I am very happy that our paper, The Geopolitical Threat Index: A Text-Based Computational Approach to Identifying Foreign Threats, has finally appeared in International Studies Quarterly. Peter and I created the Geopolitical Threat Index (GTI) that covers more than 150 years (1861 to 2017) through a computational analysis of the New York Times summaries in this study. You can see in the figure that the our index highly correlate with occurrences of important geopolitical events for the United States. We also have figures for individual countries such as Russia, China, Vietnam, Iraq, etc. in the article.
This has been my favorite project and I presented it in many places (Waseda, Innsbruck, and Musashi) over the years, because it demonstrates how easy it is both financially and technologically to perform interesting analysis: we collected the data through the free New York Times API and analyzed it using open source software (quanteda, LSX, and newsmap) on a regular laptop computer. I am sure that you will be surprised by the very short R scripts in the replication dataset.
We accomplished out analysis at low-cost by using two semisupervised machine learning tools, Newsmap and LSS, both of which are trained on the corpus and user-defined seed words. The seed words for Newsmap are historical names of countries and capital cities; the seed words for LSS are 14 words related to hostile or friendly actions.
You might wonder if this methodology applies to non-English texts. It really does as I analyzed Hebrew and Japanese newspapers successfully in terms of North Korea’s and Iran’s nuclear threats in my upcoming paper.
One thought on “New paper on historical geopolitical threats to the US”