Study of 287 Chinese cities finds AI pilot zones cut medium-skilled jobs

The designs treated the staggered rollout of the pilot zones as a quasi-natural experiment. The authors say the result held across several robustness checks and estimators built for uneven treatment effects, and the negative effects persisted over time but did not shrink or grow steadily.

Education spending and industrial upgrading both rose after the policy, yet neither coincided with better medium-skilled employment in the same period. Cities with a stronger pre-policy industrial base fared less badly, while cities with more public attention to AI before the policy fared worse.

The paper, published in the Journal of the Asia Pacific Economy, tests whether AI diffusion encouraged by Chinese government initiatives changes jobs for medium-skilled workers. It uses city-level data over 2015–2023 and compares cities by when they received pilot-zone status.

The authors report that the negative result survives multiple robustness checks and estimators that allow for heterogeneous treatment effects. Over time the effects are persistent but not steady, and different cohorts of cities experienced significantly different effects.

The authors also looked at possible offsets. Education expenditure and industrial upgrading both increased after implementation, but neither lined up with better medium-skilled employment. Local conditions mattered: a stronger industrial base before the policy softened the damage, while heightened public attention to AI beforehand deepened it.