SOCIAL SUSTAINABILITY AND INCLUSION Resilience Metrics Paola Ballon and Jose Cuesta SSI Global Unit July 2023 Outline I. Objective II. Conceptual Framework Multiple Approaches and Frameworks III. A Proposal based on the SSI-GP Framework Metric Indicators Scenarios Results IV. Next Steps: Resilience Metrics for CLD Operations 2 I. Objectives Setting the Stage A Middle Ground Moving Forward Presenting the key building Understanding Adapting blocks needed and agreeing community to develop on the merits level resilience community and limitations metrics to level resilience of those CLD metrics (with metrics operations application to Kenya) 3 II. Multiple Approaches and Frameworks • Resilience is a combination of preparedness and vulnerability • Kusumati et al, (2014). Int. Journal of Disaster Risk Reduction (10) 4 II. Multiple Approaches and Frameworks • Resilience is a combination of preparedness and vulnerability We focus on vulnerability to hazards and its links with social and economic dimensions, and community/institutional capacity. 5 III. A proposal based on the SSI-GP Framework Definition: “Resilient societies are those where everyone, including poor and marginalized groups, are safe and can withstand shocks and protect the integrity of their culture� (Barron et. al, 2023, p 26) Resilience is the ability, capacity, and flexibility to prepare for, cope with, recover from, and adapt to shocks over time, it interacts with other social phenomena: without resilience, one see exclusion, eroding trust and voice, more unrest. 6 III. A proposal based on the SSI-GP Framework Framework: Resilience is a driver (and is being driven by) of social sustainability • SSI, WB.(2023). 7 III. Metric Based on this definition we propose to compute a metric of resilience at the community/county level that addresses two elements: • Levels of resilience: individual and community (‘county’) levels • Dimensions of resilience: • Economic capacity • Social cohesion • Institutional capacity • Hazards: climate and shocks. The Resilience Index is computed as the geometric mean of the four-dimensional subindices. Each subindex is the average of the two indicators that compose it. The methodology of this calculation is based on Human Development Index Methodology (HDR, 2023). 8 III. Indicators Our choice of indicators is based on the availability of data to satisfy the following criteria: Breadth: How well can we proxy the construct/dimension? Coverage: How easy it is to gather data for all communities (counties – Kenya)? Replicability: Over time: How feasible would be to replicate this metric over time to track changes? Across countries: How would it be to compute the metric for other countries or globally? 9 III. Indicators Dimension Level Indicators Community Per capita real GDP Wealth score, based on asset ownership (% of household with tv, radio, mobile, Economic Capacity Individual/household and computer) and quality of housing indicators (% of household with adequate walls, roof, and floor) Proportion of population who thinks trust in their community has rised or stayed Social Cohesion Individual/household the same Proportion of population who feels safe in their community Budget execution Institutional Capacity Community Number of hospital beds per 10000 inhabitants Anomaly in mean temperature, median projection under a neutral scenario of Community Hazards: Climate & Shocks greenhouse gas emissions Individual/household Proportion of households affected by a shock (any type) 10 III. Results: Resilience and Poverty Counties such as Turkana and Kitui that are water scarce are Resilience is associated with poverty. Counties also the less resilient, while Nairobi and counties located in that are less resilient also exhibit higher poverty the center of the Kenya are more resilient. 11 III. What is driving resilience in Kenya? A comparison of the cumulative density distributions of the four subindices shows that resilience in Kenya is driven by social cohesion (further right curve), while economic capacity and institutional capacity are weaker drivers, but hazards is the weakest. Counties in Kenya aiming to improve resilience to climate shocks should improve their economic and institutional capacity to address them. 12 IV. Resilience metrics for CLD operations • Building on the metrics of resilience for Kenya we propose to test the methodology to measure resilience at the project level to support locally led climate action. Climate Outcomes Hazards • The testing framework will map climate hazards (ex: frequent droughts) to outcomes (such as food insecurity or livelihood loses) considering levels of social cohesion and the institutional and economic Institutional capacity capacity in communities where the project takes place. Economic capacity • To do we will apply a three-step approach: Social Cohesion • Identify outcomes of interest in projects and define indicators. • Select projects for testing. • Measure the impact of projects on these resilience The result is a set indicators that capture the outcomes. extent to which CLD operations influence resilience in the community 13 IV. Resilience metrics for CLD operations Next steps: • Develop a joint work program across SSI GSGs comprising Social Dimensions of Climate Change, Social Cohesion, and Data and Analytics • Conduct a literature review of initiatives within and outside the Bank aiming to measure resilience at the project level (ex: Hallegate’s work on ‘project-resilience’) • Adapt the SSI-GP conceptual framework to CLD metrics of resilience • Propose a list of indicators and metrics at the project level capturing the extent to which the project leads to resilient outcomes • Specifically for IP populations, adjust project level indicators to be meaningful to IP self-defined concepts • Develop a technical note for TTLs on how to integrate resilience into operations 14 Thank you! Appendix 16 A. Scenarios One indicator per dimension Dimension Baseline I Variant I Economic Capacity - Per capita real GDP - Per capita real GDP - Proportion of population who thinks trust in their - Proportion of population who thinks trust in their Social Cohesion community has rised or stayed the same community has rised or stayed the same Institutional Capacity - Number of hospital beds per 10000 inhabitants - Number of hospital beds per 10000 inhabitants - Anomaly in mean temperature, median projection Climate & Shocks - Proportion of households affected by a shock under a neutral scenario of greenhouse gas emissions 17 A. Scenarios Two indicators per dimension Baseline II Variant II - Per capita real GDP - Per capita real GDP Economic Capacity - Wealth score* - Wealth score* - Proportion of population who thinks trust in their - Proportion of population who thinks trust in their community has rised or stayed the same community has rised or stayed the same Social Cohesion - Proportion of population who feels safe in their - Proportion of population who feels safe in their community community - Number of hospital beds per 1000 inhabitants - Number of hospital beds per 1000 inhabitants Institutional Capacity - Budget execution - Budget execution - Anomaly in mean temperature, median projection - Anomaly in mean temperature, median projection Climate & Shocks under a neutral scenario of greenhouse gas emissions under a neutral scenario of greenhouse gas emissions - Proportion of households affected by a shock * Composite computed on the basis of ownership of assets: tv, radio, mobile, and computer; and quality of housing indicators: % of household with adequate walls, roof, and floor. 18 B. Robustness Analysis To assess the sensitivity of results we consider two scenarios. The first considers one indicator per dimension, and the second two indicators per dimension. For each scenario we test a variant by changing the hazard indicator (see Appendix A). Figure B.1: reports the box plots for these scenarios and variants. Table B.1 and B.2 report the agreement in the ranking of counties between pair of scenarios, and the transitions (number of positions a county changes). From these statistics we conclude that a metric based on two indicators per dimension is preferable as it shows greater agreement in the raking of counties and fewer transitions. 19 B. Robustness Analysis Figure B.1: Resilience Index: Scenario Box plots Resilience index 20 B. Robustness Analysis Table B.1: Agreement between scenarios Rank correlation between RI measures Baseline 1 Trial 1 Baseline 2 Trial 2 Baseline 1 1 Trial 1 0.7627*** 1 Baseline 2 0.7806*** 0.8704*** 1 Trial 2 0.6270*** 0.9278*** 0.9322*** 1 Significance level: ***p<0.01, **p<0.05, *p<0.1 Table Transition Matrix B.2: baseline 1 vs baseline 2 vs Number of positions trial 1 trial 2 0 to 4 difference in positions 51.06% 65.96% 5 to 9 difference in positions 23.41% 27.66% 10+ difference in positions 25.53% 6.38% 21