The impact of household registration policy liberalization on urban economic growth based on the quantitative analysis of difference-in-differences and large language model – take Shenzhen (China) as an example

The reform of household registration system is the core link of China’s new urbanization strategy, but its economic effect remains controversial. This article uses the panel data of nine cities from 2018 to 2024 and comprehensively uses the multi-period difference-in-differences (DID) and large language model (LLM) text-quantification method to evaluate the impact of household registration liberalization on urban economic growth. The results show that (1) did benchmark results show that the policy effect is reflected in the expansion of scale rather than efficiency improvement; (2) the regression results of LLM policy intensity score show that there is a significant ‘dose effect’ in the policy effect and the traditional 0/1 dummy variable underestimates the policy effect due to ignoring the intensity difference; (3) the dynamic effect shows that the policy effect lags behind 2 years and the duration is limited; (4) heterogeneity analysis shows that megacities (Shenzhen and Guangzhou) have a significant negative effect (−1.41%), while ordinary large cities have a positive effect but not significant, revealing that the effect of policy is highly dependent on urban carrying capacity. This article provides an empirical basis for the differentiated household registration reform and shows the methodological value of LLM in policy evaluation.