Incentivizing Commuters to Carpool: A Large Field Experiment with Waze

How to cite

Cohen, Maxime; Fiszer, Michael-David; Ratzon, Avia; Sasson, Roy. 2019. "Incentivizing Commuters to Carpool: A Large Field Experiment with Waze." AEA RCT Registry. AEARCTR-0004822. https://doi.org/10.1257/rct.4822-1.0

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

Maxime Cohen* , Michael-David Fiszer , Avia Ratzon , Roy Sasson

* Primary investigator

Registry ID
AEARCTR-0004822
DOI
https://doi.org/10.1257/rct.4822-1.0
First registered
Last updated
Status
Completed
Trial dates
2019-06-10 – 2019-07-03
Countries
United States of America (California), United States of America (Georgia), United States of America (Massachusetts), United States of America (Washington)
Registry record
http://www.socialscienceregistry.org/trials/4822

Abstract

Traffic congestion is a serious global issue. A potential solution, which requires zero investment in infrastructure, is to convince solo car users to carpool. In this paper, we leverage the Waze Carpool service and run the largest ever digital field experiment to nudge commuters to carpool. We find a strong relationship between the affinity to carpool and the potential time saving through an high-occupancy vehicle (HOV) lane. Specifically, we estimate that mentioning the HOV lane increases the click-through rate and conversion rate by 133--185% and 64--141%, respectively relative to sending a generic message.

Intervention

We sent in-app notifications to Waze users with an offer to try the carpool service. We tested different framings: (A) mentioning the HOV lane and the potentially (high) time saving, (B) mentioning the HOV lane, (C) using a generic carpool offer, and (D) not sending anything.

Experimental design

We identify three types of users: (i) users who can save a significant commute time by carpooling through the use of an high-occupancy vehicle (HOV) lane, (ii) users who can still use the HOV lane but with a low time saving, and (iii) users who do not have access to an HOV lane on their commute. For each user, we leverage the Waze data and algorithms to carefully estimate the potential time saving had this user used the HOV lane for his/her commute. Our experimental population comprises 537,370 users across four U.S. states (our analyses are aggregated over hundreds of thousands of users and are meant to be interpreted as statistical averages). We then apply the following set of interventions:

- For commuters with a high time saving---who could save on average 6-40 minutes if they would carpool and use the HOV lane---we randomly split them into four conditions. Each condition involves sending the user an in-app notification with an invitation to try the carpool service. Specifically, we use the following four framings: (A) mentioning the HOV lane and the potentially (high) time saving, (B) mentioning the HOV lane, (C) using a generic carpool offer, and (D) not sending anything.

- For commuters with a low time saving (i.e., users who could save on average 2-5 minutes if they would carpool and use the HOV lane), we randomly split them into the same four conditions.

- For commuters who do not have access to an HOV lane on their commute, we randomly split them into three conditions: (A) mentioning a monetary incentive (receiving $10 as a welcoming bonus to try the carpool service), (B) using a generic carpool offer, and (C) not sending anything. We further carefully select the commuters in this category to be similar to the commuters in the other two categories.

By comparing the different framings used in our experiment, we aim to understand what are the successful triggers that can persuade commuters to carpool. It also allows us to study the tradeoff between saving commute time and earning compensation.

Design details

Randomization method
Randomization done in office by a computer.
Randomization unit
We have three different groups of users: (i) users who can save a significant commute time by carpooling through the use of an high-occupancy vehicle (HOV) lane, (ii) users who can still use the HOV lane but with a low time saving, and (iii) users who do not have access to an HOV lane on their commute. For each group, we randomize the users between the different treatments and the control.
Clusters
Three different groups of users: H users: users who can save a significant commute time by carpooling through the use of an high-occupancy vehicle (HOV) lane, L users: users who can still use the HOV lane but with a low time saving, and N users: users who do not have access to an HOV lane on their commute.
Observations
537,370
Arms
84,798 users of type H 224,518 users of type L 228,054 users of type N
Primary outcomes
Click-through rate and carpool on-boarding rate for treated and controlled users

Cited by

No citation data available for this registration yet.

Keywords