by Jince Chen,
Abstract
This paper provides a causal evaluation of China’s plea-leniency reform, a prosecutor-led
negotiated criminal procedure piloted in 18 cities from 2016 to 2018 before its nationwide
adoption. In common law systems, plea bargaining is found to generate sentence discounts
and efficiency gains; in civil law jurisdictions, where bargaining is more constrained and
courts retain an active duty of fact-finding, whether analogous plea-based procedures produce
similar causal effects remains less well understood. The Chinese reform, introduced
to encourage defendant cooperation and improve judicial efficiency, provides a unique opportunity
to study this question. I exploit the pilot as quasi-experimental variation in a
difference-in-differences design. The analysis uses a newly constructed dataset of over 300,000
case-defendant-charge observations for dangerous driving, intentional injury, and theft, extracted
with large language models from nationwide criminal judgment texts across pilot
and comparison cities. I find no evidence that the reform broadly reduced nominal sentence
length or court-stage processing time. Its clearest effect appears in intentional injury cases,
where it shifted punishment toward suspension, changing the mode of execution rather than
the nominal sentence itself, with custodial sentences imposed but not executed unless the
defendant reoffends or violates the conditions of suspension. By contrast, the estimates for
dangerous driving are close to zero and the theft results show no clear evidence of shifts in
sentencing outcomes. The results suggest that in a constrained civil-law setting, the effects
of plea-based reform depend on offense structure and on which margins of leniency remain
institutionally available.
Michael Pollmann, Faculty Advisor
