Does Cost Information Shift Carsharing Adoption?

How to cite

Ciccone, Alice; Johnsen, Åshild; Andreassen, Gøril. 2026. "Does Cost Information Shift Carsharing Adoption?." AEA RCT Registry. AEARCTR-0018317. https://doi.org/10.1257/rct.18317-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Alice Ciccone* , Åshild Johnsen , Gøril Andreassen

* Primary investigator

Registry ID
AEARCTR-0018317
DOI
https://doi.org/10.1257/rct.18317-1.0
First registered
Last updated
Status
In development
Trial dates
2026-04-13 – 2027-12-31
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/18317

Abstract

This paper examines whether personalized cost information can shift behavior from private car ownership to carsharing. We conduct a randomized field experiment among car-owning households in Bergen, Norway, where treated participants are invited to use an online calculator comparing the full cost of car ownership with carsharing costs and benefits. We adress the following research questions: To what extent does tailored cost and convenience information comparing car ownership with carsharing affect (i) carsharing discount coupon-requesting behavior, (ii) coupon redemption, (iii) carsharing usage, (iv) stated preferences about carsharing and car ownership, and (v) actual car ownership status?

Intervention

This project examines whether providing tailored information about own economic costs of car ownership and the potential costs and practical benefits of carsharing can influence carsharing adoption behaviour and preferences among car-owning households in Bergen, Norway. Approximately one third of the city’s adult population is invited to participate in a short survey, where the identified target group is randomly assigned to either a treatment or control group.
The intervention consists of an online calculator that compares the total cost of owning a private car with the cost and convenience of carsharing. Participants in the treatment group are invited to use the calculator, while the control group instead answers unrelated questions designed to equalize time and effort. Finally, both groups are offered the opportunity to request a discounted coupon to try carsharing.
We implement a randomized encouragement design in which car owners are randomly exposed to the calculator. Using requesting carsharing coupons as the primary outcome, random assignment identifies the effect of the offer, and a two-step instrumental variable analysis isolates the effect among individuals whose exposure to calculator content was changed by the offer.
In addition, we study how the information provided by the calculator affects preferences about carsharing and car ownership using a within-subject survey design. Finally, we examine longer-term behavioural responses by tracking whether individuals redeem the coupons and try carsharing, and whether the intervention leads to changes in car ownership status over time.

Experimental design

Treatment (calculator offer): After completing the baseline survey, treated households receive a link to an online calculator delivering four types of information:
(a) a generic cost comparison between car ownership and carsharing shown to all visitors;
(b) a personalized total cost report for own vehicle, obtained by entering own license plate number, or alternatively chose from six representative vehicles;
(c) information on the geographic availability of carsharing in Bergen; and
(d) practical information on how carsharing works.

Control: Control households do not receive a calculator link. Their survey is extended with additional non-treatment-related questions (about freight transport and investment in infrastructure) to equalize total response effort across arms.

Design details

Randomization method
in office by computer using stata. Within each of the four strata (high/low carsharing access × female/not female) eligible individuals are randomly assigned 55% to treatment and 45% to control using a pre-committed randomization algorithm.
Randomization unit
The unit of randomization is the individual.
Clusters
We stratify on gender and access to carsharing, defined at the number of carsharing cars available in 2025 in all (142) postcodes in Bergen. Before randomization, postcodes are classified into two strata split at the median: high and low carsharing access (the median is in the low category). The median number of cars available in each postcode is 4. This variable is frozen before any data collection and is not altered during the study. 4 strata (high/low carsharing access × female/not female)
Observations
The sampling frame consists of all adults aged 18–75 in Bergen municipality reachable by email: approximately 75,000 individuals, accounting to about one third of the city’s adult population. After eligibility criteria and drop out, we assume: 9000 responses to be divided in 2 groups
Arms
T= 4950 C=4050
Primary outcomes
Our main outcome variable measures whether respondents request a coupon to try carsharing. • Y_req = 1 if the participant submits the coupon request; • Y_req = 0 otherwise, including for non-completers.
Secondary outcomes
Willingness to try carsharing(secondary) 0–100 slider: ‘How likely are you to try carsharing in the next 12 months?’ Asked at T0 (both arms) and T1 (treatment completers only). Car ownership necessity (secondary) Likert 1–5: ‘Owning a car is a necessity for me in the foreseeable future.’ Asked at T0 (both arms) and T1 (treatment completers only). Adinistrative data. Any carsharing usage (exploratory) Car ownership at 12m (exploratory)
Sponsors
Name: The NOrwegian Research Council Url: https://www.forskningsradet.no/

Cited by

No citation data available for this registration yet.

Keywords