ID4D GLOBAL DATASET Volume 1 | 2021 Global ID Coverage Estimates © 2022 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington, DC 20433 Telephone: 202-473-1000; internet: www.worldbank.org Some rights reserved This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy, completeness, or currency of the data included in this work and does not assume responsibility for any errors, omissions, or discrepancies in the information, or liability with respect to the use of or failure to use the information, methods, processes, or conclusions set forth. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http:// creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: Clark, J., Metz, A., and Casher, C. 2022. ID4D Global Dataset 2021: Global ID Coverage Estimates. Washington, DC: World Bank.” Washington, DC: World Bank. License: Creative Commons Attribution CC BY 3.0 IGO. Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third-party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to reuse a component of the work, it is your responsibility to determine whether permission is needed for that reuse and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; e-mail: pubrights@worldbank.org. ID4D GLOBAL DATASET Volume 1 | 2021 Global ID Coverage Estimates Julia Clark Anna Metz Claire Casher CONTENTS ABSTRACT V ABOUT ID4D VI ACKNOWLEDGEMENTS VII EXECUTIVE SUMMARY VIII 1. INTRODUCTION 1 2. BACKGROUND AND APPROACH 4 3. METHODOLOGY 6 Data 7 Global Estimation Strategy 12 ID4D-Findex Analysis 16 4. RESULTS 18 Global ID Coverage Estimates 18 Insights from the ID4D-Findex Survey 23 5. DISCUSSION 32 Key Findings and Limitations 32 Policy Recommendations 35 REFERENCES 37 APPENDIX 1. DATA AVAILABILITY 40 APPENDIX 2. ADULT ID COVERAGE BY COUNTRY 42 ID4D-Findex Survey Data 42 Selected Administrative Data 48 APPENDIX 3. ALTERNATIVE SPECIFICATIONS AND ROBUSTNESS CHECKS 49 Birth Certification vs. Registration Rates 49 Coverage for Older Children 52 Exploring Changes from 2018 to 2021 55 Findex Sample Restrictions 57 APPENDIX 4. EXCLUDED COUNTRIES 60 2021 Exclusions 60 2018 Criteria 61 APPENDIX 5. SUMMARY OF CHANGES FROM 2018 62 Methodology Changes 62 Errata 63 APPENDIX 6. ID4D FINDEX QUESTIONS 64 APPENDIX 7. ID AUTHORITY QUESTIONNAIRE 66 APPENDIX 8. FINDEX REGRESSION RESULTS 68 iii FIGURES Figure 1. Data Availability and Disaggregation by Age 8 Figure 2. Population Without an ID by Age, Income Group and Region 19 Figure 3. Estimates by Model, Region, and Income Group 21 Figure 4. Adult ID Ownership Rates by Economy, Income Level, and Demographic Group 23 Figure 5. ID Ownership Rates by Economy, Income Level, and Demographic Group 24 Figure 6 Individual-level Predictors of ID Ownership in Lower-Coverage Countries 25 Figure 7. Countries with Largest ID Gaps for Women 26 Figure 8. Changes in the Gender Gap Over Time 27 Figure 9. Barriers to ID Ownership: Cumbersome, Bureaucratic Journeys 28 Figure 10. Impact of Not Having an ID: Difficulty Accessing Services, Rights, and Opportunities 30 Figure 11. Data Availability by Country, Metric, and Year 40 Figure 12. Birth Registration vs. Certification Rates, by Income Group 50 Figure 13. Countries Gaps between BCR and BRR Greater Than 25 Percentage Points 50 Figure 14. Example Country-Level Trends in BRRs for Children Under 5 54 Figure 15. Global ID Ownership with Restricted vs. Full Findex Sample 58 Figure 16. Impact of Findex Age Restriction on Estimates of Adults without ID 58 TABLES Table 1. Global Data Indicators and Sources for ID Coverage 7 Table 2. Data Selection Models for Global Estimates 13 Table 3. Adult Data Sources Based on Model Selection 14 Table 4. 2021 Global ID Coverage Estimates, by Region and Income Group 19 Table 5. 2021 Global ID Coverage Using Alternate Data Selection Models 20 Table 6. Interrelated Barriers for “Low Demand” Users 29 Table 7. ID4D-Findex Adult ID Ownership, by Gender (2021, 2017) 43 Table 8. Administrative Data Used in Primary Global ID Coverage Estimates 48 Table 9. Using Birth Certification in the Global Estimates 52 Table 10. Comparing Yearly BRRs for Children Under 5 from MICS and DHS Data 53 Table 11. Estimated Effect of Changes in Data Values 56 Table 12. Estimated Effect of Methodology Changes 57 Table 13. Impact of Findex Age Restriction on Global Estimates 59 Table 14. Countries Excluded from 2021 Estimates 60 Table 15. Countries that Would Meet 2018 Exclusion Criteria in 2021 62 Table 16. Methodology Changes for Global Estimates, 2018-2021 63 Table 17. Predictors of ID Ownership 69 iv | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES ABSTRACT This paper aims to estimate the number of people globally who do—and do not—have government-recognized proof of identity (“ID”). This work is an update to the most recent estimate produced in 2018, which found that just under 1 billion people did not have an ID. The World Bank’s Identification for Development (ID4D) Initiative collected new data for this analysis: two rounds of survey-based data on ID ownership for adults (in partnership with the Global Findex Survey), as well as new administrative data acquired through outreach to ID authorities. By incorporating this data with the data sources used in 2018 and updating the methodology accordingly, we find that as of 2021 there are just under 850 million people globally without official proof of their identity. Around half are children, and the vast majority live in lower-income countries in Africa and South Asia. Analysis using individual-level survey data demonstrates that these 850 million are at a high risk of exclusion from basic services and economic opportunities and are among the most marginalized in their communities. Bridging this ID ownership gap is thus critical for ending extreme poverty, promoting shared prosperity, and realizing the global commitment to “Leave No One Behind.” This paper is the first in a series that will accompany the release of the 2021 ID4D Global Dataset; while it estimates global access to ID, subsequent papers and data will focus on the quality and characteristics of ID and civil registration (CR) systems worldwide. For more, see http://id4d.worldbank.org/global-dataset. v ABOUT ID4D The World Bank Group’s Identification for Development (ID4D) Initiative harnesses global and cross-sectoral knowledge, World Bank financing instruments, and partnerships to help countries realize the transformational potential of identification (ID) systems, including civil registration (CR). The aim is to enable all people to exercise their rights and access better services and economic opportunities in line with the Sustainable Development Goals. This is especially important as countries transition to digital economies, digital governments, and digital societies, where inclusive and trusted means of verifying identity are essential to ensure accessibility and data protection. ID4D operates across the World Bank Group with global practices and units working on digital development, social protection, health, financial inclusion, governance, gender, and data protection, among others. To ensure alignment with international good practices for maximizing development benefits and minimizing risks, ID4D is guided by the 10 Principles on Identification for Sustainable Development, which have been jointly developed and endorsed by the World Bank Group and over 30 global and regional organizations (see http://idprinciples.org). ID4D makes this happen through its three pillars of work: 1. Thought leadership, research, and analytics to generate evidence and fill knowledge gaps 2. Global public goods and convening to develop and amplify good practices, foster collaboration across regional and global stakeholders, and support knowledge exchange 3. Country and regional action through financial and technical assistance to realize inclusive and trusted ID and civil registration systems The work of ID4D is made possible through support from the Bill & Melinda Gates Foundation, the UK Government, The French Government, The Norwegian Agency for Development Cooperation (Norad), and the Omidyar Network. To find out more about ID4D and access our other publications, visit www.id4d.worldbank. org. vi | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES ACKNOWLEDGEMENTS The 2021 Identification for Development (ID4D) Global Dataset update was a significant undertaking that would not have been possible without the support and dedicated efforts of a large team. This work was led by the authors of this report, Julia Clark, Anna Metz, and Claire Casher, under the guidance of ID4D Practice Manager Vyjayanti Desai. The authors wish to acknowledge, with deep gratitude, the many people who contributed to the Dataset project as well as this report on the Global ID Coverage Estimates, which is the first in a series to accompany the Dataset release. This includes Cem Dener, Lucia Hanmer, Dorothe Singer, and Emi Suzuki (World Bank), as well as Claudia Cappa, Bhaskar Mishra, Nicole Petrowski (UNICEF), and Romesh Silva (UNFPA), for their insightful peer reviews and feedback on the methodology and presentation of the global estimates. In addition, we thank Saniya Ansar, Leora Klapper, and Mansi Vipin Panchamia (World Bank), for their partnership on the ID4D-Findex survey and helpful reviews and input on data analysis. Many ID4D team members, including Estefania Calderon, Nay Constantine, Faher Elfayez, Jonathan Marskell, Reina Ntonifor, Vasiliki Papagianni, Ana Quiroz, Aliaksandra Tyhrytskaya, and Emmanuel Vassor, also made significant contributions to the broader administrative data collection, validation, and publication process. Finally, this report and the 2021 edition of the ID4D Global Dataset would not be possible without continued support from ID4D’s Multi-Donor Trust Fund partners, including the Bill & Melinda Gates Foundation, the UK Government, The French Government, The Norwegian Agency for Development Cooperation (Norad), and the Omidyar Network. vii EXECUTIVE SUMMARY Being able to prove who you are matters for equitable, sustainable development.1 Yet many people—particularly those living in lower-income countries2 and who are part of marginalized and vulnerable groups—are unable to obtain official forms of identification (ID) that provide proof of their legal identity.3 To fully tackle this challenge, we must understand its scope: the number of people globally who do—and do not—have government-recognized proof of identity (henceforth “ID”). To this end, the World Bank’s Identification for Development (ID4D) Initiative published the first estimate of global ID coverage in 2016. These figures were updated in 2017 and 2018 to incorporate new sources of data, including administrative data on ID registration and credentials collected by ID4D through country-level surveys. In 2018, ID4D estimated that just under 1 billion people did not have an ID, representing nearly 1 in 8 people globally (World Bank 2018). This paper provides updated estimates of global ID coverage for 2021 that take advantage of new data sources and a revised methodology. This includes the incorporation of two rounds of survey-based data on ID ownership for adults collected by ID4D in partnership with the Global Findex survey and new administrative data gathered by ID4D via direct outreach to ID authorities. The availability of new data has also led to necessary updates to the calculations and created opportunities for improving the global coverage estimate methodology. As of 2021, we estimate that there are just under 850 million people globally without official proof of their identity. 1 As recognized, inter alia, through Sustainable Development Goal (SDG) Target 16.9: “to provide legal identity for all, including birth registration” by 2030. See also: https://id4d.worldbank.org/guide/good-id-supports- multiple-development-goals. 2 The term country, used interchangeably with economy in this paper, does not imply political independence but instead refers to any territory for which authorities report separate social or economic statistics. 3 “Official” ID is provided by, on behalf of, or recognized by governments, and can include both “legal” ID (which provides proof of legal identity) and “functional” ID required for a specific sector or purpose (e.g., voting, travel, social security, etc.). For more information, see World Bank (2022a). viii | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES METHODOLOGY Given the absence of a single metric to measure ID coverage available across all countries, we follow a similar approach to 2018 and other previous estimates. This approach uses a combination of metrics that take advantage of available data and align with the changing nature of ID access throughout a person’s lifetime. Coverage rates are calculated across 194 countries with available data, and then summed to arrive at the global total. To calculate coverage for each country, we first divide the population into “children” and “adults” according to a cutoff age determined by the data sources (typically between 15-18). We estimate coverage for children by applying the under-5 birth registration rate—the official indicator for measuring progress toward Sustainable Development Goal (SDG) 16.9—to the population below the cutoff age. To estimate coverage for adults, we select between available metrics that either directly measure or provide a reasonable proxy of the country’s foundational ID for adults, which is typically a national ID or similar credential. For 129 countries, we use newly available data from the ID4D-Findex survey, which directly measures adult ID ownership in 2017 and 2021. For the remaining countries, we use a combination of administrative data on registrations collected directly from ID agencies, voter registration rates, and birth registration rates, depending on data availability and country income levels. In revising the methodology for the 2021 global ID coverage estimates, our primary aim was to integrate new ID4D-Findex survey data and improved administrative data, and to maximize the proportion of the world’s population covered in the estimates. We have also adopted more transparent and uniform rules for selecting among secondary data sources to provide reasonable global-level estimates while minimizing the need for subjective judgement calls for each country. The expanded set of metrics and data sources allow for the application of multiple models for estimating coverage to better triangulate the ID coverage gap and account for uncertainty. Although we have identified a primary estimation model—which we believe represents a reasonably conservative estimate of the global ID coverage gap—we capture remaining uncertainty by providing a range of estimates using alternate metrics and data selection rules. We also publish multiple indicators and replication data that may be used to calculate adult and child coverage estimates. ix KEY FINDINGS • We estimate that just under 850 million people do not have official proof of ID— around 1 in 9 globally. Over half of those without proof of their identity are children whose births have not been registered. Over 90 percent (some 760 million) live in low- income countries (LICs) and lower-middle-income countries (LMICs), over half (around 470 million) live in Sub-Saharan Africa, and 1 in 4 (over 200 million) live in South Asia. Altogether, around one-third of adults in LICs do not have an ID. • A number of countries have made progress in closing gaps in birth registration and identification. However, the change from the 2018 estimate—just under 1 billion people without ID—represents a mix of improvements in ID coverage (potentially between 100-200 million), methodology changes and the addition of new data sources (notably, the ID4D-Findex survey data). • Within countries, the gap in ID ownership primarily affects marginalized and vulnerable groups. Although some improvements have been made in reducing women’s gap in ID ownership, some 35 percent of women living in LICs still do not have an ID, compared with 27 percent of men, a gap of 8 percentage points. Similarly, we see gaps based on age, income, education, employment, and rural versus urban location. For example—in countries with lower levels of ID coverage—adults with only a primary education are about 9 percentage points less likely to own an ID than those with secondary or higher education, controlling for other demographic characteristics. • Onerous procedures, inefficiencies, and documentary requirements remain a significant barrier to obtaining an ID for people in many countries. Nearly 40 percent of adults without an ID globally report that they do not have an ID because they lacked the necessary documents. This number is even higher in LICs, where people are more likely to lack birth certificates and other prior documentation, and millions are at risk of statelessness. Worldwide, applying for an ID remains too expensive for approximately 36 percent of adults without one, either due to direct and/or indirect costs. A large portion of these indirect costs may also be due to long travel times to apply for, obtain, or correct an ID (reported as a barrier by approximately 40 percent of adults without ID), which increase transportation and opportunity costs due to lost work. • Not owning an ID prevents hundreds of millions of people from accessing services and fulfilling rights. Globally, around 1 in 3 adults without an ID reported difficulty using financial services, receiving financial support from the government, or applying for a job. Nearly 40 percent of adults without an ID reported difficulties obtaining a SIM card or mobile phone service, while around 25 percent had problems receiving medical care. Beyond access to basic services and economic opportunities, around a third of adult without an ID reported this as a barrier to being able to participate in elections. • Beyond bureaucracy, improvement in service quality and public trust are essential. In addition to the difficulty of the process itself, we find that just under 1 in 5 adults without an ID (approximately 18 percent) report that they “do not feel comfortable x | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES giving their personal information” as a reason. There may be multiple reasons for this discomfort, including general levels of mistrust in the institutions or systems or concerns about how that data is being collected or used, and/or discomfort with the processes involved in providing their data. However, it highlights the importance of ensuring high service standards and building trust in ID and CR systems. LIMITATIONS The 2021 global estimates reflect our best understanding of current global and regional ID coverage gaps, based on the data that is (or was) currently available. However, they have a number of limitations, including potential errors or biases in the measurement of survey and administrative data, challenges in combining multiple years and sources of data, and in the variable nature of ID ownership across countries and over a person’s lifetime. In addition, the change in estimates from 2018 to 2021 should not be treated as a time series. While this paper attempts to assess how much ID coverage has improved since the last estimates, the change from 1 billion to 850 million represents a combination of progress on closing the ID coverage gap, changing demographics, and improvements to data and methodology including the incorporation of the ID4D-Findex survey data. Furthermore, while access to ID is essential, it is also crucial that the systems created to provide identification are trusted, well-governed, and fit for purpose to support development goals and protect people’s data and rights.4 Although we estimate that around 850 million people do not have the basic identity credentials they need, that is not to say that the other 7-plus billion people around the globe have good ID, or that ID practitioners should be motivated by ID and civil registration (CR) system coverage alone. To allow for further exploration of ID system features beyond coverage, forthcoming papers will provide additional analysis along with the publication of a new set of qualitative data. POLICY RECOMMENDATIONS Despite notable progress in many countries, ensuring universal access to identification (ID) and civil registration (CR) is an essential right and priority for equitable, sustainable development. This is particularly urgent as lack of ID disproportionately affects people in lower-income countries and the most vulnerable groups in society, and without ID or CR people may not be able participate fully in social, economic, and political life. Key policy recommendations include: • Governments and other stakeholders must deliberately work to reduce or eliminate barriers that continue to prevent people from obtaining official or legal 4 The Principles on Identification for Sustainable Development at http://idprinciples.org, which have been endorsed by over 30 organizations including the World Bank Group, set out essential guidelines and characteristics for ID systems that aim to support the achievement of development outcomes. xi proof of their identity. This includes removing inequalities, onerous documentary requirements, and fees for basic documents and services, both by reforming relevant laws and regulations and improving business processes and customer services standards. It requires finding ways to make ID services more convenient and user- friendly, including simplifying procedures and locating service points closer to where people live or work. • Proactive, comprehensive engagement and communication with communities, local leaders, and civil society organizations is also essential. Where ID providers do not have a good understanding of people’s needs and the barriers they face in terms of registration, coverage is likely to be low. Robust information and education campaigns, ongoing feedback during implementation, and sensible grievance redress mechanisms are needed to build trust and help people take advantage of the opportunities having official proof of identity can provide. Transparent and frequent involvement with civil society and community-based organizations—particularly those representing the interests of marginalized and vulnerable groups—can help identify and unlock key bottlenecks to boosting accessibility and enhancing coverage. • Monitoring and improving access to ID requires better, regular data collection. This paper highlights how data availability and comparability impede efforts to assess the scale of the global ID coverage challenge, and the same applies within countries. This is particularly the case for understanding ID and civil registration coverage by age, particularly for older children. Countries and development partners must invest in improving data collection on identification systems through multiple channels. Including ID indicators in censuses and other national and subnational survey efforts would help produce improved estimates at the country level about which groups are at the highest risk of exclusion. There is also more work to be done by ID authorities in defining and monitoring key indicators that can allow them to effectively track trends in registration, credential issuance, and system performance with direct relevance to inclusion objectives. ID4D and the World Bank are committed to helping implement the above recommendations, both through our direct support for countries implementing or improving ID and civil registration systems, and our global and country-level data and research. This includes continued updates to the Global ID Coverage estimates, which we expect to release every three years to align with Findex data collection. We also welcome new ideas and partnerships for improving data collection and analysis to ensure that countries and the global community have the information they need to build inclusive and trusted ID systems. xii | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES GLOBAL ID COVERAGE IN 2021 1 INTRODUCTION Being able to prove who you are matters To meet this challenge, is it important to assess for equitable, sustainable development.5 the number of people globally who do—and do Yet many people—particularly those living not—have government-recognized proof of identity (henceforth “ID”). However, while under-5 birth in lower-income economies and who registration is the Sustainable Development Goal are part of marginalized and vulnerable (SDG) indicator for target 16.9 (legal identity for groups—are unable to obtain official forms all)7 and provides an essential metric for children, of identification (ID) that provide proof measurement of ID coverage for the adult population of their legal identity.6 Such IDs are often is complex and has historically lacked standardized required for participation in economic, indicators and data collection. To help fill this gap, social, and political life and the fulfillment the World Bank’s Identification for Development (ID4D) Initiative published the first estimate of global of rights; without them, people may not ID coverage in 2016, using under-5 birth registration be able to access social assistance, legal rates (BRR) for children and voter registration rates protection, education, or healthcare, open (VRR) as a proxy of ID coverage for adults. These a bank account or obtain a SIM card, secure figures were updated in 2017 and 2018 to incorporate employment in the formal sector, operate new sources of data, including administrative data a business, or hold land or other assets in on ID registration and credentials collected by ID4D their own name. through country-level surveys.8 As of 2018, ID4D estimated that just under 1 billion people did not 5 As recognized, inter alia, through Sustainable Development Goal (SDG) Target 16.9: “to provide legal identity for all, including birth registration” by 2030. See also: https://id4d.worldbank.org/guide/good-id-supports-multiple-development-goals. 6 “Official” ID is provided by, on behalf of, or recognized by governments, and can include both “legal” ID (which provides proof of legal identity) and “functional” ID required for specific sector or purpose (e.g., voting, travel, social security, etc.). For more, see the Principles on identification (World Bank 2022). 7 Indicator 16.9.1 is the proportion of children under age 5 whose birth is registered with a civil authority. 8 A new set of estimates was initially planned for 2019 but postponed to allow for the inclusion of 2021 Findex survey and a more robust data collection effort to also include qualitative indicators on ID systems globally (forthcoming). 1 have an ID, representing approximately 13 percent of The survey data also give insights into the reasons the world’s population (World Bank 2018). why people do not have an ID and the difficulties they face as a result. Among adults living in low-income This paper provides new estimates of global ID countries (LICs), the most reported reason for not coverage for 2021 that take advantage of new data having an ID are documentary requirements, distance sources and a revised methodology. This includes to registration points, and high costs. Such issues the incorporation of two rounds of survey-based are compounded when cumbersome registration data on ID ownership for adults collected by ID4D procedures require multiple visits by design or due as part of the Global Findex survey (Demirgüç- to problems with understaffing or absenteeism, Kunt et al. 2022, Demirgüç-Kunt et al. 2018) and long-wait times, unclear policies or procedures, or supplemented by new administrative data gathered technical failures.9 About 1 in 3 people in LICs report via direct outreach to ID authorities. As detailed in the difficulty accessing financial services, receiving sections below and in the appendices, we conduct financial support from the government, applying for multiple robustness checks to assess the validity a job, and participating in elections without an ID. of our analysis and the impact of methodological As more countries reach near-universal ID coverage, changes and new data sources. Replication data those who remain without access are more likely to and code for this paper will be published via ID4D’s be left behind in the absence of proactive inclusion website (http://id4d.worldbank.org), and readers measures. are encouraged to download and customize these materials for their own use. The 2021 global estimates reflect our best understanding of current global and regional ID As of 2021, we estimate that approximately 850 million coverage gaps, based on the data that is (or was) people in the world do not have an official ID, and currently available—however, they should not be over 90 percent of this total represents people used as a time series. The change from the 2018 living in lower-middle-income and low-income estimates of just under 1 billion to the 2021 estimates countries. In addition, around half of this 850 million are of approximately 850 million people without ID children, and half live in Sub-Saharan Africa. In line represents a combination of progress on closing with World Bank (2018) and Metz and Clark (2019), the ID coverage gap, changing demographics, and analysis using the ID4D-Findex survey data also improvements to data and methodology including show that the remaining ID coverage gap is largely the incorporation of the ID4D-Findex survey data. concentrated among potentially disadvantaged Still, the data show that coverage gaps in many groups, including women, younger people, less- countries are narrowing over time, and a handful of educated people, rural dwellers, and those living in countries have made large gains in ID ownership and poverty. Although the existing data has limitations, birth registration rates. these estimates provide a big-picture assessment of the scale of ID access globally and help focus our However, while access to ID is essential, it is efforts on the areas and people with the greatest also crucial that the systems created to provide inequalities. identification are trusted, well-governed, and fit for purpose to support development goals and protect people’s data and rights.10 Although we estimate 9 See, for example, Hanmer et al. 2021. 10 The Principles on Identification for Sustainable Development at http://idprinciples.org, have been endorsed by over 30 organizations, including the World Bank Group. They set out essential guidelines and characteristics for ID systems that aim to support the achievement of development outcomes. 2 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES that around 850 million people do not have the The remainder of this paper provides a brief basic identity credentials they need, that is not to background followed by an overview of data and say that the other 7-plus billion people around the methodology before presenting the global ID globe have good ID, or that ID practitioners should coverage estimates. In addition, it includes a more be motivated by ID and civil registration (CR) system in-depth analysis of the correlates of ID ownership coverage alone. To allow for further exploration of for individuals using the ID4D-Findex survey data. ID system features beyond coverage, forthcoming It concludes with a discussion on the interpretation papers will provide additional analysis along with the and limitations of the results, as well as core policy publication of a new set of qualitative data. implications and directions for future research. INTRODUCTION | 3 2 BACKGROUND AND APPROACH Interactions with government and some importantly, over a person’s lifetime. Children are private sector services often require an primarily identified through birth registration and “official” or “legal” form of identification birth certificates. Birth registration is well established as essential for child protection, fulfillment of (ID) that provides proof of a person’s rights, access to education and other services, and identity as recognized under the law. The the ability to obtain other IDs later in life (UNICEF objective of this paper is to estimate the 2019a). In addition, the proportion of children under number of people who do not have such age 5 whose birth is registered with a civil authority forms of ID, and as a result are potentially is the official indicator for SDG Target 16.9. excluded from full participation in economic, political, and social life. These For adults, the picture is more complex. Many adults— particularly in low- and middle-income countries— estimates are designed to provide the never had their births registered and rely instead global community with a reliable indication on other forms of official documentation (Gelb and of the scale of this problem and the people Diofasi Metz 2018). Even for those whose identity most likely to be affected. was established and documented at birth, other IDs are frequently needed later in life to interact with As noted above, measuring ID ownership is not a government and the formal economy (e.g., a tax straightforward task. People may use a variety of identification number for filing taxes, a social security documents and credentials to prove who they are card for collecting benefits, a voter ID for casting for official purposes or when transacting in the ballots, or a passport for traveling). People may have private sector, and the precise nature of which ID a constellation of IDs that serve one purpose but not is required for which purposes can vary across and another, each with varying degrees of formality and within countries, by service provider or agency, and legal recognition (Gelb and Clark 2013). 4 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Given these heterogeneous realities and measurement many of the other IDs that people use for specific challenges, there is no single metric that perfectly functions (such as passports, drivers’ licenses, or captures the totality of ID ownership within and voter IDs).12 across countries. Despite this complexity, however, nearly all countries in the world have one or Our overall approach is therefore to measure more “foundational” systems that provide legally coverage of (a) birth registration for children and recognized identity credentials, such as certificates, (b) ownership of the country’s foundational ID for ID numbers, and cards.11 While the name and adults (e.g., a national ID or similar credential). characteristics of these ID systems varies by country, This follows the approach of previous editions of they are often national ID systems, population the ID4D Global Dataset and acknowledges the registers, civil registration systems, or other similar changing nature of ID across a person’s lifetime. systems. Enrollment in many of these foundational Where a direct measure of a country’s foundational systems is mandatory, and they typically provide ID is not available—or in the few countries where legally recognized credentials that may be accepted one does not exist—we use either voter registration and/or required for accessing a broad range of or birth registration as a proxy measure for adult ID government and private sector services. In addition, ownership, as described in more detail below. they are also often required or helpful to obtain 11 All countries in the world have established civil registration systems (UNICEF 2022a, UNSD 2022) and over 80 out of 198 countries included in the forthcoming ID4D Global Dataset have a national ID system, population register, or similar foundational system that provides a legally recognized ID to adults and in some cases also children. The exceptional countries which do not have a national ID system or population register include primarily North American and Commonwealth countries—including Australia, Canada, Ireland, Jamaica, New Zealand, the United Kingdom, and the United States—and some smaller island nations. In most of these countries, other “functional” or sector-specific ID systems have become the de facto identifier or ID credential across purposes (e.g., the social security number in the United States). 12 For a more detailed discussion of various definitions of ID, including official, legal, foundational, and functional, see World Bank (2019 and 2022). BACKGROUND AND APPROACH | 5 3 METHODOLOGY This section describes the data sources In addition, the expanded set of metrics and data and methodology used for the global sources allow for the application of multiple models estimates and individual-level analyses of for estimating coverage to better triangulate the ID coverage gap and account for uncertainty. Although ID ownership. In revising the methodology we have identified a primary estimation model, for the 2021 global ID coverage estimates, which we believe represents a reasonable and our primary aim was to integrate new conservative estimate of the global ID coverage gap, ID4D-Findex survey data and improved we capture remaining uncertainty by providing a administrative data, and to maximize range of estimates using alternate metrics and data the inclusion of a majority of the world’s selection rules. We also publish multiple indicators population in the estimates. We have also and replication data that may be used to calculate adult and child coverage estimates. 13 We hope adopted more transparent and uniform this will improve the usability of the data by other rules for selecting among available data researchers, while highlighting the fact that no single sources that provide reasonable global- indicator can paint a complete picture of the gaps in level estimates while minimizing the need ID coverage in any country or globally. for subjective judgement calls for each country. The availability of different indicators and data sources by country is shown in Appendix 1. Updates to the methodology and the different models employed are further described below and 13 In previous years, ID4D published a single spreadsheet that included results and calculations for all countries, along with additional qualitative data (such as the name of the ID system, level of digitalization, etc.). This year, we have made multiple improvements to increase the usability of data for analysis and improve replicability and ease of interpretation. This includes aggregating an ID series in the World Bank DataBank (available at http://databank.worldbank.org) to consolidate country-level metrics related to ID, including birth registration, birth certification, ID4D-Findex data, and publishing replication code and files for the global estimates analysis, so other researchers can more easily extend this work. Additional qualitative data will be released subsequently as part of a series of papers. 6 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES validated through a series of robustness checks coverage” as a generalization to cover various detailed in Appendix 3. Unless specifically noted, the measurements of individual ID ownership, birth methodology described is the same as that used in registration and certification, administrative data on the previous estimates (World Bank 2018). ID registration, and voter registration represented by the data. However, with the addition of the ID4D- Findex survey data, the majority is now survey- DATA based and directly measuring people’s reported ID ownership or birth registration. While there are Figure 1 and Table 1 provide an overview of the limitations with survey data (as discussed below), primary data sources used to estimate global this represents a leap forward in data availability ID coverage and their availability across 198 and a significant improvement over the 2018 edition countries.14 As in 2018, this includes a mix of survey of the estimates, which used voter registration as a and administrative data, and we use the term “ID proxy of ID ownership for adults for most countries. Table 1. Global Data Indicators and Sources for ID Coverage Countries Metric Description Source(s) Data Years available1 Under-5 birth The share of children under Survey and administrative data compiled by Varies: 1951 registration 5 whose births have been UNICEF (2022a); Multiple Indicator Cluster 2000–2021 rate (“BRR”). registered with a civil Surveys (MICS) for Kosovo and Malawi, and authority (as reported NFHS-5 for India. For the few countries where by their parent or by the above is unavailable, we use UNSD (2022) authorities) data on “birth registration completeness” as a proxy (see discussion below) Birth The share of children under 5 Survey data compiled by UNICEF (2022a, Varies: 144 certificate for whom a birth certificate 2022b); MICS (Kosovo and Malawi) and 2006–2021 prevalence rate has been issued. NFHS-5 (India) (“BCR”) Adult ID The share of adults 15+ who ID4D-Findex survey, collected in partnership 2017, 2021 130 ownership report personally owning the with the Global Findex2 rate (“ID4D- foundational ID (national ID Findex”) or similar credential).3 ID system The share of people ID4D questionnaires completed by relevant 2019, 79 registration registered in the authorities based on administrative data. 2021/2022 rate (“Admin”) foundational ID system. Voter The share of adults above IDEA indicator on number of registered voters Varies: 187 registration the voting age registered to as reported by administrative sources (IDEA 2006–2022 rate (“VRR”) vote. 2022). ACE Electoral Knowledge Network as the source of voting age eligibility data (ACE 2022). 1 Out of 198 countries included in the ID4D Global Dataset. See References for full citations and URLs of non-ID4D sources. UNICEF data on BRR are available for 181 countries; however, some are not included in these 198 countries. 2 For the full Findex survey, see Demirgüç-Kunt et al. (2018, 2022). 3 The ID ownership question was asked of all respondents ages 15 and older; however, we exclude observations where the respondent had not yet reached the age of eligibility for obtaining the ID. For more details, see Appendix 3. 14 As in previous years, we attempt to include 198 countries in the ID4D Global Dataset, excluding smaller nations and territories for which data are generally unavailable or which lack independent ID systems. In our primary models, four economies are excluded from the calculations due to lack of data. See Appendix 4 on exclusion criteria. METHODOLOGY | 7 As illustrated in Figure 1, while these data cover the administrative data for those countries that reported full range of ages, our ability to disaggregate by age disaggregated figures. The indicators and data sources is limited (except for birth registration data from are described in more detail in the following sections, UNICEF which is disaggregated for each year 0–4).15 along with strategies for determining the “cutoff age” Furthermore, there is low availability of data measured between children and adults and selecting between at ages 5 to 15: this age group includes only ID4D multiple secondary metrics when available. Figure 1. Data Availability and Disaggregation by Age Birth Registration and Certification (DHS),16 administrative data reported in the UNSD Vital Population and Vital Statistics Report or To estimate the share of children without proof collected from civil registration and vital statistics of identity, we use the official indicator for SDG (CRVS) authorities, and other sources including 16.9 compiled by United Nations Children’s Fund census and other national surveys. (UNICEF) on the “Percentage of children under age 5 whose births are registered (by sex)”, also commonly UNICEF BRR data are available for 181 countries referred to as the “under-5 birth registration rate” covering 80 percent of the global population and 179 and abbreviated as “BRR” throughout this paper of the countries included in the ID4D Global Dataset.17 (UNICEF 2022a). This includes three types of data: In the case of India, Malawi, and Kosovo, we take survey data from Multiple Indicator Cluster Surveys estimates directly from recently published surveys (MICS) and Demographic and Health Surveys that were released after the SDG reporting period 15 Although ID4D-Findex data can be disaggregated by year according to the respondents’ age, the sample size is not sufficient to generate reliable estimates for smaller age ranges; for this reason, we disaggregate by only two age categories: 15–24 and 25 plus. Birth certification rates reported directly from the MICS/DHS surveys are also often disaggregated by year. 16 The MICS and DHS surveys measure the percent of children ages 0–59 months whose births were registered with civil authorities at the time of the survey as reported by their mothers or guardians. This differs from data derived from CRVS systems (including UNSD data) that typically report the “proportion of live births that were registered within a year or the legal time frame for registration applicable in the country” (UNICEF 2022a). By capturing registrations that measures provide a more accurate estimate of the percentage of children under five whose births were registered as of the data collection date. 17 Three countries with UNICEF BRR data are not included in the ID4D Global Dataset: Montserrat, Turks and Caicos, and Cook Islands. 8 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES was closed and thus did not appear in the UNICEF data. See Appendix 3. Alternative Specifications and dataset as of July 2022 (International Institute for Robustness Checks for additional discussion and Population Sciences and ICF 2021; National Statistics analysis of this approach. Office 2021; Kosovo Agency of Statistics and UNICEF 2020).18 For 15 countries not included in the UNICEF Newly in 2021, our main model also uses BRR to data, we use the UNSD measure of birth registration estimate coverage for the adult population for high- completeness as in previous years of the ID4D Global and upper-middle-income (HIC and UMIC) countries Dataset (UNSD 2022).19 that lack ID4D-Findex or administrative data, and other countries that lack any other type of data. In In addition, UNICEF has recently compiled data on most HICs and UMICs, BRR is estimated to be at or birth certification rates for children under the age of 5 close to 100 percent, and so applying this rate to for 144 countries (UNICEF 2022a), while additional the entire population adds little to the global total. data is available for Kosovo directly from the MICS. However, it allows us to eliminate the 2018 exclusion Birth certification rates (BCR) are significantly lower criteria that removed HICs and countries with no than birth registration rates in many countries, with foundational ID systems from the global ID coverage differences of more than 50 percentage points in estimation.20 As a result, the 2021 estimates include the most extreme cases, including in Rwanda, Solomon data from countries representing over 99 percent Islands, and Sierra Leone. In alignment with the official of the global population, compared with around 66 SDG indicator for target 16.9, and to avoid inflating percent in 2018. In the case of LICs and LMICs without the ID coverage gap, we therefore continue to use ID4D-Findex or administrative data, we continue to the under-5 BRR for our primary estimates. However, use VRR in our primary model, as described below. we run alternative models using the BCR to measure coverage for children for the countries where this data is available (see Appendix 3). If this metric were used ID4D-Findex Survey Data instead of BRR where available, it would increase the ID coverage gap estimate by approximately 150 million The World Bank’s Global Findex is a nationally vis-à-vis our primary estimation model. representative survey conducted by Gallup, Inc. that measures access to financial services and other Because there is no globally available data on birth core indicators covering 91 percent of the world’s registration coverage for older age groups, we also population. In 2021, the Findex survey was carried apply the under-5 BRR to children above age 5 and out in 123 countries, with approximately 1,000 up to the cutoff age (discussed below), as done in respondents in each. Samples are representative of 2018. While this may either over- or under-estimate the population over age 15 and designed to generate the number of children without ID, depending on the reliable gender-disaggregated insights (along with context, it remains the most viable and consistent other individual age-, income-, and schooling-related option in the absence of other systematic sources of analysis) at the country level.21 18 Kosovo is not included in the UNICEF dataset and therefore the numbers are taken directly from the most recent MICS report (Kosovo Agency of Statistics & UNICEF 2020). For India, the recently published National Family Health Survey (NFHS-5; 2019-21) reports a BRR rate of 89 percent, while UNICEF (2022a) reports data from the NFHS 2015-16 (80 percent). 19 UNSD (2022) data measure the completeness of birth registration. While related, these represent a slightly different measure than the under-5 birth registration rate. However, following previous editions of the dataset, we use these to be able to include some larger countries (e.g., China, South Korea, Malaysia, Libya, and Kuwait) that do not have UNICEF data and would otherwise be excluded from the estimates. In some cases, UNSD data are presented as a range (e.g., 90-99 percent); in such cases, we use the median of the range (e.g., 94.5 percent). This is a small change from the 2018 methodology, which used the minimum of the range (e.g., 90 percent). 20 See Appendix 5, for a description of the exclusion criteria used in 2018. 21 For more information on the Findex survey and a full discussion of the 2017 and 2021 methodologies, see Demirgüç-Kunt et al. 2017 and 2022. METHODOLOGY | 9 Through a collaboration between ID4D and Global where IDs are lost or stolen over time—administrative Findex, the 2017 and 2021 rounds of this survey records may frequently be out of date, and surveys included multiple questions related to identification may provide a more dynamic measure of actual that we use for both global coverage estimates and ownership. At the same time, there are also limitations individual analysis on correlates with ID ownership and to this data, including the potential for biased results barriers to access. To measure ID ownership, the survey or not fully representative samples. These limitations asked respondents whether they personally owned are discussed in more detail in Section 5. the country’s foundational ID.22 We have data on this question from 112 countries in 2021, and 98 in 2017, for 130 countries total. For more details on the ID4D-Findex Administrative Data survey data and survey instrument, see Appendix 2 and Appendix 6. For countries where 2021 Findex data was As part of the 2021 ID4D Global Dataset, we fielded not collected, we use 2017 estimates where available. 23 a questionnaire to ID agencies24 in 126 countries In addition to the global estimates, we use the 2021 to collect and validate information about these Findex to analyze individual-level trends, as detailed in systems. As in 2018, this included requests to Section 4 Results and Appendix 8. provide administrative data on (1) the number of people registered in the system by age group and On average, we consider the ID4D-Findex survey gender, and (2) the number of people who had been data to be the most reliable global measure available issued with the primary credential (e.g., a national for adult ID coverage for multiple reasons. First, it ID card). These questionnaires were fielded by the closely approximates the concept we are trying to ID4D team and World Bank country offices between measure: the number of adults with proof of their August 2021 and May 2022, including a follow-up identity, represented by ownership of the country’s period to clarify and validate responses.25 A similar de jure or de facto legal ID. Second, as with birth questionnaire was fielded in 2019, although the ID4D registration rates, survey-based measures are likely Global Dataset was ultimately not published that to better reflect people’s current ID ownership year to allow for a more comprehensive review and relative to administrative reports, and they overcome update.26 For 2021/2022, information on ID system some of the data quality issues associated with the registration was received from 51 countries, and on latter (noted below). For IDs in particular—when credential issuance from 43 countries.27 In 2019 we people may have registered but not received an ID, or received data on registrations and credentials from 22 For each country, the survey used the actual term for the foundational ID in the local language; in most cases, this is the “national ID card” or similar. For two countries without a foundational ID, respondents were asked about a functional ID that has become the de facto or primary general-purpose ID (in Canada, this was the social insurance number; in Jamaica, this was the voter ID, which people use predominantly and colloquially refer to as the “national ID”). See Appendix 2 for terms. 23 Due to the COVID crisis, data collection for the 2021 Findex was delayed or transitioned to phone interviews in multiple countries; in a few cases, phone-based data collection was determined to be infeasible or invalid. For this reason, Findex data were not available as of 2022 in a handful of countries. An additional survey round is planned in 2022 for the following countries, with expected release in 2023: Azerbaijan, Botswana, Chad, Comoros, Democratic Republic of Congo, Eswatini, Ethiopia, Gambia, Guatemala, Lesotho, Madagascar, Mauritania, Mexico, Niger, Turkmenistan, and Yemen. When available, data for these countries will be added to the ID4D indicator series in the World Bank’s DataBank and available at http://id4d.worldbank.org/global-dataset. However, new data from these countries is not expected to affect the global estimates in a significant way. Using the data we have—including 2017 Findex, voter registration, and birth registration—the total adult ID coverage gaps in these countries are estimated at around 61 million. Even if coverage in these countries changed by 10 percent on average (much greater than the average change we have seen in other countries), this would change the global estimates by only plus or minus 6 million. 24 Authorities responsible for the country’s foundational ID system(s). 25 As a result of the data collection schedule and variation in reporting between countries, the administrative data received in these questionnaires represents a range of dates from 2020 (month not specified by respondent) to April 2022. 26 A new set of global estimates were initially planned for 2019, continuing the trend of yearly updates. However, this release was postponed to include the 2021 Findex survey results and complete a more robust data collection effort to enhance qualitative indicators. 27 In total, we received 59 questionnaires in 2021/22 (a relatively high response rate of just under 50 percent), but not all had ID registration or credential numbers. 10 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES 76 countries in total; together, this is a significant To calculate an administrative coverage rate, increase in response rate compared to the 2018 reported registration figures are divided by the total survey, when data was only available for 36 countries. population above the cutoff age in the data year (either 2019, 2020, 2021, or 2022). Administrative Like the ID4D-Findex survey data, registration coverage rates used as part of primary estimation totals are directly related to our targeted concept models are included in Appendix 2; however, they of ID ownership; however, this administrative data are not directly comparable with the ID4D-Findex also has limitations that make it a less consistent or rates or across countries. As shown in Appendix 3, reliable source than survey data in some cases. While these rates are on average higher than ID ownership some countries’ ID systems can report the number rates calculated using the ID4D-Findex survey data. of unique and living people that have registered or currently own an ID based on detailed administrative Voter Registration data, many more are only able to count the cumulative total registrations or credentials issued The original (2016) edition of the ID4D Global Dataset over time. As confirmed by a number of questionnaire used voter registration rates as the primary indicator respondents, these total registration figures can for adult ID coverage. We continue to use this often include people who are now deceased or live measure as a proxy in countries where neither survey outside the country, and/or duplicates of the same nor administrative data is available. It reasonably person. This is often more likely in cases where ID approximates adult ID ownership in many countries, systems do not receive regular updates from CR as voter ID cards sometimes serve as accepted forms systems when a person has died. of official identification, particularly in countries with low-coverage foundational ID systems. In addition, Furthermore, not everyone who is registered may registering to vote often requires some other be in possession of a (valid) physical ID credential. form of identification. However, there are important They may no longer have their credential (if it limitations with voter registration rates as a proxy for has been lost, stolen, or expired), or they may not measuring ID coverage. have received it yet (there may be a lag between registration and credential issuance). Similarly, As with administrative data, voter registration administrative data on credentials issued is often not numbers may also exceed the resident population a reliable source on the number of valid credentials size in cases where deceased or duplicate voters associated with a unique person, but instead reflect are not removed or where they include people living the totality of credentials issued over a given period, abroad. In addition, whether or not a physical voter without accounting for lost, stolen, expired, or re- identity credential is issued—and if so, whether or issued IDs. For these reasons, we use the ID4D- not it constitutes an officially recognized identity Findex survey data as our primary source for adults credential—varies by country, as do the requirements and use administrative data on total registrations for voter registration. In cases where voting in when this is not available. 28 elections is mandatory, people may be registered to vote automatically with little involvement in the process or ID proofing (Rosenberg and Chen 2009; 28 In a departure from 2018, we use only registration totals for estimation rather than a mix of registration or credential numbers depending on the country. While the number of credentials issued would be more comparable to the ID4D-Findex data (asking whether a person owns an ID), in practice we have found that this data has been less consistent over the years, particularly when countries are rolling out new credentials (e.g., an updated smartcard) that have not yet reached high coverage, but the population still owns and uses older forms of the ID. In addition, credentials are frequently re-issued when they are lost and expired, which means that these can exceed the population as with registration numbers that include migrated or deceased persons. We have, therefore, uniformly adopted registration for the 2021 estimate as the more reliable number on average and to improve consistency across countries, and because the 2021 questionnaire did not include disaggregation of credential numbers by gender or age. METHODOLOGY | 11 Schumacher and Connaughton 2020). Conversely, low Increases in data availability allow us to take a more voter registration rates often reflect broader realities nuanced approach for the 2021 estimates than in of the political system or country demographics beyond 2018 or previous years. Rather than providing a single the issue of identification. For example, in countries estimate, we run multiple alternative models to better with relatively high shares of residents who are capture and account for measurement uncertainty ineligible to vote (for example, migrant workers), using and diverging metrics. This section describes the voter registration rates as a proxy is likely to lead to data source selection method for our primary global an overestimation of the number of people without ID.29 estimates model and alternate models, setting a cutoff age for adult versus child populations, and To calculate voter registration rates, we obtain the then calculating aggregate estimates. number of registered voters in the most recent parliamentary or presidential election from the International Institute for Democracy and Electoral Data Selection Assistance (IDEA) database on voter turnout. This is available for 187 countries, with election years Given that the ID4D-Findex survey data provide the spanning 2006 to 2022 (IDEA 2022).30 To calculate most direct measure of adult ID ownership, these voter registration rates, we divide the number of data anchor our estimates, supplemented by other registered voters by the number of eligible voters in data sources when they are unavailable. For countries that year using a list of legal voting ages obtained where multiple metrics of adult ID coverage are from the ACE Electoral Knowledge Network (ACE available, our primary selection model (1a) chooses 2022). This is an update of the 2018 methodology, metrics in the following order of preference based on which simply subtracted the number of registered availability: (i) ID4D-Findex survey data (also called voters from the most recent election (e.g., 2016) from “Findex” below), (ii) administrative data (“Admin”), the eligible population in the dataset year (2018).31 (iii-a) voter registration (“VRR”) for LICs and LMICs, or (iii-b) birth registration rates (BRR) for UMICs and HICs. If voter registration is not available for a LIC or GLOBAL ESTIMATION LMIC, BRR is used instead. For each metric, we use STRATEGY the most recent year of data available. The primary goal of this paper is to provide an We believe this model (1a) is the most reliable given updated estimate of global ID coverage and gaps at the improved accuracy of the survey-based data the global level. As in years past, we calculate these in directly measuring ID ownership among adults. estimates by applying the various metrics described Furthermore, although the potential errors related to above to child and adult populations at the country administrative and voter registration data are largely level and then sum these for world, regional, and the same—e.g., that records may not be unique or not income-level totals.32 updated to reflect deceased persons or migration— 29 In the 2018 dataset, voter registration rates were adjusted for Gulf Cooperation Council (GCC) countries to account for the large shares of the population that are comprised of non-national foreign residents who are ineligible to vote, using estimates on “International migrant stock” from UNDESA. In 2021, we no longer make this adjustment due to the use of BRR to estimate adult coverage rates for HICs (including GCC countries) and UMICs for which neither ID4D-Findex survey nor administrative data is available. 30 Somalia is included in the IDEA voter registration database but excluded from our analysis given the extreme age of the data (it’s most recent elections are from 1984). Of the 187 economies with election data, all but 8 are from 2017 or later. Of the 16 countries for which we use voter registration rates in our primary estimation model, years range from 2015–2022. 31 Applying a consistent rate based on the year for which data is available is better aligned with the methodology used for other indicators such as birth registration rates and ID4D-Findex survey data. This method avoids overestimating ID coverage in countries where more time has passed since the last election year. See Appendix 4 for a discussion of this change. 32 We provide results by gender only for the countries with ID4D-Findex data, as gender-disaggregated birth registration and ID administration data are only available for some countries and are not available in the IDEA voter registration database. 12 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES we select administrative data from ID agencies as a BRR is used as the third choice for all countries that second-choice measure under the assumption that it is do not have Findex or administrative data. While a more direct measure of ID ownership in most cases. our primary model is intended to reduce potential overestimates of the ID coverage gap, these alternate Whether or not voter registration or birth registration models represent a less conservative approach— rates are a better proxy in the absence of our first and in the case of model 1b, an approach similar to and second-choice data depends, of course, on the 2018 with the addition of Findex data. In addition, country. However, for HIC and UMICs where Findex we include two models that select Findex first, and and administrative data are available, we can see where this is unavailable, then either the metric with that ID ownership rates tend to be close to universal the highest value (model 2a) or the lowest value (greater than 98 percent), and therefore typically (model 2b). Together, these models help account closer to their BRR rates (typically 100 percent) than for uncertainty by providing reasonable lower- and their voter registration rates. This is particularly the upper-bounds to our primary estimates. case in richer countries where voter registration is not mandatory or automatic (such as in the US) or Table 3 summarizes the result of the selection where there are large shares of the population who models in terms of the distribution of various data may not be eligible to vote (such as in countries sources used. See also Appendix 2 for the full table with high levels of migrant labor). Conversely for of ID ownership rates by country and administrative LMICs and particularly LICs, BRR rates are often data used in the primary model. significantly lower than ID ownership rates, whereas voter registration may offer a more realistic view of those with access to government-recognized ID. Cutoff Age and Population To test these assumptions, we run a series of alternate We use age- and sex-disaggregated population models for selecting the adult metric, shown in Table 2. estimates from the UN’s World Population Prospects In model 1b, only voter registration is used as the (WPP) to apply the various ID metrics to the global third choice for all countries that do not have Findex population (UNDESA 2022a).33 Following the 2018 or administrative data; conversely under model 1c, approach, each country’s population is divided 33 As the latest edition of WPP was released in 2022, figures for 1950–2021 are estimates; for the few cases of administrative and election data where 2022 figures are required to calculate administrative data or voter registration rates, we use the “medium scenario” projections, which are the projections deemed to be most likely by the UN and assume “a decline of fertility for countries where large families are still prevalent, a slight increase of fertility in several countries where women have fewer than two births on average over a lifetime, and continued reductions in mortality at all ages” (UNDESA 2022b, p. 28). Table 2. Data Selection Models for Global Estimates Model Adult Metric Selection Rule Child Metric 1a. Primary – income-based Findex > Admin > Voter (if LMICS, LICs) OR BRR (if UMICs, HICs) BRR 1b. Voter as third choice Findex > Admin > Voter > BRR BRR 1c. BRR as third choice Findex > Admin > BRR > Voter BRR 2a. Lower bound Findex > maximum {Admin, Voter, BRR} BRR 2b. Upper bound Findex > minimum {Admin, Voter, BRR} BRR Note: Findex refers to ID4D-Findex survey data, Admin to administrative rates from ID agencies, VRR to voter registration rates, and BRR to birth registration rates. Selection is based on World Bank country classification, where LIC is low income, LMIC is lower-middle income, UMIC is upper-middle income, and HIC is high income. METHODOLOGY | 13 Table 3. Adult Data Sources Based on Model Selection Number of Countries by Adult Coverage Metric Model Voter (most BRR (most Total Findex 2021* Findex 2017 Admin 2021 Admin 2019 recent) recent) 1a (primary) 111 18 8 6 16 35 194 1b 111 18 8 6 46 5 194 1c 111 18 8 6 0 51 194 2a 111 18 4 3 22 36 194 2b 111 18 3 1 29 32 194 *2021 ID4D-Findex survey data is available for 112 economies; however, Taiwan, China, is dropped from the estimates due to the lack of birth registration data. See Appendix 4 for more information on exclusion criteria. into two groups using a “cutoff age,” with the child varies depending on the adult data source and does population equal to the number of people below the not always equal the actual age of majority within cutoff age, and the adult population equal to the the country;34 the terms “child” and “adult” as used number of people at or above the cutoff age (see in this paper are therefore approximations. Equation 1). Due to data limitations, the cutoff age Equation 1. Cutoff Ages Child population Adult population a < c a ≥ c a = age 0 c = cutoff where cutoff c varies by economy and is based on the adult metric used, the age when a person is eligible to get the ID (IDage ), and the age when a person is eligible to vote (Voterage): Findex: c = IDage or 15, whichever is greater Admin: c = IDage or 5, whichever is greater Voter: c = Voterage BRR: c = 15 For ID4D-Findex survey data, the cutoff is the age at data are used, the cutoff is also set to the ID age, which people are eligible to register for or obtain the unless this age is less than 5 years old—for example, ID (henceforth the “ID age”),35 unless this is less than if people are eligible to register from birth—in which the minimum age of Findex respondents (15 years case it is set to 5.36 Where voter registration is used, old) in which case 15 is used. Where administrative we used the age of eligibility for voting, which is 18 34 Ideally, we would be able to measure the number of people above and below the age of majority in a country and globally (typically 18 years old) that do and do not have ID. However, as noted in Figure 1, there is limited data available for the 5–15 age range and disaggregating by age is not possible with every metric. As with data selection, our goal is to adopt a general rule that is feasible for most countries and produces a reasonable estimate of the scale of ID ownership globally. 35 Data on the ID age come from the 2021 ID4D Global Dataset. Cutoffs by country are listed in Appendix 2. 36 For administrative data in countries where people are eligible to apply for the ID from birth, we set the ID age to 5, applying birth registration rates for the population 0–4, and adult ID rates to the population over 5. This is done in recognition of the fact that birth registration is the primary pathway for children to establish their legal identities and birth certificates are often the required documents for school enrollment (UNICEF 2019a, Blitz et al. 2014). However, it is important to note that in many countries where the ID age is 0, it is because there is a population registration system that combines civil registration (CR) and ID, typically issuing a unique identity from birth. In these cases, the BRR and ID registration rates for the 0–4 population should be identical or very similar. Future work can explore this in more depth using countries where administrative data is available and disaggregated by age. 14 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES in nearly every country. Finally, for countries where For ID4D-Findex, we calculate country-wide average birth registration is used for adults, the cutoff age is ID ownership rates in the most recent year (2021 set to 15 to match the majority of other countries;37 or 2017) using the economy-level survey weights effectively, however, this means under-5 BRR is available in the Global Findex Dataset. For ID applied to the entire population. Appendix 3. provides administrative data, we divide the reported number additional analyses using alternate cutoff ages. of people registered by the population above the ID age in the data year (typically the year it was submitted to ID4D, ranging from 2019–2021); for Calculations voter registration, we similarly divide the total number of registered voters for the most recent We calculate adult coverage rates for each country i, election by the voting age population in that year.38 in data year j using the metrics selected for each Where ID administrative data or the voter registration country and model, as shown in Equation 2. This data show that the number of people registered is requires different approaches depending on the larger than the population above the cutoff age, the metric, as our main data sources include both survey coverage rate is censored to 100 percent.39 data (ID4D-Findex) and administrative data (ID and voter registration totals); for birth registration, we used the rates provided by UNICEF or UNSD. Equation 2. Coverage Rates for Each Country by Metric Findexi = HasIDij where HasIDij is the survey-weighted mean ID ownership in economy i in data year j Admini = RegisteredIDij PopIDij where RegisteredIDij is the number of people registered in the ID system, and PopIDij is the number of people at or above the ID age in economy i in data year j Voteri = RegisteredVotersij PopVotersij where RegisteredVotersij is the number of people registered to vote, and PopVotersij is the number of people at or above the voting age in economy i in data year j BRRi = BRRi where BRRi is the most recent under-5 birth registration rate (for UNICEF) or reported birth registration completeness (if UNSD) statistics in economy i Using the appropriate cutoff age (see Equation 1), we Finally, we (c) sum these totals across countries to then (a) divide the 2021 population of each country i arrive at the global totals. If no data is available for and (b) apply the selected metric to the adult a country to estimate either the "child" or "adult" population, and the birth-registration rate to the ID coverage rate, the country is excluded from the child population for each, as shown in Equation 3. global coverage estimate calculation (4 out of 198 in 37 If we set the cutoff age to zero, this would result in the same total number of people without ID, but they would all be classified as “adults,” skewing the distribution of those without ID toward adults and away from children. 38 This is a change from the 2018 methodology, where ID registration and voter registration totals were subtracted from the 2018 population to arrive at the number of unregistered people. In this edition, by first converting these totals into rates using the population in the data year, we avoid underestimating coverage in cases where populations have grown significantly since data was collected. However, this also assumes that registration rates have remained constant since the data were collected. 39 ID system records and/or voter registration numbers may exceed the population for multiple reasons, including the existence of deceased people in the registry, duplicate records, large shares of the country’s registered population living abroad, and inaccuracies in the underlying population estimates. METHODOLOGY | 15 total). This is a substantial change from 2018, which we also aggregate the estimates at the regional level applied multiple exclusion criteria based on income and based on the World Bank’s country income and and ID coverage rates. 40 In addition to global totals, lending classifications (World Bank 2021).41 Equation 3. Calculating Children and Adults without ID For each economy i, define: PopChildi = ∑ & , < & PopAdulti = ∑& , ≥ & where Popi is the 2021 population, and c is the cutoff for economy i based on the metric selected Then calculate: NoIDChildi = 1 − & ×& where BRRi is the most recent under-5 birth registration rate (for UNICEF) or reported birth registration completeness (if UNSD) NoIDAdulti = 1 − & ×& where is either Findexi , Admini , Voteri , or BRRi (see above for definitions) Finally, sum across countries to arrive at the global totals: ) Estimated global = population f(& + & ) without ID &'( ID4D-FINDEX ANALYSIS • Has to travel too far to apply • Owns another form of identification issued In addition to the global coverage estimates, we use by the government the 2021 ID4D-Findex survey data to examine the • Does not need an ID for any purpose correlates of ID ownership, the barriers people face to • Does not feel comfortable giving personal obtaining an ID, and the difficulties they face accessing information services without one. As noted above, this analysis relies on a series of additional questions in the Findex 2. Whether or not the respondent reported survey, which are detailed in Appendix 6. This includes: difficulty with the following because they did not have the ID: 1. Whether or not the respondent cited the following as a reason they did not have the ID: • Receiving financial support from the government • It is too expensive • Using financial services • Does not have the necessary documents • Obtaining a SIM card/mobile phone service 40 In 2021, Eritrea; the Federated States of Micronesia; and Taiwan, China are dropped due to lack of BRR in either the UNICEF or UNSD datasets. Somalia is dropped due to lack of reliable adult data, and because the BRR is so low that applying it to the entire country’s population would inflate estimates of the number of people without ID. In 2018, 47 out of 198 countries were excluded from the estimates based on the following criteria: (a) HICs with BRRs greater than 99 percent; and (b) countries with no foundational ID system and a BRR greater than 95 percent. For more details see Appendix 4. 41 Note, we use the 2021/22 income classifications because they best align with a majority of our data collection. However, there are updated classifications in place beginning 1 July 2022; chances from 2021/22 are summarized here: https://blogs.worldbank.org/ opendata/new-world-bank-country-classifications-income-level-2022-2023. 16 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES • Participating in elections For each question, we subset the sample to include • Applying for a job respondents above the ID age and calculate survey- • Receiving medical care weighted means at the global, and/or regional and income levels, and in some cases also by gender. 3. Demographic and socioeconomic characteristics: To examine the relationship between ID ownership and various demographic characteristics we run • Age logit models predicting whether a respondent has • Marital status an ID based on demographic and socioeconomic • Workforce status characteristics, with country-level fixed effects and • Education design-based standard errors. We limit this analysis • Income only to countries where total ID ownership is below • Location in urban or rural area 90 percent to ensure a sufficient sample size of those with and without ID.42 The full results of these models are included in Appendix 8. 42 Although the Findex survey is representative at the national level and for gender and other demographic groups, the sample size is not large enough to meaningfully examine correlates of not owning an ID or the barriers and difficulties faced by those without ID in countries where only a few respondents report not having an ID. For this reason, analysis of barriers and difficulties without an ID focuses only on countries with lower levels of ID ownership, where the sample of respondents without an ID is sufficient for statistical inference. METHODOLOGY | 17 4 RESULTS Using our primary model, we estimate that GLOBAL ID COVERAGE there are just under 850 million—or 1 in ESTIMATES 9—people globally without official proof Table 4 and Figure 2 show the contribution of of their identity. Around half are children, different regions and income level categories to and the vast majority live in the developing the global number of people without ID for the world. Analysis using individual-level survey primary estimation model (model 1a, see discussion data demonstrates that these 850 million in Methodology). Overall, we estimate that some are at a high risk of exclusion from basic 850 million people in the World do not have ID. services and economic opportunities and Over 90 percent (some 760 million) live in lower- are among the most marginalized in their middle income and low-income countries. Over half (56 percent or over 470 million) live in Sub-Saharan communities. Bridging this ID ownership Africa, while around 1 in 4 (over 200 million) live in gap is thus critical for ending extreme South Asia. poverty, promoting shared prosperity, and realizing the global commitment to “Leave No One Behind.” 18 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Table 4. 2021 Global ID Coverage Estimates, by Region and Income Group Estimated People without ID in 2021 (Millions) Percent of N Total Children Adults Total World 194 426 417 843 100% Region East Asia Pacific (EAP) 31 45.0 30.4 75.4 8.9% Europe and Central Asia (ECA) 53 0.7 20.6 21.4 2.5% Latin America and Caribbean (LAC) 33 8.2 24.7 32.9 3.9% Middle East and North Africa (MNA) 21 10.9 27.9 38.8 4.6% N. America 2 - 0.7 0.7 0.1% South Asia (SAR) 8 129.6 72.5 202.1 24.0% Sub-Saharan Africa (SSA) 46 231.9 240.0 471.9 56.0% Income* High-Income Countries (HICs) 60 0.2 16.2 16.4 1.9% Low-Income Countries (LICs) 25 164.9 99.1 264.0 31.3% Lower-Middle Income Countries (LMICs) 54 236.6 264.1 500.7 59.4% Upper-Middle Income Countries (UMICs) 54 23.2 37.3 60.4 7.2% * In the 2021 World Bank lending groups, Venezuela does not have a classification and is therefore not included in the income results. Note: Calculations based on data from the 2021 and 2017 ID4D-Findex survey, administrative data collected by ID4D in 2019-2021, birth registration data (UNICEF 2022b, UNSD 2022), voter registration data (IDEA 2022), and World Population Prospects (UNDESA 2022a). The cutoff between “adults” and “children” varies by country, according to the data source used. Typically, it is 15 or the year of eligibility for obtaining the adult ID, whichever is higher. Figure 2. Population Without an ID by Age, Income Group, and Region RESULTS | 19 Estimating the Coverage Range The models that use either voter registration (model 1b) or birth registration (model 1c) for all Table 5 gives the results of alternative models for countries without ID4D-Findex or administrative selecting adult data to bound our estimates. Our data, estimate between 30 and 50 million more primary model gives an estimate of just under people without ID than our primary model due to 850 million people without ID, which is slightly on the trends discussed above. As further illustrated in the conservative side compared to other selection Figure 3, we can see the resulting variation between models. At the high end—if we chose the lowest of these models comes from LICs and LMICs (primarily the other available metrics for countries without across Sub-Saharan Africa where birth registration survey data as in model 2b—the estimate of those rates are lower on average than voter registration rates), without ID would be around 100 million higher, at and in HICs (where birth registration rates are typically nearly 950 million. Conversely, if we chose the universal but voter registration is lower). We take this highest of the other available metrics for countries as reasonable support for our preferred model, which without survey data, as in model 2a, the estimated minimizes the possibility of overestimating the gap in number of people without ID would be about ID coverage by largely accounting for the correlation 50 million lower, just shy of 800 million. between coverage metrics and income group. Table 5. 2021 Global ID Coverage Using Alternate Data Selection Models Data Selection Model N 1a (primary) 1b 1c 2a 2b World 194 843.2 893.5 875.0 790.2 947.0 Region East Asia Pacific (EAP) 31 75.4 80.6 80.5 75.2 86.1 Europe and Central Asia (ECA) 53 21.4 28.2 21.4 21.4 28.2 Latin America and Caribbean (LAC) 33 32.9 33.7 32.9 32.9 33.8 Middle East and North Africa (MNA) 21 38.8 43.5 38.8 38.3 47.3 N. America 2 0.7 33.5 0.7 0.7 33.5 South Asia (SAR) 8 202.1 202.1 202.0 202.0 202.2 Sub-Saharan Africa (SSA) 46 471.9 471.9 498.7 419.7 515.9 Income* High-Income Countries (HICs) 60 16.4 63.6 16.4 16.4 66.9 Low-Income Countries (LICs) 25 264.0 264.0 291.4 262.8 291.5 Lower-Middle Income Countries (LMICs) 54 500.7 500.7 505.1 448.9 522.9 Upper-Middle Income Countries (UMICs) 54 60.4 63.5 60.4 60.3 64.0 * In the 2021 World Bank lending groups, Venezuela does not have a classification and is therefore excluded from the income grouping. For each model, adult data selected in the following order, based on availability: 1a: ID4D-Findex > ID4D administrative data > voter registration if low- or lower-middle income; birth registration if upper-middle or high-income 1b: ID4D-Findex > ID4D administrative data > voter registration > birth registration 1c: ID4D-Findex > ID4D administrative data > birth registration > voter registration 2a: ID4D-Findex > maximum value of {ID4D Administrative data, voter registration, and birth registration} 2b: ID4D-Findex > minimum value of {ID4D Administrative data, voter registration, and birth registration}” Sources: 2021 and 2017 ID4D-Findex survey, administrative data collected by ID4D in 2019-2021, birth registration data (UNICEF 2022b, UNSD 2022), voter registration data (IDEA 2022), and World Population Prospects (UNDESA 2022a). 20 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Figure 3. Estimates by Model, Region, and Income Group RESULTS | 21 There are, of course, other potential models Using the counterfactual approach described in and assumptions, a few of which are further Appendix 3, we apply the 2021 methodology to data explored in the appendices. For example, that was available in 2018--including both the actual adding birth certification rate (BCR) into the data used in the 2018 estimation, as well as the 2017 selection set would increase the estimates of Findex data—to control for changes in methodology. children and adults without ID by approximately For the 124 countries where the same metrics 150 million (see Appendix 3). And while we are used in both years, we find a decrease in the believe our primary estimates are a reasonable estimated number of people without an ID of around approximation of the scale of the global gap in 157 million. Around 100 million of this representing ID ownership, the appropriateness and accuracy higher birth registration rates for children, while about of various metrics to understand coverage 50 million from increased ID coverage rates for within each country may require a more tailored adults. approach. We therefore encourage readers to explore the data and replication code available However, most of this 157 million difference is the via ID4D’s website (http://id4d.worldbank.org/ result of big leaps forward in either ID ownership or global-dataset). birth registration in a handful of larger countries.43 Of the 124 countries with the same data metric in both years, coverage improved in 73 but decreased in 38. In Examining Changes from 2018 to 13 countries, changes in coverage rates were so small 2021 they had no real effect on the estimates of people with or without ID in the country. This analysis does As noted above and in more detail in Appendix 5, the not provide a complete picture of changes in global change from the estimate of just under 1 billion people ID coverage since 2018, given the partial sample of without ID in 2018 to 850 million people without ID countries that have consistent metrics over time, in 2021 represents a mix of methodological changes, and the fact that many birth registration and voter increases in data availability and actual changes in ID registration figures used in 2018 were significantly coverage. Therefore, it is difficult to identify precisely older (e.g., in the case of birth registration in India). how much of this change represents improvements However, it provides an indication that while not every in ID coverage—that is to say, the number of people country has made progress, global ID coverage has who did not have an ID in 2018 but have one now. increased in the last 5 years, likely on the order of However, we can look at countries with multiple 100 to 200 million. measurements of the same indicators over time to provide some indication of how much coverage rates have improved. 43 For example, India’s birth registration increased from 71.9 percent as of the 2013-2014 (the figure used in the 2018 estimates), to 80 percent as of the NFHS 2015-16, and 89 percent as of the NFHS-5; 2019-21. Given lags in measurement such as this, it is difficult to assess the precise changes from 2018-2021. However, we estimate that changes over the past five years are on the order of 100- 200 million. 22 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES INSIGHTS FROM THE ID4D- have an ID, compared with 10.9 percent in LMICs, and FINDEX SURVEY 1–3 percent in UMICs and HICs (Figure 4). Around 1 in 4 adults in Sub-Saharan Africa (SSA) does not Using available ID4D-Findex (2021) data, we also have an ID; of the 31 countries with an estimated explore ID ownership rates for adults across regions ID coverage of less than 90 percent, 19 are in SSA. and income groups. On average, 31.5 percent— We also see persistent gap in ID ownership among approximately one-third—of adults in LICs do not women and other groups, which are discussed in more detail below.44 Figure 4. Adult ID Ownership Gaps by Region, Income, and Gender (ID4D-Findex 2021) Correlates of ID Ownership income, education, and geographic location (Metz and Clark 2019). The 2021 round allows us to update The 2017 ID4D-Findex data provided strong evidence this analysis and provide additional insights on that unequal access to ID is highly correlated with barriers and difficulties for those without ID. demographic characteristics including gender, 44 The set of countries for which Findex data is available has changed between 2017 and 2021. There are 32 countries, mostly HICs, for which data on ID coverage is newly available in 2021. There are 18 countries for which data on ID coverage was available in 2017, but not in 2021, including Azerbaijan, Belarus, Botswana, Chad, Ethiopia, Guatemala, Haiti, India, Lesotho, Madagascar, Mauritania, Mexico, Montenegro, Niger, Rwanda, Trinidad and Tobago, Turkmenistan, and Yemen. For this reason, the regional and income group estimates for 2021 are not directly comparable with 2017. RESULTS | 23 Figure 5. ID Ownership Rates by Economy, Income Level, and Demographic Group In 2021, we again find strong correlations between live in rural areas (see Figure 5). These same gaps ID ownership and various demographic factors, are also statistically significant in LMICs, although primarily concentrated in lower-income countries. except for age, they are much smaller. In higher- Adults in LICs are less likely to have an ID when they coverage UMICs and HICs, gaps in ID ownership are below 25 years old, have only primary schooling by these demographic groups are virtually non- or less, are out of the workforce, are female, are in the existent.45 bottom 40 percent of the income distribution, and 45 Differences in ID ownership in each of these six demographic categories are statistically significant at the 95 percent level or above in LICs and LMICs. For HICs, the only statistically significant gaps are for age and rural versus urban location; for UMICs, only age is significant. 24 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Figure 6. Individual-level Predictors of ID Ownership in Lower-Coverage Countries In countries with ID coverage below 90 percent, with a married person), live in rural areas (compared each of these demographic factors remains a with an urban resident) or belong to the poorest statistically significant predictor of ID ownership, 40 percent of household (compared with adults in even after controlling for other variables (see the richest 60 percent). In addition, women in such Figure 6).46 For example, a person under ages 15–24 is countries are about 3 percentage points less likely to around 20 percentage points less likely to have an ID own an ID than men. than a person 25 or older, even if they have the same level of education, gender employment status, marital status, income-level, and location. Similarly, all else Gender Gap in Access to ID equal, the average adult in a country with lower levels of ID coverage are about 9 percentage points less likely Although gender is not the largest predictor of to own an ID than those with secondary education average ID ownership globally, it continues to be or higher; between 4–6 percent points less likely to a key factor in access to ID in many lower-income own an ID if they are out of the workforce (versus countries.47 in LICs alone, for example, an estimated a person in the workforce), unmarried (compared 35 percent of women living in LICs do not have an 46 Figure 6 plots the coefficients of logit models that regress ID ownership on various demographic characteristics and include country- level fixed effects and design-based standard errors. For a table of the regression results, see Appendix 8. 47 The broader Global Findex survey uses a binary (male/female) classification for gender. However, based on Lebbos et al. (2021), Totapally et al. (2019), and other research we expect that gender minorities including transgender and non-binary people may also have lower levels of ID coverage in many countries. This may be due to discrimination, persecution, and/or burdensome documentary procedures. RESULTS | 25 ID in 2021, compared with 27 percent of men, a gap 3—have gaps for women greater than 5 percentage of 8 percentage points statistically significant at the points and statistically significant at the 90 percent 95 percent confidence level (see Figure 4). Of the level or above (see Figure 7). 45 LICs and LMICs with data in 2021, 14—or nearly 1 in Figure 7. Countries with Largest ID Gaps for Women Despite these persistent issues, many countries have between 2017 and 2021. A few of the countries with made significant progress in increasing relative ID large gaps in 2017 (including Ethiopia, Niger, and coverage for women since 2017 (see Figure 8). In Madagascar) have had delays in 2021 Findex data others, however, the disparity in ID coverage between collection due to COVID-19, and so we are unable to men and women has stagnated or slightly worsened assess progress in this paper. 26 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Figure 8. Changes in the Gender Gap Over Time RESULTS | 27 Barriers to Obtaining an ID by approximately 40 percent of adults without ID), which increase transportation and opportunity costs In 2021, we improved the measurement of the due to lost work. Findex question on why people did not have an ID to better capture specific dimensions of the Many of these barriers are even higher in lower- process that make it difficult. As shown in Figure 9, income countries, with 46 percent—or nearly 1 in onerous documentary requirements remain a 2—of adults in LICs without an ID reporting that significant barrier to obtaining an ID for people they do not have the documents required to apply. in many countries. Globally, nearly 40 percent of Based on previous work, we also know that within adults without an ID reported that they lacked the countries, these burdens are most likely to fall necessary documents. In addition, obtaining the ID on disadvantaged women and other commonly remains too expensive for approximately 36 percent marginalized groups, such as persons with disabilities, of adults without one, either due to direct and/ low-literacy, minority language speakers, and sexual or indirect costs. A large portion of these indirect orientation and gender identity (SOGI) minorities costs may also be due to long travel times to apply (Hanmer et al. 2021, Lebbos et al. 2021, World Bank for, obtain, or correct an ID (reported as a barrier 2021b, World Bank 2020). Figure 9. Barriers to ID Ownership: Cumbersome, Bureaucratic Journeys 28 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES In addition to the difficulty of the process itself, with “low demand” for the country’s main ID also we also find that adults in many countries report face other difficulties. that they “do not feel comfortable giving their personal information” as a reason for not owning For example, although around a third of people gave an ID (approximately 18 percent). The reasons for not needing any identification or having another this discomfort may be multiple, including concerns ID as one reason they did not own their country’s about what data is collected and how it be used by foundational ID, only 16 percent said that these were ID authorities, and/or discomfort with the processes the only reason(s) they did not own it. As shown in involved in providing their data, such as poor Table 6, respondents who reporting having another ID treatment or harassment by registration officers still mentioned other barriers to obtaining identification (Hanmer et al. 2021, Lebbos et al. 2021). This question at similar rates to the full sample. Users who state they might also be capturing general levels of mistrust have no need for the ID also report lacking documents, in the institutions or systems. While unpacking this the ID being too expensive, or too far away at relatively new indicator will require deeper investigation that is high rates. People with self-proclaimed low demand context-specific, it highlights the needs for countries for the ID also appear to be less comfortable registering to close existing trust and/or knowledge deficits. than the global average, suggesting a potential relationship between low levels of demand and low Finally, as in 2017, we also see more “demand-side” trust to be explored further in future research. reasons for not having an ID, including people reporting that they perceive no need for the ID (29 percent of those without ID globally) or that they Difficulties without ID have another ID (around 30 percent). Both responses are more common among wealthier countries where In 2021, we also included a new series of questions ID coverage is close to universal, and/or other forms asking people if they had ever faced challenges of trusted ID, such as passports, are likely to be more accessing various services, opportunities, or rights widely held. This data does not fully reveal whether as a result of not having an ID. Average responses or not these other IDs respondent own are sufficient to these questions are shown in Figure 10. Globally, to access to all the services they likely to need over around 1 in 3 of those without an ID reported difficulty their lifetime and all the rights to which they are using financial services, receiving financial support entitled (including proof of legal identity). However, from the government, applying for a job, and/or it does provide some indication that many people voting in elections as a result of not having one. Nearly Table 6. Interrelated Barriers for “Low Demand” Users Respondents Who Respondents Who All Respondents Other Reported Reasons for Not “Have Another ID” (%) “Have No Need” (%) without ID (%) having the Foundational ID … n 5 1596 n 5 1440 n 5 9505 No documents 42.4 50.6 38.6 Too expensive 36.2 39.4 36.3 Too far 38.2 47.6 39.5 Not comfortable 22.8 29.5 17.5 Note: Multiple responses possible; means calculated with global weights. Source: ID4D-Findex (2021). RESULTS | 29 40 percent of those without an ID reported difficulties relatively similar for LICs and LMICs. Given the global obtaining a SIM card or mobile phone service, while estimate of 850 million people without ID, this implies around 25 percent had problems receiving medical that not having an ID has been a constraint on access care. Although there are some differences, the relative to services and the fulfillment of rights for hundreds of ranking of difficulties across type of activities is millions of people. Figure 10. Impact of Not Having an ID: Difficulty Accessing Services, Rights, and Opportunities The difficulty of accessing financial services without than 40 percent of unbanked adults cited lack of an ID is also reflected in the Global Findex 2021 documentation as a barrier (Demirgüç-Kunt et al. results. For example, 27 percent of unbanked adulted 2022). reported lacking the documentation needed to open an account at a financial institution. Unbanked More in-depth data collection and analysis are needed adults were more likely to cite these barriers in to contextualize these results locally. However, these economies such as Colombia (43 percent), Tanzania findings underscore the critical role that access to (50 percent), and Uganda (50 percent). Lack of ID identification plays in unlocking access to services also hampers access to mobile money accounts. in both the public and private sectors and enabling Some 30 percent of unbanked adults in Sub-Saharan people to exercise their rights. It furthermore points Africa reported they do not have the documentation to the urgency of closing remaining gaps in access needed to open a mobile money account. In Liberia, to identification and ensuring that future generations Mozambique, South Sudan, and Tanzania more can easily prove who they are from birth. 30 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES 5 DISCUSSION Counting the number of people who do KEY FINDINGS AND not have proof of identity worldwide LIMITATIONS is a challenging task. Fortunately, the availability and quality of data on ID These new data sources and improved methodology coverage are steadily improving, even as have enabled a more precise estimation of global ID coverage gaps for 2021, with less reliance on voter some areas of uncertainty remain. As of registration as a proxy indicator. While no single 2021, there are two rounds of ID4D-Findex indicator provides a perfect measure at all ages, in survey data on ID ownership available from combination these give a reasonable estimate of 2017 and 2021, covering the population the scale of the global identification gap and help aged 15+ across 130 countries. In addition, focus efforts on the areas with the greatest need. the amount of administrative data on ID In particular, this paper advances our knowledge on coverage collected by ID4D in 2019 and multiple topics: 2021 has more than doubled since 2018. • Scale and location of the ID coverage gap. While Birth certification rates are also now there has been some improvement over time, 1 in compiled by UNICEF for nearly three- 9 people globally still lack proof of their identity. quarters of the world’s countries, providing Even as the estimates change over time, we see a more complete picture of children’s that the problem of access to identification is access to identity documentation. overwhelmingly concentrated in the developing world, and mostly in lower-income economies in Sub-Saharan Africa and South Asia. At the same time, while most people in higher-income countries have an ID, many millions do not. The 2021 update to the estimates is able to better capture this using data from 194 countries. 31 • Who still does not own an ID, and why. As system data, and voter registration). While this with most development indicators, there is a provides the most complete estimate at the strong correlation between vulnerability and global scale, given the lack of uniform data, it marginalization, and the probability of not having means that comparisons between countries with an ID. Although some improvements have been differing data sources are not possible. Similarly, made, large gaps in ID ownership still remain we are combining different years of data. While with regard to gender, age, income, education, we use the most recent year of available data employment, and rural versus urban location. available, these vary across countries and This is the result of multiple, complex barriers, metrics (e.g., findex data vs. BRR). In addition, including onerous documentary requirements, while Findex and ID administrative data were the need to travel long distances, and direct collected within the last year, the latest birth and indirect costs. Other barriers—including registration data in 93 countries is more than potentially lack of trust—also reduce demand for 5 years old; of the countries that rely on voter and ownership of ID. registration as a proxy for adult ID coverage, there are 9 countries where the last election • Impact of not having an ID for individuals. was more than 5 years ago (2016 or before). Not owning an ID compromises hundreds of Coverage estimates for adults and children also millions of individuals’ rights and access to often have different source years, with adult services. Globally, around 1 in 3 adults without coverage estimates typically being more recent an ID reported difficulty using financial services, (see Appendix 1). receiving financial support from the government, or applying for a job. Nearly 40 percent of adults • Mixed set of indicators. In addition to combining without an ID reported difficulties obtaining a survey and administrative data, the coverage SIM card or mobile phone service, while around estimates include indicators measuring both 25 percent had problems receiving medical care. registration in a system (i.e., birth registration Beyond access to basic services and economic rates, in ID systems,48 voter registration rates) as opportunities, around a third of adult without well as ownership of a credential (i.e., Findex ID an ID reported this as a barrier to being able to ownership rates). In cases where obtaining a birth participate in elections. certificate or an ID automatically results from registering in a system, measuring registration As noted above, these results are also subject to a few alone may be a sufficient indicator of who limitations that may lead to over- or underestimation does (or does not) have proof of their identity. of the global ID gap. These include: However, this is not always the case, particularly in developing countries (e.g., see Bhatia et al. • Data sources with mixed methodologies and 2017). To the degree that owning IDs does not time frames. The estimates contain a mix of flow directly from registration, our estimates— survey data (birth registration rates taken from particularly for children—may be underestimating surveys and Findex) and administrative data the number of people without ID. (birth registration rates derived from UNSD, ID 48 As part of ID4D’s survey of ID agencies, we collect data on the number of people issued credentials. In the 2018 edition of the dataset, whether or not the number of people registered or issued with a credential was used varied by country. For 2021, we have used only registration figures to improve standardization because of larger credential data limitations. 32 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES • Potential measurement errors in ID data. As WPP data (UNDESA 2022a) is based on the discussed in Section 2, there is the potential de facto population and often makes adjustments for measurement error in both administrative to under-counting in censuses, it can often mean and survey-based data and the underlying higher population numbers than those in official population estimates from the UN’s World statistics. For this reason, our estimates of voter Population Prospects (WPP). To the extent registration and administrative registration rates that administrative data includes deceased or may be lower than those calculated by ID agencies non-resident populations or duplicate entries, themselves. numbers than based on country it may be underestimating the number of analyses. people without ID. Survey data can also have inaccuracies, including social desirability bias49 In addition to the above challenges, this paper or lack of knowledge50 (UNICEF 2013; Reed et al. also highlights persistent weaknesses in measuring 2021) that could result in inaccurate estimates access to ID globally: of coverage. To the degree that surveys are biased or unable to reach the poorest or • Birth registration and ID ownership for older most marginalized groups (due to remoteness, children. Although it is not yet available for residing in a conflict-affected area, respondents every country, the ID4D-Findex data represents being unwilling to talk to people not from within a significant advancement in measuring and the community)—i.e., those who are least likely comparing ID ownership for those 15 years and to have ID—they may also contribute to older. However, we still lack robust, standardized underestimating the ID coverage gap. In data on BRR for children ages 5-14 for many countries where phone ownership is heavily countries. This is particularly important for those correlated with having an ID (e.g., due to SIM countries where birth registration remains low, and registration requirements), phone surveys may rates in ownership may vary significantly by age. be systematically over-estimating ID ownership. The opposite could be true if surveys had a • ID ownership for non-nationals. States have harder time reaching people who are the most an obligation to provide legal identification to likely to have ID, such as face-to-face surveys all people who reside on their territory, and to where wealthy people living in gated community register the births of all children.51 In practice, (although this source of bias may be less likely). however, some countries allow all residents— including non-nationals—to register in the • Population data. Finally, there are potential foundational ID system (e.g., a national ID), while inaccuracies in the underlying population data, others maintain separate systems to provide particularly in cases where it might rely heavily documentation for non-nationals. ID ownership on modeling assumptions. Furthermore, because data from the latter are not captured here. 49 For example, respondents may report that they have registered a child’s birth or own an ID themselves—even if this is not the case— to conform with real or perceived expectations. 50 For example, research by Reed et al. (2021) (needs reference) in Tanzania found that of 2,500 women surveyed following childbirth at a major hospital in the capital, nearly half incorrectly believed that the birth notification form was the birth certificate. 51 See, inter alia, the 1951 Convention on the Status of Refugees, Article 27; the 1954 Convention on the Status of Stateless Persons, Article 27; and the Convention on the Rights of the Child (CRC) Article 7. DISCUSSION | 33 Furthermore, residents excluded from a country’s these potential shocks and lasting influence. main ID system may have other official IDs, such More work is needed to better understand the as passports or national IDs issued by their impact of this crisis on individual’s access to ID country of origin, residence permits, or other and the operation of ID and CR systems, and to IDs issued by the host country. In some cases, help build more resilient systems in the future. these IDs may enable them to access basic services or rights; in other cases, they may not be widely accepted IDs, and non-nationals POLICY RECOMMENDATIONS may struggle to participate in the country or obtain legal protections. These issues are most Although data indicates that ID coverage gaps in often faced by forced or irregular migrants, some countries are narrowing both among children including those displaced during conflict (Manby and adults over time, more must be done. Closing 2016). Given the complexity of these issues, the remaining gaps requires sustained commitment a combination of these factors may result in from policymakers and the global community. This is an over- or underestimation of the country- particularly urgent as lack of ID disproportionately level coverage gap depending on the specific affects people in lower-income countries and the individual or country-level factors. For example, most vulnerable groups in society, and without ID if individuals have IDs from their country of people may not be able to participate fully in social, origin and their host countries, they could be economic, and political life. counted in our data twice if administrative data from the country of origin include people who In addition, in countries that have recently concluded have migrated. More systematic data is needed large-scale registration efforts, such as, it will be both on non-national ID policies and on coverage important to ensure continuous access to civil numbers for people with various legal statuses registration and identification for children and adults, within a country. so that ID coverage gaps do not begin to grow over time. Additional efforts may also be needed to • Impact of COVID-19 on services and data address dips in civil registration and identification collection. We have seen the pandemic impact ID due to COVID-related office closures and mobility and CR services in multiple ways and potentially restrictions—whose impact may not be fully reflected opposing directions depending on the country in the currently available data—so that temporary or locality. Since early 2020, restrictions on disruptions do not turn into permanent coverage travel and movement, office closures, long lines and accessibility gaps. Key recommendations for and backlogs, and safety concerns have often stakeholders include: delayed or prevented people from registering births and obtaining identity credentials. At the Governments and other stakeholders must same time, government response to COVID— deliberately work to reduce or eliminate including emergency assistance—has often barriers that continue to prevent people— involved registration in benefits systems that particularly those living in LICs and groups have linked to, facilitated, or motivated people such as women, children and young adults, to obtain new identity documents, potentially low-income individuals, and those living driving up coverage in other cases (World in rural areas—from obtaining official or Bank 2022b). Unfortunately, the pandemic has legal proof of their identity. This includes also had a major impact on implementation of removing inequalities, onerous documentary routine and ad-hoc surveys and censes, which requirements, and fees for basic documents have created difficulties in accurately measuring 34 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES and services, by reforming relevant laws and Monitoring and improving access to ID regulations and improving business processes requires better, regular data collection. This and customer service standards. It requires paper highlights how data availability and finding ways to make ID services more comparability impede efforts to assess the convenient and user-friendly, including scale of the global ID coverage challenge, simplifying procedures and locating service and the same applies within countries. This is points closer to where people live or work. particularly the case for understanding ID and Analysis of the ID4D-Findex survey data civil registration coverage by age, particularly shows that these specific barriers are actively for older children. Countries and development preventing people from obtaining an ID; partners must invest in improving data identifying and targeting them should collection on identification systems through therefore be a priority. In addition, integrating, multiple channels. Including ID indicators in linking, or co-locating ID and civil registration censuses and other national and subnational (CR) services can help streamline processes survey efforts would help produce improved for individuals and ensure access to identity estimates at the country level for groups at throughout a person’s lifetime. the highest risk of exclusion. There is also more work to be done by ID authorities in defining and monitoring key indicators that Proactive, comprehensive engagement and can allow them to effectively track trends in communication with communities, local registration, credential issuance, as well as leaders, and civil society organizations is system performance with direct relevance to also essential. Where ID providers do not have inclusion objectives. Such measures must be a good understanding of people’s needs and able to be disaggregated by gender and, where the barriers they face regarding registration, possible, other demographics to ensure equal coverage is likely to be low. Robust information access and performance across potentially and education campaigns, ongoing feedback vulnerable and marginalized groups. during implementation, and sensible grievance redress mechanisms are needed to build trust and help people take advantage of ID4D and the World Bank also are committed to the opportunities that having official proof supporting these aims and helping implement the of identity can provide. Transparent and above recommendations through our direct support frequent involvement with civil society and for countries implementing or improving ID and civil community-based organizations—particularly registration systems, and our global and country- those representing the interests of marginalized level data and research. This includes continued and vulnerable groups—can help identify updates to the Global ID Coverage estimates, which and unlock key bottlenecks to boosting we expect to release every three years aligned with accessibility and enhancing coverage. Given Findex data collection. We also welcome new ideas that significant shares of the global ID4D- and partnerships for improving data collection and Findex respondents who do not have ID gave analysis to ensure that countries and the global ‘I don’t need it’ and/or ‘I’m not comfortable community have the information they need to build giving my personal information’ as reasons inclusive and trusted ID systems. why, the potential benefits of improved engagement and communication for overall This paper is the first in a series drawing on new ID coverage are significant. data collected for the 2021 ID4D Global Dataset. Ensuring universal access to civil registration and DISCUSSION | 35 identification is a core step toward unlocking the registration and ID systems and related research potential of good identification to benefit people, are steadily growing, more detailed quantitative public administration, service delivery, and economic and qualitative work is needed on the qualities and transformation. However, other aspects, such as characteristics of other aspects of these systems. ease of use, safeguarding people’s privacy and data, To that end, the ID4D Global dataset also includes a achieving operational and financial sustainability are rich set of indicators on key features of ID systems also critical aspects of ID system success. Therefore, across the world. Future papers and data releases while the volume and quality of data about civil will explore these results in depth. 36 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES REFERENCES AbouZahr, C., de Savigny, D., Mikkelsen, L., Setel, P. W., Lozano, Hanmer, L. C., Esquivel-Korsiak, V., and Rohini P. 2021. R., Nichols, E., Notzon, F. and Lopez, A. D. 2015. Civil Understanding the Gender Gap in ID : Key Research registration and vital statistics: progress in the data Findings and Policy Lessons from Nigeria - revolution for counting and accountability. Lancet Evidence Note. Washington, D.C. World Bank (London, England), 386(10001), 1373–1385. https://doi. Group. http://documents.worldbank.org/curated/ org/10.1016/S0140-6736(15)60173-8. en/632411615961644201/Understanding-the-Gender- Gap-in-ID-Key-Research-Findings-and-Policy-Lessons- ACE. 2022. Voter Registration: What is the legal voting from-Nigeria-Evidence-Note age in the national elections? ACE Electoral Knowledge Network. https://aceproject.org/epic-en/ IDEA. 2022. The Number of Registered Voters, Voter Turnout CDTable?view=country&question=VR001 (accessed 15 Database. https://www.idea.int/data-tools/question- April 2022). view/439 (accessed 15 April 2022). Bhatia, A., Ferreira, L. Z., Barros, A., & Victora, C. G. 2017. Who International Institute for Population Sciences (IIPS) and ICF. and where are the uncounted children? Inequalities in 2021. National Family Health Survey (NFHS-5), 2019-21: birth certificate coverage among children under five India. Mumbai: IIPS. https://dhsprogram.com/pubs/pdf/ years in 94 countries using nationally representative FR375/FR375.pdf. household surveys. International Journal for Equity in Kosovo Agency of Statistics. 2012. “Population by sex, size- Health, 16(1), 148. https://doi.org/10.1186/s12939-017- class of locality and age 2011.” https://askdata.rks-gov. 0635-6. net/pxweb/en/ASKdata/ (accessed 15 April 2022). Blitz, B., Apland, K., Lagaay, M., Yarrow, E., Lakshman, R. & Kosovo Agency of Statistics and UNICEF. 2020. 2019–2020 Hamilton, C. 2014. Birth Registration and Children’s Kosovo Multiple Indicator Cluster Survey and 2019– Rights: A Complex Story. Plan International: Surrey, 2020 Roma, Ashkali and Egyptian Communities Multiple UK. https://www.planinternational.be/sites/default/ Indicator Cluster Survey, Survey Findings Report. files/2017-09/birth_registration_and_childrens_rights_ Prishtina, Kosovo: Kosovo Agency of Statistics and full_report_1.pdf (accessed 12 December 2022). UNICEF. https://www.unicef.org/kosovoprogramme/ Bradley, P. M. 2017. “The Invisibles: The cruel Catch-22 of being media/1871/file/Kosovo%20MICS%202020.pdf poor with no ID.” The Washington Post. https://www. (accessed 31 March 2022). washingtonpost.com/lifestyle/magazine/what-happens- to-people-who-cant-prove-who-they-are/2017/06/14/ Lebbos, T. J., Esquivel-Korsiak, V. Clark, J. 2021. ID Systems fc0aaca2-4215-11e7-adba-394ee67a7582_story.html and SOGI Inclusive Design, Washington, DC: World (accessed 9 May 2022). Bank. https://documents1.worldbank.org/curated/ en/803981634587620467/pdf/ID-Systems-and-SOGI- Demirgüç-Kunt, Asli, Leora Klapper, Dorothe Singer, Saniya Inclusive-Design.pdf. Ansar, and Jake Hess. 2018. The Global Findex Database 2017: Measuring Financial Inclusion and the Fintech Manby, B. 2016. Identification in the Context of Forced Revolution. Washington, DC: World Bank. Displacement. Washington, D.C.: World Bank Group. https://documents1.worldbank.org/curated/ Demirgüç-Kunt, Asli, Leora Klapper, Dorothe Singer, and en/375811469772770030/pdf/Identification-in-the- Saniya Ansar. 2022. Global Findex Database 2021: Context-of-Forced-Displacement-Identification-for- Financial Inclusion, Digital Payments, and Resilience Development-ID4D.pdf in the Age of COVID-19. Washington, DC: World Bank. doi:10.1596/978-1-4648-1897-4. Metz, A. and Clark, J. 2019. Global ID Coverage, Barriers, and Use by the Numbers: An In-Depth Look at the 2017 Gelb, A. and Clark, J. 2013. Identification for Development: The ID4D-Findex Survey. Identification for Development Biometrics Revolution. CGD Working Paper. Center for Washington, D.C.: World Bank Group. http://documents. Global Development: Washington, DC. worldbank.org/curated/en/727021583506631652/ Gelb, A. and Diofasi Metz, A. 2018. Identification Revolution: Global-ID-Coverage-Barriers-and-Use-by-the-Numbers- Can Digital ID Be Harnessed for Development?. Center An-In-Depth-Look-at-the-2017-ID4D-Findex-Survey. for Global Development: Washington, DC. REFERENCES | 37 National Statistical Office. 2021. Malawi Multiple Indicator UNICEF. 2019b. “Despite significant increase in birth Cluster Survey 2019–20, Survey Findings Report. Zomba, registration, a quarter of the world’s children remain Malawi: National Statistical Office. https://mics-surveys- ‘invisible’”. UNICEF Australia press release, 11 December prod.s3.amazonaws.com/MICS6/Eastern%20and%20 2019. https://www.unicef.org.au/about-us/media/ Southern%20Africa/Malawi/2019-2020/Survey%20 december-2019/despite-significant-increase-in-birth- findings/Malawi%202019-20%20MICS%20Survey%20 registration (accessed 9 May 2022). Findings%20Report_English.pdf (accessed 31 March 2022). UNICEF. 2022a. UNICEF Global Databases. Received from Pelowski, M., Wamai, R. G., Wangombe, J., Nyakundi, H., UNICEF by email correspondence on June 9, 2022. Oduwo, G. O., Ngugi, B. K., and Ogembo, J. G. 2015. Why UNICEF. 2022b. “Indicator 16.9.1: Proportion of children under Don’t You Register Your Child? A Study of Attitudes and 5 years of age whose births have been registered with a Factors Affecting Birth Registration in Kenya, and Policy civil authority.” SDG Indicators Database. United Nations Suggestions. The Journal of Development Studies, 51:7, Department of Economic and Social Affairs. https:// 881–904, https://doi.org/10.1080/00220388.2015.1010156 unstats.un.org/sdgs/dataportal/database (accessed 31 Reed, S., Shabani, J., Boggs, D. et al. 2021. Counting on birth March 2022). registration: mixed-methods research in two EN-BIRTH United Nations. 1989. Convention on the Rights of the Child. study hospitals in Tanzania. BMC Pregnancy Childbirth 21 U.N. General Assembly Resolution 44/25. https://www. (Suppl 1), 236. https://doi.org/10.1186/s12884-020-03357-1. ohchr.org/en/instruments-mechanisms/instruments/ Rosenberg, J. S. and Chen, M. 2009. Expanding Democracy: convention-rights-child Voter Registration Around the World. Brennan Center for Justice: New York, NY. https://www.brennancenter. UNSD. 2022. Population and Vital Statistics Report: Vol. LXXIII org/sites/default/files/legacy/publications/Expanding. (January 2022). United Nations Statistics Division. Democracy.pdf (accessed 9 May 2022). https://unstats.un.org/unsd/demographic-social/ products/vitstats/ (accessed 31 March 2022). Schumacher, S., and Connaughton, A., 2020. “From voter registration to mail-in ballots, how do countries around University of Sydney. 2015. “No identity: One in five Aboriginal the world run their elections?”. Pew Research Center births unregistered in western Australia.” ScienceDaily, website, 30 October 2020. https://www.pewresearch. 3 July 2016. www.sciencedaily.com/releases/2016/07/ org/fact-tank/2020/10/30/from-voter-registration-to- 160703115732.htm (accessed 9 May 2022). mail-in-ballots-how-do-countries-around-the-world-run- World Bank. 2018. ID4D Global Dataset. Available at: https:// their-elections/ (accessed 9 May 2022). datacatalog.worldbank.org/int/search/dataset/0040787, Totapally, S., Sonderegger, P., Rao, P., Gosselt, J., and Gupta, World Bank. 2019. ID4D Practitioner’s Guide. Identification G. 2019. State of Aadhaar Report 2019. Dalberg, 2019. for Development. Washington, D.C.: World Bank https://stateofaadhaar.in/download-reports.php Group. http://documents.worldbank.org/curated/ UNDESA. 2022a. World Population Prospects 2022, Online en/248371559325561562/ID4D-Practitioner-s-Guide. Edition. United Nations, Department of Economic and World Bank. 2020. Creating Disability Inclusive ID Systems Social Affairs, Population Division. https://population. (English). Identification for Development Washington, un.org/wpp/Download/Standard/Population/ (accessed D.C.: World Bank Group. http://documents.worldbank. 23 August 2022). org/curated/en/967741605683569399/Creating- UNDESA. 2022b. World Population Prospects 2022: Summary Disability-Inclusive-ID-Systems of Results. United Nations, Department of Economic and World Bank. 2021a. “World Bank Country and Lending Social Affairs, Population Division. https://www.un.org/ Groups.” https://datahelpdesk.worldbank.org/ development/desa/pd/sites/www.un.org.development. knowledgebase/articles/906519-world-bank-country- desa.pd/files/wpp2022_summary_of_results.pdf and-lending-groups. (accessed 25 August 2022). UNICEF. 2013. Every Child’s Birth Right: Inequities and trends in World Bank. 2021b. People’s Perspectives on ID and Civil birth registration. UNICEF: New York. https://data.unicef. Registration in Rwanda: Research Report. Identification org/resources/every-childs-birth-right-inequities-and- for Development Washington, D.C.: World Bank trends-in-birth-registration/ (accessed 7 September 2022). Group. http://documents.worldbank.org/curated/ en/279741611941779893/People-s-Perspectives-on-ID- UNICEF. 2015. Unequal at Birth: Disparities in Birth Registration and-Civil-Registration-in-Rwanda-Research-Report in East Asia and the Pacific. UNICEF EAPRO: Bangkok. https://www.unicef.org/eap/reports/unequal-birth#:~:text= World Bank. 2022a. Principles on Identification for Sustainable Highlights,education%2C%20social%20services%20 Development. http://idprinciples.org. and%20protection (accessed 9 May 2022). World Bank. 2022b. The Role of Digital in the COVID-19 UNICEF. 2019a. Birth Registration for Every Child by 2030: Are Social Assistance Response. Washington, D.C.: we on track?, UNICEF, New York. https://www.unicef. World Bank Group. http://documents.worldbank. org/media/62981/file/Birth-registration-for-every-child- org/curated/en/099830009302217091/ by-2030.pdf (accessed 9 May 2022). P1731660f8c52f062092ac00d53c648bac7 38 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES APPENDICES APPENDIX 1. DATA AVAILABILITY Figure 11 shows the availability of ID coverage metrics by source and data year, organized by region. Metrics include birth registration rate (BRR), birth certification rate (BCR), ID4D-Findex ID data from 2021 and 2017 (F21 and F17, respectively), ID4D administrative data from 2018–2021 (A21, A19, A18, respectively), and voter registration rates (VRR). Figure 11. Data Availability by Country, Metric, and Year 40 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES APPENDICES | 41 APPENDIX 2. ADULT ID COVERAGE BY COUNTRY ID4D-FINDEX SURVEY DATA The ID4D-Findex Series can be downloaded for 2021 and 2017 from the ID4D website (https://id4d. worldbank.org/global-dataset) and directly from the World Bank’s DataBank (http://databank.worldbank.org). For convenience and ease of reference, Table 7 reproduces this data, disaggregated for men and women. 42 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Table 7. ID4D-Findex Adult ID Ownership, by Gender (2021, 2017) ID Ownership (%) 2021 ID Ownership (%) 2017 Economy ID Age* Total Women Men Total Women Men Afghanistan Tazkira (‫)تذکره‬ 15+ 87.4 76.7 98.5 71.4 48.4 94.2 Albania Identity Card (Leternjoftim) 16+ 97.0 97.1 96.8 92.8 92.1 93.5 Algeria Carte nationale d'identité (CNIBE) 15+ 96.7 97.3 96.1 95.1 94.5 95.7 Argentina Documento Nacional de Identidad (DNI) 15+ 98.6 98.2 99.1 99.3 99.4 99.2 1 Azerbaijan Identity Card (şəxsiyyət vəsiqəsi) 15+ 98.8 99.0 98.5 Bangladesh National Identity Card (NIC) 15+ 86.6 86.9 86.4 82.9 81.3 84.6 Belarus Паспорт (Passport) 15+ 98.9 99.2 98.6 Belgium National Identity Card (Identiteitskaart/carte d'identite/ 15+ 99.1 99.6 98.6 Personalausweis) Benin Carte nationale d’identité 15+ 47.0 41.7 52.6 45.9 36.9 55.1 Bolivia Cédula de Identidad 15+ 97.9 98.2 97.7 97.7 97.6 97.7 Bosnia and Herzegovina Identity Card (Lična karta) 16+ 96.8 96.0 97.6 94.5 95.2 93.7 1 Botswana National ID Card (Omang) 16+ 96.8 97.2 96.4 Brazil State-Issued ID Card (RG) 15+ 96.2 97.6 94.9 98.4 98.3 98.5 Bulgaria Identity Card 15+ 99.3 99.0 99.6 99.3 100.0 98.5 Burkina Faso Carte Nationale d'Identité Burkinabe (CNIB) 15+ 83.1 80.5 86.2 80.8 74.9 86.2 Cambodia Khmer Nationality Identity Card (អត្តសញ្ញាណប័ណ្ណ 15+ 89.6 92.5 86.3 88.6 88.9 88.3 សញ្ជាតិខ្មែរ) Cameroon Carte nationale d'identité (CNI) 18+ 82.8 79.7 86.1 85.3 83.7 86.9 2 Canada Social Insurance Number 15+ 97.8 96.5 99.2 1 Chad Carte nationale d'identité (CNI) 18+ 39.0 21.0 58.7 Chile Cédula de Identidad 15+ 99.2 99.2 99.1 98.8 99.1 98.5 China 居民身份证 (Resident Identity Card) 15+ 99.7 99.6 99.8 98.6 98.4 98.8 Colombia Cédula de Ciudadanía 18+ 96.9 97.3 96.4 98.9 99.2 98.6 Congo, Rep. Carte nationale d’identité 16+ 63.2 59.3 67.0 59.8 56.7 62.9 Costa Rica Tarjeta / Cédula de identidad 18+ 94.6 94.0 95.2 97.8 97.4 98.3 Côte d'Ivoire Carte nationale d’identité 15+ 71.8 67.7 75.7 68.0 68.7 67.4 Croatia Identity card (Osobna Iskaznica) 15+ 99.6 99.8 99.4 99.4 99.3 99.5 APPENDICES | 43 ID Ownership (%) 2021 ID Ownership (%) 2017 Economy ID Age* Total Women Men Total Women Men Cyprus Identity Card 15+ 87.4 87.3 87.5 Czech Republic Civil Card (Obcansky prukaz) 15+ 99.1 99.9 98.3 98.5 99.2 97.8 Denmark CPR number (Personnummer) 15+ 98.5 98.4 98.7 Dominican Republic Cédula de identidad y electoral (CIE) 16+ 91.0 90.9 91.1 89.9 90.7 89.2 Ecuador Cédula de Identidad 15+ 97.5 97.4 97.6 99.5 99.4 99.6 Egypt, Arab Rep. (National ID Card) ‫بطاقة الرقم القومي‬ 16+ 97.3 97.7 96.9 93.7 90.2 97.2 El Salvador Documento único de identidad (DUI) 18+ 97.0 96.6 97.4 96.3 95.8 96.9 Estonia ID Card (ID-kaart) 15+ 98.0 99.1 96.6 98.6 98.2 99.0 1 Ethiopia Kebele ID Card 18+ 64.2 54.1 74.8 France Carte nationale d’identité (CNI) 15+ 93.8 94.2 93.4 Gabon Carte nationale d’identité 16+ 73.0 69.0 77.4 69.9 67.2 72.7 Georgia National ID card 15+ 95.1 97.0 93.0 94.1 94.1 94.1 Germany Identity Card (Personalausweis) 15+ 95.7 95.9 95.4 Ghana National Identity Card (GhanaCard) 15+ 86.7 84.7 88.9 Greece Identity Card (ΔΕΛΤΙΟ ΤΑΥΤΟΤΗΤΑΣ) 15+ 97.7 98.2 97.2 92.8 94.4 91.1 44 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES 1 Guatemala Documento personal de identificación (DPI) 18+ 93.0 93.3 92.7 Guinea Carte Nationale d'Identite 15+ 58.6 51.6 65.7 44.6 39.4 49.6 Haiti Carte d’identification nationale 18+ 79.5 83.0 75.8 Honduras Tarjeta de Identidad (DNI) 18+ 92.8 92.8 92.7 93.2 91.6 94.9 Hong Kong SAR, China Hong Kong Identity Card (HKID) 18+ 98.1 98.0 98.4 Hungary Identity Card (Személyazonosító igazolvány) 15+ 98.9 99.7 97.9 99.7 99.6 99.8 Iceland National ID number (Kennitala) 15+ 99.7 100.0 99.3 India Aadhaar Number 15+ 96.8 96.7 96.8 Indonesia Identity Card (KTP) 17+ 96.7 97.7 95.7 95.8 95.7 95.9 Iran, Islamic Rep. National Identity Card 15+ 86.6 89.8 83.5 (Kart-e-Melli) Iraq National Identity Card 15+ 94.2 94.0 94.4 Israel Identity Card (Teudat Zehut) 16+ 98.2 98.6 97.7 99.1 99.7 98.5 Italy Identity Card (Carta d’identità) 15+ 97.9 98.6 97.3 3 Jamaica National Identification Card or National ID (Voter ID) 18+ 82.7 85.4 80.0 ID Ownership (%) 2021 ID Ownership (%) 2017 Economy ID Age* Total Women Men Total Women Men Jordan (Civil Status ID Card) ‫بطاقة األحوال المدنية‬ 18+ 96.8 97.9 95.8 95.3 95.4 95.3 Kazakhstan National passport/ID card 16+ 98.9 99.2 98.5 93.8 92.7 95.0 Kenya National ID card 18+ 91.4 88.9 93.9 94.3 93.0 95.6 Korea, Rep. Resident Registration Card (주민등록증) 17+ 96.8 97.4 96.3 Kosovo ID card 16+ 95.2 94.2 96.2 92.7 94.2 91.2 Kyrgyz Republic National ID Card / National passport 16+ 93.6 93.1 94.1 93.2 92.8 93.6 Lao PDR Identity Card 15+ 55.3 55.3 55.4 40.7 40.8 40.7 Latvia Identity Card / passport (Personas Aplieciba) 15+ 98.8 99.5 97.9 98.9 98.8 99.1 Lebanon (ID Card) ‫بطاقة الهوية‬ 15+ 97.0 96.8 97.2 96.7 96.8 96.6 1 Lesotho National Identity Card 16+ 70.9 69.1 72.7 Liberia National ID Card 15+ 30.1 26.9 33.5 Lithuania Identity Card (Asmens Tapatybės Kortelė) 15+ 92.4 92.1 92.6 81.5 80.7 82.3 1 Madagascar Carte d’identité nationale (CIN) 18+ 85.4 80.5 90.9 Malawi National Identity Card (NIC) 16+ 85.1 86.8 83.3 16.0 16.1 16.0 Malaysia Identity Card (MyKad) 15+ 95.9 95.7 96.0 94.2 94.3 94.1 Mali Carte NINA 15+ 69.3 66.3 72.5 70.7 63.1 78.5 Malta Identity Card (Karta Tal-Identità) 15+ 98.6 98.4 98.8 1 Mauritania Carte d’identification 15+ 89.1 88.4 89.8 Mauritius National Identity Card 18+ 98.9 99.1 98.6 Mexico1 Clave Única de Registro de Población (CURP) 15+ 89.2 88.3 90.2 Moldova Identity Card (Buletin de identitate) 15+ 98.8 98.5 99.2 97.1 96.9 97.3 Mongolia Citizen Identity Card of Mongolia (Иргэний үнэмлэх) 16+ 98.1 98.4 97.8 97.6 98.3 96.9 Montenegro1 Identity Card (Lična karta) 15+ 93.8 95.6 92.0 Morocco Carte d'Identite Nationale 15+ 93.8 92.4 95.3 92.7 92.9 92.6 Mozambique Bilhete de identidade (BI) 15+ 58.1 51.2 65.5 58.2 51.5 65.3 Myanmar Citizen Scrutiny Card (နိုင်ငံသားစိစစ်ရေးကတ်) 15+ 87.8 89.2 86.3 88.8 87.2 90.6 Namibia National ID Card 16+ 91.3 91.8 90.7 91.9 93.4 90.1 Nepal Citizenship certificate 16+ 88.0 86.2 90.0 Netherlands Dutch Identity Card (Nederlandse Identiteitskaart) 15+ 95.1 96.8 93.3 APPENDICES | 45 ID Ownership (%) 2021 ID Ownership (%) 2017 Economy ID Age* Total Women Men Total Women Men Nicaragua Cédula de Identidad 15+ 90.2 87.9 92.8 89.8 89.5 90.1 1 Niger Carte nationale d’identité 15+ 44.8 32.0 57.1 North Macedonia National Identity Card 15+ 97.9 97.3 98.5 94.0 95.1 92.8 Norway National Identity Number 15+ 99.5 100.0 99.0 Pakistan Computerized National ID Card (CNIC) 18+ 88.3 77.1 98.7 86.5 77.5 95.1 Panama Cédula de Identidad Personal (CIP) 18+ 97.8 98.1 97.5 95.3 96.9 93.6 Paraguay Cédula de Identidad 15+ 99.3 99.4 99.2 99.6 99.2 100.0 Peru Documento Nacional de Identidad (DNI) 17+ 97.6 98.4 96.7 98.9 98.9 99.0 Poland Identity Card (Dowód osobisty) 15+ 98.4 98.8 97.9 93.3 94.5 91.9 Portugal Cartão de cidadão / Bilhete de Identidade de Cidadao Nacional 15+ 95.6 96.4 94.8 Romania Identity Card (Carte de Identitate) 15+ 99.2 99.6 98.9 99.0 98.7 99.4 Russian Federation Internal Passport (Внутренний Паспорт) 15+ 98.7 99.4 97.8 97.9 99.2 96.5 Rwanda National Identity Card (NID) 16+ 92.9 91.5 94.4 Saudi Arabia National ID card 15+ 98.6 97.3 99.4 Senegal Carte nationale d'identité 15+ 82.5 79.4 85.9 72.6 72.1 73.1 46 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Serbia Identity card (Lična karta) 15+ 99.3 99.6 98.9 96.7 97.4 95.9 Sierra Leone National Identity Card 15+ 61.5 64.1 58.7 Singapore National Registration Identity Card (NRIC) 15+ 96.9 97.7 95.9 94.7 92.1 97.5 Slovak Republic Identity Card (Občiansky preukaz) 15+ 99.6 99.8 99.3 99.4 99.5 99.4 Slovenia Identity Card (Osebna izkaznica) 15+ 95.4 96.7 94.0 South Africa Smart Card ID / Green Barcoded ID 16+ 94.1 94.2 94.0 92.4 91.0 93.9 South Sudan National ID Card 17+ 13.2 11.7 14.9 21.7 11.6 32.0 Spain Documento nacional de identidad (DNI) 15+ 96.0 96.2 95.7 Sri Lanka National Identity Card (NIC) 15+ 92.5 94.4 90.5 92.1 89.7 94.9 Sweden Personal Identity Number (Personnummer) 15+ 100.0 100.0 100.0 Switzerland Identity Card 15+ 95.0 95.1 94.8 Taiwan, China National Identification Card 15+ 98.8 99.4 98.2 Tajikistan National Passport / ID card 16+ 85.6 84.1 87.3 Tanzania National ID Card (Kitambulisho Cha Taifa) 18+ 60.3 58.8 61.9 Thailand Thai National ID Card 15+ 98.9 99.8 98.0 99.5 99.0 100.0 Togo Carte nationale d'identité 15+ 40.3 36.0 44.9 39.6 31.6 47.7 Trinidad and Tobago National Identification Card 15+ 94.9 93.9 96.0 ID Ownership (%) 2021 ID Ownership (%) 2017 Economy ID Age* Total Women Men Total Women Men Tunisia Carte D'Identité Nationale (CIN) 18+ 98.8 98.9 98.7 96.5 95.2 97.8 Türkiye Türkiye Identification Number (Türkiye Cumhuriyeti Kimlik 15+ 96.2 97.2 95.1 96.7 95.7 97.6 Numarasi) Turkmenistan1 паспорт (Passport) 16+ 98.4 98.2 98.6 Uganda National ID Card 15+ 72.5 72.8 72.2 81.4 80.5 82.3 Ukraine Ukrainian National Passport / ID card 15+ 99.0 98.9 99.0 97.3 98.4 95.8 United Arab Emirates Identity Card (Emirates ID) 15+ 89.8 94.1 88.2 Uruguay Documento de identidad 15+ 100.0 100.0 100.0 99.7 100.0 99.5 Uzbekistan National Passport 15+ 92.3 92.1 92.6 94.5 96.0 92.8 Venezuela, RB Cédula de Identidad 15+ 98.8 99.0 98.7 98.3 98.1 98.6 Vietnam Citizen ID Card (Căn cước công dân, CCCD) / ID Card 15+ 97.0 96.9 97.0 94.1 95.9 92.1 West Bank and Gaza Identity card (hawiyya) 16+ 96.1 95.9 96.3 1 Yemen, Rep. Identity Card 15+ 48.9 27.4 74.0 Zambia National Registration Card (NRC) 16+ 93.9 93.6 94.2 89.0 87.2 90.9 Zimbabwe National Identity Card (NID) 16+ 85.5 86.9 84.0 86.2 84.9 87.7 Notes: * The Findex survey includes respondents age 15+; for the purpose of this analysis, we exclude respondents who are above 15 but below the age of eligibility for obtaining the ID. This avoids artificially deflating the ownership rates by including those who are not yet eligible to obtain it. 1 Data collection in these countries was delayed due to COVID-19 but is expected to be available in 2023. 2 Canada does not have a foundational ID (e.g., a national ID), so the survey asked about ownership of the social insurance number which is widely used across the government and private sector for identification. 3 Jamaica did not have a de jure national ID at the time of survey data collection; instead, the data measure ownership of the voter ID, which is the most commonly held and used ID and was colloquially referred to as the ‘national ID’ during the survey period. APPENDICES | 47 SELECTED ADMINISTRATIVE DATA In addition to the data presented in Table 7, our primary model for estimating Global ID coverage uses administrative data for 14 countries without ID4D-Findex survey data, shown in Table 8. This data was obtained through questionnaires fielded to ID agencies by ID4D between 2019–2022. Table 8. Administrative Data Used in Primary Global ID Coverage Estimates Data Reported Group Calculated Economy ID System Age1 Note Year Registrations2 Population3 Coverage (%)4 Angola Bilhete de 2019 6 22,027,509 25,596,346 86.1 * identidade (BI) Armenia National ID System 2021 5 3,467,144 2,608,055 100.0 Bahrain Civil Registration 2020 5 1,370,570 1,375,340 99.7 System (CRS) Cabo Verde Bilhete de 2019 5 569,720 525,416 100.0 * Identidade/ Cartão Nacional de Identificação Central African National ID System 2021 18 893,145 2,411,966 37.0 * Republic Equatorial Guinea National ID System 2022 15 646,000 1,030,216 62.7 * Guyana Guyana 2019 14 663,365 583,708 100.0 * Identification Card Kuwait Kuwait ID System 2021 5 4,348,807 3,991,248 100.0 Maldives National ID Card 2019 5 459,360 465,820 98.6 * Nigeria National Identity 2021 5 66,740,457 178,570,524 37.4 Management System (NIMS) São Tomé and Civil Identification 2021 5 89,408 192,729 46.4 Principe System Seychelles National 2021 5 155,687 97,959 100.0 Population Database (NPD) Tonga National Identity 2019 14 65,546 70,088 93.5 * Card Vanuatu Vanuatu National 2019 5 254,900 259,834 98.1 * ID card Footnotes: 1 The cutoff age, set to the minimum age of eligibility for registering in the ID system or 5 years old, whichever is higher. 2 Either the total number of registrations or the number of registrations at or above “age” (see *), as reported by the ID agency. In some cases, these numbers may include deceased persons, duplicates, or those no longer resident in the territory. 3 Total population in the territory at or above “age” in the data year (2019, 2020, 2021, or 2022), from UNDESA (2022a). 4 Coverage rate is reported registrations, divided by the group population at or above “age.” It is censored to 100 percent if the number of registrations is greater than the population. * Disaggregated registration numbers by “age” are unavailable; instead “reported registrations” are the total number of registrations in the system. Calculated coverage rates assume that all registrations are above the eligible age. 48 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES APPENDIX 3. ALTERNATIVE SPECIFICATIONS AND ROBUSTNESS CHECKS This appendix provides a series of alternative specifications and robustness checks for the global estimate methodology described above. This includes deeper analysis related to (1) the use of birth certification rates instead of birth registration rates, (2) using birth registration rates for older children, (3) the changes between the 2018 and 2021 estimates, and (4) restricting the ID4D-Findex sample to those above the ID age. As with the main estimates, data and replication code for these analyses will be available at http://id4d. worldbank.org/global-dataset. BIRTH CERTIFICATION VS. REGISTRATION RATES Birth certification is an important aspect of children’s ID coverage, but historically birth certification rates have not been comprehensively compiled in a cross-country manner. To our knowledge, the most comprehensive analysis of this type has been by Bhatia et al. (2017), in which the authors used household survey data from 94 countries to calculate birth certification and birth registration rates and examined inequalities in coverage. UNICEF (2019a) also provides global and regional estimates and country-level data for a handful of Eastern and Southern African countries. Both Bhatia et al. and UNICEF find significant gaps between birth registration and birth certification rates. UNICEF, in its 2022 Database has now made available BCR data for 144 countries, which have also been included into the ID4D DataBank series available at http:// databank.worldbank.org. The primary source for data on birth certification rates are DHS and MICS surveys, which employ slightly different question structures for this topic. Both ask caregivers to report whether a child has been registered and whether a birth certificate was issued. This creates three categories of response: “birth registered, with a certificate,” “birth registered, no certificate,” and “birth not registered.” The DHS combines all three outcomes as responses to question, whereas the MICS question is multi-step and probes further by asking the respondent to show the birth certificate if they can (in an attempt to mitigate confusion with other documents). The 2022 UNICEF Database includes BCR for 144 countries, and we obtain the same data directly from the Kosovo survey report. These are derived from the most recent DHS and MICS surveys, allowing us to perform an updated analysis of the global trends identified by Bhatia et al. (2017). This 2022 data also covers an additional 51 countries, compared to that paper’s data. Our findings mirror previous analyses that birth certification significantly lags behind birth registration rates in nearly all countries. Figure 12 and Table 9 illustrate this relationship. Figure 13 displays the nine countries with gaps between birth registration rates and birth certification rates greater than 25 percentage points, out of the 145 countries with BCR data available. Three countries—Rwanda, Solomon Islands, and Sierra Leone—have gaps greater than 50 percentage points. APPENDICES | 49 Figure 12. Birth Registration vs. Certification Rates, by Income Group Figure 13. Countries Gaps between BCR and BRR Greater than 25 Percentage Points 50 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Birth certification is difficult to measure and there may be multiple reasons for the gap between birth registration and birth certification rates. The simplest explanation is that while the birth was registered, the birth certificate was either never collected, lost, or misplaced. Other reasons include measurement errors, which may affect both birth registration and certification rates, such as a false positive response if the respondent mistakes another document (i.e., a birth notification form prepared by the health facility or a receipt printed during birth registration) for an official birth certificate. Social desirability bias may also come into play if respondents believe that an affirmative response is what is expected of them. If these measurement errors are indeed prevalent, this would suggest that the global ID coverage gap could be significantly larger among children than previously thought. Given the extremely low BCR rates in many countries, using this data for our global ID coverage estimates would significantly increase the number of people without an ID, particularly given that we would be applying this rate not only to children under 5, but also to those age 5 to the cutoff age. If BCR data were used instead of BRR data for our primary estimation model—including as a proxy for the 5–17 population—the estimate for the global ID coverage gap would increase to 975 million. This would be an approximate 150 million increase over the estimates using BRR. This jump in the coverage gap is primarily driven by countries like Bangladesh, India, Kenya, Niger, Pakistan, the Philippines, and Rwanda which are (relatively) populous and have a significant difference between BRR and BCR rates. Table 9. Using Birth Certification in the Global Estimates A. Primary Model Using BCR for B. Primary Model Using BCR for Children (Millions) Children and Adults (Millions) N Children Adults Total Children Adults Total World 194 573 417 989 573 417 990 High-Income Countries (HICs) 60 0.2 16.2 16.4 0.2 16.2 16.4 Low-Income Countries (LICs) 25 206.7 99.1 305.8 206.7 99.1 305.8 Lower-Middle Income Countries (LMICs) 54 338.9 264.1 603.0 338.9 264.1 603.0 Upper-Middle Income Countries (UMICs) 54 25.4 37.3 62.6 25.4 37.3 62.7 * In the 2021 World Bank lending groups, Venezuela does not have a classification and is therefore excluded from this table. Note: Calculations are done using our primary model, but panel A uses birth certification rate (BCR) for children where this is available, instead of birth registration rate (BRR), while panel B uses BCR where available for children and for adults in countries where BRR would have been used as the adult metric. BRC data come from UNICEF (2022a). APPENDICES | 51 COVERAGE FOR OLDER CHILDREN Data on birth registration coverage for children over 5 (or adults) are not commonly or uniformly available across countries. For this reason, we apply the under-5 BRR rate for all children in the population between 0 and the cutoff age, as in previous versions of the global estimates. While we employ this method due to the unavailability of data, it will not always provide an accurate picture of ID coverage for children in countries that have not had near-universal birth registration for the past few decade years—primarily LICs and LMICs.52 This bias can occur in two directions: A. For countries that have recently improved birth registration rates, more recent BRRs would be higher than older ones (i.e., the probability of timely birth registration for children 0–4 would be higher than for those 5–14 years, the most common cutoff age). In this case, applying the under-5 BRR to older children could underestimate the number of people without ID. B. Conversely, birth registration rates can increase a child’s age as parents register births late, for example, by waiting until a birth certificate is required for school enrollment or other purposes (AbouZahr et al. 2015, Pelowski et al. 2015). In countries with lower levels of timely birth registration, this could mean differences in BRRs between younger and older children (i.e., the rate for 5–14 would be higher than the rate for 0–4). In this case, applying the under-5 rate to older children would overestimate the number of people without ID. In the absence of birth registration data for older children that are available for all countries, it is difficult to check which of these opposing effects is stronger. Furthermore, we do not have sufficient administrative data disaggregated for the 5–15 age group to compare BRR vs. ID coverage for older children. Still, we can get a hint of some potential trends by looking at yearly registration rates for ages 0–4 where disaggregated data is available. For 104 countries, data derived from MICS and DHS surveys provides disaggregated BRRs for 0–11 months, 12–23 months, 24–35 months, 36–47 months, and 48–59 months. Among these countries, the overall population weighted average BRR for 0–11 months is approximately 4.4 percentage points lower than the rate for 48–59 months, which translates into an under-5 BRR that is 0.9 percentage points lower than at 48–59 months (see Table 10). For LICs, where birth registration rates are lowest, the difference between 0–11 months and 48-59 months is 2.3 percentage points, with an under-5 rate that is 0.1 percentage points lower than for children 48–59 months. As shown in Figure 14, however, these averages hide some significant variation at the country level, where we see trends of both increasing and decreasing BRR for children under 5. 52 Nearly all countries for which we use birth registration for adults in our primary estimates are HICs and UMICs with recorded 100 percent BRRs, and so the effect of this bias in these cases is quite limited. 52 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Still, these findings indicate that on average, birth registration rates do increase slightly over the first 4 years of life and that the may continue increasing to some degree after the child reaches the age of 5. This would suggest that our methodology of applying the most recent under-5 BRR to older children may somewhat overestimate the number of children without ID. However, given that (a) no systematic data on coverage for older children is available cross-nationally, and (b) most other methodological choices in our estimates are on the conservative side (e.g., see Section 4 and other sections of Appendix 3, and Appendix 4), we believe this continues to be a reasonable approach. As more data become available—or for those countries where there is age-disaggregated data on birth registration and/or ID coverage for older children, improvements to this methodology may be possible. Table 10. Comparing Yearly BRRs for Children Under 5 from MICS and DHS Data Months Difference between Under-5 rate under- 5 and 0–11 12–23 24–35 36–47 48–59 48–59m World 67.6 71.9 72.2 71.8 71.9 71.1 20.8 SSA 42 47 48 47.6 47.9 46.5 −1.4 SAR 69.4 73.2 73.2 72.5 72.5 72.1 −0.4 EAP 86.1 91.2 91.3 91.1 91.8 90.3 −1.5 LAC 89.9 97 97.5 98 98.2 96.3 −1.9 ECA 97.7 99.1 99.6 99.4 99.4 99 −0.4 MNA 89.5 91.2 91.6 91.7 91.6 91.1 −0.5 LIC 45 48.4 48.2 47.2 47.3 47.2 −0.1 LMIC 68.6 73.2 73.6 73.3 73.4 72.4 −1 UMIC 93.9 97.8 98.3 98.5 98.4 97.5 −0.9 HIC 94.7 99.6 99.2 99.9 100 98.8 −1.2 Note: Table shows population-weighted averages. Includes 104 countries with age-disaggregated birth registration rates and population data. Source: UNICEF (2022b), UN World Population Prospects (2019), Kosovo (2020), Malawi (2021). APPENDICES | 53 Figure 14. Example Country-Level Trends in BRRs for Children Under 5 54 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES EXPLORING CHANGES FROM 2018 TO 2021 As noted in the introduction, the 2018 and 2021 coverage estimates should not be treated as a time series—that is to say, the difference between them does not represent changes in ID coverage alone. Rather, it represents a mix of actual coverage improvements, changes in data sources due to the availability of new data (e.g., the ID4D-Findex data, and new administrative data), and other methodological changes summarized in Appendix 5. This includes the use of birth registration data for adults in some HICs and UMICs—which in turn led to the inclusion of more countries in the calculations—the changes in how we calculate voter registration and administrative coverage rates instead of using raw totals and applying the BRR to the 0–4 population in all countries, including where the ID age is 0. Some of these factors are also inter-dependent. For example, a country where the source of data changes from voter registration in 2018 to another source in 2021 will also likely have a change in the cutoff age that leads to a change in the composition of the population to which the child and adult rates are applied. In this section, we run two tests to unpack these numbers using counterfactual scenarios. The first is to apply the 2021 methodology to the data available in 2018, which gives us an idea of the total impact of the changes in data availability and coverage rates. The second is to apply the 2018 methodology to the data we are using for 2021, which this provides an approximation of the total impact due to the changes in methodology. Neither comparison is perfect, given the interdependence and multi-facet nature of these changes, however, they help validate the 2021 methodology and provide some useful data on how we should interpret our results. COUNTERFACTUAL 1: 2021 METHODOLOGY WITH 2018 DATA If we had applied the 2021 methodology to data available in 2018, including both the data used in the actual 2018 calculations, as well as the 2017 round of the Findex, the 2018 estimate would have been 969 million. This is slightly lower but on par with the actual 2018 estimate of 987 million, but significantly (126 million) higher than our 2021 estimates. Comparing these counterfactual 2018 estimates with our primary model in 2021 reflects both actual improvements in actual coverage, as well as the effect of newly available data, holding the methodology constant. These effects are disaggregated in Table 11. For 125 countries using the same metrics in both 2018 and 2021, there has been a decrease in the estimated number of people without ID by around 157 million. In parallel, the availability of better data—including newly available ID4D-Findex and administrative data in 2021—reduces the estimates by approximately 24 million. Finally, there are 28 countries that would have used the same metrics in 2018 and 2021 but have not had any changes in value, either because coverage rates were already 100 percent (in the case of birth registration rates in many HICs), or because there is no more recent data. We still see a slight difference of 8 million in the number of people without an ID in these countries, but this is an artifact of applying constant rates to increasing populations. APPENDICES | 55 Table 11. Estimated Effect of Changes in Data Values Difference (Actual – Counterfactual) Change Description Total Adults Children Value of metric only Change in value from 2018 −157 −57 −100 Which metric is used New metric available for 2021 +24 +33 −9 None Same data and/or value from 2018 +8 +4 +4 Total −125 −19 −106 Difference columns provides approximations of the impact of changes in the underlying data on our estimates of the number of people without ID in 2021. These calculations use a counterfactual scenario where we apply the 2021 methodology to data that was available in 2018—including that used in the original 2018 calculations, and the 2017 round of the ID4D-Findex survey—and compare this our primary 2021 estimates. This holds constant methodological changes and allows us to partially identify the magnitude of changes on the estimated number of people without an ID resulting from improved data sources and actual improvements to ID coverage. COUNTERFACTUAL 2: 2021 DATA WITH 2018 METHODOLOGY In the second counterfactual analysis, we explore the impact of various methodological changes by applying the 2018 methodology to the 2021 data. This counterfactual excludes ID4D-Findex data (using administrative and then voter registration for all countries), uses a cutoff age of 5 even when the ID age is 0, and subtracts voter and administrative registration totals from the 2021 population instead of calculating rates. Under this scenario, our 2021 estimate would have been 902 million, or around 58 million higher than our primary model. As shown in Table 12, we can disaggregate these results based on the type of methodological change, holding the data values constant. A large portion of this change comes from the addition of new ID4D-Findex data, and particularly the use of survey instead of administrative data. This is a result of the fact that ID ownership rates indicated by the survey data are lower, on average, than the coverage rates provided by administrative data for the reasons outlined in Section 3. In addition, we see that changes in how we calculate voter registration rates (VRR) and administrative coverage rates, have reduced the estimates slightly as expected, reducing the probability that these calculations were inflating results. A large impact also comes from the inclusion of additional countries, as a results of using BRR for HICs and UMICs, and removing other 2018 exclusion criteria—if we had applied these criteria in 2021, 58 countries would have been excluded from the calculations, accounting for an additional 34 million people without ID. 56 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Table 12. Estimated Effect of Methodology Changes Difference (Actual – Counterfactual) Change Description Total Adults Children Which data is used Switch from Admin to Findex −61 −135 +74 Switch from Voter to Findex −2 +8 −9 Switch from Voter to BRR −3 −3 −0 Calculations Change in VRR calculations −10 −10 +0 Change of admin cutoff age to 5 −15 −35 +20 Change in admin calculations −2 −2 +0 Exclusion criteria Additional countries included +34 +20 +14 Total −58 −157 +98 Difference columns provides approximations of the impact of various methodological changes on the 2021 dataset estimates, using a counterfactual scenario where we apply the 2018 methodology to the 2021 data. This holds constant the data values themselves, and allows us to partially identify the magnitude of these changes on the overall estimated number of people without an ID. FINDEX SAMPLE RESTRICTIONS The Findex survey includes respondents aged 15 and older, and analysis of this data in the financial sector typically includes this full sample (Demirgüç-Kunt et al. 2022). However, following Metz and Clark (2019), we restrict the sample for individual-, country-, and global-level calculations to include only those respondents who are over the age when people become eligible to obtain the ID.53 Respondents under this age would generally not be expected to have the ID, and by dropping these observations, we avoid potentially underestimating adult ID ownership rates among eligible adults.54 This restriction, therefore, provides more realistic snapshots of the ID gap in specific countries, and better aligns with the global estimates methodology of selecting cutoff ages based on the data source and country’s ID age. 53 In total, this drops just under 1,400 observations in 2021 (around 1.2 percent of the data), and 1,680 observations in 2017 (around 1.6 percent of the data) from our analysis. 54 In both 2017 and 2021, a minority of respondents younger than the ID age report that they have the ID. The fact that some people do report obtaining the ID before the age of eligibility may be due to discretion or exceptions in policy (e.g., some countries allow minors to obtain the ID in certain circumstances with parental permission), or errors in reported age or ID ownership. APPENDICES | 57 Figure 15. Global ID Ownership with Restricted vs. Full Findex Sample Restricting the Findex sample to only those above the ID age has only minor effects on globally-weighted Findex averages (see Figure 15). However, limiting the sample significantly impacts national-level ID ownership rates for a small set of individual countries. In Cameroon, for example, dropping observations for respondents below age 18 (the minimum to apply for the national ID card) increases the 2021 estimates of adult ID coverage by 7.2 percentage points, from 75.5 to 82.8 percent. Figure 16 shows cases where restricting the sample changes estimates of adult coverage by 1 million or more. Figure 16. Impact of Findex Age Restriction on Estimates of Adults without ID 58 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES As summarized in Table 13, aggregating across countries using the restricted rates reduces the number of adults without an ID by approximately 45 million compared with the full sample. However, in addition to changing the estimates for adults, restricting the Findex sample to those above the ID age also shifts the overall cutoff age upwards in a handful of countries, increasing the size of the child population to which BRR is applied. For example, if the ID eligibility age is 18, the Findex rate would be applied to the population 18+, with the BRR applied to ages 0–17 (instead of 0–15 if the ID age was set to 15). As a result of restricting the Findex sample, the total number of children without ID is 26.4 million higher than it would be if the full sample was used. Combining this with the adult figures, the net difference is approximately 18.6 million people, with the restricted sample providing a slightly more conservative estimate of the ID gap. Table 13. Impact of Findex Age Restriction on Global Estimates Estimated Population without ID in Economies with Findex Data Findex Sample Adults Children Total Full 309  311 620 Restricted 264 337 601 Difference −45  26 −19 Note: Includes 130 economies where Findex data is used in the global estimates. Calculated by summing the number of adults and children without ID across these economies alternately using the "full" sample of respondents ages 15 and older, or the "restricted" sample that drops respondents who are younger than the age of eligibility for obtaining the ID. APPENDICES | 59 APPENDIX 4. EXCLUDED COUNTRIES The 2018 ID4D Global Dataset provided information on 198 countries,55 but used only 151 countries for the global ID coverage estimates due to a lack of reliable data on adult coverage for HICs and some instances of missing (proxy) data for children and/or adults. For the 2021 estimates, our goal has been to maximize the number of countries included, which has been possible with the addition of Findex data and the change in methodology to use birth registration rates (BRR) for higher-income countries for which we do not have Findex or administrative data. This Appendix specifies those countries excluded in 2021 and discusses the removal of the 2018 exclusion criteria in more detail. 2021 EXCLUSIONS Four of the 198 countries covered by the ID4D dataset are excluded from the 2021 estimates (Table 14). Eritrea; Taiwan, China; and the Federated States of Micronesia are excluded because they do not have birth registration data from the UNICEF or UNSD datasets. In addition, Somalia is excluded because its under-5 BRR is extremely low (5.9 percent) and we do not have alternate coverage metrics for adults. Because Somalia does not have a national ID or similar foundational system, adults use a variety of identity documents in their daily lives; while data suggests that coverage of these is low, it is likely to be higher than 5.9 percent.56 Therefore, we exclude Somalia to avoid inflating the ID coverage gap by applying the BRR to the entire population. Table 14. Countries Excluded from 2021 Estimates Under-5 ID Ownership Voter Reg. 2021 Global Country Income Birth Reg. (%)1 (%)2 (%)3 Population4 Pop. (%) Eritrea LIC – – – 3,620,312 0.046 Taiwan, China HIC – 98.8 98.6 23,859,912 0.302 Micronesia, Fed. Sts. LMIC – – 100 113,131 0.001 Somalia LIC 5.9 – – 17,065,581 0.216 Total 44,658,936 0.565 Notes: LIC = low-income; LMIC = lower-middle income; HIC = high-income. 1 UNICEF (2022). 2 ID4D-Findex (2021). 3 IDEA (2022). 4 UNDESA (2022a). 55 The following territories and jurisdictions that may have some of the data sources used in this paper have not historically been included in the ID4D Global Dataset or coverage estimates given their small population size, lack of a distinct ID system, and/or lack of data: Åland Islands, American Samoa, Anguilla, Aruba, Bermuda, Bonaire, Sint Eustatius and Saba, British Virgin Islands, Cayman Islands, Channel Islands, Cook Islands, Curaçao, Falkland Islands (Malvinas), Faroe Islands, French Guiana, French Polynesia, Gibraltar, Greenland, Guadeloupe, Guam, Holy See, Isle of Man, Martinique, Mayotte, Melanesia, Montserrat, Netherlands Antilles, New Caledonia, Niue, Northern Mariana Islands, Polynesia, Puerto Rico, Réunion, Saint Helena, Saint Pierre and Miquelon, Sint Maarten, St. Martin, Svalbard and Jan Mayen Islands, Turks and Caicos Islands, US Virgin Islands, Wallis and Futuna Islands, and Western Sahara. 56 Previous editions of the dataset used voter registration from Somalia’s 1984 elections (available in the IDEA, 2022). However, these numbers do not provide an accurate picture of current ID ownership in the country (for example, even if all 1984 voters were issued a voter card, this would only be held by the portion of the population 55 or older in 2021). 60 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES Together, these countries represent less than one percent of the global population, so their omission is not likely to have major effects on the estimates, particularly because the largest of the three—Taiwan, China— has a near universal ID ownership rate. However, we can use the mean ID ownership rates for low-income countries (68.5 percent) to estimate the order of magnitude. If we applied this rate to the total populations in Eritrea and Somalia, this would increase the global estimate of people without ID by approximately 6.5 million. 2018 CRITERIA As noted above, 47 countries were excluded from the 2018 coverage estimates due to lack of reliable data on adult ID coverage. These exclusions were made due to a lack of data and to avoid over-inflating the estimated number of people without ID by over-reliance on voter registration rates, which are typically lower than assumed ID ownership rates in high-income countries (HICs) in particular. Two criteria were applied: 1. HICs with BRRs greater than 99 percent. A total of 44 HICs without administrative data and with birth registration rates greater than 99 percent were excluded to avoid overestimating the ID coverage gap by applying the voter registration rate.57 2. Countries with no foundational ID and BRRs greater than 95 percent. Similarly, non-HIC countries without foundational ID systems—and therefore, no administrative data—were excluded from the global estimates if they had a BRR higher than 95 percent; in 2018, this criterion was applied to two countries (Turkmenistan and Marshall Islands).58 In addition, China was excluded as neither administrative nor voter registration data was available to measure adult ID coverage in 2018. In 2021, we now have Findex data as a primary source of adult ID coverage for many of the 47 countries excluded in 2018 estimates, including a number of HICs, China, and Turkmenistan. Table 15 shows the remaining countries that would continue to meet the 2018 criteria.59 For this paper, we have taken a different approach to maximize the number of countries included in the global coverage estimates while also minimizing the likelihood of over- or underestimation. Rather than excluding them from the estimates, we apply under-5 birth registration rates for both the adult and child populations in these countries. Given reported birth registration rates of 100 percent in most of these countries, this does not affect the estimated number of people without ID globally; however, it means that the ID coverage estimates reflect data from a larger share (over 99 percent) of the global population. 57 This criterion was initially adopted to avoid overestimating the ID coverage gap in HICs like Australia, Germany, France, Switzerland, and the United States, where voter registration rates would often suggest a 10 percent or even larger ID coverage gap, even though the true share of adults without proof of identity is believed to be significantly lower. Had these countries been included using voter registration in 2018, the global ID coverage gap would have been about 80 million higher (i.e., closer to 1.1 billion). 58 See Appendix 5, Errata for exceptions. 59 Updated data for the Marshall Islands indicate a BRR of 83.8 percent (UNICEF 2022a)—compared with 95.9 percent from the 2007 DHS that was used in the 2018 dataset—so it no longer meets the second exclusion criterion. APPENDICES | 61 Table 15. Countries that Would Meet 2018 Exclusion Criteria in 2021 Under-5 Birth Voter Reg. 2021 Country Income Foundational ID Reg. (%)1 (%)2 Population3 Andorra HIC – 100 43.0 79,034 Australia HIC – 100 83.1 25,921,089 Austria HIC Identity Card 100 85.1 8,922,082 Brunei Darussalam HIC Smart Identity Card 99.9 – 445,373 Finland HIC Identity Card 100 100.0 5,535,992 Ireland HIC – 100 93.4 4,986,526 Japan HIC Individual Number 100 98.7 124,612,531 Card Liechtenstein HIC Identity Card 100 63.3 39,040 Luxembourg HIC Identity Card 100 53.0 639,321 Monaco HIC Monegasque 100 23.0 36,686 Identity Card New Zealand HIC – 100 90.8 5,129,728 Oman HIC Omani ID Card 100 22.3 4,520,471 Notes: HIC = high-income. 1UNICEF (2022a), 2IDEA (2022), 3UNDESA (2022a). Importantly, it is also extremely unlikely that any country has achieved perfect continuous coverage of birth registration or the adult ID system. However few they may be, people in high- and upper-middle-income countries without proof of their identity are likely to face significant barriers to political, economic, and social participation, and to be among the most marginalized.60 For these reasons, it is important to continue tracking progress on access to identity credentials across all countries. We hope that including more HICs in the Findex data collection is the first step toward this goal, and highlights that the need for better data is worldwide. 60 See, for example, a report from University of Sydney (2016) on significantly lower birth registration rates among Aboriginal groups in Australia. 62 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES APPENDIX 5. SUMMARY OF CHANGES FROM 2018 METHODOLOGY CHANGES Table 16 summarizes broad changes to the global coverage estimate methodology from the 2018 version of the dataset, and the general impact that this has on the 2021 estimates, using 2021 data. As a result of these changes, the 2018–2021 estimates should not be considered as a time series—i.e., the differences between 2018 and 2021 represent a mixture of methodology changes and changes in ID coverage. See Appendix 3 for more analysis of potential alternative methodologies and the impact of methodological changes. Table 16. Methodology Changes for Global Estimates, 2018-2021 Effect on 2021 Parameter 2018 Method 2021 Method Rationale for Change Coverage Gap Estimates Sources for adult Primary: Primary: Survey data Findex survey offers Around 70 million less ID coverage Administrative from Findex; improved measurement over than they would have data from ID4D Secondary: administrative data and now been if Findex and questionnaire Administrative has sufficient coverage to be BRR were not used for Secondary: data from ID4D the default measure of adult ID adults voter registration questionnaire coverage. Using BRR for HICs rates Tertiary: voter and UMICs, where BRR has registration rates been historically much higher, (LICs and LMICs); is a better proxy for adult ID birth registration rates coverage than VRR. (UMICs and HICs) HICs with BRR Exclude from Include in estimates To maximize inclusion of Around 50 million greater than estimates when countries for better global more that it would 99.99% and administrative representation; with more have been if these Countries with data is unavailable Findex and administrative criteria were still no ID and BRR (see Appendix 4) data, and the use of BRR applied greater than 95% for adults in HIC and UMICs, there is less concern about underestimating coverage. Calculating Subtract the Calculate Enable incorporation About 2 million administrative total number administrative rates of multiple years of lower than if total coverage of people by dividing the number administrative data while registrations were registered in the of people registered accounting for population subtracted from the ID system from in the ID system by growth; create parallel metrics 2021 population total population the population above to match birth registration, above the cutoff the cutoff age in the ID ownership, and voter age in the dataset data year. (2019, 2020, registration rates. year (2018). 2021, or 2022). These rates are then applied to (multiplied by) the adult population in the dataset year (2021). APPENDICES | 63 Effect on 2021 Parameter 2018 Method 2021 Method Rationale for Change Coverage Gap Estimates Cutoff age for Cutoff age equals Cutoff age equals ID Ensure BRR is the primary About 10 million less administrative ID age. If people age or 5, whichever metric used for all children than if the cutoff data are eligible to is higher. Rather than under 5. age was kept at 0 register at 0 years applying the 0+ admin for countries with ID only, the rate to the entire registration from birth administrative data population, BRR is is applied to the always applied to the entire population 0–4 age group. Calculating voter Subtract the Calculate voter rates Improve accuracy of estimate About 10 million registration total number of by dividing the number by accounting for population less than if we had coverage registered voters of registered voters by growth; create parallel metrics subtracted the number in the election the population above to match birth registration of registered voters year (e.g., 2016) the voting age in the rates and ID ownership rates. from the 2021 adult from population election year. population above the voting These rates are age in the dataset then applied to year (2018). (multiplied by) the adult population in the dataset year (2021). Countries with Take the lower Take the mean of the To better account for Slight decrease in UNSD BRR number of the range (e.g., 94.5) uncertainty and avoid number of people recorded as range (e.g., 90%) underestimating coverage. without ID a range (e.g., 90–99%) ERRATA When replicating the 2018 methodology and using this data for cross checks, a few errors and updates were identified: 1. Belize. Belize does not have a foundational ID system for adults and had a BRR of 95.7 percent as of 2015. It therefore met the exclusion criteria applied in 2018 but was included in the calculations. This had little impact on the overall totals, adding 43,994 to the global estimate of people without ID. 2. Turkmenistan. In 2018 and prior ID4D Datasets, Turkmenistan was classified as not having a foundational ID system for adults. Because it had a birth registration rate of 99.6 (MICS 2015-2016) it was excluded under criteria two (see above). However, Turkmenistan has a “domestic passport” system that issues widely-held credentials commonly used as proof of official ID, and in the 2021 Dataset is classified as a foundational system. 3. Corrected ID ages. We have revised the ID mandatory and/or minimum eligible ages for 11 countries thanks to better quality data. 64 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES APPENDIX 6. ID4D FINDEX QUESTIONS The following ID-related questions were included in the 2021 Findex survey and asked in 113 countries. For a full discussion of the Findex methodology and data characteristics, see Demirgüç-Kunt et al. (2022). FIN46. Do you personally have [a/an insert local terminology for foundational ID]? CIRCLE ONE RESPONSE: Yes 1 No 2 (I have applied, but not yet received it) 3 (DK) 4 (Refused) 5 (If code 1 FIN46/WP21399, Skip to FIN50/WP21414; If code 3 in FIN46/WP21399, Skip to FIN49/Text before WP21408; Otherwise, Continue) FIN48. Please tell me whether each of the following is A REASON why you do not have [a/an insert local terminology for foundational ID]? Is it because…? (Read Items, FIN48A first then random order) Yes No (DK) (Refused) FIN48A You have another form of identification issued by 1 2 3 4 the government FIN48B You don't need an ID for any purpose 1 2 3 4 FIN48C It is too expensive 1 2 3 4 FIN48D You don't have the necessary documents 1 2 3 4 FIN48E You need to travel too far to apply 1 2 3 4 FIN48F You do not feel comfortable giving your personal 1 2 3 4 information APPENDICES | 65 FIN49. Have you ever had any difficulty doing any of the following because you did not have [a/an insert local terminology for foundational ID]? (Read Items, random order) Yes No (DK) (Refused) FIN49A Receiving financial support from the government 1 2 3 4 FIN49B Using financial services 1 2 3 4 FIN49C Obtaining a SIM card/mobile phone service 1 2 3 4 FIN49D Participating in elections 1 2 3 4 FIN49E Applying for a job 1 2 3 4 FIN49F Receiving medical care 1 2 3 4 66 | ID4D GLOBAL DATASET 2021: GLOBAL ID COVERAGE ESTIMATES APPENDIX 7. ID AUTHORITY QUESTIONNAIRE As part of the 2021 ID4D Global Dataset, we fielded a questionnaire to ID agencies responsible for the primary foundational ID system to collect and validate key information about these systems. Section A of this questionnaire included requests to provide administrative data on (1) the number of people registered in the system by age group and gender, and (2) the number of people who had been issued with the primary credential (e.g., a national ID card). The remaining sections concerned the features of the system’s credential(s), registration process, and data management. These questions and the responses received will be featured in a forthcoming paper. These questionnaires were fielded by the ID4D team and World Bank country offices between August 2021 and May 2022, including a follow-up period to clarify and validate responses.61 The questionnaires were prepared in English and translated to one of the other World Bank Group official languages (Arabic, Chinese, French, Portuguese, Russian, and Spanish) and other languages, on demand, when appropriate for the country context. The parts of the questionnaire pertaining to ID coverage are as follows: Please answer the Questionnaire about the [SYSTEM NAME] (henceforth “ID system”) and [CREDENTIAL NAME] (henceforth “ID”), which we understand to be the primary ID system and government-issued credential for adults in [COUNTRY]. If this is incorrect, or there are other related systems or identity credentials managed by your agency or department, please provide this information below: The questions in this section are intended to understand the coverage of the [SYSTEM NAME] across [COUNTRY]’s existing resident population. If possible, please provide figures that include only unique, living persons currently residing in the territory, excluding (a) deceased persons and/or (b) those who currently reside outside of the territory. If this is unknown, or these figures include deceased persons or non- residents, please note this information under the “2021 Update” column. 61 As a result of the data collection period and variation in reporting between countries, the administrative data received in these questionnaires represents a range of dates between 2020 (month not specified) and April 2022. APPENDICES | 67 PREVIOUS INFORMATION 2021 UPDATE QUESTIONS Provided in 2019 or earlier Please provide updated figures A1. What is the total number of [Previous figure included here, if known] people currently registered in the system? Please also note the date (e.g., July 31, 2021) as of which this data is correct A2. What is the total number of people registered by gender? Female [Previous figure included here, if known] Male [Previous figure included here, if known] Other [Previous figure included here, if known] A3. What is the total number of people registered by age group? Less than 5 years old (0-4) Unknown 5 - 17 Unknown 18 - 25 Unknown 26 - 65 Unknown Over 65 Unknown A4. Of those registered, how [Previous figure included here, if known] many have been issued with a/n [CREDENTIAL NAME]? APPENDIX 8. FINDEX REGRESSION RESULTS Table 17 below shows the results of logit regression models using the 2021 ID4D-Findex survey data to predict ID ownership based on demographic covariates, illustrated in Figure 6. See Methodology and Appendix 6 for a more detailed description of the variables. Table 17. Predictors of ID Ownership Respondent Has an ID ID Ownership <90 Percent Low-Income Countries (LICs) (1) (2) (3) (4) Female -0.155*** -0.214*** -0.165** -0.208*** (0.040) (0.042) (0.060) (0.061) Bottom 40% of income -0.295*** -0.265*** -0.474*** -0.407*** (0.040) (0.043) (0.061) (0.062) Out of workforce -0.301*** -0.384*** -0.422*** -0.454*** (0.043) (0.046) (0.067) (0.067) Primary school or less -0.606*** -0.603*** -1.079*** -0.954*** (0.043) (0.046) (0.068) (0.069) Under 25 -1.335*** -1.328*** -1.182*** -1.176*** (0.045) (0.048) (0.070) (0.070) Unmarried -0.317*** -0.370*** -0.243*** -0.314*** (0.044) (0.046) (0.066) (0.066) Rural -0.298*** -0.507*** (0.045) (0.069) Constant 3.599*** 3.897*** 4.048*** 4.337*** (0.141) (0.147) (0.157) (0.166) Country Fixed Effects Y Y Y Y Observations 29,931 24,919 10,787 10,787 Log Likelihood -13,504.860 -11,576.380 -5,331.762 -5,291.130 Akaike Inf. Crit. 27,083.730 23,218.750 10,697.520 10,618.260 Note: *p<0.05; **p<0.01; ***p<0.001 Logit models using survey weights and design-based standard errors. Models 1-2 are restricted countries with ID coverage less than 90 percent; models 3-4 include low-income countries (LICS) only. Includes respondents ages 15 plus who are also over age of eligibility to obtain the ID. Income groups are based on the World Bank’s 2021 classification. Information on rural vs. urban location only available for the subset of countries where data collection was done face-to-face. Source: ID4D-Findex Data (2021). APPENDICES | 69