Poverty and Inequality Monitoring: Latin America and the Caribbean COVID-19 Geospatial Vulnerability Indices and the GeoLAC repository for Geospatial Data The GeoLAC repository contains three maps of the LAC region that aim to inform policy (governments and the World Bank) around the COVID-19 crisis. These maps present three new vulnerability indeces that tackle the following topics: (1) Expected employment losses in local labor markets; (2) Structural socioeconomic vulnerabilities at subnational levels to various forms of shocks (economic, health, etc.); and (3) Socioeconomic vulnerabilities at subnational levels specific to the COVID-19 crisis. Index of Expected Employment Losses (IEEL): The current crisis has direct effects on economic activity and labor markets, but the mix of industries in a given region can determine both the magnitude of the negative effects and the speed of recovery. For example, it is expected that regions with high concentrations of industries such as hospitality would lose more jobs and take more time to recover. In contrast, sectors like the public sector and utilities may recover faster since job losses have been less prevalent. Thus, to understand the evolution of the labor markets, it is essential to capture both the heterogeneity labor markets have and their behavior over time. Building and mapping a sub-national index to measure the expected economic effects of the crisis at the local level could be a useful tool to predict the evolution of the crisis as well as target the available resources. The Index of Expected Employment Losses (IEEL) is a measure of the expected impact that the current crisis could have on employment in 2020 at subnational-levels. 1 The index is based on country-industry- level unemployment projections and the industry mix of subnational regions. In particular, it is calculated by multiplying county-industry-level projections of job losses times the employment industry-shares at the subnational-level obtained using representative household surveys. Plotting the IEEL in the web mapping application of the GeoLAC repository shows both the heterogeneity of the effects across regions as well as concentration of the effects in some areas (Figure 3). The map shows the sub-national index classified in deciles, where darker shades of red mean a worse expected labor market effect. Peru alone concentrates most of the sub-national regions in the worst decile, along with some sub-regions in Brazil. On the other hand, countries like Colombia, Chile, and Mexico show some heterogeneity in part because not all subregions are equally exposed to the economic shocks. It should be noted that measuring the IEEL at various points in time could depict the evolution of the labor market. The exercice can be repeated using projections or data for several years thus showing the past and expected changes of local labor markets conditions. It should be noted, however, that this index 1 In the spirit of the Chmura RTI index for the US: http://www.chmuraecon.com/interactive/covid-19-economic- vulnerability-index/ does not directly capture the shocks (i.e., virus expansion/reduction) but aims to track the changes in the labor markets. Figure 3: Index of Expected Employment Losses Socioeconomic Vulnerability Index (SEVI) The pandemic has exacerbated the need for a measure that can help governments to target policies to those most affected and increase effectiveness. However, tools that capture the vulnerabilities at a more granular level are still scarce and, in some countries, have not been developed. The socioeconomic vulnerability index aims to measure the vulnerability of subnational geographical areas to different types of shocks, such as disease outbreaks or natural disasters. It is inspired by the CDC's Social Vulnerability Index and the OPHI Multidimensional Poverty Index, and its objective is to identify areas that are most vulnerable to shocks. In particular the SEVI is organized in dimensions, which are constructed using different indicators that are available from various sources. Within a dimension, each indicator is given a weight equal to the inverse of the number of indicators in the dimension. Then dimensions are re-weighetd in a similar manner considering the number of total dimensions in the index. However, shocks vary in their intensity and type; thus, the index can be complemented with other sources of information specific to the shock under analysis. Dimensions of the SEVI 1) Socioeconomic status. It is well known that economically disadvantaged populations can be disproportionately affected by shocks. Households at the lower part of the income distribution are less equipped to face shocks and thus are more likely to take the hardest hits in the case of disease outbreaks or natural disasters. Therefore, this dimension aims to capture this source of vulnerability. The set of variables that aim to capture socioeconomic status are the following: median wage, informality rate, job losses (unemployment), the share of agriculture (or urban) and percentage of people with less than high school (17 or less). a) Arguably, income, proxied by mean wage, is the first mechanism to absorb the shocks and recover from shocks. High-income populations may suffer higher household losses in absolute terms, yet find their overall position mitigated by insurance policies, financial investments, and stable employment (Bolin and Stanford 1998; Tierney 2006). On the contrary, low-income populations are at a disadvantage since they will generally not have insurance policies or financial investments to leverage. b) Likewise, most vulnerable workers in Latin America will be part of the informal sector, characterized by low-quality jobs with low productivity, not protected, regulated, and sometimes not well-recognized of valued. 2 Moreover, informal workers report lower satisfaction than their peers in the formal sector. 3 c) Another important socioeconomic factor is related to vulnerability is education. While the relationship between education and vulnerability to shocks is not well understood, lower levels of education denote population that is a higher risk of re-gaining spaces in the labor market and the practical and bureaucratic hurdles to cope with and recover from shocks prove increasingly difficult to surmount (Morrow 1999). Moreover, people with higher levels of education are likelier to have access to and act upon varied hazard information from preparation to recovery (Tierney 2006). 2) Household Composition: Another source of household vulnerability has to do with the structure of the households. Households with a larger number of elder and children, with people with disabilities or single-headed, would have larger dependency ratios and, therefore, would be more vulnerable in the face of shocks. People in any of these categories are likelier to require financial support, transportation, medical care, or assistance with ordinary daily activities during shocks. Children and elders are the most vulnerable groups in the face of different shocks (Ngo 2001; Cutter et al. 2003:251). Thus, the set of variables that aim to capture household composition are percent male or female householder, no spouse present, with children under 18, percent persons 65 years of age or older and percent of the population with five years. a) Children, especially in the youngest age groups, cannot protect themselves during a shock because they lack the necessary resources, knowledge, or life experiences to cope with the situation effectively. Perhaps because parental responsibility for children is assumed, children are rarely incorporated into shock scenario exercises (Martin et al. 2006). b) Elders and people of any age having physical, sensory, or cognitive challenges are also likely to be more vulnerable to health shocks and other disasters (Eidson et al. 1990; Schmidlin and King 1995; Morrow 1999; Peek-Asa et al. 2003; White et al. 2006; McGuire et al. 2007; Rosenkoetter et al. 2007). Family members or neighbors who would ordinarily look in on an elder, or a caretaker responsible for the welfare of a disabled person, might be less able to do so during a crisis or may find the magnitude of the task beyond their capability. 3) Exclusion: The social and economic marginalization of certain groups, the female population, and the poor have rendered these populations more vulnerable at all stages of a shock (Morrow 1999; Cutter 2 OECD/ILO (2019), Tackling Vulnerability in the Informal Economy, Development Centre Studies, OECD Publishing, Paris, https://doi.org/10.1787/939b7bcd-en. 3 OECD (2019), Key Indicators of Informality based on Individuals and their Household et al. 2003). The poor are less likely to have the income or assets needed to prepare for a possible shock or to recover after one (Morrow 1999; Cutter et al. 2003). a) Gender gaps and inequalities have a strong positive or negative effect on the vulnerability and capabilities of people exposed to threats. The number of women who die from natural threats is higher than that of men; this as a consequence of socioeconomic status uneven women. Furthermore, women face different levels of risk. They have different vulnerabilities and coping abilities, originated by a series of inequalities and political, cultural, and socioeconomic based on gender. Furthermore, women are inadequately represented in decision-makers, and sociocultural attitudes and norms hinder their participation in decision-making. b) Poverty: Although the monetary value of their property may be less than that of other households, it likely represents a larger proportion of total household assets. For these households, lost property is proportionately more expensive to replace, especially without homeowner's or renter's insurance (Tierney 2006). 4) Housing: Housing quality is an important factor in evaluating shocks' vulnerability. It is closely tied to personal wealth; that is, poor people often live in more poorly constructed houses (Eidson et al. 1990; Morrow 1999; Peek-Asa et al. 2003; Daley et al. 2005; De Souza 2004; Tierney 2006). a) Multi-unit housing in densely populated urban areas also poses a heightened risk for tenants (Cutter et al. 2003). Population densities of cities are much higher than those of suburban or rural areas. People living in high-rise apartments are particularly vulnerable to overcrowding when funneled into a limited number of exit stairwells. Furthermore, large numbers of people exiting in the street can make safe and orderly evacuation of everyone difficult and dangerous. Crowding within housing units exacerbates these difficulties (Tierney 2006). b) Electricity: Access to cleaner and affordable energy options is essential for improving the livelihoods of the poor in developing countries. The link between energy and poverty is demonstrated by the fact that the poor in developing countries constitute the bulk of an estimated 2.7 billion people relying on traditional biomass for cooking and the overwhelming majority of the 1.4 billion without access to grid electricity. (Brodman and Kimani, Global energy Assessment 2018) c) Water: Access to water and vulnerability are intertwined for consumers and producers of goods, without safe water and sanitation during the day, workers and customers have to leave their job or market to find water and a place to go to the bathroom. Employers who can provide safe drinking water and adequate sanitation facilities for their employees can retain healthier and more productive employees 4. d) Internet: The new developments and technology push for the massive use of the internet. Thus, access to new capabilities and the net is key for the development of regions. Vulnerabilities are closely related to the lack of access to the internet, especially in the face of situations where people can not move from their households. 4 https://lifewater.org/blog/water-poverty/ COVID-19 Socioeconomic Vulnerability Index (COVID-19-SEVI): The COVID-19-SEVI aims to measure the vulnerability of subnational geographical areas to the COVID- 19 crisis. The index uses a general Socioeconomic Vulnerability Index as an input. While the Socioeconomic Vulnerability Index captures structural vulnerabilities, it fails to capture specific vulnerabilities related to shocks such as natural disasters or health shocks (such as the covid19 pandemic). Therefore, it is useful to complement this index with components that capture those other factors that are not considered. Additional dimensions of the COVID-19-SEVI 5) Covid-19 vulnerability: The main vulnerability that countries face in this pandemic is related to the health systems. The identification of the most vulnerable regions/states/municipalities to the pandemic is important to target resources. While several indicators could be included in an index, information at the sub-national level for the region is still limited. Inspired in the COVID-19 Community Vulnerability Index (CCVI) 5 vulnerability index, we construct a sub-national index to be mapped that will measure the health vulnerability to the COVID-19 crisis. In essence, this index takes the Socioeconomic index as the base and complements it with this new health-COVID component. a) COVID cases: While there is limited data that allows understanding the expansion of the virus at a subnational level, most countries did a significant effort to track the developments. Thus, the rate of contagion could be proxied by the number of cases in each of the regions. b) COVID deaths: Another important indicator to track the development of the pandemic is the share of deaths by region. There is likely less miss-reporting on COVD-19 related deaths compared to reported COVID-19 cases. Moreover, regions with a high share of deaths reflect poor health systems, which is, by itself, a source of vulnerability. 5 https://precisionforcovid.org/ccvi Figure 4: Socioeconomic Vulnerability Index to general shocks Figure 5: Socioeconomic Vulnerability Index to COVID-19 ANNEX 1. Direction and Standardization of Variables The direction (sign) and normalization of the variables that make up the indicator are two processes designed to facilitate the interpretation and aggregation of the variables in the indicator. Direction This is the first stage that must be carried out in the construction of the indicator. Through direction, we aim that all the variables that enter the indicator have a positive association with it. This means that the higher (lower) value of the variable, the indicator will tend to grow (decrease), indicating greater (lower) prosperity. Normalization of the variables Once the variables are directed, the next step in the process of constructing the indicator is to add them. This is achieved through the normalization process, which makes it possible to make the values of the variables comparable. The normalizations used depend directly on the type of measurement of each of the variables originally proposed and how they affect the indicator. The final product of the normalization is to generate new variables without units of measurement, which take values between zero and 100. The advantage of this transformation is that it allows the aggregation of the new variables in the indicator. As the value of normalized variables increases, the level of prosperity tends to increase. The normalizations used in the indicator construction process are presented below. Non-standardization. When the variable is measured in percentage, that is, with values ranging from zero to one hundred, and its values directly affect the indicator, no normalization is used. Classical normalization For variables that have some unit of measurement, and for which there are no objective values, the following standardization is used: Let X be the observed value of the variable and its normalized value. So Where Max( X ) and Min( X ) are the maximum and minimum observed values of X, respectively. If a region/state has a value of X greater than Max( X ) the value assigned by normalization will be 100. If a region has a value of X, less than Min( X ) the value assigned by normalization will be 0. Weighting Aggregation The indicator is designed in dimensions. Each dimension is defined by a group of sub-dimensions, and each of these generally includes a different number of variables. Once the variables have been directed and normalized, using the procedures outlined in the previous section, the construction of the indicator implies defining a methodology that allows the information of these variables to be added to a new variable, which will be the indicator that will allow comparing regions. This leads to the definition of a nested scheme of weights, utilizing which the weights of the dimensions in the indicator of the sub- dimensions within the dimensions and of the variables within the sub-dimensions are established. The weighting scheme In the literature, there are different ways of obtaining the weighting scheme (See, for example, OECD, 2008). The weighting scheme proposed follows the proposal of Alkire and Foster (2011) for the construction of a multidimensional poverty indicator. The following scheme is used for the weightings: a) The dimensions have equal weight in the indicator. b) The sub-dimensions have an equal weight within their dimension. c) The variables have an equal weight within its sub-dimension. This weighting scheme makes explicit the assumption that all chosen dimensions are equally important in the concept of vulnerability. This assumption applies similarly to the sub-dimensions of each dimension and variables within each sub-dimension. The construction of the indicator Based on the direction, standardization procedures, and the definition of the weighting scheme, the construction of the CPI indicator follows the steps presented in the following diagram. 1. Direction 2. Normalization 3. Weighting and aggregation Style: Notes on bivariate map displays: https://vallandingham.me/multivariate_maps.html Additional references Alkire, S., Foster, J.E. (2011) "Counting and Multidimensional Poverty Measurement." Journal of Public Economics, Vol 95(7), pp 476-487. Alkire, S., Foster, J.E. (2011) "Understandings and Misunderstandings of Multidimensional Poverty Measurement", OPHI Working paper no. 43. OECD (2008) Handbook of Constructing Composite Indicators, Methodology and user Guide. ANNEX 2. The GeoLAC Repository The GeoLAC is a Geospatial repository created by the LAC Stats Teams that allows users to visualize, analyze, download, and store geospatial data (Figures 1 and 2). GeoLAC leverages on the user-friendly interface and capabilities of geowb/, the spatial data platform at the World Bank to share and map data. Users with different degrees of access can perform different tasks. For instance, content creators, such as the LAC Stats Team, can store and share geospatial data, as well as create visualizations of specific content through web apps. However, all users with access to the web apps (by clicking on the web app access link) can view, analyze, and download the data used for visualization. Further capabilities also include uploading the user’s own spatial data to the visualization to perform further analysis. Figure 1: Web app for covid-19 vulnerability indexes created by the Stats Team Figure 2: Documentation in the GeoLAC repository of each layer used for the visualization Annex 3. About This Document This document was produced by the Latin America and Caribbean Team for Statistical Development (LAC TSD) in the Poverty and Equity Global Practice of the World Bank (July 2020). The note was led by Gustavo Canavire, Javier Romero, and Diana Marcela Sanchez Castro. The team worked under the guidance of Carolina Díaz-Bonilla and Laura Moreno Herrera, under the overall supervision of Ximena del Carpio (Practice Manager, ELCPV). The LAC Stats Team also includes Pamela Gunio, Giselle del Carmen, Natalia Garcia-Peña Bersh, Karen Barreto Herrera, and Sandra Segovia. The LAC Stats Team works in partnership with CEDLAS of the Universidad Nacional de La Plata in Argentina. The numbers presented in this brief are based on a regional data harmonization effort known as SEDLAC, a joint effort of the World Bank and CEDLAS at the National University of La Plata in Argentina (see Annex 4 for the list of surveys used in this brief). They increase cross-country comparability of selected findings from official household surveys. For that reason, the numbers discussed here may be different from official statistics reported by governments and national offices of statistics. Such differences should not be interpreted in any way as a claim of methodological superiority, as both sets of numbers serve the same important objectives: regional comparability and the best possible representation of the facts of individual countries. Indicators for LAC are calculated using data from Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, and Uruguay (LAC-17). Indicators for LAC-18 use projected Haiti 2012 in addition LAC-17 countries. Indicators for LAC-24 applied projections of LAC-18 to Dominica, Guyana, Jamaica, St. Lucia, St Vincent and Grenadines, and Suriname.