Repression and Cooptation in Authoritarian Judiciary: Field Experiment with Chinese Lawyers

How to cite

Zu, Ziwen. 2025. "Repression and Cooptation in Authoritarian Judiciary: Field Experiment with Chinese Lawyers." AEA RCT Registry. AEARCTR-0016205. https://doi.org/10.1257/rct.16205-1.0

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

Ziwen Zu*

* Primary investigator

Registry ID
AEARCTR-0016205
DOI
https://doi.org/10.1257/rct.16205-1.0
First registered
Last updated
Status
In development
Trial dates
2025-06-23 – 2026-01-01
Countries
China ()
JEL codes
C93 ; D72 ; D73 ; K42 ; P26
Registry record
http://www.socialscienceregistry.org/trials/16205

Abstract

This study tests how different kinds of official-looking messages influence the day-to-day choices of lawyers in China. Around 20000 licensed lawyers who use a large legal-services app will receive a short weekly newsletter for ten weeks. Each newsletter is identical except for one short headline that is randomly varied:
• A Risk headline warns that certain public-interest cases are under state scrutiny.
• An Honor headline praises lawyers who take those cases for their professional reputation.
• A Peer headline says many colleagues are already involved.
• A State-Broker headline invites lawyers to work alongside government legal-aid offices.
• A Neutral headline simply lists new case opportunities.
Because the headline is assigned by chance, any systematic difference in what lawyers do next can be traced back to the message they saw. The app automatically records whether a lawyer opens the newsletter, clicks to view case details, agrees to represent a public-interest client, or joins a relevant discussion group. Before the first newsletter and again after ten weeks, participants also fill out a brief survey about their work, views on the legal system and willingness to do public-interest cases.
The main goal is to learn whether warnings of repression deter civic-minded lawyering and whether appeals to prestige, peer norms or state partnership encourage it. All data are stored in de-identified form on secure servers, and the study has Institutional Review Board approval. Findings will shed light on how carrots (honor, collaboration) and sticks (risk warnings) deployed by an authoritarian state can steer the behavior of professionals who play a key role in upholding—or challenging—the rule of law.

Intervention

Intervention (public description)
Each participant receives one short legal-industry newsletter every Monday for ten consecutive weeks, delivered simultaneously by in-app notification and back-up SMS. The newsletter always contains four standard items (new regulations, typical cases, compliance tips, and sector news). Only the fifth item—labelled 本期专题 (“focus of the week”)—differs across experimental arms:
1. Risk frame – warns that certain public-interest or rights-defense cases are under heightened state scrutiny and may carry professional danger.
2. Honor frame – highlights the prestige and professional recognition lawyers can earn by handling public-interest cases.
3. Peer-norm frame – reports that many colleagues in the same region have recently taken similar cases and invites the reader to join them.
4. State-broker frame – presents public-interest work as an opportunity to collaborate with official legal-aid channels and strengthen ties with government partners.
5. Neutral frame – simply lists new consultation opportunities without persuasive language (serves as the comparison condition).
The frame assigned to a lawyer is determined by computerized randomization before the first newsletter and remains the same for all ten weeks. Embedded tracking links record whether the lawyer opens the newsletter, clicks to view a case invitation, or joins a themed WeChat group. No other aspect of the platform or messaging changes.

Experimental design

This study is a randomized field experiment designed to test how different types of official-style messaging influence the behavior of licensed lawyers in China. The intervention is embedded in a legal-services platform that provides professional tools and marketing resources to tens of thousands of lawyers nationwide.

Participants are invited to subscribe to a new weekly legal newsletter. After consenting and completing a short baseline survey, each lawyer is randomly assigned to receive one of five versions of the newsletter for ten consecutive weeks. All newsletters are identical in structure and timing, but vary in one short framing message embedded in a section called “This Week’s Focus” (本期专题). This message is designed to simulate a subtle signal of either risk (e.g., warning about case sensitivity), professional recognition (e.g., honor framing), peer norms, state collaboration, or a neutral message with no persuasive content.

Randomization is conducted at the individual level and stratified on region, gender, seniority, and legal specialization. Participants remain in the same frame for the entire 10-week period. Outcomes are passively recorded through platform logs and include whether the lawyer opens the newsletter, clicks on case invitations, accepts a case, or joins legal discussion groups. A follow-up survey is conducted at the end of the study to measure attitudinal changes and confirm behavioral self-reports.

The trial is fully anonymized, approved by an Institutional Review Board, and designed to assess how state-style messaging can influence civic and professional decision-making under authoritarian conditions.

Design details

Randomization method
Randomization is conducted by computer using a reproducible algorithm. Specifically, lawyers are individually assigned to one of five treatment arms using computer-generated random numbers, implemented through the `blockTools` package in R. The randomization is stratified by four pre-specified variables: region, gender, professional seniority, and legal specialization. The process ensures balanced representation across arms and is executed prior to the start of the intervention. Treatment assignment is linked to each lawyer’s anonymized ID and uploaded to the platform backend to control message delivery.
Randomization unit
The unit of randomization is the individual lawyer. Each participating lawyer is independently and randomly assigned to one of five treatment arms. There is no group- or cluster-level randomization. All randomization occurs at the individual level, stratified by region, gender, seniority, and legal specialization to ensure balance across treatment conditions.
Clusters
Each lawyer constitutes a single randomization unit; there are no higher-level clusters such as firms or regions.
Observations
20,000 individual lawyers.
Arms
Risk frame: approximately 4,000 lawyers Honor frame: approximately 4,000 lawyers Peer Norm frame: approximately 4,000 lawyers State Broker frame: approximately 4,000 lawyers Neutral (baseline) frame: approximately 4,000 lawyers Total planned sample size: 20,000 individual lawyers Final counts may vary slightly due to stratified randomization and any post-consent exclusions.
Primary outcomes
1. Newsletter engagement (click-through). For each lawyer we record whether they open any of the ten newsletters. If the tracking link embedded in at least one issue is clicked, the outcome is coded 1; otherwise 0. Data come directly from the platform’s server logs. 2. Sensitive-case uptake. We track whether a lawyer accepts at least one consultation invitation classified as politically sensitive (for example, rights-defense or anti-corruption cases) during the ten-week window. Clicking “take case” or scheduling a consultation registers the event. The outcome is coded 1 if such an acceptance occurs, 0 if not. 3. Public-interest network participation. Each newsletter contains a link to join a dedicated “Public-Interest Law” WeChat group. The outcome is coded 1 if the lawyer joins this group at any time during the study period, 0 otherwise. Join events are captured automatically through the platform’s API. All three measures are binary, rely on unobtrusive digital traces, and are linked to participants only through anonymized IDs.

Cited by

No citation data available for this registration yet.

Keywords