Permutation-Based Testing of Topic Co-occurrence: A Network Analysis of Reddit Debates on DOGE, Tariffs, and the Big Beautiful Bill

Abstract

We study how policy arguments are structured in Reddit discussions of three initiatives from the early period of Trump's second presidency: the Department of Government Efficiency, the Liberation Day Tariffs, and the Big Beautiful Bill. We identify recurring policy considerations using a dictionary-based approach that is validated through manual coding and applied to full-text submissions. To distinguish meaningful associations from chance, we use a permutation-based test that compares observed co-occurrence counts for each pair of considerations to counts from randomly shuffled data with the same overall frequencies. From the pairs that pass this test, we build networks for each policy and describe their main features, including central topics, clusters of related considerations, and bridge topics that link clusters. The results show that the three debates have different argument structures and illustrate a simple way to add statistical testing to topic co-occurrence analysis.