A Text Convolution Network-Based Policy Uncertainty Index : An Application for China

Newspaper text has become one of the critical data sources for measuring economic policy uncertainty. In some cases, newspaper discussions of policy uncertainty are expressed in subtle and implied ways. Hence, it is necessary to mimic market practitioners to process and identify perceived, concept-based uncertainties. Using the text convolution network, this study develops an EPU index for China’s real estate market. The index effectively captures the characteristics of policy uncertainty in the market under media censorship and conservative cultural context from the perceived uncertainty angle. We highlight that the approach can also be widely applied to other concept-based text analyses