Global Flood Risk in the 2026/27 El Niño: An Empirical Estimate from the Historical Relationship

We estimate the response of recorded flood activity to the Niño3.4 sea-surface-temperature anomaly for 104 countries, using a monthly country panel built from EM-DAT (1980–2025) and the NOAA Niño3.4 series. Identification comes from within-country, within-calendar-month variation in the index, net of a linear trend in reporting intensity. The headline outcome is the expected number of recorded floods, estimated by quasi-Poisson regression; a logistic model of flood occurrence is reported alongside and gives near-identical answers. El Niño redistributes flood activity but does not change significantly the world total. The analysis finds an increase in the number of floods in 44 countries, with this increase significant in 10 countries. Country effects are large and geographically coherent, from 2.26 per °C in Uruguay to 0.46 in Togo. Applying the ECMWF Niño3.4 outlook against a neutral counterfactual, the expected number of recorded floods in these countries rises by 2.05× in 2026 and 2.88× in 2027. In parallel, the analysis suggests a decrease in the number of floods in 60 countries, with this decrease significant in 19 countries. In absolute terms, the countries with a positive signal are expected to record +42.9 additional floods in 2026 and +57.1 in 2027 on the main scenario, while the countries with a negative signal are expected to record 31.0 fewer floods in 2026 and 30.6 fewer in 2027. The two largely offset, leaving net changes of +11.8 and +26.5 events. As a comparison, the estimates agree in sign with the model-based El Niño flood signal of Ward et al. (2014) for 86% of the 29 countries whose estimated effect is significant. A reduced expected flood count does not imply reduced risk. The estimates describe a shift in the expected rate at which countries enter the disaster register; they place no bound on the severity of individual events. Damaging floods are recorded in every ENSO phase, and in countries with high floodplain exposure a single event within a suppressed season can exceed the annual average loss by a large margin. The results do not support any reduction in preparedness.