Economic policy uncertainty indexes derived from People's Daily articles using neural topic modeling

This dataset presents economic policy uncertainty indexes for China generated through deep learning analysis of historical news text. The source corpus comprises over two million articles from the People's Daily published between 1946 and 2024. By employing a neural topic modeling approach, the data captures semantic context to identify five distinct dimensions of uncertainty, including financial, macroeconomic, exchange rate, housing, and market policy uncertainty. The dataset provides normalized probability scores aggregated at monthly and yearly intervals to reflect the intensity of policy-related discussions over time. These domain-specific indexes offer a granular alternative to traditional keyword frequency counts and allow for the differentiation of policy risks. The data can be integrated with corporate financial reports or macroeconomic indicators to support research into the effects of policy volatility on firm behavior and market dynamics.