Divvy Usage Changes
The past couple of years have seen major changes in how the Divvy bike system is used. Chicago signed a contract with Lyft to run the bike share system and, starting in 2020, Lyft introduced e-bikes to the system.
But as Mary Wisniewski points out, this is a mixed blessing. People like e-bikes since they're easier to ride and they don't need to be returned to docks. But the system is a lot more expensive.
How has the public responded? I've downloaded the data from Divvy (available here) and crunched it. For the three years, there are almost 13M ride records. Here are some graphs and my thoughts about what seems to be happening.
The chart above shows how many rides were taken in 2019, 2020, and 2021.
2020 started well, with ridership above that of 2019, but dove from April through July; only in August did ridership numbers begin to exceed those of 2019. Overall, there was a 6% decrease in the total number of rides from 2019 to 2020 (3.82M to 3.47M).
In 2021, ridership was started slow, but starting in March, ridership was much higher than either 2019 or 2020. In fact, the total number of rides in 2021 was about 58% higher than in 2020 (5.60M).
Divvy / Lyft was happy to brag about this and in a September tweet noted that they had exceeded the total number of rides for prior years -- the rest of the year would be setting new records.
But who is riding? Who is responsible for all these new rides?
There has been a substantial shift in who is using Divvy. The number of member / subscriber rides (who pay an annual $108 fee to have unlimited 45 minute rides on non-ebikes) was around 3M in 2019, dipped substantially (to 2.2M) in 2020, and returned to around 3M in 2021.
Where Divvy has seen tremendous ride number growth is among casual users, folks who purchase either a $15 day pass or a single $3.30 30-minute trip.
Starting in July, 2020, Divvy introduced e-bikes. These are more expensive to use than the traditional bikes. The charges depend on where the ride begins or ends (as of Aug 15, 2022; map from the Divvy website).
For rides that start in zone 1 (yellow), Divvy charges $0.15 or $0.20 per minute, plus $2 if the bike not returned to a docking station.
For rides that start or end in zone 2 (blue), the first 45 minutes are free; additional minutes cost $0.15 or $0.20 per minute; there is no charge for out-of-station parking.
Divvy has done this to address complaints that it was slow to provide service to lower income / less-white neighborhoods. Note that the density of docking stations continues to be much lower in the blue areas of the map.
As noted above, the number of 'member' rides dropped substantially in 2020, but the trend in 2021 is similar to that of 2019, at least until September, 2021. Starting in September, the number of rides grew to be substantially higher than in 2019.
While the number of 'member' rides was generally lower in 2020 than in 2019, the number of 'casual' riders grew substantially in 2020 and even more in 2021.
The first month with a substantial increase in rides from the prior year's values is July, 2020. The rest of 2020 (Aug through Dec) saw similarly large jumps from the 2019 values. Since e-bikes were introduced in July 2020, it seems likely that the ebikes may have contributed to this increase.
It is also worth noting that 'casual' rides have generally peaked between April and October, with the highest values in June to August. This was true before the introduction of e-bikes, but is more pronounced in 2021 than in prior years.
Users are responsive to Divvy's economic incentives. This chart shows the number of greater than 1 hour, 2 hours, etc., as well as the percent of total rides this represents.
Members have unlimited rides of less than 45 minutes, so it makes sense that 99.3% of all member rides are less than 1 hour : only 56K of 8.1M rides are longer than 1 hour and only 0.19% are longer than 2 hours.
Casual riders can opt for a single 30 minute ride or can get an all-day pass:
30% (1.14m) of casual rides are less than 30 minutes.
While 12% (576K of 4.75m) of rides are longer than 1 hour, only 0.7% are longer than 4 hours.
Casual and member rides follow different usage patterns. Members average ride length is 10 to 15 minutes, as this graph shows. For the whole year, mean ride length was 14:19 (2019), 15:44 (2020), and 13:38 (2021). Ride length is generally pretty consistent, with most being 10 to 15 minutes in length.
I'm not sure why there was such a jump in April-Sept of 2020. It seems plausible that this was a reaction to Covid-19 shutdowns and/or fear of using public transit. But even at the longest, the rides only jumped to a bit over 20 minutes long.
Casual rides follow a VERY different pattern. The average ride is longer: the mean ride length in 2019 was 57:01, 47:25 (2020), and 32:26 (2021). Note that the y-scale on this graph has a much wider range than the prior graph.
The values in the graph show greater variance in the length of the ride over the course of the year.
The extreme values in Jan and Feb result from a small number of rides and a few very long rides (perhaps bikes that were not docked properly at the end of the ride?).
These maps show how member ride patterns have shifted from 2019 to 2021. They show the 15 most popular docking stations to start a trip and to end a trip.
Two notable patterns:
The geographic spread of use expands over the time period, eventually reaching further to the north and beginning to get a bit south of downtown, especially around the University of Illinois - Chicago.
The usage becomes more even. In 2019, there are several high use stations; in 2020 and 2021, the usage is more evenly distributed.
Casual users follow different patterns.
First, there is a lot of start and stop activity at docks near major tourist sites, such as Navy Pier.
Second, in 2020 and 2021, there seems to be a pattern of people picking up bikes in the Loop area and riding them out to the north or near south side.
Finally, I wanted to explore how who is using the new e-bikes. My assumption was that they were more popular with 'casual' users, especially since they tend to take longer rides. This is only somewhat true.
This chart is hard to read, but it shows the proportion of each type of bike used by casual and members.
Early 2020, only classic (non-E-bikes) were available. In the winter months, members did 80-90% of the rides. This shifted into the spring, presumably as the weather got better and more 'casual' riders began riding (orange and blue bars).
When E-bikes were introduced (green and red bars), members started using them immediately and have, consistently used them more than casual riders. But this is because members use the Divvy system more , across the board, than casual users.
What is clear from this chart is that Mary Wisneiwski is correct: the classic, non E-bikes, are declining in popularity. Classic bikes had a substantial lead until the end of 2021, when the e-bikes took over.
This pattern is easier to see if we look the type of users individually. The top chart shows that the proportion of members' use of e-bikes rose in Nov and Dec of each year. This is probably due to a smaller number of rides overall, while the number of e-bike rides was more constant. But throughout 2021, the proportion of e-bike rides never fell below 25%.
Casual users have, consistently, had higher rates of e-bike usage, by a few percentage points.
Do e-bike users move around the city differently than classic bike users?
This is a hard question to answer. E-bikes can be docked at special e-bike stations or parked "out of station" to "a public bike rack, signpost, light pole, or retired parking meter." In the 2021 data, about 36% of all Divvy bike use is e-bikes; of those rides, 70% start or end not at a station.
While the data provides lat/long data for each ride, a lot of it seems to be dummy values: over 34K 'starts' are recorded for (41.79, -87.60), which is a brownstone on Blackstone Avenue.
Eliminating the rows with the (41.79, -87.90) values provides a better map, but there are blank areas with NO reported starts. These are odd and suspicious, especially on the north side of the city, where other data shows biking is popular. In short, I don't think the Divvy data is clean enough to allow an answer to this question.
Conclusions
It will be interesting to see how Divvy usage changed in 2022 -- did the switch to ebikes continue? Did attempts to spread biking further west and south succeed?
Last updated, 6 Oct 2022
Anyone who wants to see how I did these analyses can check out my Github repo.
The most useful files will be the Jupyter notebooks: "SimplifiedDivvyAnalysis", "CovidEffectsOnDivvyUse","Station Analysis", "20and21_analysis", "Percent of Long Rides", and "MappingEBikeUsage".
Note that I did some data cleaning to produce the files that were loaded for analysis. This can be seen in the "DataMunging" notebook.