EM-DAT and quantitative analysis

EM-DAT, the international disaster database, is perhaps the most comprehensive one available. It is helpful, needed, and welcome, with gratitude and kudos those who set it up, maintain it, and support it. They even removed the category "natural disaster" to keep it consistent with disaster science. For understanding quantitative trends and patterns, do the data suffice?

EM-DAT includes disasters from 1900 until the present, providing "country level human and economic losses for disasters with at least one of the following criteria:

-10 fatalities.

-100 affected people.

-A declaration of state of emergency.

-A call for international assistance."

They are appropriately careful in not claiming that the database covers all disasters. Since it is among the most comprehensive global disaster databases, but could never be comprehensive, could challenges emerge for quantitative analyses?

The ability and legal scope to declare a formal emergency and to call for international assistance have changed substantially since 1900. Neither has ever necessarily been based on actual disaster impacts. Formally calling for help and making declarations are mainly political decisions. Leaders can be reticent, despite palpable need, because they might lose face and look incompetent. Conversely, they could request help to seem decisive or to take credit for incoming aid. Another quirk is whether or not "international" would include aid between a colonial power and its colony (with aid flowing either way), whereas it is definitely "international" post-independence.

Consequently, the baseline is always changing for including or excluding disasters on the basis of "A declaration of state of emergency" or "A call for international assistance". Quantitative analyses of such disasters would have to account for these ever-shifting reasons or show that the results are not sensitive to these changes.

"People affected" has been defined as "requiring immediate assistance during a period of emergency". Disasters can involve heritage loss. If a cemetery but nothing else is ruined, requiring immediate assistance to respect those who are buried there, and to clean up and rehabilitate the site, then would that mean everyone or no one in the adjacent city is affected? "Period of emergency" is also typically a political definition.

The numbers "10" and "100" are arbitrary. 11 and 99 respectively, or 99 and 11 respectively, could serve. More insidious is how these thresholds deny some disaster experiences. A village of 90 people lacking the power to declare a state of emergency could have 9 deaths in a lightning strike and get through the situation with nearby aid. It would not be included in EM-DAT, despite this calamity. Or the "nearby aid" could be a neighbour over an international border, suddenly satisfying the criterion "A call for international assistance", even if informal. Theoretical and practical analyses show how the overall cumulative effect of allegedly "small" but more frequent disasters can outweigh the long-term impacts of "big" ones matching EM-DAT's criteria.

Another disaster database, DesInventar, was set up in 1994 to improve understanding of local impacts and disaggregated data. It is not clear that this timeframe suffices to reflect disaster patterns and trends. Meanwhile, as worldwide information becomes more easily available and findable, more disasters will (quite rightly) be reported. Simultaneously, worldwide misinformation and disinformation becomes more easily available and findable. Baselines for the numbers are moving, which does not preclude trend or pattern analysis, but which does present hurdles to overcome. Irrespective, this database remains invaluable for its data, its method, and its self-critique.

The same is true for EM-DAT. The situation has always been complementarity, never competition, among EM-DAT, DesInventar, and other sources. Nor should the limitations, barriers, and inconsistencies ever be a reason or excuse to undermine or stop either database. Instead, we ought to support both databases alongside many more data sources, recognising and explicitly expressing the advantages and limitations of each.

For now, a major limitation remains the adequacy of the data for analysing quantitative trends and patterns in disasters. And normalisation of disaster numbers has not even been addressed! We need much more before reaching conclusions about how disaster numbers are or are not changing.