Information Flows within the US State Department

What factors are important for information sourcing within the US foreign policy bureaucracy? Are these factors generally agreed upon, or do they vary depending across different parts of institutions? One important place where this information is sourced is through the US State Department which provides large quantities of information and analysis to the rest of the US government. Information sourcing matters not just in terms of which Embassies send the most information back to Washington DC, but also in how the Embassies provide information to each other and therefore give context to their analysis. If more information is passed between Embassies in countries that are at war with each other, then Embassies will be better situated to react to security events and if more information is sent between Embassies in countries that have strong economic ties, then the Embassies will be better able to react to economic events. In this paper, we use social network analysis on a dataset of 400,000 diplomatic cables sent within the US State Department in 1973 and 1974 to look at which factors influence the flow of information between US Embassies and from the central State Department to these Embassies. We find that while security issues are associated with high cable traffic across between a dyad (a pair of Embassies), the vast majority of variation in communication levels can be explained by economic relationships between countries. We also test whether the ordering of priorities differ in the central State Department's behavior and that of the Embassies. The idea that diplomats might "go native" and have a different set of priorities to the central State Department has a long history. A difference of opinion between the parts of the bureaucracy would be important for explaining how new issues get onto the agenda. However, we find that there is a high degree of similarity in how issues are weighed in the communication patterns from the center and between Embassies in the periphery. To validate these findings, we use the network models to predict dyads most likely to include content on particular issues, and cross-check the original cables finding that messages sent across these dyads do make heavier reference to the relevant issues.