Paper detail

America Tweets China: A Fine-Grained Analysis of the State and Individual Characteristics Regarding Attitudes towards China

The U.S.-China relationship is arguably the most important bilateral relationship in the 21st century. Typically it is measured through opinion polls, for example, by Gallup and Pew Institute. In this paper, we propose a new method to measure U.S.-China relations using data from Twitter, one of the most popular social networks. Compared with traditional opinion polls, our method has two distinctive advantages. First, our sample size is significantly larger. National opinion polls have at most a few thousand samples. Our data set has 724,146 samples. The large size of our data set enables us to perform state level analysis, which so far even large opinion polls have left unexplored. Second, our method can control for fixed state and date effects. We first demonstrate the existence of inter-state and inter-day variances and then control for these variances in our regression analysis. Empirically, our study is able to replicate the stylized results from opinion polls as well as generate new insights. At the state level, we find New York, Michigan, Indiana and Arizona are the top four most China-friendly states. Wyoming, South Dakota, Kansas and Nevada are most homogeneous. At the individual level, we find attitudes towards China improve as an individual's Twitter experience grows longer and more intense. We also find individuals of Chinese ethnicity are statistically more China-friendly.