Bike ride systems around the world experience seasonable variability . For 2014, the Twin Cities Nice Ride system saw an average daily ridership of 611 rides per day in April. Ridership peaked at 2,984 in July and then declined to 653 in October. This seasonality is presumed to be driven by weather, but could also be driven by school schedule, changes in general levels of tourism, or the timing of festivals and other events that might motivate people to use Nice Ride.
But in general, both members and casual users prefer the middle of the summer, but that preference is more pronounced for casual users.
Here’s the first of many charts, showing that pattern:
This seasonality had a profound effect on Nice Ride’s usage and financial viability. Over the summer, the system saw 417,169 rides. If every month had been like July, there would have been about 200 days multiplied by 3,000 rides per day, which totals 600,000 rides. So the system is losing a third of its ridership to seasonal effects!
Daily Weather Data and Ridership
To understand the effect of weather, I downloaded daily weather data from Weather Underground and combined that with the available 2014 Nice Ride data. I mashed the data together in Tableau, and published it to Tableau Public. Feel free to browse the data and graphs, or download the Tableau workbook and do your own analysis.
First, a few general observations. There are roughly the same number of rides by members as by casual non-members. Members use the system more during the week, whereas casual usage is mostly on weekends. This is fortuitous for the system because it evens out the demand.
By hour, casual usage has a nicely unimodal distribution with a peak at 3pm; rides by members have three bumps at 7 am, 12 noon and 6 pm.
The data from Weather Underground included three variables that I investigated as predictors of ridership: the max temperature for the day, the amount of precipitation, and the presence of weather events, such as thunderstorms. This blog post looks at the first two.
Out of the 204 days that Nice Ride operated, 160 had less than 0.1 inches of rain, which I categorized as “Dry”. Casual riders were more discouraged by rain than were members. Unfortunately, a disproportionate number of the weekend days in June of 2014 were rainy, and the ridership among casual users dropped from 2334 rides per day for dry days to 1048 rides per day for rainy days.
(It would be interesting to get hourly weather data to understand whether the rainy days with good ridership can be attributed to the rain occurring during the night when Nice Ride usage is low.)
The effect of temperature on ridership is striking. One might expect to see a threshold effect with temperature, that a person will only ride once a certain temperature is reached. While that may be somewhat true for each individual, the aggregate effect depends on the how that threshold varies in the population.
This graph shows frequency of days falling into 5 degree bins and the average daily ridership for days in each bin. With the exception of one day with a high over 90, the effect of temperature is nearly perfectly linear.
Next, lets break that down by month, user type, and work day versus weekend. The trends remain remarkably linear regardless of user type or day type. Even the data for May, which span a broad range of temperatures, fall along the lines.
The data files from Nice Ride list rides, not distinct users or passes, but it is tempting to estimate the economic impact of the temperature effect. A reduction in rides by casual users probably means correspondingly fewer day passes being purchased. What if we estimate how much income is lost?
Of the days that Nice Ride was operating in 2014, 160 had little or no precipitation. Some of those are weekends and some weekdays, but on aggregate, each degree in temperature changes ridership by casual users by 36.
If we could change behavior so that users would behave as if it were 5 degrees warmer, we would gain (5*160*36=) 28,800 rides over the course of the summer.
It is hard to tell exactly how much additional income is represented by nearly 30,000 rides, but if we assume a three-to-one ratio of rides to users, those 30,000 rides represent 10,000 daily passes. That means that $60,000 of fees would be added to Nice Ride’s income if we theoretically shifted behavior by 5 degrees.
Conclusions and Next Steps
No one should be surprised that rain and cold diminish ridership. The models quantify the effect, and show that even at moderate temperatures, such as highs in the 70s, Nice Ride is losing ridership to temperature in a predictable fashion.
If we take this loss as a fact of life, days with reduced ridership can serve as opportunities to shift staff time from rebalancing to maintenance activities. Alternatively, quantifying the loss in ridership provides a rationale for expenditures that could shift people’s behavior. Market research through surveys or focus groups should be used to understand the reasons that cooler weather discourage bicycling. Nice Ride could consider outfitting some bicycles with fairings, windshields or hand protectors. It might also be worth encouraging vendors serving warm food or beverages near popular nice ride destinations on cold days, as is common in Europe. Perhaps some Nice Ride windbreakers would encourage more rides.
 : A Review of Recent Literature, Transport Reviews, DOI: 10.1080/01441647.2015.1033036 Available at http://www.tandfonline.com/doi/pdf/10.1080/01441647.2015.1033036