The Ashbury Scooter Pilot Numbers
Ashbury ran a 90-day e-scooter pilot with 250 scooters. The reported results: 41,000 trips; average trip length 1.3 miles; 62 percent of surveyed riders said they would otherwise have driven or taken a car; 3 injuries reported, 1 serious; 118 scooters impounded or pulled from the river; helmet use observed at 12 percent. Complaints to the city's sidewalk hotline rose 22 percent during the pilot.
At the council meeting, Councilmember Reyes argues: “Sixty-two percent of 41,000 trips means roughly 25,000 car trips were removed from our streets in three months. That is a real reduction in traffic and emissions. I move to extend the pilot and expand to 600 scooters.”
A staff member notes that the 311 app, which lets residents file sidewalk complaints from a phone, launched during the pilot's second month. Another notes that the rider survey had a 19 percent response rate and was offered only through the scooter company's app.
1. Evaluate the claim that “roughly 25,000 car trips were removed.” Identify what the figure assumes and why the survey cannot support it.
Read the worked answer
The figure is arithmetic, not evidence. It multiplies 41,000 by 0.62 and treats the result as if it were a measurement of traffic removed.
Assumption 1: riders accurately report a counterfactual. The survey asks people what they would have done. That is a hypothetical, and respondents routinely overstate the virtue of their alternative. Someone who would have walked 1.3 miles may answer “driven” because driving feels like the realistic alternative on a rainy day.
Assumption 2: a substituted trip equals a car trip removed. Even if every rider had driven, some of those trips would have been combined with other errands or taken on a route with spare capacity. And some scooter trips replace walking, transit, or other scooter trips - all of which remove zero cars.
Assumption 3: the sample represents riders. A 19 percent response rate drawn only from app users is a self-selected group; people who answer surveys pushed through an app are not a random sample of riders.
So “25,000 car trips removed” should read “an unknown number, possibly substantial, possibly much smaller.”
2. Reyes is challenged on the 22 percent rise in sidewalk complaints. Using only the setup, explain how that number should be treated and what conclusion is still defensible.
Read the worked answer
The 22 percent figure cannot be read as a clean measure of scooter-caused sidewalk problems, because the city launched a 311 app during the pilot's second month - the same window in which the complaints were counted.
The rise therefore has at least two candidate causes: more people obstructing sidewalks, or the same number of problems now being reported because reporting became a few taps on a phone. Reporting-rate changes are a classic confound: the underlying condition can be flat or even improving while recorded complaints climb. The right response is not to dismiss the complaints, since some are presumably real, but to stop treating 22 percent as a magnitude.
What remains defensible: complaints exist and should be investigated, and the pilot's second month is not comparable to its first. Reyes could defensibly say that obstruction reports rose during a period when reporting also got easier, and commit to checking complaint volume against the pre-app baseline and against actual sidewalk inspections. The honest version is weaker than the slogan but survives scrutiny.
3. Reyes wants to strengthen the case for expansion. Name the single best piece of additional data and explain why it is more valuable than the other available options.
Read the worked answer
The strongest addition is observed mode substitution at scooter parking areas - short intercept surveys at docking locations, asking departing riders where they came from and how they would have made that specific trip, ideally paired with counts of car traffic on the same corridors before and during the pilot.
Why this beats the alternatives:
- More rider surveys through the app repeat the flaw: same self-selected pool, same hypothetical question. A bigger biased sample is still biased.
- Helmet-use data speaks to safety, and Reyes is arguing about traffic and emissions. It cannot support the 25,000 figure.
- Injury counts likewise address a different claim, and 3 injuries is too small a base for much inference either way.
- Complaint volume is confounded by the 311 launch.
Street counts before and during the pilot address the actual claim - did cars decline? - and do not depend on anyone's self-report. That is the difference between measuring an outcome and asking people to estimate one.