Human and economic impacts of natural disasters: can we trust the global data?

Reliable and complete data held in disaster databases are imperative to inform effective disaster preparedness and mitigation policies. Nonetheless, disaster databases are highly prone to missingness. In this article, we conduct a missing data diagnosis of the widely-cited, global disaster database, the Emergency Events Database (EM-DAT) to identify the extent and potential determinants of missing data within EM-DAT. In addition, through a review of prominent empirical literature, we contextualise how missing data within EM-DAT has been handled previously. A large proportion of missing data was identified for disasters attributed to natural hazards occurring between 1990 and 2020, particularly on the economic losses. The year the disaster occurred, income-classification of the affected country and disaster type were all significant predictors of missingness for key human and economic loss variables. Accordingly, data are unlikely to be missing completely at random. Advanced statistical methods to handle missing data are thus warranted when analysing disaster data to minimise the risk of biasing statistical inferences and to ensure global disaster data can be trusted.카지노사이트

As the global effects of climate change are felt more intensively, so too are the human and economic consequences of catastrophic disaster events. In 2020 alone, disaster events attributed to natural hazards affected approximately 100 million people, accounted for an estimated 190 billion US$ of global economic losses and resulted in 15,082 deaths1,2. In light of COP-26 and recent topical events including, but certainly not limited to, the Haitian Earthquake (2021), Central European Floods (2020) and Australian Bushfires (2020), a renewed urgency has been granted to the research of, preparedness to and mitigation of disasters attributed to natural hazards.

Comprehensive historical data held in disaster databases are central for numerous purposes across both the public and private domain to inform emergency disaster relief management; configure catastrophe risk assessment models; and conduct cost-benefit analyses of disaster risk reduction policies3. However, inconsistencies in the reporting of disaster events and methodological difficulties quantifying their impacts, mean that disaster databases are prone to gaps in data availability3,4. As a result, the scope and reliability of statistical inferences which can be made from disaster data are reduced5. Systematic reporting of disaster events is required to minimise missing data. However, to achieve this is a major challenge. Instead, the use of valid and often complex statistical methods to account for missing data are required6,7.

Missing data is a common issue across all research areas. Standard practices are adopted in randomised clinical trials to combat patient attrition8,9 and in survey-based observational studies to handle unit- and item-non response10,11. However, within the disaster literature, there is little insight as to how missing data should be handled.바카라사이트

To date, there are six disaster databases which have global coverage: the Emergency Events Database (EM-DAT); NatCatSERVICE; Sigma; GLIDE; GFDRR; and BD CATNAT Global12. This analysis utilises EM-DAT data alone, as it is the only publicly available, global disaster database and is widely cited; an initial search of the terms: ‘EM-DAT’, ‘CRED’, ‘Emergency Events Database’ and ‘International Disaster Database’ in Google Scholar returned 21,000 search results spanning numerous disciplines. EM-DAT was founded in 1988 by the Université catholique de Louvain (Belgium) with support from the United States Agency for International Development (USAID), the World Health Organisation (WHO) and the Belgium Government13. Data are collated from sources including the United Nations, reinsurance firms, research institutions and the press. As well as reporting the occurrence of major disastrous events attributed to natural, technological and complex hazards, its meta-data captures the human and economic impacts to assist national and international humanitarian action.

In this study, we conduct a missing data diagnosis to assess the extent of missing data in EM-DAT and to identify observable factors associated with missingness. We restrict our analysis to disaster events attributed to natural hazards occurring between the years 1990 and 2020. In addition, through a review of highly-cited empirical literature utilising EM-DAT data, we illustrate how missing data has previously been dealt with. We conclude by providing a discussion on the potential methods to handle missing data within disaster databases.

The state of missing data in EM-DAT
The advent of digital technologies in academia and research from the 1980s initiated a proliferation in the collection of data and subsequently, the adoption of data governance and data quality tools to ensure its reliability14. In the disaster space, technological advances in disaster surveillance and a progressive global agenda towards standardising disaster reporting, embedded in the 2015–2030 Sendai framework15, favour a climate for complete data reporting. Despite this, there was a high proportion of missing data in EM-DAT for disaster events attributed to natural hazards occurring between 1990 and 2020 (Fig. 1). This was particularly evident for the reporting of economic losses: data were missing for 96.2% of disaster events on reconstruction costs, for 88.1% on insured damages and 41.5% on total estimated damages. In the three months following a disaster event, when EM-DAT collates the majority of its data, precise information on reconstruction costs and insured damages is likely to be sparse. Reporting of human losses were more complete, with proportions of missing data ranging from 1.3% to 22.3%. Given that the volume of external disaster aid hinges predominantly on death tolls and injury counts, this finding is unsurprising. Of note, there were substantial inconsistencies in the reporting of both human and economic losses, where in both cases, aggregate variables (total deaths and total estimated damages) were better informed. In particular, missing data on total deaths were negligible (1.3%). Hence, in this case, there is little risk of missing data biasing statistical inferences.온라인카지노

Leave a Reply