Natural Hazards

, Volume 35, Issue 2, pp 211–228

How Accurate are Disaster Loss Data? The Case of U.S. Flood Damage


    • Environmental and Societal Impacts GroupNational Center for Atmospheric Research
    • Center for Science and Technology Policy ResearchUniversity of Colorado

DOI: 10.1007/s11069-004-4808-4

Cite this article as:
DOWNTON, M.W. & PIELKE, R.A. Nat Hazards (2005) 35: 211. doi:10.1007/s11069-004-4808-4


Policy makers need accurate disaster loss data for decisions about disaster assistance, policy evaluation, and scientific research priorities. But loss estimation is difficult in a disaster situation, and initial loss estimates are seldom evaluated in comparison with actual costs. This paper uses the example of historical flood damage data in the U.S. to evaluate disaster loss data. It evaluates the accuracy of historical flood damage estimates from two federal agencies. The U.S. National Weather Service (NWS) has compiled annual flood loss estimates for each state since 1955. Comparison of the NWS data with similar estimates from five state emergency management agencies reveals substantial disagreement between estimates from different sources. The Federal Emergency Management Agency (FEMA) began in the 1990s to systematically collect damage estimates and cost data associated with its disaster assistance programs. Comparison of early damage estimates with actual expenditures in a California flood disaster reveals large errors in some estimates for individual counties, but no statistically significant tendency to underestimate or overestimate. Positive and negative errors tend to average out and the total damage estimate for the state approximates the final expenditures. Both comparisons indicate that damage estimates for small events or local jurisdictions often are extremely inaccurate. On the other hand, estimates aggregated over large areas or long time periods appear to be reasonably reliable; that is, this study finds that independent estimates for events with losses greater than $500 million disagree by less than 40. The paper suggests ways of interpreting and using such loss estimates to reduce the likelihood of misinterpretation.


disaster lossloss estimationflood damagecost estimates

Copyright information

© Springer 2005