Measuring the socioeconomic impact of last-mile infrastructure development in Latin America and the Caribbean

Puig Gabarró, Pau et al. Working Paper Measuring the socioeconomic impact of last-mile infrastructure development in Latin America and the Caribbean IDB Working Paper Series, No. IDB-WP-1326 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Puig Gabarró, Pau et al. (2022) : Measuring the socioeconomic impact of last-mile infrastructure development in Latin America and the Caribbean, IDB Working Paper Series, No. IDBWP-1326, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0004326 This Version is available at: https://hdl.handle.net/10419/290041 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/igo/legalcode IDB WORKING PAPER SERIES Nº IDB-WP-1326 Measuring the Socioeconomic Impact of Last-Mile Infrastructure Development in Latin America and the Caribbean Pau Puig Gabarró Raúl Katz Hernán Galperin Fernando Callorda Enrique Iglesias Rodríguez Antonio García Zaballos Marcos Robles Ramiro Valencia Inter-American Development Bank Institutions for Development Sector June 2022 June 2022 Measuring the Socioeconomic Impact of Last-Mile Infrastructure Development in Latin America and the Caribbean Pau Puig Gabarró Raúl Katz Hernán Galperin Fernando Callorda Enrique Iglesias Rodríguez Antonio García Zaballos Marcos Robles Ramiro Valencia Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 AttributionNonCommercial-NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license (http://creativecommons.org/licenses/by-nc-nd/3.0/igo/ legalcode) and may be reproduced with attribution to the IDB and for any non-commercial purpose, as provided below. No derivative work is allowed. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the UNCITRAL rules. The use of the IDB's name for any purpose other than for attribution and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this CC-IGO license. Following a peer review process, and with previous written consent by the Inter-American Development Bank (IDB), a revised version of this work may also be reproduced in any academic journal, including those indexed by the American Economic Association's EconLit, provided that the IDB is credited and that the author(s) receive no income from the publication. Therefore, the restriction to receive income from such publication shall only extend to the publication's author(s). With regard to such restriction, in case of any inconsistency between the Creative Commons IGO 3.0 Attribution-NonCommercial-NoDerivatives license and these statements, the latter shall prevail. Note that the link provided above includes additional terms and conditions of the license. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Measuring the socioeconomic impact of last-mile infrastructure development in Latin America and the Caribbean / Pau Puig Gabarró, Raúl Katz, Hernán Galperin, Fernando Callorda, Enrique Iglesias Rodríguez, Antonio García Zaballos, Marcos Robles, Ramiro Valencia. p. cm. — (IDB Working Paper Series ; 1326) Includes bibliographic references. 1. Infrastructure (Economics)-Latin America. 2. Infrastructure (Economics)-Caribbean Area. 3. Information technology-Social aspects-Latin América-Econometric models. 4. Information technology-Social aspectsCaribbean Area-Econometric models. 5. Information technology-Economic aspects-Latin America. 6. Information technology-Economic aspects-Caribbean Area. 7. Digital communications-Economic aspectsLatin America. 8. Digital communications-Economic aspects-Caribbean Area. I. Puig Gabarró, Pau. II. Katz, Raúl. III. Galperin, Hernán. IV. Collorda, Fernando. V. Iglesias Rodríguez, Enrique. VI. García Zaballos, Antonio. VII. Robles, Marcos. VIII. Valencia, Ramiro. IX. Inter-American Development Bank. Connectivity, Markets and Finance Division. X. Series. IDB-WP-1326 http://www.iadb.org Copyright © 2022 2022 Measuring the Socioeconomic Impact of Last-Mile Infrastructure Development in Latin America and the Caribbean Pau Puig Gabarró, Raúl Katz, Hernán Galperin, Fernando Callorda, Enrique Iglesias Rodríguez, Antonio García Zaballos, Marcos Robles, and Ramiro Valencia  JEL codes: G18, G28, L96, L86, L42 Keywords: digital infrastructure, connectivity, inclusion, digital economy Abstract The objective of this study is to estimate the socioeconomic impact of the deployment of last-mile digital infrastructure in Latin America and the Caribbean. To measure the impact of the economic and social aspects of this type of infrastructure, the analysis differentiates according to the geographic context (urban and rural), gender, and educational level, and details the effects and channels that link the deployment of last-mile infrastructure with socioeconomic benefits. The results of this study show that broadband improves job creation, the passage to formality, and salaries for the entire population. The findings indicate that the difference between the higher-skilled and lower-skilled segments of the population is considered in terms of the level of impact. The results also reveal that broadband deployment can generate an increase in inequality between genders, between the urban and the rural population, and between individuals with more years of formal education and individuals with fewer years of formal education if it is not accompanied by public policies that allow access equal use of this technology. This evidence confirms findings in previous studies that highlight the complementarity between broadband and skill levels in estimating benefits. For this reason, the contribution of public policies should be considered as a compensatory mechanism to counteract unintended effects. The set of results constitutes a rich base of empirical information that could help the governments of the region to make policy decisions, taking into account the importance of extending last-mile deployment to the rural context. Authors He holds an MBA in international business management from the Universidad Internacional Menéndez Pelayo and a master's degree in Telecommunications from the Universidad Pompeu Fabra. He is a Telecommunications Specialist at the Inter-American Development Bank (IDB), where he provides support to governments in Latin America and the Caribbean for the reform of public policies in digital technologies and the planning and execution of investments in telecommunications infrastructure. Previously, he held similar positions at the World Bank. Pau Puig Gabarró He holds a PhD in political science and business administration, a master of science in communications technology and policy from the Massachusetts Institute of Technology (MIT), a master's degree and a bachelor's degree in communication sciences from the University of Paris, and a master's degree in political science from the University of Paris-Sorbonne. He spent 20 years at Booz Allen & Hamilton as the lead partner of the Americas Telecommunications Practice and a member of the firm's management team. He is president of Telecom Advisory Services, LLC, and director of Business Strategy Research at the Columbia Institute for Tele-Information, Columbia Business School, as well as a visiting professor in the Telecommunications Management Program at the Universidad de San Andrés. Raúl Katz He holds a PhD and master's degree in communications from Stanford University and a bachelor's degree in sociology and economics from the University of Buenos Aires. He has been a professor in the Department of Social Sciences and director of the Master's Program in Information Technologies and Telecommunications at the Universidad de San Andrés. He is currently Associate Professor and Associate Dean of the Annenberg School of Communications at the University of Southern California and Director of the Annenberg Research Network on International Communication. Hernán Galperin He holds a master's degree and a bachelor's degree in economics from Universidad de San Andrés. He is a project manager at Telecom Advisory Services, LLC; researcher at the National Network of Public Universities of Argentina; and professor of Political Economy at the National University of La Matanza (UNLAM). Before joining Telecom Advisory Services, LLC, he worked as an analyst for the Argentine Congress and as an auditor at Deloitte. Fernando Callorda He holds a master's degree in banking and financial markets from Universidad Carlos III and a master's degree in telecommunications from Universidad Autónoma de Madrid. He is a telecommunications specialist in the Connectivity, Markets, and Finance Division of the IDB, where he has supported governments in Latin America and the Caribbean in the development of broadband and digital economy agendas through technical assistance and lending operations. Previously, he worked as a strategy and operations consultant in Madrid, where he provided services to leading telecommunications companies in Latin America and the Caribbean and Europe. Enrique Iglesias Rodríguez He holds a doctorate in economics from Universidad Carlos III. He is a professor of Finance Applied to Telecommunications at the Instituto de Empresa, and of Economic Regulation at American University and Johns Hopkins University. He is the author of several publications on economic and regulatory aspects applied to the telecommunications sector and is a leading specialist in telecommunications for the Management of Institutions for Development of the IDB, as well as coordinator of the IDB's broadband platform. He has extensive experience in the telecommunications sector, where he has carried out his professional activity in different positions of responsibility. At Deloitte Spain, he led the regulatory practice for Latin America and the Caribbean. Prior to that, he served as chief economist of the Cabinet of Economic Studies of Regulation at Telefónica de España and deputy director of Economic Analysis and Markets at the Spanish telecommunications regulator. Antonio García Zaballos . Throughout his professional career he has advised regulators, telecommunications operators, and governments in countries such as Argentina, China, the Czech Republic, the Dominican Republic, Ecuador, Paraguay, Poland, and Saudi Arabia. He is a member of various technical expert committees, including the World Economic Forum (WEF), within the Internet for All initiative, and the United Nations Broadband Commission. He holds a master's degree in economics from the Centro de Investigación y Docencia Económicas de México. He is a research economist in the Social Sector at the IDB. Previously, he worked in the IDB's Poverty and Inequality Unit and the Research Department. He has provided technical support to Bank projects related to poverty, inequality, and targeting of social programs for Ecuador, the Dominican Republic, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, and Peru. Prior to joining the IDB, he was an advisor to the National Institute of Planning and the National Institute of Statistics, and Manager of Quantitative Methods at Maximize Consulting in Peru. He has been technical coordinator of the Program for the Improvement of Surveys and the Measurement of Living Conditions in Latin America and the Caribbean (MECOVI) of the IDB, the World Bank and the Economic Commission for Latin America and the Caribbean (ECLAC) and of the Budget and Social Expenditure Project of the United Nations Development Programme (UNDP) and the United Nations Children's Fund (UNICEF) in Paraguay. He has also been a professor of Econometrics and Quantitative Methods at several universities in Mexico and Peru. His research focuses on issues related to poverty, inequality, and social assistance. Marcos Robles He holds a master's degree in economics from the Latin American Faculty of Social Sciences (FLACSO), Ecuador, and is an electronics and telecommunications engineer from the National Polytechnic School of Ecuador. He currently works as a consultant at Telecom Advisory Services, LLC. Before that, he worked for nine years at the Ministry of Telecommunications and the Telecommunications Regulation and Control Agency of Ecuador, where he held the position of Director of Studies, Statistical and Market Analysis. Ramiro Valencia 6  Although broadband deployment generally leads to a positive variation in the income of the entire rural population, the use of the service or equipment promotes a greater increase in all the variables studied. The differentiation of a lower impact in non-metropolitan or rural areas with high broadband availability, compared to the same types of areas with low service availability, could be explained by the fact that in counties with high coverage, households without service are not waiting for broadband access, while in counties with low coverage, the increase in availability causes an increase in income, since there is a population that is waiting for broadband access to be able to generate economic benefits. 7  The evidence generated by previous research has made it possible to formalize 11 working hypotheses to be evaluated in the context of Latin America and the Caribbean. These can be grouped into five categories. Study Hypothesis Aggregate economic impact H1: Fixed broadband deployment generates a positive impact on total household and labor income. H2: Broadband deployment is associated with an increase in the employed population and an increase in labor formality. Comparative urban-rural impact H3: Urban areas tend to benefit more than rural areas in economic terms (increase in income, total and labor income) as a consequence of broadband deployment, since they have the most transaction-intensive and information-intensive industrial sectors (e.g., financial services or professional activities). H4: Urban areas tend to benefit more than rural areas in terms of employment generation and labor formality, since they have the industrial sectors with the highest intensity of transactions and use of information. 8  Comparative impact by gender H5: Internet use contributes to the reduction of the income gap between men and women, since access to the service allows women to obtain better-paying jobs. H6: The use of the internet contributes to the reduction of the gender employment gap, since access to the service can especially help women to access better-paying jobs. Comparative impact by educational level H7: The economic impact of broadband access is higher for the more educated population, since they have a higher level of digital literacy. H8: The economic impact in rural areas varies according to the level of human capital and digital skills: the higher the level of education, the higher the impact on employment. Temporal impact H9: The impact on total and labor income may grow over time due to an increase in the experience of using the service. 9  H10: The economic benefit in labor terms generates an increase, in the short term, in labor formality and, in the long term, in the generation of new jobs. H11: The lower benefit in rural areas in relation to urban areas is also short term: the impact on income in rural areas appears in the medium and long term compared to the economy as a whole. The causality may be crossed by a temporal factor, in which the deployment of last-mile infrastructure does not generate benefits simultaneously or in the short term; rather, they only appear in the medium or long term. $ $ 10  The objective of the regional analysis has been to apply the difference-in-differences methodology to a consolidated base of sub-sovereign unit (parishes, municipalities, and regions) in Latin American and Caribbean countries. This allows us to compare the economic effect (increased income, job creation, and increased labor formality) resulting from the treatment that certain sub-sovereign units receive (i.e., when they move to fixed broadband access), compared to those that do not change their status (those that do not benefit from the deployment of last-mile fixed broadband). To evaluate the impact of last-mile infrastructure deployment on income, a difference-in-differences model is specified using Equation 1. This is a simple regression, which determines the effect on income generated by residing in an area where there is the possibility of accessing broadband service at home. Methodology of Analysis Ln (Y ) = β + β βTreatment Area. Year+ it it i 0 1 2 β+ 3tβ .X+ 4µ+ itit (1) Áreas with broadband in the ome, defined as areas where at least 10 percent of the households in the survey adopt the service. 1: Areas with no broadband in the home, defined as areas where less than 10 percent of the households in the survey adopt the service. 0: Where: Y : Income. Treatment : This is the variable that distinguishes the groups. it it 11  Year : Corresponds to a fixed effect for each year between 2008 and 2019. X : This is a matrix of other independent variables used as controls in the specifications, such as urban and rural area, gender, and years of education. it t i µ : It is the error term. it Different econometric models are applied to the dependent variable, considering both total income (which also includes non-labor income, such as rents or remittances) and exclusively labor income. Different specifications of the econometric model corresponding to Equation 1 are made for the independent variables for each analysis. First, the direct relationship between the treatment and income is analyzed. Subsequently, understanding that both years of education and area of residence are factors that affect income, an additional control for these factors is included. Finally, a third model is added with an additional control for gender. In all specifications we include controls for year fixed effect (a binary variable for each year included in the regression) and geographic area (a binary variable for each sub-sovereign unit included in the regression). To assess the impact of last-mile infrastructure deployment on employment metrics (percentages of employed population, inactive population, unemployed population, and ratio of formal to informal workers),2 a difference-in-differences model is specified, according to the following equation: Percentage of population by group Treatment .Year=+ + + it t β0β2Area +i β3.X +it itit β4 β1µ(2) 2 The percentage of formal employees is calculated by dividing the number of formal jobs by the total number of jobs (formal plus informal). Area : Corresponds to a fixed effect for each geographic area (subnational unit) included in the regression. 12  This is a simple regression, which determines the effect on the percentages of each group, generated from residing in an area where there is the possibility of accessing broadband service at home. Where: For the independent variables used in each analysis, different specifications of the econometric models are made. The first model evaluates the direct relationship between treatment and the percentage of the population, by labor group. The second, under the assumption that expected income can affect labor participation decisions, includes a control for total income; and, in a third model, another control for labor income is added. All specifications include controls for year fixed effect (a binary variable for each year included in the regression) and geographic area (a binary variable for each sub-sovereign unit included in the regression). Percentage of population by group : Percentage of employed, inactive, and unemployed population, and ratio of formal to informal workers. Treatment : This is the variable that distinguishes the two groups. Areas with broadband in the home, defined as areas where at least 10 percent of the households in the survey adopt the service. 1: Areas with no broadband in the home, defined as areas where less than 10 percent of households adopt the service. 0: it it Year : Corresponds to a fixed effect for each year between 2008 and 2019. Area : Corresponds to a fixed effect for each geographic area (subnational unit) included in the regression. X : A matrix of other independent variables used as controls in some specifications. it t i µ : It is the error term. it 13  The regional analysis was carried out based on data on broadband adoption generated by national household surveys in Latin American and Caribbean countries, included in the IDB's harmonized database. Based on the information available in the database, 16 countries were included in the analysis (Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Jamaica, Mexico, Panama, Paraguay, Peru, and Uruguay) between 2008 and 2019. The lack of availability of panel data at the household/individual level was solved by means of pseudo-panels with sub-sovereign units.3 The next step was to calculate, for each year and sub-sovereign unit, the average (weighted by the weight of each individual observation) of the indicators of interest (internet ownership, total income, labor income, years of education, gender, urban population, and rural population). Thus, we ended up with 2,159 observations for the analysis for the years 2008 to 2019 (see Table 12). 3 This was the methodology applied in the case of Ecuador to solve the lack of panel data and to be able to perform the long-term analysis. 14  The results of the regional analysis, in terms of the hypotheses to be considered, allow the following conclusions to be drawn: Analysis of Results 4 For the purposes of this study, all figures will be presented in U.S. dollars at purchasing power parity. Aggregate Economic Impact C1: The hypothesis that fixed broadband generates a positive impact on total income and labor income is confirmed. As a consequence of the deployment of the service, there is a positive and significant impact on total and labor income of 6.92 percent (US26.46 according to purchasing power parity [PPP])4 and 7.43 percent (US22.38), respectively. C2: The hypothesis that broadband generates incentives to join the labor force is confirmed. As a consequence of the deployment of the service, the percentage of inactive population decreases 0.80 percentage points, which generates a positive effect on the employed population of 0.84 percentage points. In addition, the hypothesis that broadband produces a positive effect on higher quality jobs, which is reflected in an increase in labor formality, is confirmed. In particular, it is observed that, on average, labor formality increases 0.66 percentage points, which implies an increase of 1.84 percent. CONCLUSIONS 15  5 In urban areas (Table 33), the employment rate is 54.58 percent and the inactivity rate is 41.30 percent. In rural areas, the distribution is similar (see Table 35): 53.81 percent and 44.31 percent, respectively. Unemployment is higher in urban than in rural areas, with levels of 4.12 percent and 1.88 percent, respectively (Tables 34 and 36). However, in neither case is a statistically significant impact found. C3: The hypothesis that, as a consequence of broadband deployment, urban areas tend to benefit more than rural areas in total and labor income is confirmed. In particular, we find that, in general, broadband provision in urban centers has a positive impact on total and labor (monthly) income of 4.33 percent (US19.46) and 4.96 percent (US17.63), respectively, while in rural areas the impact is not significant, except under certain conditions (see H8). C4: The hypothesis that urban areas tend to benefit more than rural areas in terms of employment (employment generation and labor formality) as a consequence of broadband deployment is confirmed to the extent that the industrial sectors with the highest volume of transactions and information are concentrated there. Indeed, in urban centers there is a migration from the inactive population (0.43 percentage points) to the employed population (0.44 percentage points). This effect does not appear in rural areas.5 Labor formality in urban areas increased by 1.55 percentage points, while in rural areas the impact is significant but lower, at 0.97 percentage points. Considering temporality, in rural areas there is no impact of migration of the inactive population to the employed population, maintaining the effect on the improvement of labor formality. Comparative urban-rural impact 22  Table R2. Comparison of the Results of the Study, by Analytical Module (continued) HypothesisCategory Regional Brazil Ecuador El Salvador Jamaica Greater impact on the employed population and labor formality in the more educated population. Greater impact of labor formality on the more educated population. H8: The economic impact of broadband in rural areas varies according to educational level and digital skills. The impact on income in rural areas is positive and significant for the population with more than 11 years of formal education. The hypothesis could not be verified because the data do not differentiate between urban and rural areas at the educational level. Results suggest a greater impact among those with higher education. The hypothesis could not be verified because the data do not differentiate between urban and rural areas at the educational level. The hypothesis could not be verified because the sample is too small to differentiate between urban and rural areas. Temporary impact H9: The impact on total income and labor income grows over time. Higher impact on total and labor income, employed population, and labor formality for early adopters. Superior impact on GDP per capita for early adopters. Impact on labor income grows over time, although not monotonically. Major impact on total income and labor income for early adopters. The hypothesis could not be verified since the sample is limited in time. H10: The impact on employment and labor formality grows over time. Increase in labor formality in the short term and new jobs in the long term. Higher impact on GDP per capita for early adopters, which is assumed to be accompanied by better or more employment. Increased labor income over time, which is assumed to be accompanied by a combination of increased employment and formality. Increase in labor formality in the long term. The hypothesis could not be verified since the sample is limited in time. 23  Table R2. Comparison of the Results of the Study, by Analytical Module (continued) HypothesisCategory Regional Brazil Ecuador El Salvador Jamaica The increase in income of the rural population is slower than in the economy as a whole, but it is durable and sustainable in the long term. The hypothesis could not be verified since a small rural sample is available to evaluate the effect of time. H11: The impact on income in rural areas appears in the medium/ long term. The hypothesis could not be verified because the data do not differentiate between urban and rural areas. The long-term analysis does not generate significant results on the income impact of rural areas. The hypothesis could not be verified since the sample is limited in time. Source: Authorsʼ elaboration. Confirmation of hypotheses Non-confirmation of hypotheses Impossible to verify hypotheses due to lack of data In terms of consistency, the following conclusions can be drawn from the results of the analytical modules: CONCLUSIONS C1: Last-mile infrastructure generates a positive impact on total household income and labor income. C2: The increase in the employed population and, especially, in labor formality is associated with the implementation of this infrastructure. C3: Urban areas tend to benefit more from last-mile infrastructure deployment than rural areas. 24  C4: The positive impact generated by the implementation of broadband is greater among men than among women, and among the more educated population, which may accentuate the inequality between these groups. C5: The impact of broadband deployment grows over time. The greatest impact occurs in Ecuador and Brazil (4.6 percent), which are the countries with the highest level of fixed broadband adoption and the highest quality of service (measured by fixed broadband speed), indicating the existence of a return to scale already pointed out in the literature (Koutroumpis, 2009; Katz, Avila, and Meille, 2010; Katz and Jung, 2021). The impact is lower in El Salvador (2.9 percent), where adoption only reached 31.17 percent by the end of the analysis period and (broadband) speed is less than half that of Brazil (29 Mbps vs. 12 Mbps). Furthermore, and beyond the directional consistency of the analyses, the degree of impact per country is related to the level of fixed broadband adoption and the quality of connections in the period under analysis: In El Salvador and Ecuador, countries with low levels of unemployment during the period analyzed, there was a significant impact on the increase in labor formality (El Salvador) and adequate employment (Ecuador).8 8 The Jamaican study was not taken into account for the comparison, as it required a different analysis methodology (impact of increased adoption rather than service introduction, since both fixed and mobile technologies were considered for the analysis). $ 25  The body of evidence is presented as a rich empirical basis for the formulation of last-mile digital infrastructure deployment strategies and the reduction of demand gaps in Latin America and the Caribbean. In particular, these results show that broadband deployment can generate an increase in inequality at three levels (between genders, between urban and rural populations, and between individuals with more and less formal education) if it is not accompanied by public policies that allow access to equal use of such technology. This evidence is consistent with the results of previous studies that highlight the complementarity between broadband and the levels of training and skills in the estimation of benefits. The results of this study show that broadband improves job creation, the transition to formality, and the wage level for the entire population; likewise, the difference between the more-skilled population and the less-skilled population is posed in terms of level of impact. This is why the contribution of public policies should be considered as a compensatory mechanism to counteract unexpected effects. In view of the above, four public policy axes should be considered to complement connectivity infrastructure deployment programs: Public Policy Implications The results highlight the need to carry out digital literacy actions in rural areas to support the use of broadband in the productive fabric. Digital literacy programs should focus not only on communicating available services, but also on developing reliability in use and explaining the benefits of digital connectivity and the conditions necessary to ensure privacy and security. Programs can be organized into three areas of intervention: Incorporation of digital literacy content in formal education programs, both for students and teachers. Deployment of programs aimed at specific segments of the population, including the elderly, the unemployed, people with disabilities, and others. Implementation of generic programs to support the population in all community centers (libraries, cultural centers, clinics, etc.). 26  The lower impact on rural areas, a topic widely covered in the literature surveyed, requires recognition that conventional rural development programs aimed at the creation of new ventures represent an adequate complement to the development of digital infrastructure with universal reach. The results of the study suggest a greater impact on those who actually use the internet. In other words, the results suggest a spillover effect towards the entire population of the sub-sovereign unit, which may, however, imply an increase in income inequality between users and non-users. Therefore, there is a need to implement public policies to encourage the adoption of broadband service to close the demand gap in the localities that receive connectivity. This may be reflected in the growing gender inequality and differences in terms of educational level. With respect to the growing gender inequality, the greater short-term impact on men's employment (due to the network building effect), and the lack of impact on women's labor participation in the long term indicate the need to act on online employment opportunities in sectors with higher labor participation among women, such as services, health, and education. 27 Introduction The objective of this study is to estimate the socioeconomic impact of last-mile digital infrastructure deployment in Latin America and the Caribbean. For this task, the study considered five analytical modules: Four econometric studies for the same number of countries (Brazil, Ecuador, El Salvador, and Jamaica) (Puig Gabarró et al., 2022a, 2022b, 2022c, 2022d) in which the deployment of last-mile infrastructure is analyzed in a quasi-random manner to examine a causal link between the infrastructure and certain socioeconomic indicators. The selection of these four countries was determined by the availability of data and because of their different socioeconomic and technological profiles (see Table 1). A regional analysis, based on a consolidated panel of data from 16 countries for the purpose of constructing correlations between last-mile deployment and socioeconomic impact, and 28  This document reviews the academic research conducted on the subject and the results of the regional analysis. Section 1 explains the need to study the deployment of last-mile digital infrastructure for the development of countries in the region. Section 2 presents the evidence from the academic literature on the differentiated socioeconomic impact of broadband. Its objective is to explore the research that differentiates the impact on income and employment according to variables such as geographic area, population with access to technological devices, educational level, and gender. Sections 4 and 5 present the methodology, data, results and discussion of the results of the regional analysis in relation to income and employment, respectively. Finally, Section 6 examines the implications of these results for public policy and presents four axes to complement last-mile broadband infrastructure deployment programs. Table 1. Socioeconomic and Technological Profile of the Countries Studied BrazilCountry Year GDP per capita (current US$) 2007 $12.550$12.550 2018 $15.020$15.020 Ecuador 2011 $9.858$9.858 2019 $11.851$11.851 El Salvador 2008 $6.063$6.063 2019 $9.147$9.147 Jamaica Source 2014 $8.545$8.545 IMF (2019)IMF (2019) 2018 $9.969$9.969 Unemployment rate 8.33%8.33% 12.33%12.33% 3.46%3.46% 3.81%3.81% 5.88%5.88% 3.96%3.96% 13.74%13.74% ILO (2021)ILO (2021)9.10%9.10% FB adoption (% households) 13.49%13.49% 51.05%51.05% 21.24%21.24% 59.61%59.61% 7.81%7.81% 31.17%31.17% 18.94%18.94% ITU (2022)ITU (2022)34.89%34.89% FB speed (Mbps) 1.111.11 28.5328.53 2.552.55 22.9622.96 1.301.30 12.4112.41 6.906.90 Ookla (2022)Ookla (2022)21.6021.60 Source: Authorsʼ elaboration. Note: Monetary figures expressed in U.S. dollars. FB: Fixed broadband. 29 1. The Nature of the Problem to Be Studied The economic impact of last-mile digital infrastructure deployment (in most cases referred to as "broadband") has been studied in the aggregate at the national level in numerous research studies over the last three decades. Analyses have evolved from a purely correlational methodology toward the development of structural models aimed at demonstrating the economic value of fixed or mobile broadband adoption (Crandall, Lehr, and Litan, 2007; Czernich et al., 2009; Koutroumpis, 2009; Ferrés, 2010; Katz and Koutroumpis, 2012a, 2012b; Atif, Endres, and Macdonald, 2012; Gallego and Gutiérrez, 2013; Katz and Callorda, 2020; Katz and Jung, 2021). In most of these studies, the methodology used was based on the analysis of independent variables at the national level (e.g., fixed or mobile broadband penetration) and dependent variables, such as gross domestic product (GDP) per capita and job creation. The overall conclusion was, with few exceptions, that broadband leads to a number of positive externalities, including economic growth, job creation, and increases in per capita income, productivity, and entrepreneurship development. In parallel, broadband deployment and adoption over the last 15 years in Latin America and the Caribbean has rapidly evolved. In the aggregate, fixed broadband penetration per household has grown from a weighted average of 24.87 percent in 2010 to 56.47 percent in 2020, although, as expected, adoption by country shows certain asymmetries (see Table 2). 30  Table 2. Latin America and the Caribbean: Fixed Broadband Household Penetration (percentage), 2020 Country Argentina Barbados Bolivia Brazil Chile Colombia Costa Rica Dominican Republic Ecuador El Salvador Guatemala Honduras Jamaica Mexico Panama Paraguay Peru Trinidad and Tobago Uruguay Venezuela 78.61 98.58 34.67 57.96 74.63 55.26 67.10 33.93 53.20 42.75 19.17 24.95 45.72 71.98 56.20 26.12 44.44 84.88 76.72 42.73 Percentage Sources: Regulators' reports; extrapolation of ITU estimates. 31 As shown in Table 2, despite the progress recorded at the aggregate level, the continent still shows marked differences between advanced countries (Argentina, Barbados, Chile, Costa Rica, Mexico, Trinidad and Tobago, and Uruguay), countries in transition (Brazil, Colombia, Ecuador, El Salvador, Jamaica, Panama, Peru, and Venezuela), and somewhat more backward countries (Bolivia, Dominican Republic, Guatemala, Honduras, and Paraguay). The same trend can be detected in the case of mobile broadband, where the weighted penetration of unique subscribers9 for the region between 2010 and 2020 has increased from 19.78 percent to 56.82 percent, respectively (see Table 3). Table 3. Latin America and the Caribbean: Mobile Broadband Penetration (unique subscribers as a percentage of the population), 2020 Country Argentina Barbados Bolivia Brazil Chile Colombia Costa Rica Dominican Republic Ecuador El Salvador Guatemala Honduras Jamaica Mexico Panama 68.60 56.27 44.83 61.60 67.99 51.50 62.12 57.62 48.71 46.44 41.26 38.97 48.20 59.42 65.37 Percentage  9 The "unique subscribers" indicator is different from the total number of connections, as it includes only those individuals who have a subscription to the service. 38 In an econometric study comparing the economic impact by region, Katz, Avila, and Meille (2010) analyzed the impacts of broadband on median household income and job creation in rural counties in the state of Kentucky. The study estimated that a 1 percent growth in broadband coverage would result in a 0.0704 percent increase in median income in rural counties adjacent to urban centers, and 0.0800 percent in that of isolated rural counties, compared to the 0.0968 percent found for median income in urban centers. This means that the observed impact on the median wage was higher for metropolitan counties than for isolated rural counties, while outlying rural counties occupied an intermediate position. On the other hand, a 1 percent increase in broadband penetration in rural counties peripheral to urban centers and isolated rural counties was associated with a reduction in the unemployment rate of 0.1953 percent,11 while the results for urban centers were not significant (see Table 6). Productivity gains in certain industries (transportation, lodging, entertainment), leading to job destruction due to the substitution effect between factors of production. Improved provision of health, education, social inclusion, and entertainment, with a consequent increase in consumer surplus. 11 The results, differentiated by type of rural county, do not generate statistically significant coefficients.  Table 6. Kentucky: Impact of a 1 Percent Increase in Broadband Availability on Average Wage and Unemployment Metropolitan Rural peripheral to urban center Rural isolated 0.0968*** 0.0704*** 0,0800*** Average salaryType of county 0.0301 -0.1953*** Unemployment Source: Katz, Avila, and Meille (2010). Statistical significance: ***p<0.01. 39  In the same vein, Mack and Faggian (2013) developed a series of spatial econometric models to examine the impact of broadband deployment on productivity in selected U.S. counties. The authors found that the variance in the impact of broadband was determined by the level of human capital, which determined that productivity gains occurred in territories with high levels of human capital and/or high-skilled workers, concentrated in urban and suburban environments, which increases levels of inequality. Akerman, Gaarder, and Mogstad (2015) deepened these findings and analyzed the complementarity between the skill level of workers and broadband access. In particular, they found that wages and employment level increase with broadband deployment among higher-skilled workers and decrease with lower-skilled workers, thereby increasing inequality. The channel of broadband's impact is through the increase in productivity and performance of those firms that benefit from the deployment of the technology. In other words, broadband adoption, according to these authors, generates a change in the use of production factors by firms, increasing the marginal productivity of the highest-skilled workers. The effect of broadband on productivity was also confirmed by Cambini, Grinza, and Sabatino (2021) in a study conducted in Italy on the deployment of fiber optics at the municipal level. Another confirmation of the heterogeneity of complex effects by region can be seen in a study conducted in Germany by Katz, Avila, and Meille (2010), which differentiates between counties with high and low broadband penetration, which corresponds to the work focused on urban and rural areas. When analyzing the temporal impact in these two geographic areas, it was observed that in urban areas where more last-mile infrastructure was deployed, there was an immediate increase in GDP and employment rate, offsetting the increase in productivity (and consequent job destruction) with the innovation effect and the growth of entrepreneurship. On the other hand, the increase in broadband penetration in rural areas had a smaller initial impact on the GDP growth rate, which increased after the technology managed to penetrate the productive fabric. On the other hand, it was also found that the impact on job creation in rural areas did not show up in the initial years of penetration, since the positive impact of broadband on productivity entailed a capital/labor substitution with no compensatory creation of new enterprises. These effects observed in the different regions can be seen conceptually in Figure 1. 40  Figure 1. Germany: Differential Impact of Broadband by Region Source: Katz, Avila, and Meille (2010). Counties with high penetration Economic impact Counties with low penetration High Low Increased broadband penetration Initial economic growth decreases over time (“supply shock” effect) Growth of the digital economy (innovation, new services) Employment GDP T+1 T+2 T+3 T+4 Economic impact High Low Increased broadband penetration Stable economic growth (effect to reach high penetration counties) Capital/labor substitution places a limit on employment growth (“productivity effect”) Employment GDP T+1 T+2 T+3 T+4 The CINTEL study (Katz and Callorda, 2011) conducted in Colombia between 2006 and 2010 on the impact of the Vive Digital Plan and internet massification confirms different effects for departments with low and high broadband penetration. Thus, it was observed that for every 10 percent increase in broadband penetration, real income per household varies between 0.035 percent for departments with low penetration and 0.025 percent for departments with high penetration. Similarly, the authors point out that there is an impact on the employment rate of 0.003 percent for every 10 percent increase in broadband penetration, which in turn is made up of 0.0029 percent for areas with low penetration and 0.0065 percent for those with high penetration, although the latter are not statistically significant. Atasoy (2013) analyzed the impact of broadband on the U.S. labor market between 1999 and 2007. To do so, he used Federal Communications Commission (FCC) deployment data in conjunction with demographic and labor market information from the Bureau of Labor Statistics census for 3,116 counties. The study was based on a county fixed effects model and found that access to service was associated with a positive impact on the percentage of the population employed of 1.8 percentage points (see Table 7). 41 Table 7. Impact of Broadband Availability in 3,116 U.S. Counties All counties Large metropolitan counties Metropolitan counties Small metropolitan counties Metropolitan micro counties Rural counties 0.0181*** 0.0123* 0.0152* 0.0172* 0.0175* 0.0224*** EmploymentType of county 0.0048 - - - - - Number of establishments -0.0476*** - - - - - Unemployment Source: Atasoy (2013). Statistical significance: ***p<0.01; **p<0.05; *p<0.1.  Similarly, the aforementioned study indicates that unemployment in the counties is reduced by approximately 4.7 percent and the number of establishments is increased by 0.48 percent, although statistical significance was not verified. In addition, the availability of broadband in rural or isolated counties reported an impact on the employed population of 2.2 percentage points, which exceeds the impact found in metropolitan counties, although the author does not state the reasons for a higher rate of increase in employment in rural areas. Beyond the comparative analyses by geographic area, the study by Viollaz and Winkler (2020) using an ordinary least squares and instrumental variables model, conducted in Jordan between 2010 and 2016, analyzed the impact of broadband by gender. The authors identified a positive effect on female labor force participation from internet adoption, but found no effect on male labor force participation. The study shows that for every percentage point increase in internet access, female labor force participation increases by 0.7 percent. In summary, comparative studies of the impact of last-mile digital infrastructure in urban and rural environments, based on the methodology of ordinary least squares econometric models, have identified six differentiated effects: 42 Urban areas tend to benefit more than rural areas, since they concentrate the sectors with the highest transaction intensity and use of information. From a temporal dimension, these benefits of urban areas, such as employment and wage impacts, occur faster in cities than in rural communities. Rural communities on the periphery of urban areas benefit more in terms of employment, wages, and entrepreneurship compared to more isolated ones, as broadband deployment facilitates the relocation of certain industrial sectors from the metropolitan center to the periphery.  Urban and suburban areas with a higher concentration of skilled workers receive more benefits from broadband deployment, mainly due to the increased productivity of more technologically advanced firms. Broadband deployment in rural areas is associated with GDP growth (albeit at a lower rate than in urban areas) and a loss of the least productive jobs in the short term, since the positive impact on productivity results in a capital/labor substitution that is not offset by the innovation and entrepreneurship effect observed in cities. The variance in the impact of broadband is determined by the level of human capital. The complementarity between the skill level of workers and broadband determines that, with the deployment of broadband, wages and employment levels increase among the most skilled workers and decrease among the least skilled, a phenomenon that increases inequality. The broadband impact channel is manifested in the increase in productivity and performance of those companies that benefit from the deployment of this technology. 43  Access to the internet would contribute to reducing the labor participation gap for women, since access to the service could change women's job search strategy. Such behavior is more frequent among young women, women with low educational levels, and single women, and assumes greater access to information as a means to explain greater labor participation. Advances in statistical methods and the gradual availability of statistical series and panels made it possible to advance in the analysis of the differentiated urban-rural impact of last-mile digital infrastructure. In addition, due to the availability of household surveys, studies could be extended beyond the environment of advanced economies. For example, Whitacre, Gallardo, and Strover (2014a) evaluated the impact of broadband on economic growth in rural communities in the United States between 2001 and 2010, in a study of 3,073 counties. The statistical series of supply and adoption of the service were compared with economic variables (e.g., median household income), and the latter were analyzed using the propensity score matching technique between treated group (associated with various broadband thresholds) and control group. The results showed a positive impact on income, especially in rural areas. It is interesting to note that the supply of the service, as opposed to adoption, shows a lower impact (see Table 8). 2.2. Studies Based on Difference-in-Differences Models 44 Table 8. Impact of Broadband in Non-Metropolitan U.S. Counties Non-metropolitan (rural) counties - Number of establishments Source: Whitacre, Gallardo, and Strover (2014a). Statistical significance: ***p<0.01; **p<0.05; *p<0.1. Counties with high broadband availability (>85%) - Counties with low broadband availability (<50%) - Counties with high broadband adoption (>60%) - -0.028*** Counties with low broadband adoption (<40%) -0.054** Income 0.017* 0.013* - 0.010* - Counties with low average speed (<3Mbps) - Employment - -0.034* - -0.0476*** - -0.096*** - - Unemployment  Table 8 indicates that increasing broadband availability (i.e., coverage) in rural counties with high coverage has a negative impact on income. Alternatively, increasing availability in counties with low coverage leads to a positive impact (see the first two rows of the table). The authors speculate that this difference could be due to the fact that, in counties with high coverage, unserved households are not waiting for broadband access. Conversely, in counties with low coverage, the increase in availability leads to an increase in income because there is a population that is waiting to access broadband to generate economic benefits. The impact on adoption confirms that the higher the penetration, the greater the increase in income and number of jobs. Along the same lines, in a working paper by the Peruvian Ministry of Transport and Communications, Aguilar et al. (2020) estimate the impact of internet services on household welfare in the country for the period 2017-2019. To do so, the authors use a quasi-experimental difference-in-differences method, in combination with the propensity score matching technique, on the national household survey panels. 45 The change in income due to fixed internet access at home reported, on average, values of S/ 298.5 per month per year, of which S/ 275.8 corresponded to urban areas and S/ 390.9 to rural areas, while the impact of internet use at the rural level amounted to S/ 212.1 per month per year. The analysis would indicate a greater impact size in the rural sector, provided that internet access through establishments other than the home is included in the analysis, such as digital kiosks, booths, and educational or work centers, whose presence is greater in rural areas than in urban areas. However, it should also be noted that the authors indicate that "although rural incomes are slightly higher, the percentage of access is still quite low." The study by Katz and Callorda (2013) estimated the economic impact of broadband deployment in Ecuador. The authors built a model based on microdata from the National Survey of Employment, Unemployment and Underemployment, and study the impact of broadband service deployment on the income of individuals at the cantonal level between 2009 and 2011, based on a difference-in-differences analysis. The regression model evaluates the impact on the treatment group, which is located in cantons where households started to have broadband, versus a control group, where households never had broadband in the period of analysis. The result is that the deployment of broadband service increases average individual labor income by 3.67 percent per year. The study generated other results of differentiated annual impact on labor income, depending on the use of computer devices and internet access. For example, if the individual uses a computer, the percentage increase in income is higher (3.92 percent). The authors also indicate that there is an annual impact on labor income for the male subsample of 3.40 percent. Among the potential channels of impact on income, the study mentions the following:  Household members can improve their job search by accessing job boards that provide an efficient means of matching supply and demand (matching platforms). Last-mile infrastructure allows household members to improve their ability to highlight and signal their capabilities (resume promotion effect). 46  Broadband access allows household members to access training platforms, which can increase their income through better-paid work. Broadband generates a positive effect on worker productivity. Then, following the classical labor economics literature, wages in competitive markets are equal to marginal productivity and, therefore, the higher the labor productivity, the better the average wages. The introduction of broadband also helps reduce job search times and enables the underemployed to obtain full-time employment or jobs with better conditions. This situation reduces periods of unemployment and generates an increase in the migration of underemployed workers to full-time jobs, which, in turn, is a source of higher labor income. Finally, Bahia et al. (2020) study the impact of mobile broadband coverage on household consumption and poverty reduction in Nigeria. The study crosses information from the general household survey with mobile broadband deployment data from Nigerian mobile operators between 2010 and 2016. The paper uses household and time fixed effects, concluding that mobile broadband coverage had a positive impact on households, both on their consumption levels (increase of about 6 percent) and poverty (4.3 percent reduction of households below the poverty line). The analysis also makes a geographical differentiation, where the impact of coverage on food consumption in rural areas is 7.7 percent, while in urban areas the result is not significant. Similarly, the presence of mobile broadband promotes a 5.2 percent reduction in poverty among rural households. To summarize, comparative studies of the impact of last-mile digital infrastructure on urban and rural areas, based on propensity score matching and difference-in-differences methodologies, have found different effects according to the universe considered: 47  Rural areas tend to benefit more from the availability of broadband, since there are more establishments, such as kiosks, telecenters, educational centers, and others that promote access to information, which would bring greater benefits to the population due to the possibility they have to promote the services and products they can offer in the market. The impact of broadband availability on skilled labor employment is higher in most rural areas compared to urban areas. In general, rural and isolated localities show a significant impact on employment; in addition, all areas show a representative benefit in income levels, unemployment rate, and number of establishments. That is, consistent with research analysis based on ordinary least squares econometric models (especially Akerman, Gaarder, and Mogstad, 2015), the population in rural areas that benefits the most from broadband deployment is the most skilled. While broadband deployment in general leads to a positive variation in the income of the entire rural population, the use of the service or equipment promotes a greater increase in all the variables studied. The differentiation of a lower impact in non-metropolitan or rural areas with high broadband availability, compared to the same types of areas with low service availability, could be explained by the fact that in counties with high coverage, households without service are not waiting for broadband access, while in counties with low coverage, the increase in availability does cause an increase in income, since there is a population that is waiting for broadband access in order to be able to generate economic benefits. In conclusion, the following study, focused on the Latin American context, aims to verify the existence of heterogeneity of effects with the deployment of last-mile broadband access, both geographically (urban-rural), and in terms of training and gender. 54  Table 11. Number of Observations by Country and Year Considered in the Regional Analysis (in thousands) Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Note: As the values are expressed in thousands, there are differences between the sums of the rows/columns and the totals due to rounding. Country 2008 0 391 0 209 46 2009 16 399 247 208 47 2010 0 0 0 206 41 2011 34 358 200 207 41 2012 32 362 0 202 39 2013 36 362 218 198 39 2014 37 362 0 197 38 2015 37 356 267 198 37 2016 39 460 0 196 37 2017 38 458 216 194 35 2018 38 0 0 191 70 2019 40 444 0 190 0 Total 345 3.950 1.149 2.395 471 83 85 18 4 0 163 0 4 91 70 85 18 5 0 0 48 5 103 74 86 18 29 0 43 0 7 102 81 82 18 29 0 0 44 21 121 117 80 17 5 6 74 43 20 120 113 88 24 24 5 0 42 31 122 114 76 24 24 6 258 41 38 134 110 75 23 21 13 0 43 35 128 0 75 22 23 16 269 41 19 136 0 74 22 22 0 0 43 18 125 919 959 204 284 50 984 345 205 1.366 79 68 0 0 5 178 0 3 92 79 83 0 98 0 0 0 3 93 60 1.129 133 1.405 132 827 131 1.304 120 1.113 128 1.376 132 1.247 121 1.465 119 1.565 118 1.529 109 1.029 108 1.106 1.410 15.097 Bolivia Brazil Chile Colombia Costa Rica Ecuador El Salvador Guatemala Honduras Jamaica Mexico Panama Paraguay Peru Uruguay TOTAL 00 0 0 0 0 0 0 0 20 20 21 62 Dominican Republic 55  The unavailability of panel data at the household and individual level prevented us from running difference-in-differences regressions at that level of disaggregation. This problem was solved by generating pseudo-panels through the sub-sovereign units.13 Thus, the next step was to calculate, for each year and sub-sovereign unit, the average (weighted by the weight of each individual observation)14 of the indicators of interest (internet ownership, total income, labor income, years of education, gender, urban population, and rural population). Thus, we ended up with 2,159 observations for the analysis for the years 2008 to 2019 (see Table 12). 14 For indicators such as internet ownership, urban population, rural population, or gender, the only available option is to use the average as a measure to quantify the percentage of the population that meets each condition in each sub-sovereign area, since these are originally binary variables. For other indicators, such as total income, labor income, or years of education, there is the alternative of using the median as the reference indicator. The average was used for consistency in the treatment of all indicators. In any case, applying logarithms to the variables reduces the sensitivity of the estimates to extreme or atypical observations. 13 This was the methodology applied in the case of Ecuador to solve the lack of panel data in the long-term analysis. Table 12. Sub-Sovereign Units by Country and Year Considered in the Regional Analysis Country 2008 0 27 0 25 7 2009 5 27 15 25 7 2010 0 0 0 25 6 2011 9 27 15 24 6 2012 9 27 0 24 6 2013 9 27 15 24 6 2014 9 27 0 24 6 2015 9 27 15 24 6 2016 9 27 0 24 6 2017 9 27 16 24 6 2018 9 0 0 24 6 2019 9 27 0 24 0 Total 86 270 76 291 68 Bolivia Brazil Chile Colombia Costa Rica 16 14 0 0 1 32 0 0 18 14 0 16 0 0 0 1 18 14 3 2 0 32 0 3 18 14 3 2 0 0 10 3 18 14 3 13 0 32 0 4 18 14 3 12 0 0 11 7 18 14 3 2 1 32 11 7 18 14 3 9 1 0 11 7 18 14 3 9 1 32 11 7 18 14 3 11 1 0 11 7 0 14 3 12 1 32 11 8 0 14 3 10 0 0 11 8 178 168 30 98 6 192 87 62 Ecuador El Salvador Guatemala Honduras Jamaica Mexico Panama Paraguay 00 0 0 0 0 0 0 0 8 5 6 19 Dominican Republic 56  Table 12. Sub-Sovereign Units by Country and Year Considered in the Regional Analysis (continued) Source: Authors' elaboration based on IDB Harmonized Household Surveys. Country 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Total 19 166 19 172 19 147 19 175 19 194 19 190 19 198 19 188 19 205 19 199 19 169 19 156 228 2,159 Uruguay TOTAL Finally, for each econometric regression, observations generated by less than 750 surveys were excluded to preserve statistical reliability at the sub-sovereign level. This situation implies the non-use of only 15 observations out of a total of 2,159 in most of the regressions.15 This restriction plays an important role in the study of the sample for rural areas, where, if such an exclusion is not made, very important changes in the adoption of the internet are observed due to the temporal variability in the number of observations of the sub-sovereign units with rural population. 16 For the purposes of this study, all figures will be presented in dollars at purchasing power parity. 15 The inclusion of these 15 observations does not significantly affect the results of the overall model. The first econometric model estimates the impact of broadband on revenues based on all available observations in the region. In the first case, where only the fixed effect per year and per sub-sovereign unit is included as a control, we find that the introduction of the service generates a positive and significant impact on total income of 6.92 percent (US$26.46, according to PPP)16 and 7.41 percent (US$22.32) on labor income. Considering the controls by area (positive and significant effect for urban areas) and years of education (positive and significant effect with more years of formal education), there is a positive and significant impact on total and labor income of 6.83 percent (US$26.11) and 7.32 percent (US$22.04), respectively. The following are some of the channels that make it possible to generate this effect: 4.3. Results 25 25 25 25 25 25 25 25 25 25 25 25 300 Peru 57  Household members can improve their job search by accessing job boards that provide an efficient means of matching supply and demand. Last-mile infrastructure allows household members to improve their ability to highlight and signal their capabilities (resume promotion effect). Broadband access allows household members to access training platforms, which can increase income by getting a better-paying job. Broadband generates a positive effect on worker productivity. In competitive markets, wages are equal to marginal productivity and, therefore, the higher the labor productivity, the better the average wages. The introduction of broadband helps reduce job search times and allows the underemployed to seek full-time employment in this way. This situation reduces periods of unemployment and generates an increase in the migration of underemployed workers to full-time jobs, which, in turn, generates higher labor income. The following is a control for gender, which is not significant, with a positive and significant impact on total and labor income of 6.92 percent (US$26.46) and 7.43 percent (US$22.38), respectively (Table 13). 58  Table 13. Impact of Broadband in the Home on Total and Labor Income of the General Population Ln total revenues Ln labor income (1) 0.0692259 *** (0.0135630) - - (2) 0.0683115 *** (0.0117090) 0.2313491 *** (0.0760247) (3) 0.0692314 *** (0.0116884) 0.2427173 *** (0.0778672) (1) 0.0741532 *** (0.0151337) - - (2) 0.0731912 *** (0.0130618) 0.2558870 *** (0.0842351) 0.0743324 *** (0.0130529) (3) 0.2699903 *** (0.0857824) - - - - -0.4959753 (0.4076116) - - - - -0.6153026 (0.4271907) - - 0.0860749 *** (0.0097619) 0.0869612 *** (0.0096396) - - 0.0877849 *** 0.0115171 0.0888845 *** (0.0113974) 2.145 236 Yes Yes 2.145 236 Yes Yes 2.145 236 Yes Yes 2.145 236 Yes Yes 2.145 236 Yes Yes 2.145 236 Yes Yes General model Offer Zone Gender Years of education Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Average income Impact on income 0.1380 $382.22 $26.46 0.6290 $382.22 $26.11 0.6323 $382.22 $26.46 0.1316 $301.06 $22.32 0.6430 $301.06 $22.04 0.6488 $301.06 $22.38 Source: Authors' elaboration based on IDB Harmonized Household Surveys. Note: Monetary figures expressed in U.S. dollars. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. 59  Subsequently, we estimate the impact of the provision of broadband in the home on income, considering only the available observations for men in the region. When only the fixed effect by year and sub-sovereign unit is included as a control, we find that the introduction of the service generates a positive and significant impact on total and labor income of 6.97 percent (US$27.09) and 7.66 percent (US$24.04), respectively. And if we consider controls for area (positive and significant effect for urban areas) and years of education (positive and significant effect with more years of formal education), we find a positive and significant impact on total and labor income for men of 6.86 percent (US$26.68) and 7.55 percent (US$23.68), respectively (see Table 14). Table 14. Impact of Broadband in the Home on Menʼs Incomes Ln total revenues Ln labor income (1) 0.0696679*** (0.0135370) - - - - 2.145 236 Yes Yes 0.1394 $388.83 $27.09 (2) 0.0686097*** (0.0116355) 0.2548879*** (0.0759164) 0.0845612*** (0.0098025) 2.145 236 Yes Yes 0.5804 $388.83 $26.68 (1) 0.0766183*** (0.0145228) - - - - 2.145 236 Yes Yes 0.1364 $313.73 $24.04 (2) 0.0754907*** (0.0125084) 0.2834128*** (0.0837913) 0.0839045*** (0.0113009) 2.145 236 Yes Yes 0.5991 $313.73 $23.68 Source: Authors' elaboration based on IDB Harmonized Household Surveys. Note: Monetary figures expressed in U.S. dollars. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. Men Offer Zone Years of education Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Average income Impact on income 60  Subsequently, we estimate the impact of the provision of broadband in the home on income, considering only the observations of women available in the region. When only the fixed effect by year and sub-sovereign unit is included as a control, we find that the introduction of the service generates a positive and significant impact on total and labor income of 6.86 percent (US$25.79) and 7.02 percent (US$20.30), respectively. Then, when considering controls for area (positive and significant effect for urban areas) and years of education (positive and significant effect with more years of formal education), there is a positive and significant impact on women’s total and labor income of 6.77 percent (US$25.43) and 6.93 percent (US$20.01), respectively (see Table 15). Table 15. Impact of Broadband in the Home on Womenʼs Income Ln total revenues Ln labor income (1) 0.0686119*** (0.0140419) - - - - 2.145 236 Yes Yes 0.1369 $375.86 $25.79 (2) 0.0676683*** (0.0123922) 0.2360093*** (0.0763537) 0.0846811*** (0.0093192) 2.145 236 Yes Yes 0.6576 $375.86 $25.43 (1) 0.0702520*** (0.0159095) - - - - 2.145 236 Yes Yes 0.1253 $288.94 $20.30 (2) 0.0692562*** (0.0139503) 0.2514879*** (0.0842806) 0.0893621*** (0.0112144) 2.145 236 Yes Yes 0.6626 $288.94 $20.01 Source: Authors' elaboration based on IDB Harmonized Household Surveys. Note: Monetary figures expressed in U.S. dollars. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. Women Offer Zone Years of education Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Average income Impact on income 61  When comparing the results for the subgroups of men and women, we find that the impact measured in dollars is 5 percent lower for women in terms of total income (US$26.68 vs. US$25.43) and 16 percent lower in terms of labor income (US$23.68 vs. US$20.01). In view of these results, the possibility of a reduction in this gender gap was studied for some population subgroup. To this end, the impact of broadband in the home on income was estimated, considering only the observations available for men with fewer than eight years of formal education. The analysis indicates that when only the fixed effect by year and sub-sovereign unit is included as a control, we find that the introduction of the service generates a positive and significant impact on men’s total and labor income of 6.82 percent (US$19.75) and 7.47 percent (US$16.80), respectively. Now, if we consider controls for area (positive and significant effect for urban areas) and years of education (positive and significant effect with more years of formal education), we find a positive and significant impact on the total income of men with fewer than eight years of formal education of 6.05 percent (US$17.53), and 6.61 percent (US$14.86) on labor income (see Table 16). 62  Table 16. Impact of Broadband at Home on the Income of Men with Fewer Than Eight Years of Formal Education Ln total revenues Ln labor income (1) 0.0681984*** (0.0139309) - - - - 2.145 236 Yes Yes (2) 0.0605406*** (0.0121982) 0.3704800*** (0.0947947) 0.1088422*** (0.0247682) 2.145 236 Yes Yes (1) 0.0746555*** (0.0147703) - - - - 2.145 236 Yes Yes (2) 0.0660551*** (0.0135686) 0.4121347*** (0.1068533) 0.0594756** (0.0294200) 2.145 236 Yes Yes Source: Authors' elaboration based on IDB Harmonized Household Surveys. Note: Monetary figures expressed in U.S. dollars. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. 0.1693 $289.52 $19.75 0.5431 $289.52 $17.53 0.1518 $225.03 $16.80 0.4923 $225.03 $14.86 Men with fewer than eight years of formal education Offer Zone Years of education Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Average income Impact on income After repeating this exercise for women with fewer than eight years of formal education, we find that the introduction of the service generates a positive and significant impact on total and labor income of 7.44 percent (US$21.31) and 8.18 percent (US$17.11), respectively. Likewise, when controls for area (positive and significant effect for urban areas) and years of education (positive and significant effect with more years of formal education) are considered, a positive and significant impact on the total income of women with fewer than eight years of formal education of 6.64 percent (US$19.03) and 7.25 percent (US$15.14) on labor income is obtained (see Table 17). 63  Table 17. Impact of Broadband in the Home on the Income of Women with Fewer Than Eight Years of Formal Education Ln total revenues Ln labor income (1) 0.0744094*** (0.0151606) - - - - 2.145 236 Yes Yes 0.1788 $286.39 $21.31 (2) 0.0664418*** (0.0122102) 0.3400318*** (0.0994367) 0.1292859*** (0.0236386) 2.145 236 Yes Yes 0,5597 $286.39 $19.03 (1) 0.0818452*** (0.0165062) - - - - 2.145 236 Yes Yes 0.1753 $209.00 $17.11 (2) 0.0724611*** (0.0141241) 0.4355596*** (0.1170145) 0.0807845** (0.0318135) 2.145 236 Yes Yes 0.4910 $209.00 $15.14 Source: Authors' elaboration based on IDB Harmonized Household Surveys. Note: Monetary figures expressed in U.S. dollars. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. This result shows that, for the subgroup of the population with fewer than eight years of formal education, the introduction of broadband generates an equalizing effect on income between genders. In particular, the gap between men and women is reduced by 9 percent and 2 percent for total and labor income, respectively. Women with fewer than eight years of formal education Offer Zone Years of education Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Average income Impact on income 70  Table 21. Impact of Broadband in the Home on the Income of the Rural Population, with Time Effect Ln total revenues Ln labor income (1) 0.0284039 (0.0200298) 0.0474584 ** (0.0204507) (2) -0.0138592 (0.0188131) 0.0613316 *** (0.0191544) (3) -0.0148846 (0.0186498) 0.0619684 *** (0.0190853) (1) 0.0224925 (0.0252178) 0.0532153 ** (0.0228879) (2) -0.0222663 (0.0228749) 0.0679077 *** (0.0212414) -0.0237783 (0.0227134) (3) 0.0688467 *** (0.0211699) - - - - -0.2534766 (0.3727859) - - - - -0.3737749 (0.4679536) - - 0.1178724 *** (0.0194828) 0.1183948 *** (0.0195359) - - 0.1248329 *** (0.0255781) 0.1256031 *** (0.0257469) 1.624 210 Yes Yes 0.1598 $253.17 $12.02 1.624 210 Yes Yes 0.5158 $253.17 $15.53 1.624 210 Yes Yes 0.5216 $253.17 $15.69 1.624 210 Yes Yes 0.1215 $196.86 $10.48 1.624 210 Yes Yes 0.5315 $196.86 $13.37 1.624 210 Yes Yes 0.5418 $196.86 $13.55 Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Note: Monetary figures expressed in U.S. dollars. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. $0.00 $0.00 $0.00 $0.00 $0.00 $0.00 Rural area and time effect Offer Offer 2008-2013 Gender Years of education Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Average income Impact on income of early adopters Impact on income of late adopters 71  These last two results generate sufficient evidence to affirm that the impact in rural areas is positive and significant in certain cases: (i) when the population has advanced education (more than 11 years of formal education), and (ii) when the service has been available for a prolonged period (more than six years). Next, we estimate the impact of home broadband on income based on observations of individuals with fewer than eight years of formal education. In the first case, where only the fixed effect by year and sub-sovereign unit is included as a control, we find that the introduction of the service generates a positive and significant impact on total and labor income of 4.54 percent (US$12.93) and 5.69 percent (US$12.19), respectively. The controls for zone and years of education have a positive and significant impact on total and labor income of 4.70 percent (US$13.37) and 5.63 percent (US$12.05), respectively. Finally, we also include a control for gender, which is not significant, resulting in a positive and significant impact on total and labor income of 4.77 percent (US$13.57) and 5.55 percent (US$11.89), respectively (see Table 22). 72  Table 22. Impact of Broadband in the Home on the Income of Population with Fewer Than Eight Years of Formal Education Ln total revenues Ln labor income (1) 0.0454221 *** (0.0115645) - - (2) 0.0469864 *** (0.0114102) 0.0727517 (0.0934542) (3) 0.0476736 *** (0.0114047) 0.0689343 (0.0938560) (1) 0.0569403 *** (0.0131150) - - (2) 0.0562852 *** (0.0131476) 0.1109037 (0.1085597) 0.0555306 *** (0.0131012) (3) 0.1150955 (0.1096035) - - - - 0.2099384 (0.2960068) - - - - -0.2305347 (0.3388226) - - 0.0765690 *** (0.0271505) 0.0772429 *** (0.0273978) - - 0.0265546 (0.0365013) 0.0258147 (0.0363368) 2.026 231 Yes Yes 0.1573 $284.65 $12.93 2.026 231 Yes Yes 0.3457 $284.65 $13.37 2.026 231 Yes Yes 0.3409 $284.65 $13.57 2.026 231 Yes Yes 0.1421 $214.06 $12.19 2.026 231 Yes Yes 0.2894 $214.06 $12.05 2.026 231 Yes Yes 0.2903 $214.06 $11.89 Source: Authors' elaboration based on IDB Harmonized Household Surveys. Note: Monetary figures expressed in U.S. dollars. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. Fewer than eight years of formal education Offer Zone Gender Years of education Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Average income Impact on income 73  Next, we estimate the impact of the provision of broadband in the home on income considering only the observations for individuals with eight to 11 years of formal education. When only the fixed effect by year and sub-sovereign unit is included as a control, we find that the introduction of the service generates a positive and significant impact on total and labor income of 4.25 percent (US$15.91) and 5.51 percent (US$16.66), respectively. Now, when we consider controls for area and years of education (which has a positive and significant effect), we obtain a positive and significant impact on total and labor income, respectively, of 3.97 percent (US$14.87) and 5.18 percent (US$15.68). Finally, we also include a control for gender, which is not significant, which shows a positive and significant impact on total and labor income of 4.13 percent (US$15.47) and 5.40 percent (US$16.33), respectively (see Table 23). 74  Table 23. Impact of Broadband in the Home on the Income of the Population with 8–11 Years of Formal Education Ln total revenues Ln labor income (1) 0.0424615 *** (0.0146144) - - (2) 0.0396858 *** (0.0143223) 0.0399957 (0.1206303) (3) 0.04129890 *** (0.0143562) 0.0431417 (0.1210785) (1) 0.0550738 *** (0.0160241) - - (2) 0.0518406 *** (0.0155893) 0.0240657 (0.1251106) 0.0540035 *** (0.0156677) (3) 0.0282842 (0.1254755) - - - - -0.1757119 (0.1973933) - - - - -0.2356054 (0.2113955) - - 0.1309928 *** (0.0434015) 0.1315349 *** (0.0432028) - - 0.1509293 *** (0.0528640) 0.1516562 *** (0.0527357) 1.430 182 Yes Yes 0.0672 $374.64 $15.91 1.430 182 Yes Yes 0.0249 $374.64 $14.87 1.430 182 Yes Yes 0.0240 $374.64 $15.47 1.430 182 Yes Yes 0.0810 $302.43 $16.66 1.430 182 Yes Yes 0.0185 $302.43 $15.68 1.430 182 Yes Yes 0.0185 $302.43 $16.33 Source: Authors' elaboration based on IDB Harmonized Household Surveys. Note: Monetary figures expressed in U.S. dollars. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. Eight to 11 years of formal education Offer Zone Gender Years of education Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Average income Impact on income 75  Subsequently, we estimate the impact of broadband in the home on income, considering only the observations of individuals with more than 11 years of formal education. In the first case, where only the fixed effect by year and sub-sovereign unit is included as a control, we find that the introduction of the service generates a positive and significant impact on total and labor income of 14.31 percent (US$88.45) and 14.45 percent (US$72.46), respectively. Likewise, when controls for area (which has a positive and significant effect) and years of education are considered, there is a positive and significant impact on total and labor income of 10.41 percent (US$64.38) and 10.59 percent (US$53.11), respectively. Finally, we also control for gender, finding a positive and significant impact on total income of 10.39 percent (US$64.23) and 10.56 percent (US$52.97) on labor income (see Table 24). 76  Table 24. Impact of Broadband in the Home on the Income of Population with More Than 11 Years of Formal Education Ln total revenues Ln labor income (1) 0.1430865 *** (0.0244684) - - (2) 0.1041392 *** (0.0245824) 1.154522 *** (0.1590141) (3) 0.1038993 *** (0.0243411) 1.118926 *** (0.1541068) (1) 0.1444644 *** (0.0262717) - - (2) 0.1058834 *** (0.0264131) 1.126660 *** (0.1679727) 0.1056131 *** (0.0261019) (3) 1.086559 *** (0.1624411) - - - - 0.5624732 * (0.2921113) - - - - 0.6336575 ** (0.2917237) - - -0.0671357 *** (0.0140683) -0.0623281 *** (0.0124095) - - -0.0671707 *** (0.0147717) -0.0617547 *** (0.0130560) 1.451 197 Yes Yes 0.0822 $618.18 $88.45 1.451 197 Yes Yes 0.3008 $618.18 $64.38 1.451 197 Yes Yes 0.3431 $618.18 $64.23 1.451 197 Yes Yes 0.0800 $501.57 $72.46 1.451 197 Yes Yes 0.2996 $501.57 $53.11 1.451 197 Yes Yes 0.3448 $501.57 $52.97 Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Note: Monetary figures expressed in U.S. dollars. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. More than 11 years of formal education Offer Zone Gender Years of education Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Average income Impact on income 77  Thus, the analysis by educational level allows us to conclude that the impact of the introduction of broadband increases with years of formal education. This is consistent with the research of Mack and Faggian (2013), which established that the variance in the impact of broadband was conditioned by the level of human capital, which determines that the increase in productivity occurs in places with high levels of human capital and/or workers. Similarly, Akerman, Gaarder, and Mogstad (2015) analyze the complementarity between worker skill level and broadband access. In particular, this study found that, with broadband deployment, wages and employment increase for higher-skilled workers and decrease for less-skilled workers, thus increasing inequality. The impact channel of broadband is the increase in productivity and performance of those companies that benefit from the deployment of this technology. Finally, we examine the possibility that for the complete sample there is a differential effect based on the timing of the introduction of the service. For this purpose, the original model is given two controls: the first one, which takes into account the cases in which the internet service (treatment) was introduced between 2008 and 2011, and the second one, which takes into account the cases in which the internet service (treatment) was introduced between 2012 and 2015. When only the fixed effect by year and sub-sovereign unit is included as a control, we find that the introduction of the service generates a positive and significant impact on total and labor income of 5.43 percent (US$20.77) and 6.25 percent (US$18.81), respectively. In this first scenario, no additional effect is found in the temporality controls. Then, in the second model, we add a control for urban area (with a positive and significant effect) and years of education (also with a positive and significant effect), finding that, in general, the introduction of the service generates a positive effect on total and labor income of 4.24 percent and 5.04 percent, respectively. It is worth noting that this effect is not the total of total income, since for early adopters (between 2008 and 2011) we obtain an additional effect of 3.23 percent on total income and, for late adopters (between 2012 and 2015), one of 3.07 percent. Thus, the impact on total income for early adopters, laggard adopters, and late adopters is US$28.56, US$27.94, and US$16.91, respectively. In relation to labor income, with a statistical significance level of 10 percent,17 the temporal effects are not significant, with an impact of US$15.19 for all periods. 17 At a statistical significance level of 15 percent, the time effects are significant for both early adopters and late adopters. 78  Finally, controlling for gender in total income also yields an overall effect of 4.19 percent, to which must be added an additional effect of 3.44 percent for early adopters and 3.22 percent for late adopters. This implies that the final impact is US$29.14, US$28.30, and US$16.00 for early adopters, laggard adopters, and late adopters, respectively. On the other hand, in relation to labor income we obtain an overall effect of 4.98 percent and, in addition, an additional effect for early adopters of 3.04 percent. The additional effect for laggard adopters is 2.96 percent, but it is not significant at the 10 percent significance level.18 This implies that the impact is US$24.17 for early adopters and US$15 for the other two groups (see Table 25). 18 The coefficient has a statistical significance of 10.1 percent. 79  Table 25. Impact of Broadband in the Home on the Income of the General Population, with Time Effect (full sample) Source: Authors' elaboration based on IDB Harmonized Household Surveys. Note: Monetary figures expressed in U.S. dollars. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. Ln total revenues Ln labor income (1) 0.0543419 *** (0.0160903) 0.0218166 (0.0176273) (2) 0.0423660 *** (0.0148879) 0.0323432 ** (0.0163262) (3) 0.0418548 *** (0.0148815) 0.0343896 ** (0.0163548) (1) 0.0624945 *** (0.0176595) 0.0175524 (0.0193465) (2) 0.0504470 *** (0.0165093) 0.0279634 (0.0181043) 0.0498276 *** (0.0164994) (3) 0.0304426 * (0.0181329) 0.0105796 (0.0175084) 0.0307452 * (0.0162150) 0.0321969 ** (0.0162247) 0.0072855 (0.0192160) 0.0277919 (0.0179810) 0.0295507 (0.0179886) - - 0.2248742 *** (0.0550016) 0.2368282 *** (0.0553580) - - 0.2503650 *** (0.0609917) 0.2648475 *** (0.0613765) - - - - -0.5407628 * (0.2967548) - - - - -0.6551388 ** (0.3290180) - - 0.0870746 *** (0.0055800) 0.0880933 *** (0.0056045) - - 0.0886761 *** (0.0061877) 0.0899102 *** (0.0062139) 2.145 236 Yes Yes 0.1358 $382.22 $20.77 2.145 236 Yes Yes 0.6299 $382.22 $28.56 2.145 236 Yes Yes 0.6337 $382.22 $29.14 2.145 236 Yes Yes 0.1296 $301.06 $18.81 2.145 236 Yes Yes 0.6443 $301.06 $15.19 2.145 236 Yes Yes 0.6506 $301.06 $24.17 $20.77 $20.77 $27.94 $16.19 $28.30 $16.00 $18.81 $18.81 $15.19 $15.19 $15.00 $15.00 Time effects (full sample) Offer Offer 2008-2011 Offer 2012-2015 Zone Gender Years of education Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Average income Impact on income of early adopters Impact on income of laggard adopters Impact on income of late adopters 86 5. Regional Analysis of Impact on Employment Indicators Similar to the analysis of the impact of broadband on income levels, the objective of the regional study that covers employment indicators focuses on the differentiated impact of broadband on employment, considering indicators such as employed population, inactive population, unemployed population, and degree of labor formality. To assess the impact of last-mile infrastructure deployment on employment metrics (percentages of employed population, inactive and unemployed population, and ratio of formal to informal workers), we use a difference-in-differences model, according to the following equation: This is a simple regression, which determines the effect on the percentages of each group generated from residing in an area where broadband service can be accessed at home, where: 5.1. Methodology Percentage of population by group .Treatment .Year=+ + + it t β0β2Area +i β3.X +it itit β4 β1µ(2) Percentage of population by group : ercentage of employed, inactive and unemployed population, and ratio of formal to informal workers. Treatment : This is the variable that distinguishes the two groups. Areas with broadband service in the home, defined as areas where at least 10 percent of the households in the survey adopt the service. 1: Areas with no broadband in the home, defined as areas where less than 10 percent of households adopt the service. 0: it it 87  Year : Corresponds to a fixed effect for each year between 2008 and 2019. Area : Corresponds to a fixed effect for each geographic area (subnational unit) included in the regression. X : A matrix of other independent variables used as controls in some specifications. it t i For the independent variables used in each analysis, different specifications of the econometric models are made. The first model evaluates the direct relationship between treatment and the percentage of the population, by labor group. The second, under the assumption that expected income can affect labor participation decisions, includes a control for total income; and, in a third model, another control for labor income is added. All specifications include controls for year fixed effect (a binary variable for each year included in the regression) and geographic area (a binary variable for each sub-sovereign unit included in the regression). The regional analysis has been performed based on the broadband adoption data used in the preceding analysis of impact on income. For the years, countries, and sub-sovereign units for which information was available, only the microdata responding on household internet ownership and labor indicators are retained. For the observations that meet the prerequisites, cross-country employment data are matched based on the harmonized base indicators. This resulted in a total of 12,430,747 observations at the regional level.21 5.2. Data Used 21 The difference in the number of observations, in relation to the regional analysis on income, lies in the fact that, in this case, only the observations with information on employment status are kept. µ : It is the error term. it 88  As in the case of the previous analysis, the unavailability of panel data at the household and individual level did not allow us to run difference-in-differences regressions at that level of disaggregation, so pseudo-panels were generated through sub-sovereign units.22 Thus, the next step was to generate for each year and sub-sovereign unit the average (weighted by the weight of each individual observation) of the indicators of interest. Finally, for the performance of each econometric regression, observations generated by less than 750 surveys were excluded, as a way to ensure statistical reliability at the sub-sovereign level. This restriction plays an important role in the study of the sample for rural areas, where, if this exclusion is not made, temporal changes in internet adoption associated with temporal variability in the number of observations of sub-sovereign units with rural population are observed. Thus, the analysis ended up with a maximum of 2,119 observations for the period 2008-19. 22 This was the methodology applied in the cases of Brazil, Ecuador, and El Salvador to solve the lack of panel data in the long-term analysis. The first econometric model estimates the impact of broadband in the home on the levels of employed population, inactive population, unemployed population, and labor formality. Considering only those individuals over the age of 18, the population can be grouped into three categories: employed population (55.38 percent), inactive population (41.23 percent) and unemployed population (3.39 percent). Within the employed population, it is possible to distinguish between the formally and informally employed, which, in the period analyzed, yields an average of 35.59 percent of formally employed (out of the total employed population). The first analysis shows that, with the introduction of broadband, there is a significant increase in the employed population by 0.84 percentage points, which means an increase in employment of 1.51 percent (Model 3). This increase in the employed population comes entirely from a population that was previously inactive; specifically, it has been found that with the broadband offer the inactive population decreases by 0.80 percentage points (1.93 percent). Since all the increase in the level of activity is absorbed by a growth in employment levels, no significant changes in unemployment levels are observed. Finally, since the introduction of broadband, an increase in the labor formality rate of 0.66 percentage points is detected, which implies a 1.84 percent rise in formality levels (see Tables 27 and 28). 5.3. Results 89  Table 27. Impact of Broadband in the Home on the Employed and the Inactive Population Employed population Inactive population (1) 1.0354750 *** (0.2481441) - - (2) 0.9792463 *** (0.2438173) 0.0114776 *** (0.0013755) (3) 0.8369154 *** (0.2405168) - - (1) -0.9791757 *** (0.2310316) - - (2) -0.9253868 *** (0.2267683) -0.0109795 *** (0.0012794) -0.7952386 *** (0.2240095) (3) - - - - - - 0.0180701 *** (0.0015719) - - - - -0.0167393 *** (0.0014640) 2.119 232 Yes Yes 2.119 232 Yes Yes 2.119 232 Yes Yes 2.119 232 Yes Yes 2.119 232 Yes Yes Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. 2.119 232 Yes Yes 0.0240 55.38 0.98 1.77 0.0239 55.38 0.84 1.51 0.0102 41.23 -0.98 -2.37 0.0525 41.23 -0.93 -2.24 0.0476 41.23 -0.80 -1.93 0.0029 55.38 1.04 1.87 General model Offer Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact Incremental percentage 90  Table 28. Impact of Broadband in the Home on the Unemployed Population and Labor Formality Unemployed population Labor formality (1) -0.0562992 (0.0753451) - - (2) -0.0538593 (0.0753664) -0.0004980 (0.0004252) (3) -0.0416766 (0.0754138) - - (1) 0.8905924 *** (0.2814423) - - (2) 0.8103909 *** (0.2734755) 0.0164102 *** (0.0015430) 0.6556249 ** (0.2719035) (3) - - - - - - -0.0013307 *** (0.0004929) - - - - 0.0213962 *** (0.0017771) 2.119 232 Yes Yes 2.119 232 Yes Yes 2.119 231 Yes Yes 2.119 231 Yes Yes 2.119 231 Yes Yes Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. 2.119 232 Yes Yes 0.0001 3.39 0.00 0.00 0.0046 3.39 0.00 0.00 0.0236 35.59 0.89 2.50 0.4850 35.59 0.81 2.28 0.4065 35.59 0.66 1.84 0.0090 3.39 0.00 0.00 General model Offer Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact Incremental percentage 91  The following econometric model estimates the impact of broadband in the home on the levels of employed population, inactive population, unemployed population, and labor formality, considering only the subgroup of men. Including in the analysis only men of legal age, it is possible to group them into three categories: employed population (67.11 percent), inactive population (29.48 percent), and unemployed population (3.40 percent). Within the employed population, it is possible to disaggregate between the formally employed and the informally employed, with an average of 36.14 percent being formally employed (out of the total number of employed) during the period analyzed. The first analysis shows that with the introduction of broadband, the level of employment among men increased significantly by 1.22 percentage points, which implies an increase of 1.83 percent in employment (Model 3). This growth comes entirely from a population that was previously inactive, so it is also observed that, with broadband offer, the inactive population decreases by 1.03 percentage points (3.49 percent). In this case, the increase in employment manages to absorb the totality of the reduction in inactivity levels. To this is added the fact that there is also a 0.20 percentage point decrease in the percentage of unemployed men (5.74 percent). In other words, there is a shift from an inactive to an active population, although not all of them find full-time employment. Finally, with the introduction of broadband, there is an increase in the rate of labor formality of 1.27 percentage points, which means an increase of 3.51 percent (see Tables 29 and 30). 92  Table 29. Impact of Broadband in the Home on the Employed and Inactive Population, Men Employed population Inactive population (1) 1.366934 *** (0.2423258) - - (2) 1.334404 *** (0.2412687) 0.0057255 *** (0.0013048) (3) 1.224940 *** (0.2388397) - - (1) -1.143824 *** (0.2097196) - - (2) -1.118443 *** (0.2090054) -0.0044673 *** (0.0011303) -1.029709 *** (0.2072065) (3) - - - - - - 0.0121703 *** (0.0014929) - - - - -0.0097808 *** (0.0012952) 2.119 232 Yes Yes 0.0069 67.11 1.33 1.99 2.119 232 Yes Yes 0.0064 67.11 1.22 1.83 2.119 232 Yes Yes 0.0087 29.48 -1.14 -3.88 2.119 232 Yes Yes 0.0236 29.48 -1.12 -3.79 2.119 232 Yes Yes 0.0206 29.48 -1.03 -3.49 Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. 2.119 232 Yes Yes 0.0024 67.11 1.37 2.04 Men Offer Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact Incremental percentage 93  Table 30. Impact of Broadband in the Home on the Unemployed Population and Labor Formality, Men Unemployed population Labor formality (1) -0.2231096 *** (0.0858070) - - (2) -0.2159610 ** (0.0857021) -0.0012582 ** (0.0004635) (3) -0.1952316 ** (0.0856048) - - (1) 1.5115990 *** (0.2867954) - - (2) 1.4255020 *** (0.2795668) 0.0151856 *** (0.0015121) 1.2676630 *** (0.2770312) (3) - - - - - - -0.0023894 *** (0.0005351) - - - - 0.0209191 *** (0.0017317) Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. 2.119 232 Yes Yes 2.119 232 Yes Yes 2.119 231 Yes Yes 2.119 231 Yes Yes 2.119 231 Yes Yes 2.119 232 Yes Yes 0.0191 3.40 -0.22 -6.35 0.0209 3.40 -0.20 -5.74 0.0525 36.14 1.51 4.18 0.4963 36.14 1.43 3.94 0.4500 36.14 1.27 3.51 0.0002 3.40 -0.22 -6.56 Men Offer Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact Incremental percentage 94  The following econometric model estimates the impact of broadband in the home on the employed population, inactive population, unemployed population, and labor formality, considering only women. Older women can be grouped into three categories: employed population (44.27 percent), inactive population (52.33 percent),23 and unemployed population (3.39 percent). In turn, if we take the subgroup of the employed population, it is possible to distinguish between the formally employed and the informally employed. The study finds that in the period analyzed the average number of formally employed (out of the total number of employees) was 35.02 percent. The first three analyses show that, with the introduction of broadband, there are no significant changes among women in the distribution between the employed, inactive, and unemployed population. Furthermore, the last econometric model shows no significant impact on women in terms of labor formality (see Tables 31 and 32). 23 It is likely that the gender gap in the inactivity rate is due to the fact that it probably includes care and household tasks that are indispensable for the family, which makes the transition to broadband-facilitated jobs more difficult. 95  Table 31. Impact of Broadband in the Home on the Employed and Inactive Population, Women Employed population Inactive population (1) 0.6775855 ** (0.3075820) - - (2) 0.6268281 ** (0.3002078) 0.0162344 *** (0.0016670) (3) 0.4697075 (0.2972760) - - (1) -0.6663388 ** (0.3052229) - - (2) -0.6153361 ** (0.2977140) -0.0163129 *** (0.0016532) -0.4611509 (0.2951129) (3) - - - - - - 0.0226225 *** (0.0019100) - - - - -0.0223298 *** (0.0018961) 2.119 232 Yes Yes 0.0395 44.27 0.63 1.42 2.119 232 Yes Yes 0.0386 44.27 0.00 0.00 2.119 232 Yes Yes 0.0083 52.33 -0.67 -1.27 2.119 232 Yes Yes 0.0680 52.33 -0.62 -1.18 2.119 232 Yes Yes 0.0615 52.33 -0.46 -0.88 Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. 2.119 232 Yes Yes 0.0052 44.27 0.68 1.53 Women Offer Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact Incremental percentage 102  Table 36. Impact of Broadband in the Home on the Unemployed Population and Labor Formality, Rural Zones Unemployed population Labor formality (1) 0.1039985 (0.0830071) - - (2) 0.0969888 (0.0834523) 0.0005241 (0.0006220) (3) 0.0965847 (0.0837022) - - (1) 1.1414150 *** (0.3964513) - - (2) 1.0299370 *** (0.3950407) 0.0083399 *** (0.0029446) 0.9717130 ** (0.3942017) (3) - - - - - - 0.0005269 (0.0007372) - - - - 0.0120655 *** (0.0034722) 497 77 Yes Yes 0.0449 1.88 0.00 0.00 497 77 Yes Yes 0.0425 1.88 0.00 0.00 495 76 Yes Yes 0.0413 20.06 1.14 5.69 495 76 Yes Yes 0.2553 20.06 1.03 5.14 495 76 Yes Yes 0.2822 20.06 0.97 4.84 Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Note: Econometric models were used for rural areas, considering seasonality. There is no impact of migration from inactive to employed population, and the effect on the improvement of labor formality is maintained. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. 497 77 Yes Yes 0.0207 1.88 0.00 0.00 Rural area Offer Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact Incremental percentage 103  This result demonstrates that broadband deployment can generate increases in inequality, especially when it is not supported by public digital literacy policies aimed at achieving its assimilation among different sectors of the population, such as people residing in rural areas. The following econometric model estimates the impact of broadband in the home on the levels of employed population, inactive population, unemployed population and labor formality, considering only the population with fewer than 11 years of formal education. Among older individuals with fewer than 11 years of formal education, it is possible to distinguish three groups: employed population (49.33 percent), inactive population (48.10 percent), and unemployed population (2.57 percent). Likewise, as mentioned above, the employed population includes both the formally and informally employed; in the period analyzed, the average number of formally employed (out of the total number of employees) was 23.50 percent. The first three analyses show that the introduction of broadband does not generate significant changes for the population with fewer than 11 years of formal education in its distribution among the employed, the inactive, and the unemployed. In contrast, the last of the models suggests that with the introduction of broadband there is an increase in the labor formality rate of 0.47 percentage points, which implies an increase of 2.02 percent (see Tables 37 and 38). 104  Table 37. Impact of Broadband in the Home on the Employed and Inactive Population with Fewer Than 11 Years of Formal Education Employed population Inactive population (1) -0.0884591 (0.2760333) - - (2) -0.1197647 (0.2740498) 0.0122088 *** (0.0023096) (3) -0.2377975 (0.2711757) - - (1) 0.1897896 (0.2631094) - - (2) 0.2159765 (0.2616858) -0.0102125 (0.0022054) 0.3104526 (0.2599669) (3) - - - - - - 0.0230463 *** (0.002686) - - - - -0.0186210 (0.002575) 2.040 232 Yes Yes 0.0001 49.33 0.00 0.00 2.040 232 Yes Yes 0.0005 49.33 0.00 0.00 2.040 232 Yes Yes 0.0081 48.10 0.00 0.00 2.040 232 Yes Yes 0.0033 48.10 0.00 0.00 2.040 232 Yes Yes 0.0007 48.10 0.00 0.00 2.040 232 Yes Yes 0.0063 49.33 0.00 0.00 Source: Authors' elaboration based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. Fewer than 11 years of formal education Offer Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact Incremental percentage 105  Table 38. Impact of Broadband in the Home on the Unemployed Population and Labor Formality, for Population with Fewer Than 11 Years of Formal Education Unemployed population Labor formality (1) -0.1013308 (0.0710284) - - (2) -0.0962120 (0.0708445) -0.0019963 *** (0.0005971) (3) -0.0726553 (0.0704097) - - (1) 0.5660362 ** (0.2350292) - - (2) 0.5339776 ** (0.2325043) 0.0125456 *** (0.0019597) 0.4735480 ** (0.2331032) (3) - - - - - - -0.0044253 *** (0.0006974) - - - - 0.0142830 *** (0.0023091) 2.040 232 Yes Yes 0.0450 2.57 0.00 0.00 2.040 232 Yes Yes 0.0298 2.57 0.00 0.00 2.035 231 Yes Yes 0.0283 23.50 0.57 2.41 2.035 231 Yes Yes 0.4357 23.50 0.53 2.27 2.035 231 Yes Yes 0.2878 23.50 0.47 2.02 2.040 232 Yes Yes 0.0023 2.57 0.00 0.00 Source: Authors' elaboration based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. Fewer than 11 years of formal education Offer Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact Incremental percentage 106  The following econometric model estimates the impact of broadband in the home on the levels of employed population, inactive population, unemployed population, and labor formality considering only the population with more than 11 years of formal education. Among older individuals with more than 11 years of formal education, it is possible to distinguish three groups: employed population (67.01 percent), inactive population (27.96 percent), and unemployed population (5.04 percent). In addition, the employed population includes the formally employed and the informally employed; in the period analyzed, the average number of formally employed (out of the total number of employees) was 52.94 percent. The first analysis shows that the introduction of broadband significantly increased the employed population in urban areas by 0.69 percentage points, which implies an increase in employment of 1.03 percent (model 3). This growth comes entirely from a previously inactive population; in particular, with broadband, the inactive population decreases by 1.49 percentage points (5.35 percent). The incentives that the inactive population had to become active were of such magnitude that the labor market did not manage to employ all of the new population. Thus, these 0.80 percentage points of excluded population generated an increase in the unemployment rate. Finally, with respect to the labor formality rate, there was an increase of 1.01 percentage points, equivalent to an increase of 1.92 percent (see Tables 39 and 40). 107  Table 39. Impact of Broadband in the Home on the Employed and Inactive Population with More Than 11 Years of Formal Education Employed population Inactive population (1) 0.9029532 ** (0.3865692) - - (2) 0.8179917 ** (0.3742232) 0.0118776 *** (0.0010516) (3) 0.6931426 * (0.3690447) - - (1) -1.6796880 *** (0.3642418) - - (2) -1.6031800 *** (0.3536471) -0.0106958 *** (0.0009938) -1.4945750 *** (0.3498592) (3) - - - - - - 0.0160643 *** (0.0011752) - - - - -0.0141733 *** (0.0011141) 2.119 232 Yes Yes 0.0918 67.01 0.82 1.22 2.119 232 Yes Yes 0.0989 67.01 0.69 1.03 2.119 232 Yes Yes 0.0247 27.96 -1.68 -6.01 2.119 232 Yes Yes 0,0862 27.96 -1.,60 -5.73 2.119 232 Yes Yes 0.0898 27.96 -1.49 -5.35 2.119 232 Yes Yes 0.0206 67.01 0.90 1.35 Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. More than 11 years of formal education Offer Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact Incremental percentage 108  Table 40. Impact of Broadband in the Home on the Unemployed Population and Labor Formality, for Population with More Than 11 Years of Formal Education Unemployed population Labor formality (1) 0.7767354 *** (0.1457779) - - (2) 0.7851886 *** (0.1455223) -0.0011818 *** (0.0004089) (3) 0.8014332 *** (0.1452969) - - (1) 1.1106810 ** (0.4341643) - - (2) 1.0719690 ** (0.4320894) 0.0054003 *** (0.0012143) 1.0139490 ** (0.4313228) (3) - - - - - - -0.0018910 *** (0.0004627) - - - - 0.0073995 *** (0.0013735) 2.119 232 Yes Yes 0.0299 5.04 0.79 15.58 2.119 232 Yes Yes 0.0274 5.04 0.80 15.91 2.114 231 Yes Yes 0.0206 52.94 1.11 2.10 2.114 231 Yes Yes 0.3540 52.94 1.07 2.02 2.114 231 Yes Yes 0.3448 52.94 1.01 1.92 2.119 232 Yes Yes 0.0409 5.04 0.78 15.42 Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. More than 11 years of formal education Offer Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact Incremental percentage 109  As is the case with gender and residence, broadband deployment can generate increases in inequality in relation to people's formal education, provided that it is not accompanied by public digital literacy policies aimed at facilitating its use among the different sectors of the population. The last of the econometric models estimates the impact of broadband in the home on levels of employed population, inactive population, unemployed population, and labor formality, considering the entire population over the age of 18 and including an additional control for seasonality. Older individuals can be grouped into three categories: employed population (55.38 percent), inactive population (41.23 percent), and unemployed population (3.39 percent). Within the employed population, a distinction can be made between the formally employed and the informally employed. In the period analyzed, the average number of formally employed (out of the total number of employees) was 35.59 percent. In particular, this model finds that, in the short term, labor formality increases by 3.62 percent, while the level of employment does not show a significant change. On the other hand, in the long term, the level of labor formality reports an increase of only 0.91 percent, due to a 2.66 percent increase in the employed population (see Tables 41 and 42). It is possible that, in the long term, the new jobs generated are informal (quantified in terms of the increase in the employed population), which explains the lower increase in labor formality. 110  Table 41. Impact of Broadband in the Home on the Employed and Inactive Population, with Time Effect Employed population Inactive population (1) 0.3044914 (0.3443026) 1.122808 *** (0.3674958) (2) 0.0061285 (0.3391940) 1.489827 *** (0.3626368) (3) -0.1309505 (0.3341858) 1.477409 *** (0.3557958) (1) -0.2742403 (0.3205007) -1.082797 *** (0.3420905) (2) 0.0112706 (0.3153748) -1.434006 *** (0.3371714) 0.1295072 (0.3111747) (3) -1.411588 *** (0.3312967) - - 0.0121294 *** (0.0013789) - - - - -0.0116069 *** (0.0012821) - - - - - - 0.0186184 *** (0.0015707) - - - - -0.0172633 *** (0.0014625) 2.119 232 Yes Yes 0.0050 55.38 1.12 0.00 2.03 0.00 2.119 232 Yes Yes 0.0266 55.38 1.49 0.00 2.69 0.00 2.119 232 Yes Yes 0.0260 55.38 1.48 0.00 2.67 0.00 2.119 232 Yes Yes 0.0122 41.23 -1.08 0.00 -2.63 0.00 2.119 232 Yes Yes 0.0553 41.23 -1.43 0.00 -3.48 0.00 2.119 232 Yes Yes 0.0499 41.23 -1.41 0.00 -3.42 0.00 Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. With time effect Offer Offer 2008-2012 Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact on early adopters Impact on late adopters Incremental percentage of early adopters Incremental percentage of late adopters 111  Table 42. Impact of Broadband in the Home on the Unemployed Population and Labor Formality, with Time Effect Unemployed population Labor formality (1) -0.0302508 (0.1047986) -0.0400110 (0.1118582) (2) -0.0173989 (0.1053128) -0.0558202 (0.1125913) (3) 0.0014435 (0.1052551) -0.0658212 (0.1120614) (1) 1.7800030 *** (0.3903615) -1.3661100 *** (0.4166504) (2) 1.3861280 *** (0.3816925) -0.8814224 ** (0.4080667) 1.2882040 *** (0.3789559) (3) -0.9655796 ** (0.4034554) - - -0.0005225 (0.0004281) - - - - 0.0160247 *** (0.0015518) - - - - - - -0.0013552 *** (0.0004947) - - - - 0.0210379 *** (0.0017812) 2.119 232 Yes Yes 0.0095 3.39 0.00 0.00 0.00 0.00 2.119 232 Yes Yes 0.0001 3.39 0.00 0.00 0.00 0.00 2.119 232 Yes Yes 0.0045 3.39 0.00 0.00 0.00 0.00 2.114 231 Yes Yes 0.0287 35.59 0.41 1.78 1.16 5.00 2.114 231 Yes Yes 0.4831 35.59 0.50 1.39 1.42 3.89 2.114 231 Yes Yes 0.4069 35.59 0.32 1.29 0.91 3.62 Source: Authors' elaboration, based on IDB Harmonized Household Surveys. Standard errors in parentheses. Statistical significance: ***p<0.01; **p<0.05; *p<0.1. With time effect Offer Offer 2008-2012 Total income Labor income Remarks Groups Effect per year Fix effect for sub-sovereign unit R2 Percentage of population Impact on early adopters Impact on late adopters Incremental percentage of early adopters Incremental percentage of late adopters 118 The minor impact on rural areas, a topic widely covered in the literature surveyed, requires recognition that conventional rural development programs aimed at the creation of new ventures represent an adequate complement to the development of digital infrastructure with universal reach. The results of the study suggest that there is a greater impact on those who actually use the internet service. In other words, the results suggest a spillover effect toward the entire population of the sub-sovereign unit, which, however, may increase income inequality between users and non-users. Therefore, there is a need to implement public policies to encourage the adoption of broadband service to close the demand gap in the localities that receive connectivity. This may be reflected in the growing gender inequality and differences in educational level. With respect to the growing gender inequality, the greater short-term impact on men's employment (due to the network building effect), and the lack of impact on women's labor participation in the long term, indicate the need to act on online employment opportunities in sectors with higher labor participation among women, such as services, health, and education. These programs can include support modules for homemakers to increase social and economic inclusion. Among the best practices for the development of such programs (Katz and Berry, 2014), international experience recommends:  119 Digital literacy and mentoring courses should be taught by women. Promote courses in places that women attend, such as clinics, markets, and schools. Consider partnerships of each program with civil society organizations involved in women's development and inclusion. Organize curricula in such a way that they focus on the specific needs of the students and not on compliance with a rigid program.  120 References Aguilar, J. , C. Gil, E. Aparco, D. Acosta, A. Cajavilca, A. Camayo et al. 2020. Economic Impact of Internet Access in Peruvian Households. Working Paper No. 01. Lima: Ministry of Transport and Communications (MTC). Available at: https://www.gob.pe/institucion/mtc/informes-publicaciones/1359614-impacto-economicodel-acceso-a-internet-en-los-hogares-peruanos. Industrial and Labor Relations Review, 66(2): 315–345. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1890709. Atasoy, H. 2013. The Effects of Broadband Internet Expansion on Labor Market Outcomes. A Panel Data Analysis of OECD Countries. Kiel: Leibniz Information Centre for Economics (ZBW). Available at: http://hdl.handle.net/10419/65419 Atif, S., J. Endres, and J. Macdonald. 2012. Broadband Infrastructure and Economic Growth: Mobile Broadband Internet: Evidence from Nigeria. TPRC48: The 48th Research Conference on Communication, Information and Internet Policy. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3757666. Bahia, K., P. Castells, G. Cruz, T. Masaki, X. Pedrós, T. Pfutze et al. 2020. The Welfare Effects of Evidence from Central Appalachia. Huntington, WV: Center for Business and Economic Research, Marshall University. Available at: https://www.marshall.edu/cber/files/2021/04/2005_07_XX_Final_Rural_Broadband_July_2005.pdf Burton, M. and M. Hicks. 2005. The Residential and Commercial Benefits of Rural Broadband: Evidence from Italian Firms. CEBRIG Working Paper No. 21-020 (December). Brussels: Free University of Brussels. Available at: https://ideas.repec.org/p/sol/wpaper/2013-334687.html. Cambini, C., E. Grinza, and L. Sabatino. 2021. Ultra-Fast Broadband Access and Productivity: Internet. Working Paper No. 20826. Cambridge, MA: National Bureau of Economic Research. Available at: https://www.nber.org/papers/w20826. Akerman, A., I. Gaarder, and M. Mogstad. 2015. The Skill Complementarity of Broadband and Employment: A Cross-sectional Analysis of U.S. Data. Issues in Economic Policy No. 6. Washington, D.C.: The Brookings Institution. Available at: https://www.brookings.edu/wp-content/uploads/2016/06/06labor_crandall.pdf. Crandall, R., W. Lehr, and R. Litan. 2007. The Effects of Broadband Deployment on Output Economic Growth. Working Paper No. 2861. Munich: CESifo. Available at: https://www.cesifo.org/DocDL/cesifo1_wp2861.pdf. Czernich, N., O. Falck, T. Kretschmer, and L. Woessmann. 2009. Broadband Infrastructure and 121  2006–2009. Lima: Regional Dialogue on the Information Society. Available at: https://docplayer.es/27760209-Adopcion-de-internet-empleo-y-pobreza-en-uruguay.html Ferrés, D. 2010. Relationship between internet adoption and high quality employment: Uruguay American Economic Review. Available at: https://www-2.rotman.utoronto.ca/~agoldfarb/internet_wages.pdf. Forman, C., A. Goldfarb, and S. M. Greenstein. 2010. The Internet and local wages: A puzzle. Analysis at the Level of Municipalities and 23 Major Cities. Lima: DIRSI. Available at: https://repository.urosario.edu.co/handle/10336/28371. Gallego, J. and L. Gutiérrez. 2013. Internet and Economic Activity in Colombia, 2007-2011: An 2020. The Impact of Digital Infrastructure on the Consequences of COVID-19 and Mitigation of Future Effects. Washington, D.C.: IDB. Available at: https://publications.iadb.org/en/impact-digital-infrastructure-consequences-covid-19and-mitigation-future-effects García Zaballos, A., E. Iglesias Rodríguez, M. Cave, A. Elbittar, R. Guerrero, E. Mariscal et al. Final Report. National Technical Assistance, Training, Research, and Evaluation Project No. 99-07-13829. Washington, D.C.: U.S. Department of Commerce, Economic Development Administration. Available at: http://cfp.mit.edu/publications/CFP_Papers/Measuring_bb_econ_impact-final.pdf. Gillett, S., W. Lehr, C. Osorio, and M. A. Sirbu. 2006. Measuring Broadband's Economic Impact. Virtual Meeting, December 13–15, 2021. Available at: https://innovalac.cepal.org/3/en –––. 2021. Third Meeting of the ECLAC Conference on Science, Innovation and ICTs | access to digital technologies to address the consequences of COVID-19. Santiago: ECLAC. Available at: https://www.cepal.org/en/publications/45939-universalizing-access-digital-technologiesaddress-consequences-covid-19 ECLAC (Economic Commission for Latin America and the Caribbean). 2020. Universalizing https://www.gsmaintelligence.com/data/. GSMA. 2021. GSMA Intelligence Database. Available at: ILO. Available at: https://ilostat.ilo.org/. ILO (International Labour Organization). 2021. ILO Statistics Database (ILOSTAT). Geneva: 122  An Evaluation of USDA's Broadband Loan Program. Growth and Change, 41(2): 165–191. Available at: https://onlinelibrary.wiley.com/doi/10.1111/j.1468-2257.2010.00524.x. Kandilov, I. and M. Renkow. 2010. Infrastructure Investment and Rural Economic Development: Washington, D.C.: Rural Cellular Association (RCA). Available at: http://www.teleadvs.com/wp-content/uploads/RCA_FINAL.pdf. Katz, R., J. Avila, and G. Meille. 2010. Economic Impact of Wireless Broadband in Rural America. Springer International Publishing. Available at: https://link.springer.com/book/10.1007/978-3-319-07197-8. Katz, R. and T. Berry. 2014. Driving Demand for Broadband Network and Services. Geneva: the Massification of the Internet in the Online Government Strategy. Government of Colombia: Ministry of Information and Communication Technologies, CINTEL (Center for Research and Development in Information and Communication Technologies). Bogotá: MINTIC. Available at: http://www.teleadvs.com/wp-content/uploads/CINTEL_Informe_Final_Impacto_VD_y_GEL__V1._10_VF.pdf Katz, R. and F. Callorda. 2011. Measuring the Impact of the Plan Vive Digital in Colombia and https://idl-bnc-idrc.dspacedirect.org/handle/10625/53797 ———. 2013. Impact of Broadband Deployment in Ecuador. Lima: DIRSI. Available at: Global Econometric Modelling. Geneva: ITU. Available at: https://www.itu.int/dms_pub/itu-d/opb/pref/D-PREF-EF.BDR-2020-PDF-E.pdf. ———. 2020. How Broadband, Digitization and ICT Regulation Impact the Global Economy: D.C: IMF. Available at: https://www.imf.org/en/Publications/WEO/weo-database/2019/October IMF (International Monetary Fund). 2019. World Economic Outlook Database. Washington, the Post-COVID-19 World - Report of the VII ITU Economic Experts Roundtable. Geneva: ITU. Available at: https://www.itu.int/pub/D-PREF-EF.POST_COVID-2021 ITU (International Telecommunication Union). 2021. The Telecommunications Industry in Available at: https://www.itu.int/en/ITU-D/Statistics/Pages/publications/wtid.aspx. ———.2022. World Telecommunication/ICT Indicators Database 2021. Geneva: ITU. 123 ITU, Broadband Commission for Digital Development. Available at: http://www.teleadvs.com/wp-content/uploads/BB_MDG_Panama_-Final1.pdf. Katz, R. and P. Koutroumpis. 2012a. The Economic Impact of Broadband in Panama. Geneva: Commission for Digital Development. Available at: https://www.eldis.org/document/A75831. ———. 2012b. The Economic Impact of Broadband in the Philippines. Geneva: ITU, Broadband the German Economy. Intereconomics, 45(1): 26–34. Available at: https://www.intereconomics.eu/contents/year/2010/number/1/article/the-impact-ofbroadband-on-jobs-and-the-german-economy.html. Katz, R., S. Suter, S. Vaterlaus, and P. Zenhäusern. 2010. The Impact of Broadband on Jobs and Public Policy Institute of California (PPIC). Available at: https://www.ppic.org/content/pubs/report/R_110JKR.pdf. Kolko, J. 2010. Does Broadband Boost Local Economic Development? San Francisco, CA: Approach. Telecommunications Policy, 33(9): 471–485. Available at: https://www.sciencedirect.com/science/article/abs/pii/S0308596109000767. Koutroumpis, P. 2009. The Economic Impact of Broadband on Growth: A Simultaneous Regional Science Review, 36(3): 392–423. Available at: https://journals.sagepub.com/doi/abs/10.1177/0160017612471191. Mack, E. and A. Faggian. 2013. Productivity and Broadband: The Human Factor. International Ookla. 2021. Speedtest Global Index. Available at: https://www.speedtest.net/global-index#mobile.  al. 2022a. Measuring the Socioeconomic Impact of Last-mile Infrastructure Development in Brazil. Washington, D.C.: IDB. Available at: https://publications.iadb.org/es/medicion-del-impacto-socioeconomico-del-desarrollode-infraestructura-de-ultima-milla-en-brasil. Puig Gabarró, P., R. Katz, H. Galperin, F. Callorda, E. Iglesias Rodríguez, A. García Zaballos et in Ecuador. Washington, D.C.: IDB. Available at: https://publications.iadb.org/es/medicion-del-impacto-socioeconomico-del-desarrollode-infraestructura-de-ultima-milla-en-ecuador. ----------. 2022b. Measuring the Socioeconomic Impact of Last-mile Infrastructure Development 124 in El Salvador. Washington, D.C.: IDB. Available at: https://publications.iadb.org/es/medicion-del-impacto-socioeconomico-del-desarrollode-infraestructura-de-ultima-milla-en-el. ———. 2022c. Measuring the Socioeconomic Impact of Last-mile Infrastructure Development in Jamaica. Washington, D.C.: IDB. Available at: https://publications.iadb.org/es/medicion-del-impacto-socioeconomico-del-desarrollode-infraestructura-de-ultima-milla-en-jamaica. ———. 2022d. Measuring the Socioeconomic Impact of Last-mile Infrastructure Development Begets Further Advantage. New York: Columbia University Press. pp. 35–52. Rigney, D. 2010. Matthew Effects in the Economy. The Matthew Effect: How Advantage Shippagan Area, a Rural Zone in Atlantic Canada. Journal of Information, Information Technology, and Organizations, 2. Available at: https://www.researchgate.net/publication/267956616_Social_Impact_of_Broadband_Internet_A_Case_Study_in_the_Shippagan_Area_a_Rural_Zone_in_Atlantic_Canada. Selouani, S. and H. Hamam. 2007. Social Impact of Broadband Internet: A Case Study in the in Kentucky. Regional Economic Development, 3(2): 88–118. St. Louis, MO: Federal Reserve Bank of St. Louis. Available at: https://files.stlouisfed.org/files/htdocs/publications/red/2007/02/Shideler.pdf. Shideler, D., N. Badasyan, and L. Taylor. 2007. The Economic Impact of Broadband Deployment Fibre Network. Ontario: Strategic Networks Group. Available at: http://www.sngroup.com/wp-content/uploads/2011/03/DTI-SD-Case-Study_Final_IssuedJune-27-2003.pdf. Strategic Networks Group. 2003. Economic Impact Study of the South Dundas Township  Pandemic: Econometric Modelling. Geneva: ITU. Available at: https://www.itu.int/dms_pub/itu-d/opb/pref/D-PREF-EF.COV_ECO_IMPACT_B-2021-PDF-E.pdf. Katz, R. and J. Jung. 2021. The Economic Impact of Broadband and Digitization through the pandemic? Evidence from SARS. Telecommunications Policy, 44(10): 102044. Available at: https://www.researchgate.net/publication/345940025_Can_digitization_mitigate_the_economic_damage_of_a_pandemic_Evidence_from_SARS. Katz, R., J. Jung, and F. Callorda. 2020. Can digitization mitigate the economic damage of a 125 in Rural Areas: Moving Towards a Causal Relationship. Telecommunications Policy, 38(11): 1011–1023. Available at: https://www.researchgate.net/publication/263847256_Broadbands_contribution_to_economic_growth_in_rural_areas_Moving_towards_a_causal_relationship. Whitacre, B., R. Gallardo, and S. Strover. 2014a. Broadband's Contribution to Economic Growth First-Differenced Regressions. The Annals of Regional Science, 53(3): 649–670. Available at: https://www.researchgate.net/publication/272008852_Does_rural_broadband_impact_jobs_and_income_Evidence_from_spatial_and_first-differenced_regressions. ———. 2014b. Does Rural Broadband Impact Jobs and Income? Evidence from Spatial and  Policy Research Working Paper No. 9183. Washington, D.C.: World Bank. Available at: https://www.researchgate.net/publication/340129978_Does_The_Internet_Reduce_Gender_Gaps_The_Case_of_Jordan. Viollaz, M. and H. Winkler. 2020. Does the Internet Reduce Gender Gaps? The Case of Jordan. 126  Annex 1. Countries and Sub-Sovereign Units Considered in the Regional Analysis Table A1. Countries and Sub-Sovereign Units Used to Measure the Socioeconomic Impact of Last-Mile Infrastructure Development in Latin America and the Caribbean Country Region Bolivia Bolivia Bolivia Bolivia Bolivia Bolivia Bolivia Bolivia Bolivia Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Chuquisaca La Paz Cochabamba Oruro Potosí Tarija Santa Cruz Beni Pando Rondônia Acre Amazonas Roraima Pará Amapá Tocantins Maranhão Piauí Ceará Rio Grande do Norte Paraíba Pernambuco Alagoas Sergipe Bahia Minas Gerais Espírito Santo Country Region Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Chile Chile Chile Chile Chile Chile Chile Chile Chile Chile Chile Chile Chile Chile Chile Chile Colombia Colombia Rio de Janeiro São Paulo Parana Santa Catarina Rio Grande do Sul Mato Grosso do Sul Mato Grosso Goiás Distrito Federal Tarapacá Antofagasta Atacama Coquimbo Valparaíso Libertador General Bernardo O'Higgins Maule Bío Bío La Araucanía Los Lagos Aysén del General Carlos Ibáñez del Campo Magallanes and Antarctica Chilena Metropolitana de Santiago Los Ríos Arica y Parinacota Not delimited Antioquia Atlántico 127 Table A1. Countries and Sub-Sovereign Units Used to Measure the Socioeconomic Impact of Last-Mile Infrastructure Development in Latin America and the Caribbean (continued) Country Region Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Colombia Costa Rica Costa Rica Costa Rica Costa Rica Costa Rica Costa Rica Costa Rica Bogotá, D.C. Bolívar Boyaca Caldas Caquetá Cauca César Córdoba Cundinamarca Chocó Huila La Guajira Magdalena Meta Nariño Norte de Santander Quindío Risaralda Santander Sucre Tolima Valle Central Chorotega Pacífico central Brunca Huetar Atlántica Huetar Norte Limón Country Region Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Dominican Republic Distrito Nacional Azua Bahoruco Barahona Dajabón Duarte Elías Piña El Seibo Espaillat Independencia La Altagracia La Romana La Vega María Trinidad Sánchez Monte Cristi Pedernales Peravia Puerto Plata Salcedo Samana San Cristóbal San Juan San Pedro de Macorís Sanchez Ramirez Santiago Santiago Rodriguez Valverde Monseñor Nouel Monte Plata 