Narrative economics using textual analysis of newspaper data: new insights into the U.S. Silver Purchase Act and Chinese price level in 1928–1936

In light of the recent advancement in economic narrative analysis, we develop a computational textual analysis method to study economic history. In this method, we collect narrative data from newspapers to measure economic trends. In particular, the popularity (frequency) of a narrative (keyword) on the newspapers is used as the proxy of the amount of economic activities associated with the narrative term; a high frequency indicates that there is a high volume of economic activities associated with the narrative term and vice versa. Regularized regression algorithms are then applied on the narrative frequency data to identify narrative terms whose associated microeconomic activities have macroeconomic impact. We apply the method to study a classic topic in Chinese economic history research: U.S. Silver Purchase Act and the Chinese price level in 1928–1936. Our results provide new insights into this controversial subject. For example, we find that the economic activity associated with the narrative term silver stock had no impact on the Chinese price level, which is contrary to previous research on the topic by Friedman and Schwartz [10]. Meanwhile, economic activities associated with the narrative terms U.S. silver purchase act and silver export are found to have a negative impact on the Chinese price level. This suggests the concerns at that time about the effects of U.S. Silver Purchase Act on the Chinese economy were not misplaced.