Belief Updating and Backfire across Social Groups: Behavioral Interventions in Chinese Gig Workers

How to cite

Wang, Tenghui; Liu, Qiao; Meng, Juanjuan; Zhang, Zheng. 2026. "Belief Updating and Backfire across Social Groups: Behavioral Interventions in Chinese Gig Workers." AEA RCT Registry. AEARCTR-0019113. https://doi.org/10.1257/rct.19113-1.0

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

Tenghui Wang* , Qiao Liu , Juanjuan Meng , Zheng Zhang

* Primary investigator

Registry ID
AEARCTR-0019113
DOI
https://doi.org/10.1257/rct.19113-1.0
First registered
Last updated
Status
Completed
Trial dates
2021-03-01 – 2021-03-12
Countries
China ()
JEL codes
C93, D91, I38, M38
Registry record
http://www.socialscienceregistry.org/trials/19113

Abstract

Many public programs, such as retirement savings, social welfare, and housing assistance, suffer from persistently low take-up despite substantial policy effort. And behavioral interventions are widely used for increasing the take-up rates. We conducted a megastudy among 21,000 Chinese gig workers, testing ten behavioral interventions on their enrollment intentions in a hypothetical public housing provident fund program. Based on the different social groups' potential benefits from the program, we tested the heterogeneous treatment effects across four social groups: local-urban, local-rural, nonlocal-urban and nonlocal-rural. We examined differential belief updating mechanism as potential explanations for the differential responses across the four social groups.

Intervention

Our study implemented ten distinct treatments, grouped into two categories: behavioral interventions (seven treatments) and program feature interventions (three treatments). Motivated by social identity, information updating, complexity and limited attention literature, we implemented resident identity, policy difference, favorable statistics, unfavorable statistics, example, formal statement, housing prices treatments. Program feature interventions include decreased loanable amount, rental subsidies, and hukou point treatments.

Experimental design

The survey experiment consisted of ten treatments and one control groups, and it was conducted on the Alipay platform in ten cities of China. We randomly distributed our questionnaire on the front page of Alipay, under the title: “Alipay sincerely invites you to complete a questionnaire about the Housing Provident Fund for gig workers.” Unlike typical online surveys on commercial survey platforms where respondents self-select in exchange for monetary incentives, we offered no financial rewards. Instead, participants were clearly informed at the outset that the survey was conducted by a research team at Peking University to understand gig workers’ views on the HPF-GW program and housing-related issues.

Our survey experiment consisted of three parts: (1) demographic background information, such as age, gender, marital status, education, occupation, income, consumption, and debt; (2) current living conditions and housing demand, including housing ownership status, living conditions, monthly rent or loan payments, and commute distance; and (3) a hypothetical HPF-GW program, but modeled on typical terms currently offered to gig workers in reality under existing policies with elicitation of willingness to participate. We provided a baseline program description to control groups, and different treatments were applied during the introduction of the program in Part (3).

Design details

Randomization method
Randomization done by Alipay.
Randomization unit
Individual. Within each of the 10 cities, eligible users were randomly assigned by the platform’s randomization procedure to one of 11 survey versions, including one control arm and ten treatment arms. Each user received only one survey version. City was used as a stratification/blocking variable, not as the unit of treatment assignment.
Clusters
Not applicable. Treatment was randomized at the individual-user level.
Observations
20,864 responses from gig workers
Arms
2176 respondents control, 2114 respondents resident identity treatment, 1801 respondents policy difference treatment, 1768 respondents favorable statistics treatment, 1841 respondents unfavorable statistics treatment, 1769 respondents example treatment, 1834 respondents formal statement treatment, 1752 respondents housing prices treatment, 2052 respondents decreased loanable amount, 1877 respondents rental subsidies, and 1880 respondents hukou point treatment.
Primary outcomes
The respondents' willingness to participate in the proposed housing provident fund program (gig workers), including “very unwilling”, “somewhat unwilling”, “neutral”, “somewhat willing”, and “very willing”.

Cited by

No citation data available for this registration yet.

Keywords