Quantifying Minsky cycles

We develop a novel sentiment measure derived from survey data to empirically vali date the Minsky-Kindleberger view on financial crises. Using survey data from multiple countries, we decompose beliefs into components explained by public information that are orthogonal to optimal machine beliefs, constructing a framework that isolates sentiment and its dispersion among individuals. We show that deviations from machine-optimized benchmarks arise from systematic misaggregation of public information. The sentiment measure is validated through its predictive relationships with financial markets and belief dynamics consistent with heterogeneous-beliefs asset pricing theory. We extend this senti ment measure historically for a panel of 78 countries using machine learning models trained on BERT embeddings of historical news articles (1903-2020). The backcasted sentiment shows that shocks in median sentiment predict credit booms in the non-tradable corporate sector, which prior research has linked to financial crises, providing the first historically large-scale empirical validation of the Minsky cycle. We further show that sentiment, which is a misaggregation of public information, is influenced by memory-related dynamics, as the time elapsed since major crises and the share of young-to-old people in the population strongly predict surges in optimism even when recent economic developments are controlled for.