Detecting Critical Junctures as They Unfold: Foundations, Measurement, and Applications from 120 Years of Reporting

Explaining why some countries grow rich while others remain poor requires knowing how — and when — institutions change. A large body of research argues that institutional change often occurs at critical junctures: moments when multiple institutional futures are possible, including a return to the status quo. Yet most empirical work examines critical junctures only in retrospect, risking the omission of pivotal moments when institutions were in flux but ultimately remained unchanged. We develop a new empirical approach to detect critical junctures as they unfold. Our key insight is to understand critical junctures as moments of uncertainty about institutions, which provides a text classifier with a theoretically motivated object to detect in contemporaneous discourse. In this chapter, we apply this idea to the study of historical critical junctures using newspaper text as a rich source of information regarding contemporary beliefs about institutional uncertainty. We translate this insight into a measurement strategy by training a large language model-based classifier, which we apply to the Times of London archive — 11.95 million articles spanning 1800 to 2019 — while focusing our main empirical analysis on 120 years of reporting from 1900 to 2019. This produces the first long-run, text-based measure of institutional uncertainty grounded in the concept of critical junctures. Compared to established indices of geopolitical and economic risk, our measure innovates on several fronts: (i) it captures high uncertainty even in periods without formal institutional change, consistent with ex-ante definitions of critical junctures; (ii) it provides earlier signals of major global convulsions, such as World War I; (iii) it evaluates candidate critical junctures using a common ex ante measure; and (iv) it opens new avenues for studying how critical junctures shape institutions and long-run development. Today, our Times-based index suggests that institutional uncertainty in the US and UK is as elevated as it was in the early 1940s — a signal missed by other leading measures of risk.