Earthquake Deaths Fall Worldwide Even as Disasters and Losses Rise, 45-Year Study Finds

A sweeping new analysis of more than four decades of global earthquake records has revealed a striking paradox at the heart of modern disaster risk: the number of recorded earthquake disasters and the cumulative economic losses they inflict have both climbed since 1980, yet deaths relative to population size have steadily declined, particularly after the early 2000s. The study, published in the journal Natural Hazards and Earth System Sciences, offers one of the most comprehensive human-Earth assessments of earthquake impacts ever attempted, weaving together seismology, geography, and socioeconomic data to explain why similar earthquakes can produce wildly different outcomes in different countries.

Researchers led by Zekang Zhang of Hebei University of Engineering, together with colleagues from the Chinese Academy of Sciences, compiled earthquake disaster records from the Emergency Events Database, known as EM-DAT, covering the period from 1980 to 2024. Unlike instrumental earthquake catalogs, which capture virtually every seismic event detected by sensors, EM-DAT records only disasters that cross specific impact thresholds, such as causing at least ten deaths, affecting at least one hundred people, or triggering a state of emergency or a call for international assistance. This impact-oriented approach allowed the team to focus not on where and when the ground shakes, but on where and when earthquakes actually devastate societies.

The temporal findings are nuanced. The number of recorded earthquake disaster events shows a statistically significant upward trend, confirmed by the non-parametric Mann-Kendall test after correcting for autocorrelation in the annual series. Yet mortality-related indicators moved in the opposite direction: the mortality rate, defined as fatalities relative to the affected population, declined significantly over the 45-year window. The researchers emphasize that this divergence does not mean earthquakes are becoming less dangerous in a physical sense. Rather, it reflects changes in how societies are exposed, how they record disasters, and how effectively they protect their populations, with several catastrophic events, including the 2004 Sumatra-Andaman earthquake and Indian Ocean tsunami, the 2008 Wenchuan earthquake, the 2010 Haiti earthquake, the 2011 Great East Japan earthquake, and the 2023 Türkiye-Syria earthquake, dominating short-term fluctuations in the data.

To test whether these patterns were artifacts of a handful of extreme tragedies, the team ran sensitivity analyses excluding the five deadliest disasters of the era. The overall temporal tendencies remained consistent, suggesting that the decline in mortality relative to population is a genuine structural feature of the record rather than a statistical shadow cast by a few outliers. The team also adjusted all economic losses for inflation using the OECD Consumer Price Index, expressing them in constant US dollars, and cross-validated their EM-DAT extraction against the independent NCEI earthquake database for overlapping years, finding consistent temporal patterns in events, fatalities, and losses.

Perhaps the most visually compelling result concerns the migration of statistical centroids, the weighted geographic centers of disaster impacts. The unweighted centroid of recorded earthquake disasters shifted eastward across three consecutive 15-year periods, jumping 1,872.4 kilometers between 1980-1994 and 1995-2009, a displacement the permutation tests found statistically significant, and then a further 770.4 kilometers into the most recent period. But when the researchers weighted the centroids by fatalities, affected population, or economic losses, the trajectories diverged dramatically. The fatality-weighted centroid, for instance, leapt roughly 11,463 kilometers between the second and third periods, from the eastern Mediterranean region to the central Atlantic, a shift driven largely by the changing geography of mass-casualty events. These divergent paths demonstrate that where earthquakes strike, where people die, and where money is lost are three spatially distinct questions.

At continental and national scales, Asia dominates nearly every cumulative measure, accounting for the largest share of recorded events, affected populations, fatalities, and economic losses, a pattern tied to its vast populations overlapping the Pacific Ring of Fire and the Alpine-Himalayan seismic belt. Yet normalized indicators tell a subtler story. Densely populated countries such as China, India, and Indonesia contribute heavily to cumulative fatalities and affected populations, while asset-rich nations such as Japan, the United States, and several European countries register disproportionately high economic losses per event. Europe, despite contributing relatively few recorded events, shows a comparatively elevated share of average economic loss per disaster, underscoring how concentrated economic exposure can amplify the financial footprint of even infrequent earthquakes.

To move beyond description, the team applied the Geographical Detector model, a statistical framework designed to quantify how well the spatial stratification of an explanatory variable matches the spatial differentiation of an outcome. For mortality measured as deaths per million residents, earthquake magnitude emerged as the single strongest factor, with a q-statistic of 0.17, followed by seismotectonic type and disaster process type, which distinguished simple earthquakes from earthquake-tsunami compound events. Socioeconomic variables such as population density and GDP per capita showed weaker but still detectable individual explanatory power. For economic losses normalized by national GDP, magnitude again led with a q of 0.19, but population density rose to prominence with a q of 0.14, confirming that economic damage tracks the geography of exposed assets as much as the physics of shaking.

The interaction analysis delivered the study’s most methodologically important insight: combinations of factors consistently outperformed any single variable. For mortality, the pairing of magnitude and population density achieved a joint q of 0.33, roughly double the explanatory power of magnitude alone, while magnitude combined with GDP per capita reached 0.27. For economic losses, the interaction between population density and GDP per capita produced a q of 0.53, the highest in the entire analysis, followed by focal depth paired with tectonic setting at 0.47. Most factor pairs exhibited bilinear or nonlinear enhancement, meaning their joint spatial stratification explained more variance than the sum or the maximum of their individual contributions. In plain terms, deadly and costly earthquakes arise where severe seismic hazard and dense human and economic exposure overlap, not where either factor is extreme in isolation.

The authors are careful to frame these results as statistical spatial associations rather than causal mechanisms, and they acknowledge the limitations inherent in disaster databases. EM-DAT’s inclusion criteria and reporting practices vary across countries and decades, and economic loss figures depend heavily on post-disaster assessments and national reporting procedures. The national scale of the analysis may also mask subnational differences in vulnerability, building quality, and emergency response capacity. Bootstrap resampling and permutation tests were used to attach confidence intervals to centroid locations and displacements, and alternative discretization schemes for the Geographical Detector analysis produced stable q-statistics, bolstering confidence in the robustness of the identified patterns.

The practical implications are considerable. Because mortality and economic loss follow different spatial logics, the study argues that earthquake risk assessment and preparedness planning should not rely on seismic hazard maps alone. Regions where strong earthquake characteristics coincide with dense populations warrant prioritized attention to human losses and emergency readiness, while areas with concentrated economic assets require strategies focused on financial resilience and infrastructure protection. As urban expansion continues to push populations into tectonically active zones, the finding that deaths are declining relative to population size offers measured encouragement, evidence that development, exposure management, and resilience planning can bend the curve of disaster mortality even in an era of rising recorded losses and mounting economic exposure.

Subject of Research: Global spatiotemporal patterns of earthquake disaster impacts and their geophysical and socioeconomic drivers, 1980-2024

Article Title: Spatiotemporal patterns of global earthquake disaster impacts from a human-Earth perspective

Article References: Zhang, Z., Qiu, Y., Ye, Y., & Xuan, C. (2026). Spatiotemporal patterns of global earthquake disaster impacts from a human-Earth perspective. Natural Hazards and Earth System Sciences, 26(10), 4807-4824. https://doi.org/10.5194/nhess-26-4807-2026

Image Credits: AI Generated

DOI: 10.5194/nhess-26-4807-2026

Keywords: earthquakes, disaster risk, EM-DAT, mortality trends, economic losses, spatial analysis, centroid shift, Geographical Detector, population exposure, seismic hazard, resilience planning, human-Earth interactions