Can e-scooters disrupt carsharing in cities?

Two shared modes. One city. Who wins the short trip?

E-scooters exploded onto urban streets around 2018–2019 with very little regulatory preparation and even less research into how they would interact with existing shared modes. Most studies asked what scooters replaced — the typical answers were walking, ride-hailing, and personal vehicles. Almost no one had looked at whether scooters could pull users away from carsharing.

This is a genuine gap: carsharing and e-scooters share a surprising amount of DNA. Both are dockless (or semi-dockless), app-based, and accessed on demand. Both target short urban trips. Both appeal to a young, urban, tech-comfortable demographic. For a trip of 2–3 km, a carsharing vehicle and a shared scooter could plausibly compete for the same user making the same journey.

The study fills this gap with a stated preference experiment in Munich — one of Europe's leading carsharing cities — targeting 18–34 year olds, who represent both the current core of carsharing early adopters and the likely first wave of e-scooter users. The key question is simple: given the same trip, and varying conditions, which would they choose?

Carsharing vs. e-scooters: comparing the modes directly

The choice model incorporates attributes specific to each mode. Understanding how they differ on each dimension helps explain the model results.

What percentage of carsharing trips shift to scooters?

The researchers ran 1,620 scenarios varying scooter cost, speed, risk, carsharing cost, and route diversion. Explore the key dimensions below — mirroring the paper's sensitivity analysis (Figures 5–6).

What the choice model tells us about decision-making

The multinomial logit model (N = 4,527 observations from 503 respondents) estimates how each attribute shapes the probability of choosing between carsharing, scooter-sharing, and "neither". Here are the key findings.

• ↑ Very strong
  Travel cost — most significant predictor for both modes Cost coefficients were extremely significant at 99% and 98% for scooters and carsharing respectively. The scooter coefficient is larger in magnitude, meaning users are more cost-sensitive for scooters. Calculated value of time: ~€6.7/hour for scooters, ~€7.9/hour for carsharing — users are willing to pay slightly more per hour for carsharing.
• ↓ Strongest for scooter
  Rain — the single biggest deterrent to scooter use The rain coefficient for scooters (−0.977) is the highest in magnitude and significance of all scooter attributes (99% confidence). By contrast, rain has a positive effect on carsharing utility (+0.159) — rain pushes users toward enclosed vehicles. This asymmetry alone could drive dramatic seasonal variation in scooter-to-carsharing competition.
• ↓ Significant
  Accident risk — only 4× risk is significantly negative The model tested scooter accident risk at 1×, 2×, and 4× carsharing risk. Only the 4× level was significant (−0.369, 99%). The 2× level was removed as insignificant. This means users are somewhat tolerant of moderately higher scooter risk, but draw a clear line at very high accident probability differentials.
• ↓ Gender effect
  Female — less likely to use both modes, much less for scooters Women are less likely to choose carsharing (−0.195, 90% significance) and significantly less likely to choose scooters (−0.344, 99% significance). The larger magnitude for scooters is consistent with city pilot reports from Calgary, Paris, and San Francisco showing 70–80% male scooter user bases. Safety perceptions and physical comfort on scooters may drive this gap.
• ↓ Moderate
  Travel time — more important for scooters than carsharing In-vehicle travel time was significant for scooters at 90% confidence; total travel time (including access and egress) was used for carsharing and was less significant. Access and egress times for scooters were surprisingly not significant — possibly because scooters are typically found nearby and the egress is immediate. This is an important finding for service design: for scooters, cutting in-vehicle time matters more than reducing the walk to the vehicle.

Distance kills scooter competitiveness — completely

The most striking finding from the scenario analysis is how brutally distance-sensitive e-scooter competitiveness is. For trips above 4 km, the model predicts the scooter's market share approaches zero — not declines, but effectively disappears.

This is consistent with observed data from five North American cities used in the model: average scooter trip distances are 1.6 km (San Francisco), 1.85 km (Portland), and 2.4 km (Chicago). Scooters at 10–22 km/h simply cannot compete with a car on a 6 km trip.

The implication is that carsharing operators with a trip profile dominated by longer journeys have little to fear from e-scooter competition. But for operators with concentrated short-trip demand in dense urban cores — particularly leisure and access trips of 1–4 km — the 13–23% scenario range is a meaningful competitive threat.

In the Munich case study, after filtering the carsharing dataset to the 0–4 km range, that 23% shift would translate to approximately 44,624 trips or 118,060 km being redirected to scooters.

What real-world pilots from six cities tell us

The paper synthesises findings from scooter pilots in Calgary, Chicago, Bloomington, Paris/Lyon/Marseille, San Francisco, and Portland. The consistent patterns across these cities inform and validate the Munich model.

Four takeaways for cities and carsharing operators