There is likely no single ideal density for bikeshare success and equity

NACTO has a "Practitioner Paper" out arguing that the ideal density for bikeshare stations is 28 per square mile. They argue that such a density will create the most successful and equitable system. But there are several problems with the analysis, starting with the definition of success and including a confusion of the terms "convenient" and "equitable", as well as the odd decisions to ignore population density and selectively ignore the impact of multi-jurisdictional systems.

How do we define success for bikeshare?

NACTO uses the metric "trips per bike per day" as the sole definition of success. I've used the term but I by no means think this is a way to measure success. It's useful for seeing how a system's use is growing, but that's about it. No city leader was saying that they would fund bike share "but only if bikes are used many times every day." 

The reasons for adding bike share are related to use, but a high number of trips per bike is not the goal. And how something achieves its goals should be the primary measure of success. A good place to look at how DC area leaders saw the goals of bikeshare is the region's 2010 Tiger II application for bikeshare expansion funding. 

The benefits they foresaw were user cost savings, user travel time savings, increased access, congestion reduction, emissions reduction, healthcare cost savings and accident reduction. They also saw benefits in getting users to purchase and use their own bikes - something that just doesn't show up in tpbpd. 

Considering these benefits, not every trip is equal. In systems with identical tpbpd which is more successful, a system that replaces mostly walking and other biking trips or one that replaces mostly driving and taxi trips? Certainly it's not unreasonable to think that a tightly packed system replaces more walking trips than a more sprawling system. 

There probably isn't one metric for measuring success. In fact tpbpd isn't even necessarily the best metric for defining use. I'd argue that miles per bike per day is - it has the added benefit of giving a direct relationship to health benefits too. But use is only one part of the equation. To measure success we'd also need to know trip replacement, initial cost, subsidy per unit of use, emissions reduction, trips where users would have stayed home instead, how much it encouraged users to bike with their own bikes, increase/decrease in transit use, reduced VMT, etc...

This analysis ignores the nuances of the other benefits. In fact by their definition, CaBi is less successful than Divvy and Citi-bike, despite Citi-bike's funding issues and Divvy's smaller use. By this logic, CaBi would be more successful if it took the stations east of the river - which have low ridership - and dumped them into the Anacostia. That should raise red flags. 

In contrast to NACTO, I would argue that any city maximizing tpbpd is leaving money on the table.

If each trip has an average benefit value to the city of, for example, $1 and each bike has a subsidy of $0.99 day, than the goal should be to maximize trips while adding stations that average at least 1 tpbpd. That means expanding the system more and more into lower use areas until all of the benefit has been captured. The benefit above cost is the "profit", and the goal is to maximize "profit".

Convenient does not mean Equitable

On the subject of an equitable system, NACTO states that

Like everyone, low-income people want convenient travel options that work for the trips they are trying to make. Bike share systems can provide this if they maintain a high station density (approximately 28 stations per square mile) across all neighborhoods in the service area, including low-income areas.

No doubt higher station density in low-income areas would be better for low-income people. But we live in a world of limited resources. Adding stations to one area to create high station density would mean removing them from somewhere else. Likely an effort to create 28-station-per-square-mile density would mean removing the worst performing stations - those in poorer neighborhoods - to create high density elsewhere. Would that be more equitable? According to NACTO, yes.

Station density and service is often worse in low-income areas. New York’s Citi Bike is an exception: while Citi Bike does not cover the whole city, the areas that are served are evenly covered with stations. As a result, bike share in New York is equally convenient throughout the program.

So, having high density in wealthy parts of the city (see the map of NYC on page 4) - lower Manhattan and Brooklyn - and none in low-income neighborhoods like Harlem, Queens and Staten Island is more equitable than spreading them out. Got it.

Conflating convenience with equity misses the point that the two are actually in conflict. Something they concede elsewhere.

systems that have lower station density in low-income neighborhoods often exacerbate equity issues as stations are too far apart to provide a real transportation option for low-income riders. Such systems attempt to achieve nominally or geographically equitable bike share coverage at the expense of service quality and utility

Right. For a given cost you can increase equity and coverage, but that comes at the expense of lower use. Or you can increase use, but that means covering a smaller service area. I find it hard to believe that the poor are better off with no bikeshare stations where they live than with only 9 per square mile.

In an ideal world, cities would have the money to add as many stations as an area needs, but they don't. Choices have to be made. Trade-offs have to be accepted. The reason that NYC's system looks different is that it is a privately-owned business, and businesses don't seek to create equity, or mitigate congestion, or improve public health, or reduce pollution. Businesses look to make money. And making money means maximizing tpbpd. Publicly-run systems have different goals and will look and operate differently as a result. In NYC, positive exteranlities are a side effect. In DC, they're the main point. So it's odd to hold NYC up as a symbol of equity.

The way to improve equity is to spend more money to add stations to low-income areas while finding ways to help low-income customers to use the system without credit cards or large upfront costs.

Station density follows population and activity density, which is likely the larger driver of use

The NACTO paper shows plots from each city showing rides per day per dock plotted against station density at each station. That certainly seems to hold up the premise that higher station density leads to higher utility. But there are two errors with this.

First of all, station density is driven by population/activity density (the old "correlation does not equal causation"). Designers place stations where there are a lot of people and activity centers, and these are exactly the places where we'd expect to have the most use. Dupont Circle gets a lot of users because a lot of people want to go there or leave there; not because there are so many stations nearby. We could bump Rock Creek park up to 28 stations per square mile, but it probably wouldn't increase use that much. I suspect that if we redid these plots, but replaced station density with population density, they would look extraordinarily similar.

The other is that the curve they chose seems to imply that further density will lead to even more rides per dock per day, with growth being exponential. So if we moved the entire CaBi system into a circle with 1/4 mile radius, ridership would go up to several thousand a day - which is ridiculous. Even if we moved the DC system to the ideal 28 stations per square mile (an ideal which these plots appear to undermine, I'll add), it's unclear that this would increase ridership, since utility has two parts - how easy is it to get a bike and return it and where can you go. Densifying the system would cause one of these to go up and the other to go down. 

CaBi, like San Francisco, isn't really one system

One thing that lowers CaBi's tpbpd is the fact that the core is surrounded by lower use "pods" in Montgomery County, east DC, and in Arlington and Alexandria. Again, that the system serves these areas is not a sign of failure. If we look at just DC by itself, the tpbpd goes up to 6.1, which would place DC at the top of the heap. If we looked at the part of CaBi just in the L'Enfant city, it would likely be even higher.

For the paper, NACTO only looked at the San Francisco portion of the Bay Area Bike Share, a similar consideration for CaBi seems appropriate, especially since each jurisdiction is responsible for placing it's own stations. There is no CaBi overlord looking to maximize tpbpd over the whole system and Montgomery County is uninterested in moving one of its stations to Arlington even if it would increase the system's average utility. But this is irrelevant since the rest of the premise is flawed anyway. 

28 stations for square mile is BS

I've been pushing back on this notion that there is one ideal station density that has to be replicated world-wide ever since people started holding up Paris' station density as the ideal.

The 28 stations every square mile "industry standard" comes from a study done by APUR, the Paris Urban Planning Agency in 2006. In that they set the guideline for station placement at one every 300 meters, which several people have translated as 28 per square mile. (ITDP places it at 36 per square mile, but i'll leave that math to someone else). Regardless, that guideline was created before Velib started operation in 2007.

Now I'm sure there are some extraordinarily smart people in APUR, but the idea that the guideline they designed before they'd installed even one bike share station, and based on no comparable experience from anywhere else, is the ideal one is a bit hard to swallow. I can't believe that with the benefit of nearly 8 years experience in multiple cities, NACTO can't do better.  The original guideline was a guestimate, for a very different city starting with a much larger system. Would 320 meters have worked better? 230? Has anyone bothered to find out? And does what work for Paris - which has a unique urban design - work for New York City with it's skyscrapers, or DC with its much lower population density? Some planners are very good, but rarely does anyone hit the bulls-eye on the first try. And rarely does someone find not only their ideal solution but everyone else's - despite different inputs. Is Paris even still using this number?

DC has two large rivers and a lot of off-limit federal land. It will never be able to achieve the station density of lower Manhattan or of Paris. Nor should it try to. The ideal station density is different for CaBi than it is for Chicago, different for DC than it is for Arlington and different for Ward 1 than it is for Ward 5.

Comments

[darren]

This is so spot-on, i'd put it on office letterhead if i could. All of those graphs could easily swap out density, centrality, or presence of bike facilities on the horizontal axis. Lots of things spatially co-vary with where planners plunk down more/less dots.

[darren]

i'm also eager to see how the 300m rule holds up in Staten Island.

[Crickey7]

You had me at "maximizing tpbpd".

[Alex B.]

"Some planners are very good, but rarely does anyone hit the bulls-eye on the first try. And rarely does someone find not only their ideal solution but everyone else's - despite different inputs."

Yep. Not only is it rare to 'hit the bullseye' on the first try, there also needs to be the realization that there are multiple bullseyes out there to be hit.

[Greenbelt]

I love the comment about how a made up estimate done before any experience somehow becomes a standard.

Sort of like with parking minimums. I mean you can go out and see the empty parking lots, day after day, hour after hour, but somehow the parking minimum requirement never changes to reflect what our eyes can plainly see.

[washcycle]

Greenbelt, that's a great analogy. I was trying to think of one that wasn't a path dependency issue. I feel like there is one about Roman roads and how people continued to follow their design guidelines into the 20th Century, even after cars were being used, but I can't find it, and I'm not even sure it is real.

[Uptowner]

"I find it hard to believe that the poor are better off with no bikeshare stations where they live than with only 9 per square mile."

There are a bunch of stations east of the river that literally get 0 trips per day. I'd love to hear how these stations are improving the lives of poor people.
http://bikes.oobrien.com/washingtondc/#zoom=14&lon=-76.9748&lat=38.8838

[DE]

"I'd love to hear how these stations are improving the lives of poor people."

Well, the option is there for them, and options are good to have, even if not used much. You can only really give people the opportunity to do something; you can't force them, and as I've said before, outside of enabling use without a credit card, it's hard to find any way to blame CaBi here.

[Uptowner]

"In an ideal world, cities would have the money to add as many stations as an area needs, but they don't. Choices have to be made. Trade-offs have to be accepted."

While choices have to be made, the trade-offs should be studied and used to make intelligent decisions that help us to meet our various goals. We should do more studies on how different station densities, systems sizes, and density of activity affect our ability to achieve the goals of user cost savings, user travel time savings, increased access, congestion reduction, emissions reduction, healthcare cost savings and accident reduction.

The author is correct to point out the flaws in this study, which are many, but I would argue that instead of defending our system as it is, let's examine it (and others) to find out if there are ways we could improve it to better meet our stated goals.

[washcycle]

There are a bunch of stations east of the river that literally get 0 trips per day. I'd love to hear how these stations are improving the lives of poor people.

This would be a fantastic question if in fact there were stations that get 0 trips per day.

[washcycle]

The author is correct to point out the flaws in this study, which are many, but I would argue that instead of defending our system as it is, let's examine it (and others) to find out if there are ways we could improve it to better meet our stated goals.

And the various agencies (DDOT, ADOT etc...) do that all the time. Not to tell tales out of school, but DDOT has a project ongoing to define a variety of measures for success (and “trips per bike per day” is not among them) for DC’s part of the system, in a holistic fashion to inform revenue forecasting, station placement, etc. Planning for a July completion.

[Uptowner]

Perhaps I was exaggerating. So I went back and looked at the data. Based on the CaBi Q3 trip data (July 1, 2014 - September 30, 2014), peak usage months, the following stations averaged less than 1 trip originating out of them per day:

34th St & Minnesota Ave SE (0.1)
Broschart & Blackwell Rd (0.1)
Crabbs Branch Way & Redland Rd (0.2)
Shady Grove Hospital (0.2)
S George Mason Dr & 13th St S (0.2)
Taft St & E Gude Dr (0.2)
6th & S Ball St (0.3)
Randle Circle & Minnesota Ave SE (0.3)
Needwood Rd & Eagles Head Ct (0.3)
Medical Center Dr & Key West Ave (0.4)
Fairfax Village (0.4)
S Abingdon St & 36th St S (0.4)
Branch & Pennsylvania Ave SE (0.5)
Washington Adventist U / Flower Ave & Division St (0.5)
Traville Gateway Dr & Gudelsky Dr (0.5)
S Kenmore & 24th St S (0.6)
Piccard & W Gude Dr (0.6)
Garland Ave & Walden Rd (0.6)
Alabama & MLK Ave SE (0.7)
Nannie Helen Burroughs & Minnesota Ave NE (0.7)
S Oakland St & Columbia Pike (0.8)
Congress Heights Metro (0.8)
Montgomery College/W Campus Dr & Mannakee St (0.8)
Fallsgrove Dr & W Montgomery Ave (0.9)
Frederick Ave & Horners Ln (0.9)
Fleet St & Ritchie Pkwy (0.9)

Taken together, these stations generate a cumulative average of 2.03 trips per day. I'm all ears as to how 26 stations producing two daily trips is helping to create real progress towards achieving user cost savings, user travel time savings, increased access, congestion reduction, emissions reduction, healthcare cost savings, or accident reduction. Keep in mind that each station costs around $10,000 to buy, and additional money must go into maintenance. That's $260,000 spent to generate 2 trips per day. If we are actually trying to achieve the stated goals, perhaps that money or those stations can be put to a more effective use.

[Uptowner]

Btw, the data is here, if anyone wants to double check my analysis:
https://www.capitalbikeshare.com/trip-history-data

Also, I'm glad to hear that DDOT is looking into defining measures for success, and I'm not trying to say "trips per bike per day" is the be all end all measure. obviously far from it. However, I am saying that low usage can be a good indicator of places where things are not working well, where the other system goals are almost certainly not being achieved. If we understand how better planning can help avoid low usage, we can help ourselves better achieve our other goals.

[washcycle]

For starters, removing these stations is unlikely to improve access. And isn't that actually 20.3 trips per day?

[DE]

"Keep in mind that each station costs around $10,000 to buy, and additional money must go into maintenance. ...perhaps that money or those stations can be put to a more effective use."

This strikes me as similar to arguments against bikeshare systems at their beginnings, when a lack of infrastructure leads to low usage--until enough stations and bike lanes are constructed to reach a tipping point where more people start using them. If you look at the map you linked to, you can see that there are far fewer stations east of the river; we already know there are fewer bike lanes. I can't imagine the solution is to do less.

[washcycle]

At 20.3 trips a day, and a 5 year life-span for the equipment, that's a cost of ~$0.70 per trip. Considering "fare box recovery" and the value of benefits, they may still be benefit-positive.

[bplanner]

Fantastic analysis. It seems that the easiest hard numbers have become the de facto measuring stick, but you point out a good suite of factors that should be evaluated for any program.

It's unfortunate that it's impossible to truly experiment with transportation systems (no way to control), or else maybe someone could test some of these questions you're raising.

[Uptowner]

My mistake, when you sum the trips, it's 12.8 trips per day in the 26 stations listed above. I'm not advocating for removing the stations, per se, but rather rethinking how we make station location and bicycle infrastructure decisions. That may mean consolidating poorly-used stations to create a more useful network, and it may mean investing in more stations and bicycle infrastructure in the area. There are budget limitations, as the author rightly points out, so we can't always just buy our way to a better performing system.

If we want bikeshare to serve people in an equitable manner, that means figuring out how to do it so it actually serves people. There is little evidence that these stations are doing much at all to meet the stated goals, and there is a strong indication (listed above) that they aren't meeting the stated goals. If we as bicycle advocates can't admit that something is not working here, it indicates that we have little interest in actually meeting the goals we claim to care about. That is a very distressing sign for bicycle advocacy in DC.

[washcycle]

Well, the process of making station location decisions is kind of new and DDOT is still working their way through that. It is relevant to point out that decisions to place and remove stations aren't just technical, but also political. Pulling all the stations out of Ward 8 is just not going to fly - no matter how the numbers add up.

The upside of having the government run bikeshare is that they're willing to consider the social benefits like air quality and public safety. The downside is that it becomes more political and at risk of rent seeking. So, y'know, as Alan Thicke wrote "You take the good. You take the bad. You take them both and there you have the Facts of Life."

Having said that, I think DDOT puts a lot of work into placing stations in a way that they do serve people. I think there is evidence that these stations are doing something to meet their stated goals in that they're being used. Are they being used enough to justify the cost? I'm not sure. I guess we could look at the calculations from the 2010 MWCOG report to see how much use a station needs to be justified by it's benefits.

I'm perfectly willing to admit when things aren't working, but the statement that "these stations are the least used" isn't the same as "these stations aren't working." It's a little like Amtrak or an airline, you need the little, lightly-used stations to feed the busier ones and if you start cutting those out then you start to kill the core too.

DDOT should experiment with placement, but that comes with its own cost. And they'd be unwise to ignore the politcal implications.

Furthermore, if PG County ever gets its act together, the stations EOTR will suddenly move from the edge to the core and will likely see more business. So removing them will only require the work of replacing them later. And of course, the stations also serve as advertising space with it's own revenue stream - consolidating stations might change the value of this.

Just because these stations are used lightly doesn't mean they aren't necessary. But then it may mean that they aren't necessary (or can be improved). It's complicated, but from what I've seen DDOT is doing what it can to wrap its arms around the complexity of this and make rational decisions.