apr 10, 2012 12:24 pm • linkreport

Can you clarify how each of the five variables impact ridership (positively or negatively)? In particular, I would imagine that ridership would grow with INCREASES in:
* Population 20--39
* Level of *white* population
* Retail density
* Proximity to metro
* *Proximity* to center

Did I get that right?

by xmal on Apr 10, 2012 12:29 pm • linkreport

yes, not including arlingon was a bit strange.

Confirms everything I've been saying abotu bikeshare EOTR.

I am pretty sure the contract with Alta says 1 free move. Then DC has to pay. Not sure about Arlington. There is obviously some nudge there. And I wonder if you can get more granular with moves (in terms of moving it catty corner, for instance)

by charlie on Apr 10, 2012 12:30 pm • linkreport

Was the proportion of white population included in the suitability map? If so, I think it should be removed as a factor. The location of CaBi station should not be determined by racial factors.

by Alan on Apr 10, 2012 12:42 pm • linkreport

Wow, this really fits in well with my own grad school research -- thanks! I'll take a closer look at the paper this week. In making future maps, though, please avoid using both red & green; those colors are difficult for colorblind people (4% of us) to distinguish.

(Maybe I should redirect that complaint to ESRI?)

by Payton on Apr 10, 2012 12:46 pm • linkreport

Sigh. First the author ignores Arlington. Then he notes that Arlington stations have relatively little use, while forgetting that the Arlington stations are at the edge of the system. But when you compare Arlington (geographically WOTR, or SOTR) with EOTR and then suddenly Arlington does not so bad.

Come on people. If you want to ignore a part of the system, fine. But be consequent and ignore it. Don't ignore it first and then still come up with speculation about it anyway.

by Jasper on Apr 10, 2012 12:49 pm • linkreport

Other than CC, I don't think you have enough data on Arlington yet. Really only this spring do you have a "complete" R-B system, and that is streching it.

by charlie on Apr 10, 2012 12:50 pm • linkreport

What about controling for having to cross a bridge to get to the center? I think that is a significant factor driving down usage in Anacostia and Arlington, rather than pure distance.

Bridges are very uninviting, especially for casual users and tourists that bikeshare caters to. It would be interesting to compare the same distance away from the center in northern DC and Maryland as compared to across the Anacostia and Potomac.

I think you could also look at highways and streets that require bikers to go on overpasses as barriers to usage. Crossing Arlington Boulevard or New York Ave is just as unpleasant or impossible as crossing some bridges.

by Tim on Apr 10, 2012 12:54 pm • linkreport

One of the more important things that you mentioned in your paper, but don't mention here is "Bicycle share planning should be highly customized to a specific geography."

by selxic on Apr 10, 2012 12:56 pm • linkreport

Did you take into account proximity to other stations? I understand your center of the city idea but feel it is remiss if not taken into account that is the densest area of stations. Some on the edge also have few stations nearby which could lessen use. Increased station density in any area could expand station use.

That said what is the threshold level of daily/monthly trips that makes a station successful?

by Sally on Apr 10, 2012 1:07 pm • linkreport

FWIW, I made another CaBi trip data animation, using the new 2012 data: Animating a Busy Day for Capital Bikeshare. It confirms the "center of gravity" you show (and also include Arlington data), on the busiest day thus far, Mar 23.

by M.V. Jantzen on Apr 10, 2012 1:08 pm • linkreport

Fascinating research. San Francisco is planning for a polycentric system along the Caltrain corridor, with each center to be physically seperate from the others. Your research would indicate that the suburban areas would see relatively little use compared to the stations in the City, and that expanding into the Bay Area's suburbs and smaller cities would be a poor idea.

by OctaviusIII on Apr 10, 2012 1:19 pm • linkreport

Having a lot of stations close together in outer areas will increase use? How? If density of people and destinations decreases by distance (which they more or less do), then why would more bike rental locations generate more traffic. Zipcar usage probably follows similar parameters although destinations would be less important than population density (which is probably a good proxy for car non-ownership).

Michael, 60% of the variance means correlations in the neighborhood of 0.75 to 0.80 ("variance accounted for" = correlation squared) which is very impressive for social/environmental research. Variance is a more conservative/useful metric than correlation coefficient; an impressive sounding correlation easily turns into an impressive amount of variance. The density/distance probably swamps just about everything else, including the white hipster factor which goes unquantified.

by Rich on Apr 10, 2012 1:30 pm • linkreport

I have a question about the analysis that is often run on CaBi trip data. I know it doesn't pertain much here since you're not analyzing individual trip data, but is it standard practice to combine trip data where it's obvious that a user checked a bike in and then immediately out in order to reset the 30 min clock?

There was a story here recently analyzing data on individual trips (how far people went, etc.) and it seemed clear to me that the trip length information did not take this into account. It seems like an important step to take, especially when analyzing trips from the outer stations.

by jyindc on Apr 10, 2012 1:37 pm • linkreport

What about places where there are multiple bikeshare stations in the same location? Does the study account for that, and if so, what impact does that have on ridership?

by Ben on Apr 10, 2012 1:41 pm • linkreport

I appreciate many of these productive comments. The larger 50-page masters project (http://bit.ly/DaddioMP) answers many of the questions here. As you might imagine, it's difficult to consolidate such a major undertaking into a blog post.

The results of this study are quite simple: People travel to get to places. The system is designed for short, less than 30 minute trips. DC's density of activity in closer-in neighborhoods far exceeds the density of activity on the periphery. The magnitude of activity in Arlington (and further out locales like Alexandria, Bethesda, College Park, Silver Spring) is much lower than DC proper. Based on this, I predict the planned suburban expansion of the system will produce dramatically lower ridership than observed at DC stations.

A couple points:
- Arlington was not included because a) the 14 original variables are not easily mapped in both Arlington and DC and b) Arlington makes up a tiny percentage of the total trips taken in any given day. Future research is needed as Arlington's system matures.
- The suitability map is designed to predict ridership, not relate equity goals of the program. The analysis shows that stations in census blocks with a higher percentage of non-white residents will experience lower ridership. It's a statistical fact.
- For the color blind, I suggest you look at Appendix B of the report when you get a chance

by David Daddio on Apr 10, 2012 1:42 pm • linkreport

@OctaviusIII
A very important factor in how people travel via bikeshare is the 30 minute time limit. This model may suggest adopting higher time limits in more disperse systems rather than predicting the lack of viability of such systems.

by Lucre on Apr 10, 2012 1:45 pm • linkreport

The only problem with extrapolating from the drawn conclusions to 'what works' is that it omits potential solutions to what 'may work' once the right set of circumstances are in place. I.e. one could conclude from the map that bikeshare doesn't work along the MD border, but that's dependent on the fact that there are no bike share destinations (lockups) on the other side of the border. I mention this as merely an example of the limitations on market research focused on 'today' and extrapolating to 'tomorrow' without considering the 'why'.

There was no market for a $600 cell phone that could go on line until it existed; now everyone has an Iphone. There may be no market for bike share in Anacostia (the region) until other issues are solved, but then suddenly there's a market.

Maybe it's a failure of understanding causation and simply limiting the data to correlation. I leave it to the author to consider this limitation.

by Name on Apr 10, 2012 1:45 pm • linkreport

@ Name: There is a huge difference between correlation and causation. Unfortunately, too many people skip correlation and dive straight into causation. This is a poor practice, because without correlation, you can never get to any causation. In fact, because the two are so easy to confuse, it's better separate them.

by Jasper on Apr 10, 2012 2:07 pm • linkreport

Similar question, some different conclusions...
http://amonline.trb.org/pap@PaperNo=12-3539

Haven't had a chance to read in detail, but didn't see mention of cross correlation of independent variables? Seems that many variables among both your selected and rejected are spatially correlated. Important qualification, esp in regard to settlement patterns of different ethnicities. And distortion of non-white population in very center of city, where resident population overall is low.

by darren on Apr 10, 2012 2:09 pm • linkreport

I have to read the paper, but I think commenters are making a mistake in how they are considering the purpose of the Arlington part of the "system" as primarily to be used to ride into DC. Think of DC as a node and Arlington as a node.

Arlington has to have a set of characteristics for station use that work for Arlington. The number of cross-jurisdictional trips are minimal comparatively speaking--they do occur, and especially they do between Rosslyn and Georgetown--but cross-jurisdictional rides aren't the point of the system, which is to serve trips in Arlington.

Similarly, the point of expanding to other suburbs is to develop bike sharing in the locations where the various factors (a la the Montgomery County heatmap -- http://greatergreaterwashington.org/post/11852/montgomery-planners-work-to-find-biking-hot-spots/ ) support the placement and development of bikesharing systems. (Note that I think the Rockville initiative will fail, but it's because it doesn't have the right characteristics of density, activity centers, proximity, etc. that support bike sharing success generally.)

I haven't gotten around to writing about going to the Arlington County Bikeshare planning meeting a couple weeks ago, which was lightly attended, but good regardless. Clearly, they know what factors shape use and they are working to develop the station footprint that supports use. Their biggest limitation is funding. They are developing some good materials for promotion (the kinds of things I suggested before I got into the business, in the paper "Ideas for Making Cycling Irresistible in DC").

One of my problems with bike sharing research in the trade--not student theses like this one--is that no one wants to be critical, so they define everything as a success, and as a result we don't learn useful stuff that can be used to make better systems going forward.

E.g., someone mentioned the SF system, and we intended to bid on it, but then they changed certain requirements at the last minute that we knew we couldn't meet, and while it's right to think of it as one system, it's really 5 subsystems. Each participating city is far enough from the others that they need to be planned and managed as single units, but where members can use the system in multiple communities.

by Richard Layman on Apr 10, 2012 2:21 pm • linkreport

The analysis shows that stations in census blocks with a higher percentage of non-white residents will experience lower ridership. It's a statistical fact.

To Jasper's point, let's separate correlation from causation. It seems highly unlikely that race is the cause of differences in bikeshare usage unless you believe that the level of pigmentation in someone's skin influence's bike usage. More likely, is that race is correlated to other factors that can be causally linked to bikeshare usage, such as income, education, age, or physical fitness.

The book The Bell Curve makes the same kind of statistical mistake.

by Falls Church on Apr 10, 2012 2:41 pm • linkreport

I'd also point out that the analysis was done on the basis of October 2011 trips. This was before the Mall stations were put into place which are some of the most highly used in the system but only rank highly on 1 of the 5 criteria (proximity to metro stations).

by Falls Church on Apr 10, 2012 2:46 pm • linkreport

FWIW, I agree with Falls Church's point that " More likely, is that race is correlated to other factors that can be causally linked to bikeshare usage, such as income, education, age, or physical fitness."

To that I would also add neighborhood proximity to the core of the city/activity centers, etc.

by Richard Layman on Apr 10, 2012 2:47 pm • linkreport

@Falls Church, and one of the stations (by the FDR/MLK memorials) isn't even close to metro, although I think Daddio's metrics use census tracts, which adds to the predicted usage of stations on the Mall (though not nearly the amount of usage they've seen).

by Jacques on Apr 10, 2012 2:55 pm • linkreport

> The suitability map is designed to predict ridership, not relate equity goals of the program. The analysis shows that stations in census blocks with a higher percentage of non-white residents will experience lower ridership. It's a statistical fact.

That may be a statistical fact, but the article implies that the map was created as a tool to drive policy. Thus by including race as a factor, you have decided that non-white neighborhood's should get fewer CaBi stations because they are non-white. The fact that you do not see this as a problem, and furthermore, are callous about the implications, is very troubling. Scientists need to be more than just automatons when it comes to discussing the implications of their research. If research showed that non-white's were less likely to use the library, would that mean we should build fewer libraries in non-white neighborhoods, thereby depriving the residents of non-white neighborhood the benefit of equal government services? I'll say it again just so you understand: race should be removed as a factor in any tool that can be used, directly or indirectly, to decide where CaBi stations should be built. Until it is removed, your study is fundamentally flawed.

by Alan on Apr 10, 2012 3:10 pm • linkreport

I think the dependent variable is problematic here.

Imagine if each station had one person leave in the morning, and all riders rode to a station downtown. Now reverse this for the evening commute. Your model wouldn't give us any useful information about how to place stations, but it would probably give you a similar regression printout with a large coefficient for being near the downtown station.

by Michael Hamilton on Apr 10, 2012 3:18 pm • linkreport

@Alan -- in bike and pedestrian planning circles, I argue that the next stage in bike planning needs to be (1) integrating infrastructure and programming plans; and (2) taking sustainable transportation planning down to sub-city levels, such as by wards/neighborhoods.

Anyway, if people don't use the library, you can either say, people don't use the library, or you can research on why people don't use the library, do a factor analysis, and respond.

The same goes for biking and walking (and transit). Some of it is spatial, some of it is access (owning a bike, which is impacted by income) but a lot of it has to do with providing the kind of programming-training-assistance that people need to routinize new behaviors, and they especially need support when the behavior is seen as non-standard, which biking is.

I'm slowly writing an entry about a couple of best practice initiatives on promoting biking as transportation within people of color communities in Manhattan and Portland, which discusses this general point.

The plan I did in Baltimore County (which hopefully will be approved in a couple weeks) has a serious set of integrated recommendations on soft infrastructure (programming) of which I am very proud.

by Richard Layman on Apr 10, 2012 3:20 pm • linkreport

"Capital Bikeshare should schedule an annual station redistribution."

I think this has to be done very cautiously. Bikeshare stations have become a part of the transit network and have thus begun to generate their own small levels of economic development. Like Metro stations, people and businesses can be attracted to an area based on the availability of nearby transit service. More and more home listings are featuring nearby bikeshare stations. If the location of bikeshare stations suddenly becomes variable, then that development potential may decrease as well. While economic development may not have been an original intent of the system, it is now a very important factor that should also guide the placement of bikeshare stations.

by Adam L on Apr 10, 2012 3:25 pm • linkreport

"Capital Bikeshare should schedule an annual station redistribution."

I could see moving stations that don't have many rides. But just moving stations around to try to optimize things is silly. Especially when you should probably be aiming for a more dense system than currently exists. Just put more stations in period.

by MLD on Apr 10, 2012 3:40 pm • linkreport

I agree with the comments on correlation vs. causation as well as spatial autocorrelation. Again, I would encourage folks with methodological questions to refer to the full 50-page report rather than rely on a 1,000 word blog posts written to a general audience.

No study is methodologically flawless. That said, this is about as robust as any public examination of Capital Bikeshare usage (of the data that has been released).

I'll emphasize again. The suitability map is designed to predict ridership if a station is placed in any given part of the city. There is no other ulterior motive. Actual ridership will vary considerably around the means presented.

This map is designed to inform policy, not "drive" policy. This blog posts and the larger report acknowledge that ridership maximization is not and should not be the paramount objective of any bicycle sharing system (or any transportation investment for that matter).

by David Daddio on Apr 10, 2012 3:42 pm • linkreport

I think it's a useful and worthwhile analysis, but the problem is not knowing where to place new stations. DDOT have already identified new locations for stations...they were supposed to go in "fall of 2011", but we're still waiting...

by renegade09 on Apr 10, 2012 3:46 pm • linkreport

@David Daddio,

Are you sure the distribution of liquor licenses matches the distribution of all retail? In any case, relying on this proxy seems unnecessary. A full set of registered properties with use codes from OTR is available from data.dc.gov, and these use codes include a wider set of classifications that ought to give you a better picture of where retail is located.

by Dan on Apr 10, 2012 3:46 pm • linkreport

@David Daddio -- by dropping out all of your independent variables of lesser significance all at once, you've drawn a correlation to one variable (non-white pop %) that parsimoniously captures a bunch of other variables with predictive value that cross-correlate (the inequitable dispersion of bike lanes in the city, the distribution of HH income, elevation changes in/out of the CBD, total household trips, total population within 1/x mile of a station, etc). Your full paper does not include data on the cross-correlation of your independent variables -- without this, the reader unfamiliar with DC is left wihtout this very necessary context, and by not eliminating your other independent variables stepwise, some of these issues of cross-correlation might be lost.

by darren on Apr 10, 2012 4:04 pm • linkreport

I would submit that if there were a density of stations in Chevy Chase, DC, Friendship Heights, the central "Ten-Frien" area and locations in Southern, central and western Bethesda, that such a "node" would be highly utilized, irrespective of trips to downtown DC.

by Andrew on Apr 10, 2012 4:10 pm • linkreport

Really no suprise but i hope that if they use this information that they remember what other effects cause a reduction in ridership. For example construction, the pennbranch cabi site was removed for some part of last year and the roads around it where often in flux.

by mp on Apr 10, 2012 4:35 pm • linkreport

Wow.

Pick a variable other than race. It is wildly inappropriate to argue for more of a government benefit based on race. While it does appear true white people in DC are more likely to use the system than black people it is still not appropriate for the the government to use that information to help decide where to put more bikes. There are plenty of other variables to use.

I'm pretty shocked GGW published something like this.

by mike on Apr 10, 2012 4:55 pm • linkreport

I don't see any suggestion that race should be considered in placement of stations, just that it may be a factor in use of a station once its placed.

by Canaan on Apr 10, 2012 5:05 pm • linkreport

Looking at the tables, race and (especially) age seem to make negligible contributions to prediction, controlling for everything else. Otherwise its location, location, location (being central and near a Metro). A lot of the comments seem to be wishful thinking, as if greater DC will become Amsterdam at some point, which is pretty unlikely. CaBi is most practical in places that are dense and destination rich and that seems to be reflected in the findings. The pricing is set-up to favor short-trips which probably constrains this further and people use it in conjunction with Metro--probably as an alternative to waiting for a bus and as a quicker mode than walking. If those conditions are true, then places like Riggs Park, Chevy Chase and Colonial Village are unlikely to become the next centers of CaBi.

by Rich on Apr 10, 2012 5:06 pm • linkreport

@David Daddio-congratulations on your degree and on completing an ambitious project. You are brave to put it up here for the critical eye of a bunch of other people who are knowledgeable about data analysis and are not your advisor.

I have not read your manuscript. I may be talking through my hat. I just wanted to echo @Falls Church and @Richard Layman that it is better to use SES variables rather than ethnicity/race as a predictor for a behavior like bike riding. Race, as we think of it, is a biologically determined genetic component that is not modifiable. Bike riding is a modifiable behavior. Everything about SES and behavior influenced by SES is modifiable. There's no biological plausibility in the association between bike riding and race/ethnicity.

I'm sorry to pile on. Congratulations again on a great project.

by Tina on Apr 10, 2012 5:29 pm • linkreport

One word ... hills!!!!! Ward 8 has alot. The center of the city's tourist sites has few. Columbia Heights (sightly away from the city center) is truely a height with a huge hill on 14th street. Where does that figure into your thoughts?

by tour guide on Apr 10, 2012 6:42 pm • linkreport

As a researcher, my responsibility is to present my findings. I sincerely appreciate the interest and engagement. I'm also glad to see that about 110 people have clicked through to read the paper.

The table on page 20 indicates that the study controls for age, non-white population prevalence, and income. This self-identified data all comes from the 2010 census. It's also largely in line with the demographic findings of a user survey conducted by Virginia Tech grad students.

I agree that there are more sophisticated ways to analyze and model the data (indeed this is not a phd dissertation), but Table A4 on page 33 indicates that (holding units constant) distance from the center of the system has far and away the most influence on the model. Then age followed by non-white population prevalence.

Distance from the center of the system captures many of the 14 variables (which @darren identifies)... which by itself explains 60% of the variation in station demand. With the other variables, this model explains about 80% of the variation in activity between stations.

by David Daddio on Apr 10, 2012 6:52 pm • linkreport

@tour guide -- I rode to the Anacostia Community Museum once, it was a b****. The hills to get up Morris Street are really tough. Plus, I felt uncomfortable riding through there at night, especially as I made some wrong turns to get over to Anacostia as it was (I missed the turn to the commercial district and got to Firth Sterling before I realized my mistake, although it was at night).

It's also an issue as you go north of the Fall Line (around Clifton Street in NW).

by Richard Layman on Apr 10, 2012 8:47 pm • linkreport

I'm pretty shocked GGW published something like this.

Yes. If only there were more censorship. Then we wouldn't have to think about things that make us uncomfortable.

by David C on Apr 10, 2012 10:27 pm • linkreport

tour guide, you make an excellent point about hills, and DC was chosen because it sits in a bowl of hills that protects it from invasion by dirty Canadians who seek to steal our plentiful beaver pelts, so it is not a trivial factor. David D's study didn't look into that, but I think topography could be included in his list of "for future study" elements.

by David C on Apr 10, 2012 10:32 pm • linkreport

I'm pretty shocked GGW published something like this.

It's really no different than the "White People Walk Faster" study that City Paper reviewed here:

http://www.washingtoncitypaper.com/blogs/housingcomplex/2012/01/23/study-districts-white-people-walk-more-faster-than-non-whites/

And that was linked to at GGW, DCist, and most other DC news blogs:

http://greatergreaterwashington.org/post/13462/breakfast-links-speedwalking/comp/13461/

That said, I ridiculed that study in the GGW comments section because it's preposterous.

by Falls Church on Apr 10, 2012 11:01 pm • linkreport

This study controls for age, race, income, and distance from the city center. The study referenced in the "White People Walk Faster" City Paper post doesn't. How are the two "no different"?

Also, do you attribute the dramatic differences in the racial makeup of CaBi users completely to income, age, destination accessibility, and infrastructure?:
http://ralphbu.files.wordpress.com/2012/01/vt-bike-share-study-final3.pdf

This study appears to show that DDOT has much more work to do in marketing CaBi to different DC communities.

by No different? on Apr 11, 2012 10:14 am • linkreport

@No different?,

that city paper study (which i haven't read, and probably won't) sounds like an observational study, which you can't introduce such control variables for, and I would hope is appropriately qualified.

There's an important difference in that the aims of those two studies are different -- one appears to explicitly be aiming to identify differences in people's physical activity, the other is to inform decision makers on placement of transportation infrastructure.

There's a fair amount of anecdotal evidence of a divide in Capital Bikeshare usage (and bicycling more generally). But when we try to attach explanatory significance to who people are and how they behave, and especially when we start discussing how that relates to how public assets are deployed, a closer examination and consideration of variables that underlie the observed behavior is necessary. In other words, and this is my main quibble with the paper, it's fine to establish a correlation between race and CaBi usage, but to do so with improper treatment of other independent variables (both troubles with the treatment of those included in the study that i noted in a prior comment, and those not included in the study at all) that may or may not impact the disparate ridership does a disservice to the people being studied. Establishing a statistical correlation isn't good enough, when talking about unchangable differences in people about which we have no reasonable basis to posit a causal impact on the outcome of the study.

Focused study is needed of the conditions that encourage and discourage bicycling in general (and bikesharing in particular) EOTR, but this requires stand-alone study, with data that DDOT either hasn't released or doesn't exist (open data plug!!)

The VT intercept survey study is a bad comparison in this debate - that study sampled a small subset (300 or so) of users, in downtown tourist stations, only 24 hour renters. The demographics of bikeshare users that meet that very narrow definition are never going to resemble the demographics of the broader city residency.

by darren on Apr 11, 2012 10:46 am • linkreport

It seems highly unlikely that race is the cause of differences in bikeshare usage unless you believe that the level of pigmentation in someone's skin influence's bike usage.

"Level of pigmentation" is being deliberately obtuse. However, using race as a cultural marker could very well point to differences in bikeshare usage.

by Juanita de Talmas on Apr 11, 2012 11:42 am • linkreport

@Darren, appreciate your comments. The study recognizes that the analysis measures association, not causation. In my experience, the distinction is lost upon the general public (much like it was lost upon GGW editorial staff).

In the course of the study, topography did not strike me as a substantial factor for DC although it certainly is a major consideration for cities like San Francisco and Seattle that are planning systems. Many variables, including employment and topography, underlie (and are accounted for by) the distance from system center variable. Distance from the system center does much more to explain ridership than race, retail density, or metrorail stations. Bike infrastructure (lanes+trails) was significant in my original unadjusted regression.

The study presents a methodology to develop and refine a tool. In the case of the tool I've developed from that methodology, the non-white population prevalence variable is one small piece. It certainly masks other variables that could be teased out. That said, I do not present the suitability map as a definitive tool to predict station demand nor decide station placement.

Also, the analysis comes out of one snapshot in time. It assumes preferences and (as a couple commenters smartly pointed out) the system pricing structure and implementation are static. Both are changeable.

I certainly encourage other researchers to take my methodology and data visualization techniques and apply more sophisticated statistical approaches. This open data is badly in need of more non-proprietary analysis.

by David Daddio on Apr 11, 2012 12:06 pm • linkreport

that city paper study (which i haven't read, and probably won't) sounds like an observational study, which you can't introduce such control variables for, and I would hope is appropriately qualified.

It was not an observational study. It was based on NHTS data.

that city paper study (which i haven't read, and probably won't)

It wasn't a City Paper study. It was a study (from the Institute of Transportation Studies, UC-Davis) that City Paper wrote an article about. I'd suggest at least reading the abstract before coming to conclusions:

http://amonline.trb.org/1sblln/1

by Falls Church on Apr 11, 2012 12:47 pm • linkreport

@Juanita de Talmas - However, using race as a cultural marker .... but Margaret Mead showed us a long time ago that race, as a genetic biological aspect, has nothing to do with adaptation of cultural norms. Thats what wrong with using race as a measure of culture. Its like using ear wax consistency as a measure of culture. (Some genetic racial groups have different ear wax consistentcies than others). Define culture. Yes, its more difficult. So what?

@D.D. the non-white population prevalence variable is one small piece. It certainly masks other variables that could be teased out.

Its more difficult but worth doing. Again, there is no biological plausibility in the association of bike riding and race. Even using the word "ethnicity" would be an improvement b/c the conotation of "ethnicity" is more about culture than genetics whereas "race" has the usage history of being more about genetics than culture.

by Tina on Apr 11, 2012 1:30 pm • linkreport

@ David Daddio; I think it is shame you didn't include Arlington, because this is tying into the debate on how to expand the system there. If I read your conclusions correctly, adding density to the system might be more useful than connecting R/B and CC.

by charlie on Apr 11, 2012 1:47 pm • linkreport

"Define culture. Yes, its more difficult. So what? "

So if you've got a limited research budget, it matters. Census and other data has info on "race" and not on meaningful cultural variables, aside from "hispanic". And if you are asking in your own survey - what are you going to do, ask "do you participate in the culture that originated among slaves in the American South" or are you going to ask "Are you black?"

Clearly race is a cultural construct, and the census race categories dont map to any particular pure "race". That said, they are useful in analyzing and predicting behavior, and spending more than half your research budget to avoid use of them, or simply relabeling them, is silly (and if you are collecting your own data, and ask someone what culture they are, and give white, black, asian, etc as choices, you will likely simply confuse the respondents"

by SocSciDatauser on Apr 11, 2012 1:48 pm • linkreport

charlie -- I don't know if you went to the Arlington Bike sharing master plan meeting. (Since I don't know what you look like...) I haven't uploaded my photos from the session (and some are a bit blurry) but I think it's fair to say that they understand that density of stations in the R-B corridor is more important than connecting to Crystal City. However, they are stuck with also having to provide more access throughout the county. (I also suggested they consider the part of Alexandria/Potomac Yard abutting Arlington as an issue where Arlington County interests justify their making recommendations, even though it's technically outside their jurisdiction.)

by Richard Layman on Apr 11, 2012 1:53 pm • linkreport

@SocScidatauser- the census collects education data in addition to income/poverty/home ownership and employment data. I agree with much of your comment. I disagree that relabeling "race" to "ethnicity" is silly and will confuse people. I think thats kind of condescending. I've worked on studies that use ethnicity instead of race in original data collection. I'm not aware of any participants expressing confusion. I would hope that anyone designing their own study would include more than "race" if the aim is to find patterns in behavior.

Biologically race is not a cultural construct. In the human species it is. It is also a fraught term and thus its worth using a different term to make clear one is trying to find out something about culture/education/etc and not biological destiny.

The author of this study indicates that race as a predictor would be washed out if other variables were used. Its those other variables that are interesting. Your comment is an indication that this is not a simple matter, or silly, contrary to the content of your comment.

by Tina on Apr 11, 2012 2:14 pm • linkreport

Hi David,

Great work! I am wondering, (how) did you take into account people who used more than two "trips" to reach a destination? I used two legs often for fear of going over the 30 minutes, so that to get from point A to B, I am using two <30 minute trips consecutively. Does your source allow you to consider this in your data? I feel like suburbs would get more credit if so. My commute from Crystal City to Dupont ends up looking like a commute from Crystal City to downtown, and then another from downtown to Dupont, but its still just me going from one place to another. Soon it'll be Alexandria to Dupont using 3 trips. Just curious if they give you enough ID info to take that into account.

Cheers,
Bryan

by Bryan on Apr 11, 2012 2:21 pm • linkreport

David,

Very interesting, I would what it would look like if you could take into account the usable paths and bike lanes in DC. I would be really interested to see how Arlington ends up looking once there is more data. I took a quick peek on why you only went with DC. I have personal experience with the lack of bike paths in DC and it definitely effects your commute, going over some of the bridges is next to impossible but I am sure cross county bike share would perk up. I am not sure about redistribution of stations as I remember reading Anacostica received stations specifically for a get moving initiative, I would like to see studies like this be use in more bike lanes and reduce barriers to entry, when you move closer to the burbs biking just isn't safe with the car culture that lives there. Though having mobile stations put up, such as now with the festival would be fantastic.

by teganann on Apr 11, 2012 2:56 pm • linkreport

NYC will shortly announce the 600 locations for some 10k shared-bikes. One detail the NYCDOT promulgates over-and-over is that the cost is fully-funded by Alta. Thanks for your subsidy to NYC riders.

by tom murphy on Apr 11, 2012 10:54 pm • linkreport

No, Alta has performed above contract spec due to the viz of CaBi, and likely taken a loss. NYC is 100% private solely because of the revenue from sponsorships and advertising will offset the capital outlay. Operational benefits of scale will flow to other Alta-operated systems as a result of the huge NYC system coming online. So really, thanks for being the valuable advert targets that you are, thus providing a subsidy to DC riders.

by darren on Apr 11, 2012 11:04 pm • linkreport

"I would submit that if there were a density of stations in Chevy Chase, DC, Friendship Heights, the central "Ten-Frien" area and locations in Southern, central and western Bethesda, that such a "node" would be highly utilized, irrespective of trips to downtown DC."

@Andrew

I totally agree with you on this. The other day I was just commenting that if there were a station at Nebraska and Connecticut and more in the Bethesda area, I would use them every day to travel to the Van Ness station and to run errands. There is Comet and Politics and Prose there, so I think that ridership would be consistent with people riding over from Tenleytown and Van Ness stations.

I live on Connecticut and there are plenty of times when the L1 or L2 is taking too long and I wish I had easy access to the bikeshare. I often wonder if they thought of the fact that everyone who lives in DC isn't travelling toward downtown, they're also travelling within their own neighborhoods to run errands as many neighborhoods are practically self-contained especially the Chevy Chase DC and "Ten-Fren" area.

by Alycia on Apr 12, 2012 9:29 am • linkreport

did you account for density of workers? If you look at frequency of trips during the day it's usually around commute time.

by roygbiv on Apr 12, 2012 9:54 am • linkreport

David .. Very interesting article. Curious if you did any primary research/interviews with individual riders about their experience, love, frustration with CapBi?

I co-founded a company - weBike - that provides station-less bike sharing for college and small communities; I am always intrigued to get more data out of CapBi's operations and listen to user feedback. Would love to hear more about what you learned!

by Allie on Apr 17, 2012 11:29 pm • linkreport

Add a Comment

Name: (will be displayed on the comments page)

Email: (must be your real address, but will be kept private)

URL: (optional, will be displayed)

Your comment:

By submitting a comment, you agree to abide by our comment policy.
Notify me of followup comments via email. (You can also subscribe without commenting.)
Save my name and email address on this computer so I don't have to enter it next time, and so I don't have to answer the anti-spam map challenge question in the future.

or

Support Us

How can our region be greater?

DC Maryland Virginia Arlington Alexandria Montgomery Prince George's Fairfax Charles Prince William Loudoun Howard Anne Arundel Frederick Tysons Corner Baltimore Falls Church Fairfax City
CC BY-NC