Inter-team status competition for ride sharing: A large-scale field experiment at DiDi

How to cite

Chen, Yan; Mei, Qiaozhu. 2018. "Inter-team status competition for ride sharing: A large-scale field experiment at DiDi." AEA RCT Registry. AEARCTR-0003537. https://doi.org/10.1257/rct.3537-1.0

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

Yan Chen* , Qiaozhu Mei

* Primary investigator

Registry ID
AEARCTR-0003537
DOI
https://doi.org/10.1257/rct.3537-1.0
First registered
Last updated
Status
On going
Trial dates
2018-10-19 – 2019-01-18
Countries
Private
JEL codes
C93, J22, L11
Registry record
http://www.socialscienceregistry.org/trials/3537

Abstract

While the gig economy benefits workers with autonomy and flexibility, the lack of strong identities from work and bonds with colleagues can lead to low productivity and higher attrition compared to those working in traditional organizations. To address these problems, we conduct a randomized field experiment which uses recommender systems to build driver teams and introduce team contest in DiDi Chuxing, the largest ride sharing company in the world. Our goal is to evaluate the extent to which contest effect on productivity and driver retention can be maintained post contest, by using a team leaderboard. Our evaluation will be conducted using a three-phase randomized field experiment.

Intervention

In collaboration with DiDi, we will send out text messages telling drivers in four cities that they can participate in a team contest. The drivers can choose to participate or opt out. If they opt out, they will not receive further messages. If they opt in, they will participate in the following multi-phase field experiment:

Phase 0: team formation (one week): Drivers will join teams using the Didi team app.

Phase 1: warm-up contest (one week): The experimenter assigns teams to contests, each of which consists of five teams. The contest runs for seven days. At the end of the week, the team with the highest cumulative revenue wins the cash prize, whereas the four losing teams receive no prize. The winning prize is kept constant within the same city, and adjusted based on city-average driver revenue across cities.

The social information regarding the daily revenue of each driver and the corresponding ranking within a team, similar statistics of competing teams are accessible through the app as within-team and within-contest leaderboards.

Phase 2: leaderboard (three weeks): This phase is designed to maintain team identity in the treatment, but not in the control. The intervention is limited to social information, without monetary incentives. It will last for three weeks. Prior to the start of phase 2, within each city, the experimenter sorts leaderboards in descending order of their total revenue during the 14 days prior to the start of the experiment, and group each 9 consecutive leaderboards into a stratum. Within each stratum, the 9 leaderboards are randomly assigned to each of the three experimental conditions, including the control group, the placebo group, and the leaderboard treatment group.

Phase 3: surprise contest (one week): After the second phase, every team is again called to participate in a surprise week-long contest for a cash prize, as in the first stage. This contest has not been announced before. This phase is designed to evaluate the effects of the leaderboard on team identity and productivity. Everything remains the same as in the warm-up contest.

Phase 4: post-experiment survey: At the end of the three contest phases, the drivers will each receive a survey which evaluates their sense of belonging to their team as well as to the organization (DiDi).

Experimental design

In collaboration with DiDi, we will send out text messages telling drivers in four cities that they can participate in a team contest. The drivers can choose to participate or opt out. If they opt out, they will not receive further messages. If they opt in, they will participate in the following multi-phase field experiment:

Phase 0: team formation (one week): Drivers will join teams using the Didi team app.

Phase 1: warm-up contest (one week): The experimenter assigns teams to contests, each of which consists of five teams. The contest runs for seven days. At the end of the week, the team with the highest cumulative revenue wins the cash prize, whereas the four losing teams receive no prize. The winning prize is kept constant within the same city, and adjusted based on city-average driver revenue across cities.

The social information regarding the daily revenue of each driver and the corresponding ranking within a team, similar statistics of competing teams are accessible through the app as within-team and within-contest leaderboards.

Phase 2: leaderboard (three weeks): This phase is designed to maintain team identity in the treatment, but not in the control. The intervention is limited to social information, without monetary incentives. It will last for three weeks. Prior to the start of phase 2, within each city, the experimenter sorts leaderboards in descending order of their total revenue during the 14 days prior to the start of the experiment, and group each 9 consecutive leaderboards into a stratum. Within each stratum, the 9 leaderboards are randomly assigned to each of the three experimental conditions, including the control group, the placebo group, and the leaderboard treatment group.

Phase 3: surprise contest (one week): After the second phase, every team is again called to participate in a surprise week-long contest for a cash prize, as in the first stage. This contest has not been announced before. This phase is designed to evaluate the effects of the leaderboard on team identity and productivity. Everything remains the same as in the warm-up contest.

Phase 4: post-experiment survey: At the end of the three contest phases, the drivers will each receive a survey which evaluates their sense of belonging to their team as well as to the organization (DiDi).

Design details

Randomization method
randomization done in office by a computer
Randomization unit
a driver team
Clusters
375 - 564 teams in each city
Observations
4,000 drivers in each city; 16,000 drivers in total
Arms
188 teams in each arm
Primary outcomes
driver's daily revenue, hours worked, whether a team wins the warm-up contest or the surprise contest
Secondary outcomes
whether a driver volunteers to be a team captain, a driver's sense of belonging to his team and to the organization, a driver's satisfaction with the contest and with his team

Cited by

No citation data available for this registration yet.

Keywords