Previously, I looked at the general behaviour of e-scooter users based on the 2019 Calgary pilot data; where do they travel (mostly in the downtown, Beltline and adjacent communities) and when do they travel (mostly evenings and weekends). I talked a little about origins and destinations, but there’s tons of more data to look into there. The first question I asked myself when I got the data was simple — where is the most popular single trip? From somewhere on 17th Avenue to Kensington? From Inglewood to Prince’s Island?
The city data is binned into a hexagonal grid, about 100m on a side, or roughly the size of a city block. There are 6000 of these, so there’s about 18 million possible origin-destination pairs — I won’t care about direction right now; a trip from the Calgary Tower to Tomkins Park is the same as a trip the other way around. With 465,000 e-scooter trips, the average origin-destination pair would have 0.024 trips. In fact, the most common origin-destination pair has over 1000 trips made in 2019. There were 2,633 trips made between the top three origin-destination pairs.
The most popular scooter trip is…
Here’s a couple of grid cells, in the Eau Claire area. DO-100 contains most of the mall itself, along with the plaza, the smokestack, Joey’s, the Barley Mill, and most of the parking lot. DO-99 is the grid cell immediately north; it contains what I’m sure is the busiest intersection in the pathway system between the north-south path onto Prince’s Island over the Jaipur Bridge (which was still open in 2019) and the east-west Bow River pathway, as well as the wading pool and amphitheatre in Eau Claire.
The single busiest origin-destination pair in 2019 was from DO-99 to DO-99; that is, starting and ending in the same cell on the north end. The second busiest was from DO-100 to DO-100; starting and ending in the same cell in Eau Claire. The third busiest? Between these two cells.
It’s important to note here that the data excludes all trips under 100m, so there are no actual very short trips being made that are included here, if someone did actually want to take an e-scooter a very short distance. The scooter data includes the distance the scooter actually travelled; here’s the distribution of the trips being made involving these cells:
The longest possible trip between points in these two grid cells is about 500m. The average trip that was actually made was 2.3 kilometres (with a duration of 19.4 minutes); over 15% of trips were over 5 kilometres (22% of trips are over half an hour — remember these scooters are billed by the minute). Short trips of under 500m were actually less common for people starting in one grid cell and ending in another than they were for people starting and ending within DO-100; my guess is because there were people trying out the scooter by doing laps in the plaza.
It’s pretty obvious that the main thing done here is actually e-scooters operating as a de facto scooter rental service, for people looking to have a nice outing on the pathway system. In this case, they are not any different than the bike rental businesses based in Eau Claire Market. It seems remarkable to me that the discussion around e-scooters has been based so entirely in mobility that I was surprised to see that a big market for e-scooters was actually… short term scooter rental.
The bigger picture
There’s a ton of different ways to cut the data, but one thing that is key in the discussion around e-scooters is their role in the transportation system; for instance, are people taking e-scooters instead of making car trips? It seems clear from the most popular origin and destination pair that at least some of e-scooter travel is purely recreational in nature.
I’ll start by pulling out the subset of trips that start and end either in the same grid cell or an adjacent grid cell. These pretty clearly don’t provide any real transportation benefit; the distances are too short; a few hundred metres at most. I’ll call them “loop” trips; they must be primarily for leisure purposes. Overall, 19.4% of trips made are loop trips; the average scooter on a loop trip travels 1.59 km, taking 14.3 minutes. Here’s where loop trips are made (from cells with 100 or more trip starts):
It’s pretty clearly primarily a recreational thing, following the river. Particularly high spots include the Crescent Road bluff, SAIT (note these trips are in the summer), and parks like Pearce Estate and Stanley Park. However, almost everywhere has at least 10% loop trips, and 30-40% of scooter trips starting in much of Marda Loop and the emerging 19 St NW corridor are loop trips.
Of the remaining trips, a key question is how many of them could plausibly be replacing a car trip. In my experience, most trips under 1 km are made by walking (especially in the inner city, with traffic, paid parking and an efficient, gridded pedestrian network). Few trips over 2 km are made by walking, although there is a tail of longer walk trips — this month, I personally walked over 6 km home from a medical appointment; it was a lovely day and my route was along the river.
There clearly isn’t a perfect breakpoint between walking distance and too far to walk, but in the interest of a simple analysis, let’s split them at 1500m. One challenge is that the scooter dataset shows how far the scooter travels, but if you remember above, the average loop trip involves the scooter going over 1500m only to end up where it started. So instead, I looked at the actual shortest path between grid cells using the road, sidewalk and pathway network from OpenStreetMap. (This is called a “network distance” in transportation engineering.) This allows me to identify trips where the destination would be a 1500m walk from the start, regardless of how far the scooter travels.
Using this 1500m break, 56.2% of trips are “short” trips, and 24.4% are “long” trips. The average distance from the origin to the destination on a short trip is 884m, although the average scooter travels 1.43 km and spends 10.6 minutes. A subject for future analysis is looking at the subset (around one quarter) of these short trips that are relatively inefficient – travelling much further than would be expected for transportation. For instance, instead of starting in Eau Claire and doing a loop and returning there, someone might have a leisurely trip on the pathway of a few km but end 800m from the start in Chinatown for bubble tea. Here’s where short trips start; note the bins are different from the other two trip types.
These are obviously centred in the, well, the centre. That core blob starts around 6 St/6 Ave and continues along 4th and 5th Streets, and east-west along 11th and 12th Avenues in the Beltline. This makes sense; this is the area with the most nearby destinations.
The “long” remainder are 24.4% of trips; these involve an average distance of 2.3 km from origin to destination, and an average scooter trip of 3.0 km over 17.1 minutes; these are much more efficient in general than the shorter trips. These are the set of trips where it begins to be plausible that a car trip is being replaced. But about half of these trips are not that much longer, falling in the 1.5km to 2km range; only 12.5% of e-scooter trips involved a network distance of over 2 km. Here’s where they start:
Here’s some of the best arguments for e-scooters as a transportation mechanism; high shares of these long trips are from areas like Inglewood, Sunalta, Bankview, 14th St S, 16th Ave and Erlton; mostly residential areas towards the perimeter of the densest parts of the city. Many of these areas have relatively good transit to the downtown; on one hand, this suggests that e-scooters might be taking trips from transit, on the other hand, it might suggest that the high cost of a transit fare for a relatively short trip is less competitive. By Marda Loop, the share of long trips has dropped somewhat, suggesting there’s a limit to how far someone will use an e-scooter.
What are they good for?
The question that arose for me was “are e-scooters a transportation system or a leisure system”? The discussion in Calgary has been almost entirely around transportation — the first thing on the Lime website is the slogan “Smart Mobility for the Urban Commuter”. Here’s the split of scooter trips by my typology again.
And here it is in a little more detail; the loop trips are split by how far the scooter travelled, the others by the network distance from origin to destination.
I think there are three groups of scooter trips:
- Leisure – trips where the point was to go and enjoy riding a scooter for it’s own sake; to see the river, feel the wind in your hair, impress your date, whatever. Most if not all of the loop trips fit here, and I’d guess something like a quarter of the short trips do too; probably around 35% of scooter trips are leisure.
- Replacing walk trips – scooter trips over short distances where walk is the most plausible alternative. I’d suggest almost all of the remaining short trips and a chunk of the long trips under 2 km are in this group; around 50% of trips overall.
- Replacing other modes – these are the longer trips; almost all of the 2 km+ and a chunk of the shorter ones; perhaps 15% to 20%. In most cases, the mode they replace would be car in some fashion — assuming the trip was still being made.
It’s important to keep these things in perspective; almost 100% of the talk about e-scooters is around replacing a car trip, but a random scooter trip is more likely to be going literally nowhere at all than to be replacing a car trip.
There’s nothing wrong with a city providing facilities for leisure; I’m sure many of the walking and cycling users of the pathway system are actually out for leisure and to enjoy the city’s parks. (I will note that active leisure in the form of walking or cycling has health benefits that standing on an electric scooter don’t offer, with much lower accident rates.) Even aside from the transportation issues, passive leisure like e-scooters is a legitimate societal benefit, and it would be a sad day if our cities only provided for the most efficient activities.
This analysis was done with all scooter trips in the 2019 Calgary e-scooter pilot data; more details and a download link in the previous post. It’s not a sample, it’s every trip over 100m, around 460K trips.
I used QGIS and specifically QNEAT3, the QGIS Network Analysis Toolbox 3, to analyze travel between the centroids of all grid cells using the complete OpenStreetMap network in the area; this provides a reasonable estimate for the actual travel distance, considering that scooters are allowed to travel on all sidewalks, pathways and so on.