We developed a machine learning algorithm to “read” China’s official newspaper — the People’s Daily — and predict its policy changes using only the information in that newspaper. The output of this algorithm, the Policy Change Index (PCI) for China, is a leading indicator of the actual policy changes in China since 1951. The nearby figure plots the PCI against the ground truth of China’s policy changes. The PCI often spikes months before policy changes take place, validating the index’s predictive power.
For details, see “Reading China: Predicting Policy Change with Machine Learning.”
A spike in the PCI signals a policy change, while a vertical bar marks the occurrence of a policy change labeled by the event.
Sign up to stay up to date on the PCI project.
Jan 15, 2019: We discussed what the latest PCI has to say about the US-China trade conflict.
Dec 3, 2018: We demonstrated the PCI design in a simulated example without using the People’s Daily text.
Nov 27, 2018: Zhong presented the PCI for China at the inaugural Policy Simulation Library meetup at AEI.
Nov 19, 2018: We talked about how to use machine learning to detect structural differences in complex data on MLconf.
Zhong is presenting the PCI project at the 2019 Strata Data Conference in London.
The PCI for China will be upgraded to a monthly index.
The PCIs for Cuba and East Germany are under way.