Public Disclosure Authorized LAC COVID-19 HIGH-FREQUENCY PHONE SURVEYS - PHASE II 2021 TECHNICAL NOTE * November 2022 Latin America and the Caribbean (LAC) were among the regions most affected by the COVID-19 pandemic in 2020 –despite being home to only 8.5 percent of the world's population, over 47 million people were infected by the virus by the end of November 2021, representing almost 18 percent of global cases–, and the socio- economic and human capital effects will be felt for many years to come. While in 2021 the region experimented an important economic recovery, with mobility almost fully restored to pre-pandemic levels and a remarkable progress in vaccination campaigns1, positive spillovers on households’ welfare have proved elusive in some countries and for some segments of the population. In an effort to continue monitoring how the COVID-19 pandemic has affected the livelihoods of households in the region2, the World Bank and the United Nations Development Programme (UNDP) joined forces to implement a second phase of High-Frequency Phone Surveys (HFPS) in over 20 LAC countries in 2021. A first wave (Wave 1) was collected between May and July, with a second (Wave 2) following between October and December 2021 -January 2022 for Peru- (Figure 1). The HFPS Phase II adds 11 countries to those originally included in phase one (2020). A total of 24 countries were surveyed in Wave 1, including: Antigua and Barbuda, Argentina, Belize, Brazil, Bolivia, Chile, Colombia, Costa Rica, Dominica, Dominican Republic, Ecuador, El Salvador, Guatemala, Guyana, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, St. Lucia, and Uruguay. Wave 2 surveyed the same set of countries with two exceptions: Antigua and Barbuda, and Brazil. The 2021 HFPS collected information on changes in employment and income (including government transfers), the prevalence of food insecurity, and the access to health, education, and financial services. In this way, it permits an analysis of how the living conditions and situations of households have evolved since phase one (2020 HFPS) for several countries. Additionally, the 2021 HFPS collected information on a series of novel modules. Information was obtained about a variety of other themes, including: the presence of persons with disabilities; access to, and the use of, internet and digital banking services; gender issues; vaccination; mental health; coping mechanisms; and childcare for children up to 5 years old. Finally, in Wave 2, a representative sample of minorities was obtained in seven countries.3 * This technical note was prepared by Carolina Mejía-Mantilla (cmejiamantilla@worldbank.org), Ana Rivadeneira (arivadeneiraalav@worldbank.org), Gabriel Lara Ibarra (glaraibarra@worldbank.org), and Carlos Castañeda Castrillón (Consultant) from the World Bank. The authors are grateful for comments from Adriana Camacho, from the United Nations Development Programme (UNDP). The overall survey project was possible thanks to financial support from the World Bank Latin American and Caribbean Regional Vice Presidency, and United Nations Development Program (UNDP) Rapid Financing Facility (RFF) initiative. 1 By the end of November 2021, on average, 60.2 percent of the population in 15 of the 24 HFPS countries had received at least one dose of the vaccine, and 50.8 percent were fully vaccinated (ranging from a low of 23.2 percent in Guatemala to a high of 82.9 percent in Chile). Source: https://ourworldindata.org/covid-vaccinations. Countries with data comprise: Argentina, Bolivia, Brazil, Chile, Colombia, Dominican Republic, Ecuador, El Salvador, Guatemala, Guyana, Jamaica, Mexico, Peru, Saint Lucia, and Uruguay. 2 In 2020, Phase I of HFPS was collected in 13 countries of the region. See: Mejía-Mantilla, C.; Olivieri, S.; Rivadeneira, A.; Lara Ibarra, G.; Romero, J. 2021. COVID-19 in LAC: High Frequency Phone Surveys - Technical Note. World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank.org/ handle/10986/35902. 3 More specifically, in Bolivia, Guatemala, Mexico and Peru indigenous populations were sampled, and in Panama, Dominican Republic, and Colombia, afrodescendents. 1 Public Disclosure Authorized LAC COVID-19 HIGH-FREQUENCY PHONE SURVEYS - PHASE II 2021 TECHNICAL NOTE November 2022 Figure 1. Data collection dates of Phase II of HFPS 2021 2022 May June July August September October November December January Country 10/25 12/20 10/18 12/27 11/29 8/30 11/22 12/13 9/20 6/28 8/23 5/24 10/11 11/15 9/27 8/16 5/10 7/26 6/14 12/6 6/21 9/13 10/4 5/31 5/17 7/19 1/10 11/8 7/12 8/9 9/6 8/2 5/3 6/7 11/1 7/5 1/3 Antigua & Barbuda Argentina Belize Bolivia Brazil Chile Colombia Costa Rica Dominica Dominican Republic Ecuador El Salvador Guatemala Guyana Haiti Honduras Jamaica Mexico Nicaragua Panamá Paraguay Perú St. Lucia Uruguay Source: LAC HFPS Team. First wave Second wave 2 Public Disclosure Authorized LAC COVID-19 HIGH-FREQUENCY PHONE SURVEYS - PHASE II 2021 TECHNICAL NOTE November 2022 General information As in Phase I, eligible respondents for the phone surveys were adults aged 18 years old and above. Only one respondent per household was interviewed, and he/she answered both individual and household-level questions. For the panel part of the sample4, the same respondent was contacted and interviewed. Over 28,000 interviews were completed across the region in each wave. An average of 1,192 interviews per country were completed in Wave 1 and 1,275 in Wave 2 (see Table 1). The panel cases –those coming from the Phase I HFPS– represented 13 percent of the total cases in Wave 1 (27 percent among the countries that were in Phase I) and 43 percent in Wave 2 (40 percent among countries that were in Phase I). Table 1. Number of completed interviews per wave and percentage by sample type W1 W2 Country Panel with respect Panel with respect Total cases New (%) Total cases New (%) to Phase I W1 (%) to Phase II W1 (%) Antigua & Barbuda 790 0% 100% N.A. N.A. N.A. Argentina 1,216 24% 76% 1,321 45% 55% Belize 816 0% 100% 898 48% 52% Bolivia 1,272 40% 60% 1,312 47% 53% Brazil 2,166 0% 100% N.A. N.A. N.A. Chile 1,212 33% 67% 1,329 36% 64% Colombia 1,221 39% 61% 1,688 40% 60% Costa Rica 805 26% 74% 905 39% 61% Dominica 861 0% 100% 879 51% 49% Dominican Republic 1,205 19% 81% 1,364 39% 61% Ecuador 951 0% 100% 1,032 52% 48% El Salvador 818 28% 72% 812 35% 65% Guatemala 1,207 22% 78% 1,521 39% 61% Guyana 785 0% 100% 875 49% 51% Haiti 2,814 0% 100% 2,361 63% 37% Honduras 1,021 24% 76% 1,004 35% 65% Jamaica 829 0% 100% 871 43% 57% Mexico 2,625 17% 83% 2,511 34% 66% Nicaragua 833 0% 100% 865 42% 58% Panama 815 0% 100% 1,335 31% 69% Paraguay 1,076 16% 84% 1,061 56% 44% Peru 1,212 30% 70% 1,724 34% 66% St. Lucia 835 0% 100% 860 48% 52% Uruguay 816 0% 100% 930 40% 60% Source: LAC HFPS Team. Note: N.A. is not applicable. 4 For 12 countries, Phase 2 Wave 1 tried to recontact households and individuals who had responded in Phase 1 Wave 1 in 2020 and added a fresh supplement sample to compensate for attrition nonresponse. For all countries, Phase 2 Wave 2 tried to recontact all respondents to Phase 2 Wave 1 (for countries present in Phase I, some panel cases remained from Phase I) and also incorporated a supplement sample. More detail in the Sampling section below. 3 Public Disclosure Authorized LAC COVID-19 HIGH-FREQUENCY PHONE SURVEYS - PHASE II 2021 TECHNICAL NOTE November 2022 As in Phase I, the surveys are representative of households with a landline or for which at least one member has a cellphone. They are also representative of individuals aged 18 years or above who have an active cellphone number or a landline at home. Thanks to an additional effort in the sampling methodology, the surveys are also representative (among the universe of households with access to a telephone) of school-age children and adolescents for education indicators (and for infants and children less than 5 years old in Wave 2). Additionally, as mentioned, in Bolivia, Colom- bia, Dominican Republic, Guatemala, Mexico, Panama, and Peru, surveys are representative of ethnic minority groups (i.e., Indigenous and Afro-descendant populations) in Wave 2. Topics covered in the LAC HFPS The questionnaires followed closely those of Phase I for the main topics, thus maintaining comparability between the 2020 and 2021 core indicators. Certain modules and questions were removed to reflect the changing reality of countries, such as the evolution of containment measures and a widespread knowledge of the COVID-19 virus and how to minimize contagion risk. Instead, new modules and questions on vaccination, gender, digital finances and migration were introduced to satisfy new data needs. This was particularly true of the unexplored effects of the pandemic on well-being (including mental health, gender roles within households, working arrangements, etc.). See Table 2 for a detailed account. Table 2. Topics covered in the 2021 LAC HFPS Included in Phase I No. Topics Description W1 W2 (2020 HFPS) General household characteristics, such as: size, location, and area. 19 questions Respondent characteristics, such Additional questions Yes (no questions Basic 1 as: gender, education level, age, 11 questions about household head's on ethnic information and marital status. In 6 countries: age, education, and minorities) affiliation to ethnic minorities gender. (self-declared). Information on access to medical services and disability at the Yes (no vaccine household level. Individual information 2 Health 13 questions 13 questions and mental health on COVID-19 testing, knowledge about questions) vaccine access and willingness to get the vaccine, and mental health issues. Questions about food insecurity at the Yes (no Food household level, 1 about the quality 3 4 questions 4 questions retrospective insecurity of food and 1 retrospective question questions) (before the pandemic). 4 Public Disclosure Authorized LAC COVID-19 HIGH-FREQUENCY PHONE SURVEYS - PHASE II 2021 TECHNICAL NOTE November 2022 Included in Phase I No. Topics Description W1 W2 (2020 HFPS) Questions about the respondent’s Yes (no questions current and pre-pandemic to separate 27 questions 4 Employment employment status, type, industry, 28 questions unemployment No relevant changes. teleworking, job formality, and other from out of labor characteristics. force) 13 questions Important changes: no questions by income Questions to assess changes (with source, only total respect to pre pandemic) in total income, government income and income from different transfers and sources: labor (wage, business, and remittances with more Yes (questions are 5 Income loss 17 questions agricultural work) and non-labor detail for 6 countries. similar to Wave 1) (remittances, government regular Changes in income are transfers, and COVID-19-related with respect to previous transfers, others). wave. Additional questions about household’s ability to meet basic needs. 23 questions Additional questions Questions about household about number of No (household composition (number of members by household members composition 6 Roster age and sex), demographic changes, 22 questions who worked before the was part of basic and complete list of children and pandemic and at the information) adolescents age 0-17 years old. time of the survey, by gender. 22 questions Important changes: For households with school-age Additional questions Partially (only children, questions about one for young children questions about 7 Education randomly selected child on enrollment, 11 questions (age 0-5) about homeschooling at attendance at school (face-to-face attending childcare/ household level) classes), and homeschooling activities. development centers and responsibility of childcare at home. Individual questions about changes in the time dedicated to housework and 10 questions 8 Gender 10 questions No children’s care and education, as well No relevant changes. as couple dynamics at home. 5 Public Disclosure Authorized LAC COVID-19 HIGH-FREQUENCY PHONE SURVEYS - PHASE II 2021 TECHNICAL NOTE November 2022 Included in Phase I No. Topics Description W1 W2 (2020 HFPS) 15 questions Questions applied only in four Additional perception 9 Migration countries about nationals' perceptions 11 questions No questions on general on immigrants in their country. issues. Questions about telephone and 7 questions about phone Digital internet access, uses, and challenges. and internet access and 10 25 questions Partially technologies Questions about access to, and use of, 4 questions on digital digital financial services. finances for 9 countries. Questions about dwelling 11 Dwelling 4 questions 4 questions Yes characteristics and household assets. Added in Wave 2. Questions about households' strategies to cover basic 12 Coping expenses during 2021, and individuals' No. 6 questions Partially perceptions on insecurity and domestic violence. Source: LAC HFPS Team Phase II questionnaires were designed in Spanish and later translated into English, French, and Portuguese; and conducted according to the main language spoken in each country.5 Some of the changes with respect to the Phase I questionnaire are the following: • Changes to the data collection on children’s education: One of the main differences between modules appearing in both Phase I and II is the way the education questions were implemented. During the 2020 HFPS, education questions were asked about all school-age children in the household. Estimation was done using household weights and indicators were interpreted at the household level (e.g., percentage of households with school-age children engaged in learning activities at home). For the 2021 HFPS, one school-age child was randomly selected in each household and all questions were about the selected child / adolescent. • New module for children 0-5: Wave 2 introduced questions for younger children, ages 0 to 5 years old. Again, one child in this age bracket was selected in each household and new questions on education and childcare were asked about this child. 5 The Brazil COVID-19 Phone Survey was developed and implemented when the LAC HFPS Phase II Wave 1 was about to conclude. Inspired by this activity, the survey collected information along multiple dimensions relevant to the welfare of the population (e.g., changes in employment and income, access to health and education services, gender inequalities, and food insecurity). However, it incorporated more detailed questions on the social protection system, coping mechanisms among households, and subjective poverty. The survey followed a Random Digit Dialing sampling methodology using a dual sampling frame of cellphone and landline numbers. However, given the availability of state calling codes, the sampling frame was also stratified by type of phone and state. Results are nationally representative for households with a landline or at least one cell phone and of individuals of ages 18 years and above who have an active cell phone number or a landline at home. For more information, see World Bank and UNDP (2021): https://microdata.worldbank.org/index.php/catalog/4533. 6 Public Disclosure Authorized LAC COVID-19 HIGH-FREQUENCY PHONE SURVEYS - PHASE II 2021 TECHNICAL NOTE November 2022 • Focus on different age groups: For households with both young (0 to 5 years old) and school-age children, two children were selected in Wave 2; one for each corresponding age bracket. In the case of panel households with more than one school-age child, a different child was selected for Wave 2. Therefore, the education results are estimated using specific weights and the indicators are interpreted for school-age or young children, respectively (e.g., share of school-age children attending school in person; share of 0-to-5-year-olds attending childcare centers). Sampling6 For 12 of the 13 countries included in Phase I, the samples for Phase II Wave 1 included two components: (i) a panel formed by respondents to Phase I Wave 1, and (ii) a fresh, supplementary sample of phone numbers to compensate for attrition between Phase I Wave 1 and Phase II Wave 1, and to slightly increase the overall sample size. Phase II aimed to achieve between 800 and 3,000 completed interviews per country. Meanwhile, the samples for the 10 additional countries and Ecuador7 were completely new. Similarly, for Wave 2, the samples included two components in all countries: (i) a panel of respondents to the previous wave (some of which came from Phase I); and (ii) a fresh supplementary sample of phone numbers to compensate for attrition between waves. The fresh samples in all countries and in both waves were obtained through the same Random Digit Dialing (RDD) dual- frame methodology as in Phase I.8 The RDD methodology produces all possible phone numbers in the country under the national phone numbering plan and draws a random sample of numbers. Telephone numbers belong to either the landline frame or the cellphone frame. First, for each country, a large sample was selected in both frames, with an allocation ranging from 0 percent landlines and 100 percent cellphones to 20 percent landlines and 80 percent cellphones. Second, these large samples went through an automated process to identify the active numbers, before then being cross- checked with business registries to identify and remove business numbers not eligible for the survey. Thus, a smaller second-instance sample was selected from the active residential numbers and was delivered to the country teams. In the cases of Argentina, Bolivia, and Mexico, the cell phone frame in each country was geographically stratified by department, province, or state. In addition, the sample of landlines was selected with a proportionate allocation among strata. A summary of the steps followed for the estimation of weights by type of countries and wave is presented in Table 3. Taken together, the Phase II surveys have several units of analysis. In Wave 1, the units of analysis are households, adult individuals (age 18 and older), and children age 6-17. In Wave 2, the units of analysis are households, adult individuals (age 18 years and older), children age 6-17, and children age 0-5. Weights were computed for each sample unit and should be used accordingly to obtain the statistical estimates of interest.9 6 A detailed description of sampling and weighing is presented in: Flores Cruz, Ramiro. 2022. High-Frequency Phone Survey (HFPS) - Phase 2: Sampling design, weighting, and estimation. 7 Ecuador is the only country from Phase I where a fully fresh sample of phone numbers is used. 8 For a full description of Phase 1 sampling design, see: HFPS Phase 1 Technical Note on Sampling, Weighting and Estimation, available at: https://documents1. worldbank.org/curated/en/336371631859678760/pdf/COVID-19-High-Frequency-Phone-Surveys-in-Latin-America-Technical-Note-on-Sampling-Design- Weighting-and-Estimation.pdf 9 As mentioned, in Wave 2, certain countries have a specific sample of minorities (i.e., self-reported Indigenous people in Mexico). However, these samples do not have a specific set of sampling weights. Indicators for these groups should be obtained by using the appropriate sampling weight of the unit of analysis while restricting the set of observations of analysis to the sub-group of interest. 7 Public Disclosure Authorized LAC COVID-19 HIGH-FREQUENCY PHONE SURVEYS - PHASE II 2021 TECHNICAL NOTE November 2022 Table 3. Steps for weight estimation of the Phase II surveys (2021 HFPS) Wave 1 Countries in both 1. Calculation of the inclusion probabilities of landline and cellphone numbers. phases 2. Computation of design weights for households and individuals. 3. Attrition non-response weighting adjustment for the panel sample component, and first-time, non-response adjustment for the supplement sample component. 4. Calibration of household, adult, and child/adolescent weights, using external ancillary data from official sources (adjusted by national phone coverage). 5. Weight trimming and recalibration. Phase II new countries In step 3, design weights are adjusted for first-time non-response only. + Ecuador Wave 2 Countries in both In step 3, non-response weighting adjustments accounted for the different response phases mechanisms observed in the three components of the overall sample. For each country, overall sample size (before fieldwork activities) comprises: I. Phase I Wave 1 overall selected sample of phone numbers, which was affected by first-time non-response in Phase I W1, plus attrition non-response in Phase II Wave 1 and attrition non-response in Phase II W2. II. Phase II Wave 1 overall supplement fresh sample, which was subject to first-time non-response in Phase II Wave 1 and attrition non-response in Phase II Wave 2. III. Phase II Wave 2 overall supplement fresh sample of phone numbers, which was only subject to first-time non-response. Phase II new countries In step 3, non-response weighting adjustments accounted for the different response + Ecuador mechanisms observed in the two components of the overall sample for each country: I. Phase II Wave 1 overall selected sample of phone numbers, which was affected by first-time non-response in Phase II Wave 1 and attrition non-response in Phase II Wave 2. II. Phase II Wave 2 overall fresh supplement sample, which was subject to first-time non-response only. Countries with ethnic • Minority populations (Indigenous and Afro-descendants) had to be sampled at minority group higher rates than primary samples in all countries except for Guatemala and Mexico. representation: Bolivia, This was achieved by executing a screening operation in each country. Colombia, Dominican • Total sample of minority respondents in the five countries is formed by two Republic, Guatemala, components: minority cases already present in the primary sample; and minority Mexico, Panama, and cases interviewed in the screening operation. Peru • Base weights of the first minority component were adjusted for attrition non- response between Waves 1 and 2, whereas the base weights of the second minority component were adjusted for first-time non-response. • Adjusted weights of the combined sample of minority respondents were calibrated to minority totals by gender, age, and educational attainment. 8