This paper sought to characterize the current state of climate and disaster risk modeling with regards to equity to identify opportunities for improvement in practice and future research. We are currently seeing a boom in the development of climate and disaster risk analytics, as actors ranging from governments to the financial sector to humanitarian and development agencies seek to leverage new technologies and increased availability of spatial data about communities and the natural and built environment to respond to increasing concerns about natural hazards. Unfortunately our research shows that, despite widespread recognition that the impacts of climate change and disaster are disproportionately experienced by the most vulnerable members of society, social vulnerability, whether linked to economic standing, gender, or otherwise is rarely accounted for in these analytics. As a result, policies and other efforts aimed at risk mitigation may be ineffective, or worse, contribute to increasing inequity.

To investigate this further, our team focused on one area of climate and disaster risk assessment where we have extensive experience, international development. We turned to several major sources of research and publications in the sector: the World Bank’s Open Knowledge Repository, the online libraries of the Asian Development Bank and Inter-American Development Bank, and PreventionWeb, a disaster-focused library which contains materials and resources produced by the United Nations and many large international NGOs. Using a consistent set of search terms and inclusion criteria, we identified 69 risk assessments from across the platforms that had been conducted by major international development organizations since 2010. We read each report carefully, and classified the extent and approaches used for addressing differentiated disaster risks faced by different demographic groups in the area of study.

What we found was both notable and distressing, if not entirely unsurprising. As shown in Table 1, the majority of risk assessments in the international development sector conducted between 2010 and 2022 did not address social inequities in the distribution of climate and disaster risk. Those that did characterize this form of inequity most often relied on qualitative descriptions or indices of vulnerability, neither of which are well suited to the design or evaluation of specific climate or disaster risk management interventions. In the end, only 11 of the 69 (~16%) assessments we evaluated incorporated quantitative measures of social inequity directly into the risk model. 

Table 1: Summary of Risk Assessment Review Results

*Categories are not mutually exclusive as some risk assessments considered more than one parameter

As a way of exploring the consequences of not including equity sensitivity into risk modeling, we re-analyzed one of the assessments we found in the first part of our study . Here we chose a multi-hazard risk assessment conducted for the country of Nepal in 2012. Following the typology described in Table 1, the original assessment was a Type 0, with no consideration of equity included in the model, or discussed in the associated report. Using more recently available data as well as damage and fatality estimations from other studies on disaster and climate risk in the region, we updated the results with a Type 3, 4, and 5 analysis that incorporate spatial distribution of vulnerable groups vis a vis hazards, estimates of differential impacts of hazards on vulnerable groups, and financial impacts of disaster weighted by economic status, respectively. 

Unsurprisingly, we found that these equity sensitive approaches uncovered substantial disparities in how disaster risk is distributed in the population – a finding that the original assessment missed. For example, our Type 3 and Type 4 analysis, displayed in Figure 1, found notable differences in the fatality rates for all groups, especially for different income groups and caste types, which the original assessment failed to identify. Similarly, our Type 5 analysis reversed the findings of the original analysis, and found that the disaster impacts are higher in poorer districts (Figure 3). Indeed, the focus on asset losses in Type 0 analysis tends to highlight high-income neighborhoods with valuable assets. In contrast, Type 5 analysis, which measures welfare loss, shifts the emphasis to low-income neighborhoods, where even minor financial losses can have a disproportionately large impact on welfare (Figure 2).

Fig 1 | Disaggregated Fatality Rates. a-e Previously modeled fatality rates disaggregated by differentiated exposure (Type 3 approach, red) or by both differentiated exposure and vulnerability (Type 4 approach, orange). Both a Type 3 and Type 4 approach were used to disaggregate fatalities by (a) gender, (b) age and (d) disability. A Type 4 approach was used to disaggregate fatalities by (c) caste type and (e) income.  Although the income analysis is technically a Type 3 approach, in practice it is a Type 4 approach as building vulnerability, which is included in the Type 3 analysis, is strongly correlated with income.  

Fig 2 | Welfare and total losses by district and income level in Nepal for a 1 in 500-year earthquake (in million euros). a, b Comparison of a Type 0 and Type 5 approach to an earthquake risk assessment initially conducted for Nepal. (a) Total losses (Type 0 approach) for a 1 in 500-year earthquake by district (high income districts are red, medium income districts are orange, low income districts are yellow) in Nepal and (b) welfare losses (Type 5 approach) for the same event. Total losses are lowest for the poorest districts and highest for the wealthiest ones. By comparison, welfare losses are highest for the poorest districts, and lowest for the wealthiest ones. District-level losses – both total and welfare – are displayed as proportional black circles. 

Fig 3 | Expected weighted and unweighted losses by district in Nepal for a 1 in 500-year earthquake (in million euros). From left to right, the districts are ordered by average income where Bajhang has the lowest average income and Manang has the highest. Unweighted losses (red; brown line of best fit) are lowest for the poorest districts and steadily increase for the wealthiest ones. When the analysis considers weighted losses (orange; blue line of best fit) as per the Type 5 welfare-loss approach, this phenomenon is reversed.  In this analysis, the poorest districts experience the greatest losses, which decrease with average district income. 

The results of our reanalysis of a risk assessment for Nepal suggest that, on aggregate, the effects of not accounting for equity in disaster risk assessments are likely to be quite impactful. In the worst case, risk assessments that do not account for equity may misallocate risk reduction investments in the wealthiest communities despite the fact that welfare losses are higher in the poorest ones. Indeed, accounting for equity in disaster risk models would support the design of more effective interventions that better identify communities where specific risk reduction measures have the most impact. Despite reasonable claims that this type of detailed examination is challenging due to a lack of disaggregated data on topics such as locations of vulnerable groups and impacts from past disasters, we show that some approaches are still possible with available information.

Taken together, our work suggests that considerable effort is required to better account for equity, in all of its forms, in disaster risk modeling. Our typology, and the corresponding analysis, confirmed that few risk assessments account for equity and our reevaluation of a risk assessment for Nepal highlights the negative impacts of this failure. Neglecting such equity considerations may thus undermine the findings, conclusion, and policy measures of current climate and disaster risk management. International agencies and organizations, such as the development banks and the United Nations, are major funders of risk modeling and are well positioned to influence the sector. Meanwhile, scientists can further develop and refine the techniques these organizations use to conduct risk assessments that account for distribution, along with other forms, of equity. To that end, with partners from the University of Michigan, the University of Colorado, Arup, and the World Bank’s Global Facility for Disaster Reduction and Recovery, our authoring team are launching a partnership to design better methods for equity-sensitive risk modeling. Although this will not address the structural issues that produce inequality in climate and disaster impacts, such models could represent a significant step towards this goal. 

You can find our full paper here: https://www.nature.com/articles/s43247-023-01039-2