The Future of US Alliances and Partnerships: A Data Science Approach

The subject of US alliances and strategic partnerships is becoming more salient in the era of strategic competition. Although policy-makers and academics devoted considerable effort to the study of US alliances during the Cold War, the topic receded in attention over the past two decades. Now, in the current environment of US threat perceptions from China and Russia, academics and policy-makers are reorienting their gaze to the great power competition that looks to dominate US international interests for the foreseeable future.

High-level officials recognize the role that alliances and partnerships serve in advancing US interests. For example, President Joe Biden traveled to Europe for his first overseas trip, which included a summit of the North Atlantic Treaty Organization (NATO) alliance. Secretary of State Antony Blinken held his first diplomatic meetings with neighboring Mexico and Canada through virtual interaction. In the Indo-Pacific region, the previous Trump administration resurrected the senior-level “Quad” discussions between the US, Japan, Australia, and India. Through these high-profile interactions, US policy-makers have signaled that alliances and partnerships will remain central to US global interests.

Because of the increased need to understand the role of alliances and partnerships in an era of strategic competition, US policy-makers need to modernize their tools for conducting analysis. This study aims to draw on rising interest in Big Data and apply data science methods to the long-standing topic of alliance management for advancing US national security interests. To this end, CNA is building linkages between academic disciplines and developing a more rigorous and traceable understanding of US alliances and partnerships to inform US policy-makers.

In this report, we provide select findings from this exploratory research effort and recommend caution when interpreting results about future relationship outcomes. This includes acknowledging the assumptions that went into model development and the subjective nature of variable selection, as with all models.
Keeping these caveats in mind, we assembled datasets with key international security indicators, built machine-learning models, and derived the following results about the future of alliances and partnerships.