Friday, September 27, 2013

Citi Bike at Age 4 (Months): More Numbers, More Bikes

Today marks the four-month anniversary of the introduction of Citi Bike. This is as good a time as any for me to revisit the program's ridership data, as I promised in my original post from the four-week mark. Long story short: the system is thriving.

Since my last post, NYC Bike Share has rolled out a System Data webpage, on which you can view and download substantial data on the system's usage thus far. The data is stored and displayed via Datawrapper; I'm not thrilled with the presentation (particularly the absence of tooltips to display exact values), but I'm surely grateful not to have to manually transcribe another three months' data off of blog posts. Data is available through Tuesday 9/24, so we have just over 17 weeks (and just under four months) to play around with.

Of course, the first step is to validate all this new data. I was dismayed to identify several glaring inconsistencies within the provided numbers. The System Data page includes both daily and total counts for trips taken, miles traveled, and annual memberships sold, plus daily sales of 24-hour and 7-day memberships – but on numerous occasions, adding a day's trips taken to the cumulative total thus far yields a number different than the next day's cumulative total. (Similar problems abounded for miles traveled and annual memberships sold.) Most egregiously, the cumulative numbers for July 18 reflect an additional 12,666 trips and 317,494 miles (= a highly unlikely 25 miles per trip!) beyond the provided daily figures. So I've used the numbers I'd collected from the blog, along with simple arithmetic and common sense, to reconcile the conflicting values; all my data is available here for the curious.

Having massaged our data into some semblance of consistency and plausibility, let's take another look at the key charts from my first post. (If you haven't read it, click here for background on why these figures are important.) It turns out that my first post, dated June 25, occurred right at the beginning of a substantial slide in membership sales. Around 6/23, users were purchasing nearly 2700 daily passes; by 7/10 that number had dropped 45% to under 1500. But at the same time, 7-day and annual membership sales made a comeback; since then, weekly memberships have continued to putter along (this is, perhaps, a category with limited demand; the new Bay Area Bike Share instead offers a three-day pass) while annual sales remain surprisingly robust for a system now beyond its infancy, and daily membership sales have stabilized around the 1500 mark. All in all, sales remain sizeable; when more information on the system's operating costs becomes available, I'll be able to assess whether they're bringing in enough business to cover costs.
System use is extremely strong. The 7-day average "high" of 29,091 trips on 6/24 has turned out to be a "low" in comparison with the system's subsequent performance: Since July 14, the average daily ridership has not dropped below 30,000. At the same time, though, the continued growth in active members has kept the "trips per active member" ratio in check. I had hoped to see that figure continue its dramatic mid-June climb, but no such luck thus far; it's hovering around 0.45 as peak cycling season comes to a close. (Expect that number to drop come winter.) If the system does expand as promised and hoped, active members across the city would be able to use Citi Bike for more trips in more areas, likely leading to a rise in the trips-per-active-member ratio. The unsmoothed data remain noisy, but there is at least one important trend they reveal. It is easy to pick out weekends – when 24-hour membership sales explode – and Mondays, when satisfied 24-hour pass buyers seal the deal by purchasing an annual membership. But in late June, the daily trip count, which had up to that point been closely correlated with 24-hour pass sales, decouples from short-term memberships and begins to stay high throughout the week. My surmise (and hope) is that by the end of the program's first month, annual members began figuring out how to integrate Citi Bike more fully into their daily transportation routines. In any case, daily trips have continued to rise even as sales of short-term passes decline with the end of peak tourist season.

(I promised earlier that I'd investigate the impact of weather on daily ridership; I've now made a preliminary survey and found that impact to be quite limited. I collected daily mean temperature, mean humidity, and total precipitation data from Weather Underground, then exported my data to run some linear regression and correlation analysis on it in R. I found that a simple linear model with total active members as the independent variable and daily trips as the dependent variable accounted for (i.e. adjusted R-squared = ) 76% of the variance in daily trips. Although correlation analysis indicated a weak negative relationship between humidity and precipitation and daily trips, a multivariate model incorporating the three weather measurements accounted for only 5% more of the variance – not a substantial increase in the model's predictive power.)

(If any readers with a strong statistics background want to chime in with a more nuanced analysis of the data, I'm all ears. The comments section is below and here's the spreadsheet.)

It's quite clear from these charts and figures that Citi Bike has been performing excellently over its first four months when compared with its first four weeks. But that is probably not a very useful context. (In the United States' first four Presidential elections, the runner-up became the Vice President.) Instead, let's see how New York's system usage stacks up against a successful and established program across the Atlantic.

London, population 8.3 million, has had Barclays Cycle Hire1 since 2010. With 8,000 bikes and 570 stations in operation, that three-year-old system logged 904,155 rides this August, split 50-50 between short-term riders and annual members; its all-time peak thus far was 47,105 rides in one day, during the 2012 Olympics. Meanwhile, a three-month-old Citi Bike, comprising 6,000 bikes and 330 stations2 in a city of the same population as London, recorded 1,115,340 rides in August, representing a whopping 6.0 trips per bike per day, 64% higher than London's usage levels. NYC's peak thus far was 44,083 rides on Saturday, August 18 – 94% of London's peak, which is not bad given the absence of any Olympic Games.

Data isn't yet available on the percentage of Citi Bike rides taken by annual members vs. temporary members (a division roughly equivalent to "locals vs. tourists" or "work vs. play"), but we can approximately assess the proportions of utility trips vs. recreational trips by examining how many trips occur on weekdays vs. weekends. (This is imperfect, of course – many locals make utility trips on weekends, and most tourists' weekday trips are still recreational in nature – but bear with me.) In May and June, weekends saw substantially more use than weekdays, but in the months since, weekday usage has equalled and often exceeded weekend ridership. Combined with the decoupling of daily rides and 24-hour pass sales we saw above, this leads me to conclude that Citi Bike is less and less dominated by joyriding out-of-towners. Will Citi Bike, having attracted more than 84,000 annual members in its first four months, manage to equal the Boris Bikes' three-year sales figure (186,762)? Almost certainly. About 250 people per day are currently purchasing new annual memberships, a rate that, if sustained, would push total membership sales past the 100,000 mark in just over two months; and when those annual memberships begin to expire next summer, count on the vast majority of users to re-up. Indeed, I am predicting here and now that by June 1, 2015, Citi Bike will have sold more than 200,000 annual memberships.

So: given Citi Bike's burgeoning ridership, its favorable performance compared to London's longer-established system, and the strong support/warm acceptance shown it by New York's mayoral candidates, it looks like bike share is here to stay.

How has bike share been treating you so far? Are there any other bits of analysis you'd like to see in future updates?


1 I always find it so funny that "bike share" has exactly zero words in common with the British English equivalent, "cycle hire." I'm also amused that what we Americans call a "system" the British call a "scheme," which always sounds slightly sinister. ("Cycle to Work Scheme: brought to you by the All-Powerful Bike Lobby")

2 Officially, the system has 6,000 bikes. However, a fabulous visualization by Jehiah Czebotar suggests the true number is currently closer to 4,100 – which would mean that New Yorkers put in almost nine trips per day on every Citi Bike this August.

No comments:

Post a Comment