Problems of Warning: Forecast Precision and Forecast Accuracy in Declassified State Department Cables


Why is there relatively little early action or preventative diplomacy to stop or mitigate civil war onset or mass atrocities? This paper discusses several explanations for this lack of early action before arguing that a dearth of early action is best explained by the limitations of intelligence. I then advance an argument about the proper way to conduct retrospective analysis of intelligence and intelligence failures, before reviewing the existing literature on early warning, intelligence, and preventive action. In the empirical section of the paper I describe the dataset I use to examine my theory about the link between intelligence and early action. This dataset is a comprehensive collection of 1.8 million declassified U.S. State Department cables sent between 1972 and 1978. Finally, I describe a machine-assisted text analysis approach to searching for early warning in the corpus, and conclude with the limitations of my study and what I conclude about early warning.

Andy Halterman
Andy Halterman
PhD Candidate

My research interests include natural language processing, text as data, and subnational armed conflict