Consistency and legal innovation in the BIT universe

International investment law is marked by the proliferation of bilateral investment agreements(BITs). Close to 3000 BITs have been concluded by 2014 and virtually every country is a signatory. In this paper we construct a novel data set of 1623 BIT texts and compute q-gram string distances between treaty pairs to analyze similarity and dissimilarity in the BIT universe. First,we identify patterns of legal coherence(close mutual distances) and find strong support for the hypothesis that Western capital exporting countries are rule-setters in the treaty-making process. Second, we examine causes of legal innovation (change of treaty distance over time for treaty-making countries) and show that being hit by an Investor-State arbitration claim does not lead to a change in treaty design, in contrast to the predictions of previous studies. Our research thereby demonstrates that text as data models of treaty design can be a powerful tool for policy-makers, arbitrators and scholars to trace consistency and innovation within and across individual country's BIT networks.