Strictly under embargo until Wednesday 22 September at 00:01 GMT (02:01 Geneva time)Key problems with disaster data include the lack of standardized collection methodologiesand definitions. The original information, collected from a variety of publicsources, is not specifically gathered for statistical purposes. So, even when the compilationapplies strict definitions for disaster events and parameters, the original suppliersof information may not. Moreover, data are not always complete for each disaster. Thequality of completion may vary according to the type of disaster (for example, thenumber of people affected by transport accidents is rarely reported) or its country ofoccurrence.Data on deaths are usually available because they are an immediate proxy for the severityof the disaster. However, the numbers put forward immediately after a disaster maysometimes be seriously revised, occasionally several months later.Data on the number of people affected by a disaster can provide some of the mostpotentially useful figures, for planning both disaster preparedness and response, butthey are sometimes poorly reported. Moreover, the definition of people affectedremains open to interpretation, political or otherwise. Even in the absence of manipulation,data may be extrapolated from old census information, with assumptions beingmade about percentages of an area’s population affected.Data can also be skewed because of the rationale behind data gathering. Reinsurancecompanies, for instance, systematically gather data on disaster occurrence in order toassess insurance risk, but with a priority in areas of the world where disaster insuranceis widespread. Their data may therefore miss out poor, disaster-affected regions whereinsurance is unaffordable or unavailable.For natural disasters over the last decade, data on deaths are missing for around onetenthof reported disasters; data on people affected are missing for around one-fifthof disasters; and data on economic damages are missing for 67 per cent of disasters.The figures should therefore be regarded as indicative. Relative changes and trends aremore useful to look at than absolute, isolated figures.Dates can be a source of ambiguity. For example, a declared date for a famine is bothnecessary and meaningless – a famine does not occur on a single day. In such cases,the date the appropriate body declares an official emergency has been used. Changesin national boundaries cause ambiguities in the data and may make long-term trendanalysis more complicated.However, in some cases, available data may differ greatly according to sources, be moreor less documented estimations and / or subject to controversies. In these cases, CREDalways compiles all available data or analysis to try to make its own documented estimation,which can be revised when more accurate data are provided.164ANNEX 1Caveats
Strictly under embargo until Wednesday 22 September at 00:01 GMT (02:01 Geneva time)Information systems have improved vastly in the last 25 years and statistical data arenow more easily available, intensified by an increasing sensitivity to disaster occurrenceand consequences. Nevertheless there are still discrepancies. An analysis of the qualityand accuracy of disaster data, performed by CRED in 2002, showed that occasionally,for the same disaster, differences of more than 20 per cent may exist between thequantitative data reported by the three major databases – EM-DAT (CRED), NatCat(Munich Re) and Sigma (Swiss Re).Despite efforts to verify and review data, the quality of disaster databases can only beas good as the reporting system. This, combined with the different aims of the threemajor disaster databases (risk and economic risk analysis for reinsurance companies,development agenda for CRED), may explain differences between data provided forsome disasters. However, in spite of these differences, the overall trends indicated bythe three databases remain similar.The lack of systematization and standardization of data collection is a major weaknesswhen it comes to long-term planning. Fortunately, due to increased pressuresfor accountability from various sources, many donors and development agencies havestarted paying attention to data collection and its methodologies.Part of the solution to this data problem lies in retrospective analysis. Data are mostoften publicly quoted and reported during a disaster event, but it is only long afterthe event, once the relief operation is over, that estimates of damage and death can beverified. Some data gatherers, like CRED, revisit the data; this accounts for retrospectiveannual disaster figures changing one, two and sometimes even three years after theevent.ANNEX 1Philippe Hoyois, senior research fellow with CRED, Regina Below, manager of CRED’sEM-DAT disaster database, and Debarati Guha-Sapir, director of CRED, preparedthe statistics annex. For further information, please contact: Centre for Research on theEpidemiology of Disasters (CRED), School of Public Health, Catholic University ofLouvain, 30.94, Clos Chapelle-aux-Champs, 1200 Brussels, Belgium, tel.: +32 2 7643327, fax: +32 2 764 3441, e-mail: contact@emdat.be, web site: www.emdat.beWorld Disasters Report 2010 – Disaster data165