4 Areas for further consideration
Given that the CRM is intended as a live resource, source inputs are likely to evolve over time – particularly as new sources of risk information emerge. For further details on diagnostics of the CRM and its outputs see Supplementary Text 3 in the Annex. While efforts have been made to ensure that methodological inputs are a robust reflection of underlying vulnerability and emerging threats conditions, a number of design options should be considered.
Below we highlight priority areas for further development of the CRM:
4.1 Indicator selection
Aligning indicators with ongoing WB monitoring tools. Where possible, the team chose indicators and thresholds aligned with (or sourced from) existing WB risk monitoring initiatives. This includes data from the Food Price Monitor (Food security), Global Economic Prospects (Macro-fiscal), DSSI (Macro-fiscal), and the FCS list (Fragility) amongst others. Going forward, the team will continue to coordinate with relevant teams to ensure that planned WB risk monitoring systems will feed directly into the CRM once launched – including for example, outputs from FCV’s Conflict Risk Monitoring Initiative and AGF’s Food Security Monitoring Hub. In doing so, the hope is to provide a holistic view of risk and minimize different country-level warnings issued across the Bank’s various risk monitoring efforts.
Identify suitable forward-looking indicators for dimensions of compound risk, particularly those with fewer source indicators. Ongoing consideration of the selection of suitable source indicators will continue as the CRM evolves, in close conjunction with relevant sectoral experts across WB units and external partners. Notably, robust sources of information have yet to be identified for a number of emerging dimensions of compound risk. Household-level socio-economic vulnerability is one example where globally-applicable real-time information of changing risk remains scant (despite availability of country-specific and regional databases). In addition, a range of other dimensions are reliant on proxies with limited scope. Efforts will continue in seeking suitable indicators to improve the Monitor. Indeed, the CRM is tailored to easily accommodate new sources of information as they come online. Should no viable proxies be found for some dimensions of compound risk, emerging threats scores will be reflected solely on the number of dimensions with adequate sources.
Refine risk threshold values: The designation of risk thresholds plays a large role in determining the rank of countries across the CRM. At present most of the thresholds are done by via percentiles. However, choice of the cut-off is subjective (typically above the 80th percentile). It also does little to recognize any systematic changes in risk conditions – for example, insights from FEWSNET suggests that medium-term food insecurity is likely to worsen for all monitored countries (implying that the food security threshold may need to be lowered compared with other dimensions of compound risk). One alternative is to assign thresholds in accordance with critical values (rather than percentiles). This could be done via expert elicitation, drawing on insights from a range of technical specialists within the Bank (as well as outside partners).
In most cases, thresholds in the CRM are aligned to classifications used internally by relevant units or programs in the Bank (such as the CRW ERF thresholds), or where there are established external high-risk thresholds used widely. In some cases, neither internal nor external thresholds are available (or readily applicable to the CRM’s methodology). If so, inputs are sought from relevant sectoral teams within the Bank in gauging suitable levels of risk as a guide. Related to issue of compatibility and comparability of indicators of indicators/thresholds within each risk component a number of steps are typically taken in selection.
The selection of thresholds will be subject to continued consultation with the relevant internal teams for further refinement.
4.2 Outputs and user interface
Isolate countries in active crisis. To facilitate the identification of countries that warrant increased vigilance among the universe of countries flagged by the CRM, the team is considering isolating countries in active crisis. This could increase the informational value of the CRM by clearly focusing on countries not currently in crisis but that warrant heightened vigilance going forward.
Link to interactive dashboard and horizon scanning exercise. A digital dashboard will be used to visualize the data-driven outputs of the CRM (see Annex Figures 10–13 for screenshots of the existing prototype). Through this dashboard, the CRM will bring value added as an aggregator of risk information produced by different Bank teams and external partners, directing users to relevant data and resources on different aspects of compound risk. In addition, inputs from the CRM will be combined with a qualitative assessment (through regular horizon scanning meetings) will help ground-truth its outputs, reflect latest developments not captured in the data (i.e. political economy developments), and explore the different likelihood and potential impacts of key risk dimensions.
4.3 Opportunities and next steps
Beyond the methodological refinements described above, there are a number of opportunities that the team will consider in future iterations to expand the scope and value of the CRM. These include:
Model and weight interactions between dimensions of compound risk. The CRM currently models each dimension separately. This is done to promote transparency and simplicity – given the considerable complexities associated with predicting interactions across different threats. Future iterations of the CRM may wish to consider the use of functions that express how two (or more) dimensions of compound risk are likely to interact (whether increasing or diminishing overall levels of risk).
Factor in sub-national risk information. The CRM is currently focused on country-level risks. With that in mind, most of the indices and indicators are carried out at the national level. However, this does not account for potential heterogeneity within a country – recognizing that risk factors may differ from one region to the next. It also makes it difficult to use data collected at high spatial resolutions – such as satellite data, of information focused solely on urban environments. One option may be to combine the CRM with sub-national data using the Dashboard (as an added map layer). A similar approach is used by WFP’s Hungermaps live.
Factor in seasonality and intra-annual components. Issues of compounding risk are often affected by seasonality or factors that vary within a year. For example, the impacts of above/below average rainfall are especially relevant during growing seasons in countries highly reliant on agriculture; the socio-economic impacts of COVID are pronounced during the timing of peak exposure (and stages of policy response). While it is difficult to factor these into the CRM holistically, it is important to consider how issues of seasonality can be incorporated. At present, this is being trialed in calculating future natural hazard risks and conflict risk – though there may be opportunities to add temporal dimensions to other dimensions of compound risk.