The undercounting of child-mother births

Chauvin, Juan Pablo; Rubião, Rafael; Talamas Marcos, Miguel Ángel Working Paper The undercounting of child-mother births IDB Working Paper Series, No. IDB-WP-1665 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Chauvin, Juan Pablo; Rubião, Rafael; Talamas Marcos, Miguel Ángel (2025) : The undercounting of child-mother births, IDB Working Paper Series, No. IDB-WP-1665, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0013407 This Version is available at: https://hdl.handle.net/10419/315931 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ The Undercounting of Child-Mother Births J uan Pablo Chauvin Rafael Rubião Miguel Ángel Talamas Marcos WORKING PAPER No IDB-WP-1665 InterA merican Development Bank Department of Research and Chief Economist February 2025 * InterA merican Development Bank ** University of California, Los Angeles The Undercounting of Child-Mother Births J uan Pablo Chauvin* Rafael Rubião** Miguel Ángel Talamas Marcos* InterA merican Development Bank Department of Research and Chief Economist February 2025 Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Chauvin, Juan Pablo. The undercounting of child-mother births / Juan Pablo Chauvin, Rafael Rubião, Miguel Ángel Talamas Marcos. p. cm. — (IDB Working Papers Series ; 1665) Includes bibliographic references. 1. Pregnant teenagers-Census-Brazil. 2. Pregnant teenagers-Census-Mexico. 3. Pregnant teenagers-Census-United States. 4. Census-Costs-Brazil. 5. Census-CostsMexico. 6. Census-Costs-United States. 7. PopulationStatistics-Brazil. 8. Population-Statistics-Mexico. 9. Population-StatisticsUnited States. I. Rubião, Rafael. II. Talamas, Miguel. III. Inter-American Development Bank. Department of Research and Chief Economist. IV. Title. V. Series. IDB-WP-1665 http://www.iadb.org Copyright ©2025 Inter-American Development Bank ("IDB"). This work is subject to a Creative Commons license CC BY 3.0 IGO (https://creativecommons.org/licenses/by/3.0/igo/legalcode). The terms and conditions indicated in the URL link must be met and the respective recognition must be granted to the IDB. Further to section 8 of the above license, any mediation relating to disputes arising under such license shall be conducted in accordance with the WIPO Mediation Rules. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the United Nations Commission on International Trade Law (UNCITRAL) rules. The use of the IDB's name for any purpose other than for attribution, and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this license. Note that the URL link includes terms and conditions that are an integral part of this license. The opinions expressed in this work are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. Abstract∗ Accurate demographic data are essential for effective policy design, yet private costs may deter individuals from truthfully reporting sensitive information. We examine this market failure and its implications in the context of child motherhood. Using administrative records from Brazil, Mexico, and the United States, along with census data from 59 countries, we identify systematic patterns of under-reporting, indicating that child motherhood is significantly more prevalent than previously thought. Births to mothers aged 10-14 are often missing from contemporary administrative records but appear in censuses conducted a decade later, with under-counting in birth registries reaching 20-30% in Brazil, Mexico, and the United States. We introduce a model where reporting decisions weigh instrumental benefits against age-dependent private costs, yielding predictions that align with observed patterns: truthful reporting increases markedly with the mother’s age, under-reporting of child-mother births decreases with the time elapsed between data collection and childbirth, and retrospective census estimates generally provide more accurate birth counts than contemporary administrative records for this age group, but not for older mothers. Our findings suggest that social costs, rather than fear of legal repercussions, are a primary driver of under-reporting. JEL classifications: J13, D10, J18, D82 Keywords: child motherhood, private information, social norms. ∗ Chauvin: Inter-American Development Bank: [email protected]. Rubião: UCLA Anderson: rafael.r[email protected]. Talamas Marcos: Inter-American Development Bank: [email protected]. We thank Giovanna Chaves and Juan Ferrer for their excellent research assistance, Livia Almeida, Michael Cardona, Shruthi Nagamani, and Vinícius Princiotti for additional research support, and two anonymous reviewers for useful comments. The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. 1 1 Introduction Accurate demographic and health data are a valuable public good critical for effective resource allocation, informed policy responses, and advocacy that addresses social problems. Producing this public good relies on institutional capacity and, at times, heavily on individuals truthfully reporting information. However, when reporting incurs private costs, such as legal exposure or social stigma, individuals may choose to withhold information or report it inaccurately, imposing a negative externality on society by reducing the overall availability and quality of this public good (Blank et al.,2009;Duflo et al.,2013). We study this asymmetric information market failure in the context of an important social issue: childmother births. Extensive research shows that childbearing at a young age harms the well-being of both mother and child (e.g., Duflo 2012;Kearney and Levine 2015;Lang and Weinstein 2015). This is especially true for younger mothers (Aizer et al.,2022), whose pregnancies are often unwanted and a result of sexual violence. Hence, accurate data on the prevalence of child motherhood is essential to understanding the magnitude of the issue at hand and keeping track of its evolution. However, this age group, for which accurate data are essential, is also likely the group for which the asymmetric information problem is the largest because social or legal repercussions tend to intensify the younger the mother is at childbirth. To guide our analysis, we propose a simple framework for the decision to report a birth with two reporting methods: administrative records and census interviews. Mothers can report births using either or both of the methods or choose not to report. Their decision depends on whether the benefits of reporting outweigh the associated costs. Young mothers, in addition to facing instrumental costs (e.g., fees, travel, or time required to report), may also encounter "social costs" (such as stigmatization or ostracism) and "judicial costs" (if the sexual encounter leading to the pregnancy is considered a crime).1These social and judicial costs decrease as the mother ages and more time passes between the birth and the mother’s age at reporting. Motivated by the testable predictions of our framework, we investigate the magnitude and drivers of the under-reporting of child motherhood in administrative records and censuses from over 50 countries. We begin by estimating the extent of under-reporting of child-mother births in administrative records in three large countries: Brazil, Mexico, and the United States. These countries have significant population sizes and the availability of comprehensive administrative and census data. By comparing birth counts from administrative records with those reconstructed from census data collected up to 15 years after childbirth, 1 Even if direct penalties affect only the father, the child-mother may experience a direct disutility from this legal exposure, either because they share an emotional, financial, or familial bond with the father, because they are vulnerable to verbal or physical retaliation from him, or because going through this legal process generates stress and suffering for the victim. 2 we can identify patterns of under-reporting across different age groups at the time of childbirth. In our analysis, we expect that, in the absence of undercounting in birth registries, the number of births recorded in those registries will exceed the births inferred from subsequent censuses because census counts depend on children living with their mothers, which is not always the case and becomes less common as the children grow older. This prediction holds for mothers aged 15 or older at childbirth. However, for mothers aged 14 or younger at childbirth, we find that in all three countries, there are more births based on the estimates from the census 10 years after childbirth than in birth registries. These gaps, which represent lower bounds for under-reporting in birth registries, are substantial: 20% in Brazil (1999), 22% in Mexico (2009), and 29% in the United States (2009). Next, we explore how child-mother birth reporting changes with the mother’s age at the time of reporting. We leverage two complementary sources of variation. First, we compare birth counts for the same cohort of mothers recovered from subsequent census rounds. By comparing reporting of the same births across different census rounds, we hold the reporting method constant while varying only the mother’s age at reporting. This comparison reveals that while birth counts for teen and young-adult mothers (aged 15-24 at delivery) are similar across census rounds, counts for child mothers are substantially lower in the baseline census—when mothers are still in their teens—compared to a decade later when they are in their twenties. Figure 1: Estimated Birth Counts Based on Subsequent Censuses Notes: The sample comprises 80 pairs of national censuses from 59 countries, with one observation per census pair, age of the mother at birth and the year of birth. We estimate the number of births by identifying the age at birth of the reported mother. Samples exclude children of foreign citizenship. Figure 1displays this striking pattern using 78 census pairs from 59 countries: while birth count estimates for the same cohort of teen and young-adult mothers are comparable across subsequent censuses (left side), 3 births to mothers aged 14 or younger (right side) show systematic under-reporting in the baseline censuses worldwide. This evidence is consistent with households not acknowledging the birth to a mother 10 to 14 years old if the mother is still in her teens at the time of the census but acknowledging it once the mother is in her twenties at the time of the census. This misreporting could take several forms, for example, omitting the presence of the child at the time of the interview or attributing the maternity of the child to another household member, for which we find evidence and later discuss. The magnitude of under-reporting is inversely related to the mother’s age at birth: baseline censuses miss approximately 65% of births to mothers aged 13 or younger, 50% of births to 14-year-old mothers, and 22% of births to 15-year-old mothers when compared to counts based on subsequent censuses. This pattern is remarkably consistent across countries and most pronounced for births occurring in the year immediately preceding the census, with the reporting gap diminishing as the time between birth and census increases. These findings suggest that previous estimates of under-registration of births—based on the possession of birth certificates among children reported in surveys and censuses occurring within a few years from childbirth—may significantly understate the true magnitude of the prevalence of child motherhood. To investigate further how child-mother birth reporting changes with the mother’s age at the time of reporting, we used Mexican administrative data from 1995 to 2020, where delayed information is available. We find that, while births to mothers aged 15-24 are typically registered within a year (80-90% of cases), births to mothers aged 10-14 show significantly longer delays, with only 50-80% registering within a year of the birth. The average registration delay decreases sharply with the mother’s age, from 40 months for 12-year-old mothers to 10 months for 15-year-old mothers. Both patterns are consistent with reporting costs—whether social or judicial—declining as mothers age. Finally, we investigate three key mechanisms underlying the observed under-reporting patterns. First, we find evidence that many births to child mothers are not entirely concealed but rather misattributed to older women in the household, particularly those beyond reproductive age. Comparing birth counts across census rounds reveals that women aged 45-55 report 80-200% more births in baseline censuses than in subsequent ones—a gap far exceeding what mortality rates alone could explain. This suggests families often report these births but attribute them to grandmothers or aunts. Second, to assess the importance of judicial costs in the reporting decision, we exploit variation in minimum age of consent laws across countries and the timing of the pregnancy relative to the mother reaching the minimum age of consent. Using a regression discontinuity design, we find no evidence of discontinuous increases in birth reporting when mothers reach the age of legal consent, suggesting that fear of legal consequences for the abusers may not be a substantial driver of under-reporting. 4 Third, we examine how under-reporting varies with social norms surrounding female sexuality and reproductive rights across countries. Using data from UN Demographic and Health Surveys on women’s autonomy in sexual decision-making, we find that, controlling for income levels, a one standard deviation increase in the share of women reporting autonomous sexual decisions is associated with a 20% reduction in under-reporting of child-mother births. This paper contributes to three bodies of research. First, we make a contribution to the literature on child and teenage motherhood. Prior studies have examined the causes and consequences of fertility among teenage girls, showing that it is associated with negative outcomes for both mothers and their children (Corbacho et al.,2012;Duflo,2012;Kearney and Levine,2015;Klepinger et al.,1999;Lang and Weinstein, 2015). This social problem has attracted considerable policy attention, and the United Nations included reducing fertility rates among girls aged 10-19 as part of its Sustainable Development Goals (SDGs). However, both the academic literature and data used for policy primarily focus on fertility among girls aged 15-19, for whom reliable statistics are widely available. In contrast, fertility data for girls younger than 15 remain scarce and incomplete (Schoumaker and Sánchez-Páez,2022;WHO,2024). We contribute by producing lower-bound estimates of under-reporting for 59 countries, demonstrating that this phenomenon is more widespread and more prominent than previously thought, also affecting highand middle-income countries. Importantly, we show that under-reporting occurs not only in census surveys but also in birth registries, which are often considered the most reliable data source for birth counts (Kearney and Levine,2012). Second, we extend prior research on the economics of information and public goods. Previous studies have shown that private costs can prevent individuals from truthfully disclosing socially valuable information in a variety of contexts, such as auditors reporting firms’ pollution levels (Duflo et al.,2013), schools reporting student enrollment to the government (Sandefur and Glassman,2015), and self-employed individuals declaring their income to tax authorities (Hurst and Pugsley,2010). In the context of vital statistics, Blank et al. (2009) show that when early marriage is outlawed, individuals have incentives to misreport, leading to administrative records being an inferior data source compared to retrospective census data for studying the effects of age-of-marriage laws. Our study finds that administrative records can under-report child-mother births, even in the absence of legal repercussions, when individuals face non-pecuniary costs stemming from informal norms and social pressure. Finally, we contribute to the literature on social norms. Prior research has shown that concerns for reputation and social image can be as influential as material incentives (Bénabou and Tirole,2006;Butera et al., 2022;Daughety and Reinganum,2009). Social pressure can encourage socially beneficial behavior in contexts such as voting (Ali and Lin,2013), workplace safety (Johnson,2020), saving (Breza and Chandrasekhar, 5 with their mothers, as questions about non-resident children are rarely included. To mitigate this concern, we estimate births only for groups where the child would be at most 15 years old at the time of the second measurement, increasing the likelihood that the child is still living with the mother. However, even with this adjustment, the second measurement of births, taken years later, is likely to be lower due to older children being less likely to reside with their mothers. Factors such as death, migration, and marriage dissolution also contribute to this decline. As a result, our estimates of births under-reporting represent a lower bound of the actual under-reporting. 4 Estimation We turn now to the empirical analysis of child-mothers’ birth reporting. Our goal is to gauge the magnitude of under-reporting and investigate, in light of our model, how the mother’s age at birth and at reporting influences reporting behavior. 4.1 Under-reporting of Child-Mother Births in Administrative Records We begin by estimating the extent of under-reporting in administrative birth records for child mothers (ages 10 to 14) in Brazil, Mexico, and the United States—three large countries where the coverage of civil registration and vital statistics (CRVS) is considered complete or nearly complete (Kisambira and Schmid,2022). We do this by comparing administrative birth counts with birth counts obtained from census data collected 11 years later. 7 Table 1reports the counts from administrative records (column 1) and the subsequent census estimates (column 2). It also includes our estimate of unrecorded births as a percentage of the census counts (column 3) for two groups: child mothers (ages 10 to 14) and teen and young adult mothers (ages 15 to 24). 8 A positive value in column 3 indicates under-reporting in administrative records relative to the census. In contrast, a negative value suggests that administrative records exceed the census estimates, which would be expected in the absence of social or judicial costs associated with reporting because not all children cohabit with their birth mothers. The table displays a pattern of sizable under-reporting for child mothers in the administrative records 7 Census counts are computed for the year prior to the most recent available census in each country: 2009 for Brazil and 2019 for Mexico and the United States. This is because census data do not capture births occurring in the same year after the census collection dates. The base year is chosen to be 11 years prior. 8 These groupings follow the standard age-brackets classification used by the United Nations and other providers of demographic statistics (United Nations DESA,2017) 12 Table 1: Under-reporting in Administrative Records Relative to Subsequent Census (1) (2) (3) Administrative Subsequent Administrative Records Census Underreporting (2)−(1) (2) Brazil (1999) Child mothers (ages 10 to 14) 27518 34372 0.20 Teen mothers (ages 15 to 24) 1737520 1244020 -0.40 Mexico (2009) Child mothers (ages 10 to 14) 9739 12558 0.22 Teen mothers (ages 15 to 24) 1136376 841056 -0.35 United States (2009) Child mothers (ages 10 to 14) 5038 7112 0.29 Teen mothers (ages 15 to 24) 1417630 839953 -0.69 Note: The table provides estimates of the number of births by age group at the time of delivery. The administrative counts reflect all births recorded within the first year after delivery, as captured in each country’s administrative registry. The "subsequent census" counts are derived from the ages of children and their mothers in the microdata from the census conducted 10 years later. The percentage of unrecorded births is calculated as the difference between the administrative and census counts, expressed as a proportion of the census counts. across the three countries. In Brazil, the number of births to child mothers recorded in administrative data is 20% lower than those captured by the subsequent census. The estimated under-reporting is 22% in Mexico and 29% in the United States. This contrasts sharply with the pattern observed for teen and young adult mothers (ages 15 to 24), where administrative records exceed census birth counts by substantial margins: 40% in Brazil, 35% in Mexico, and 69% in the United States. These patterns are evident in Figure 2, which plots the estimated birth counts by age group and year using census and administrative records data for Brazil, Mexico, and the United States. For teenage and young adult mothers, on the left side, the birth counts from the registries are always considerably higher than in the censuses—as expected due to the census limitations discussed. However, for child mothers, on the right side of Figure 2, the birth counts estimated using a census occurring 10 or more years in the future are significantly higher than that of the birth registries. For example, based on the 2020 Mexican census, there were 14,000 births to mothers aged 10 to 14 in 2010; however, in the administrative records, this number is less than 10,000. As we get closer to the census interview date (moving to the right on the x-axis), mothers are younger at the time of reporting, and the 13 (a) Births in Brazil by mother age group (1985-2020) (b) Births in Mexico by mother age group (1985-2020) (c) Births in the USA by mother age group (1985-2019) Figure 2: Total Births by Mother’s Age Group According to Census and Registry Data Note: The figure presents annual birth estimates derived from national censuses and birth registries for Brazil, Mexico, and the United States. The left panels represent mothers aged 15 to 25, while the right panels depict those aged 10 to 14. Panel (a) illustrates Brazilian birth estimates, utilizing imputed data from the 2000 and 2010 Censuses in conjunction with the Brazil Live Birth Information System (SINASC). Panel (b) portrays Mexican birth estimates, employing data from the 2010 and 2020 Censuses supplemented by the birth registry. Panel (c) exhibits U.S. birth estimates using data from the 2000 Census, the American Community Surveys (IPUMS) of 2000, 2010, 2019, and the birth registry. The 2000 survey randomly sampled 0.13% of the population, while the 2010 and 2019 surveys sampled 1%. Data from immigrant children in the United States and information from 2020, potentially affected by COVID-19 pandemic-related reporting delays, are excluded. 14 under-reporting problem increases to a point where the census recovers a lower estimate than that of the registries. This generates a single-crossing in the estimates between the census and the birth registries, which is the prediction of proposition 4. The main disadvantage of this methodology is that the "true" census count for a year close to a census (e.g., 2019) will only be available when the 2030 census is conducted. These estimates and patterns demonstrate that the under-registration of births—a well-documented issue affecting many countries, especially in the developing world (United Nations DESA,2020a)—is strikingly more pronounced among child mothers. The extent of this bias is greater than previously recognized. Using questions on registration from the Demographic and Health Survey (DHS) for five Latin American countries, Duryea et al. (2006) find that the likelihood of children under five lacking a birth certificate increases by 4.5 to 6.5 percentage points when the mother is a teenager, relative to an average of 14.5%, in Bolivia, Colombia, and Peru (but not in Brazil or Nicaragua). Similarly, Ebbers and Smits (2022), using DHS data from 40 Sub-Saharan African countries, find that children born to mothers younger than 18 have 6% lower odds of being registered. 4.2 Mothers’ Age and Reporting This section explores how the age of the mother at childbirth and at the time of reporting affects the reporting behavior. We first discuss evidence based on administrative records and then evidence based on censuses. Administrative Records Evidence If the social cost of reporting decreases with the mother’s age at childbirth and the time between reporting and the birth, it should influence reporting behavior in administrative records in two ways. First, if the mother can somewhat control the timing of reporting, we would expect delays. These delays would happen because, at the time of birth, the social cost is so high that the optimal choice is not to report. However, as suggested in Proposition 2, as time progresses and this cost decreases, it may become beneficial to report. Second, the duration of the reporting delay is a function of the social cost, which, according to Proposition 1, declines with the mother’s age at childbirth. Therefore, we would expect that, conditional on eventual reporting, this delay would be longer the younger the mother was at the time of birth. Among the administrative registries considered in this study, only Mexico provides detailed data on delayed registration, allowing us to estimate reporting delays for different maternal age groups and birth cohorts. We summarize these data in Figure 3. Panel (a) shows the share of children of mothers aged 15-24 15 (a) Share of Births Registered by Time Span since Year of Birth (b) Average delay in registration (1995-2019) Figure 3: Birth Registration Delays in Mexico, by Mother’s Age Group Notes: This figure uses birth registry data from Mexico (INEGI). Panel A plots the share of births that took place in each year that were registered up to December 2019, by maximum delay allowed. The 1 (5, 10, 20) year(s) delay line includes births occurred from January to December of each year, registered up to 12 (60, 120, 240) months from birth. For the 1 (5, 10, 20) year(s) delay line, we only plot data up to the year 2018 (2014, 2009, 1999), which corresponds to the period that we fully observe data for the next 12 (60, 120, 240) months. Panel B plots the average delay in birth registration across the age of the mother at birth in Mexico (1995-2019), conditional on the children being registered up to 15 years after birth. (left panel) and 10-14 (right panel) registered within one, five, ten, or twenty years of their birth date. 9 We expect some level of delay in all registrations due to costs such as fees or transportation. These delays have decreased over time as the Mexican government has expanded access to civil registry services and incentivized registration. For mothers aged 15-24, between 1995 and 2005, just over 80% of births to this age group were registered within one year, rising to 90% between 2010 and 2018). 9 For consistency with our 2019 data cutoff, births with up to 1 year delay are shown only through 2018. This ensures all 2018 births are captured within the 1-year window without requiring 2020 data. Similarly, births with up to 5, 10, and 20 years of delay are shown through 2014, 2009, and 1999, respectively. 16 In contrast, for child mothers aged 10-14, registering within one year was substantially less common, ranging between 50 and 60% in the earlier period (1995-2005), improving to just above 80% by 2018. Panel (b) of Figure 3illustrates how the average delay in birth registration decreases with the mother’s age, dropping from approximately 40 months for 12-year-old mothers to around 10 months for 15-year-old mothers, after which the delay stabilizes. These findings align with the work of Keskin and Çavlin (2020) in Turkey, who found that, while 98% of births to adolescent mothers were eventually registered between 2011 and 2015, only 78% were registered within the first 30 days, compared to 95% for all births. Census Evidence: Brazil, Mexico, and the United States Census surveys differ from administrative records in terms of their availability as a reporting mechanism. While the option to report a birth to the registries is continuously available, the option to report it in the census is generally available only once every decade, which could occur shortly after the child’s birth or several years later. Crucially, the timing of the census is exogenous to the child’s birth date, the mother’s age at delivery, and the mother’s age at the time of reporting. Thus, we can isolate the effect of the mother’s age at the time of reporting by comparing estimates from two different census rounds for the same mother’s age at childbirth and year, thereby holding the reporting method and age at childbirth constant. 10 This provides an additional test for Proposition 2, which posits that under-reporting should be less severe the older the mother is at the time of reporting. In line with this proposition, we find sharply different patterns between the teen and young-adult mothers group and child-mothers. When we consider mothers aged 15 to 24 at the time of delivery—a group for which we expect social and judicial costs of reporting to be smaller or non-existent— we observe that birth counts based on prior and later censuses are similar in Brazil, Mexico, and the United States (Figure 2, left side). In contrast, birth counts for child mothers—for whom we expect higher social and/or judicial costs—are consistently lower in the baseline census (when the mother is still in her teens at the time of reporting) than in the census conducted a decade later (when the mother is already in her twenties). The gap is large. Relative to the subsequent counts, the baseline counts are smaller by 40 to 90%. 10 As previously discussed, this approach allows us to identify reporting differences net of attrition between the baseline and subsequent census rounds, providing a lower bound for under-reporting in the baseline census. 17 Census Evidence: Global Patterns of Under-reporting of Child-Mother Births The under-reporting of child-mother births in Brazil, Mexico, and the United States, as documented above, raises the question of whether similar patterns of misreporting are prevalent in other countries. To address this, we now examine the extent of under-reporting of child-mother births in censuses across a broader set of countries, using data from 78 census pairs covering 59 countries. 11 While previous studies on child and teenage fertility have primarily focused on formal birth registration in administrative records (e.g., Duryea et al. 2006;Ebbers and Smits 2022;Wendt et al. 2022), under-registration in demographic censuses and surveys is important in its own right. These datasets play a critical role in shaping policies to address the issue and in producing research that informs policy decisions (United Nations DESA,2020a). Figure 4: Birth Counts in Each of the Five Years prior to the Baseline Census Notes: The sample comprises 80 pairs of national censuses from 59 countries, with one observation per census pair, age of the mother at birth and the year of birth. We estimate the number of births by identifying the age at birth of the reported mother. Samples exclude children of foreign citizenship. Figure 4presents the cross-country averages and 95% confidence intervals of birth counts (from both baseline and subsequent censuses) for each of the five years preceding the baseline census. 12 Consistent with our findings for Brazil, Mexico, and the United States, we observe systematic differences in birth counts between the baseline and subsequent censuses for child mothers (right panel), but not for older mothers (left panel). Furthermore, the undercounting of births in the baseline census is most pronounced for births 11 Estimates are derived from census microdata integrated and standardized by IPUMS (Integrated Public Use Microdata Series, IPUMS [dataset, accessed March,2024). To our knowledge, no comparable source exists for international birth registry data. 12 Appendix Table A1 reports the underlying data, along with child-mother birth undercounting estimates by the mother’s age at birth for each baseline census. 18 occurring in the year immediately preceding the census and diminishes as the time between the birth and the census increases. This provides broader empirical support for the model’s prediction that the propensity to report births by young mothers increases with the mother’s age at reporting, all else being equal (Proposition 2). Finally, we explore how under-reporting in the baseline census varies by the mother’s age at birth across our global sample. Figure 5presents cross-country averages and 95% confidence intervals for the gap between baseline and subsequent census counts, expressed as a share of the subsequent census count, for each maternal age. Under-reporting of child-mother births in contemporary censuses is widespread. Across countries, birth counts are consistently higher for mothers aged 15 or younger in subsequent censuses compared to baseline censuses conducted soon after the birth. Only for mothers aged 17 or older do baseline censuses yield the higher counts that are mechanically expected. In line with Proposition 1, the likelihood of reporting increases monotonically with the mother’s age at childbirth. Figure 5: Gap in Reported Childbirths between Consecutive Censuses by Mother’s Age Notes: The sample comprises 80 pairs of national censuses from 59 countries, with one observation per census pair, age of the mother at birth and the year of birth. We estimate the number of births by identifying the age at birth of the reported mother. The gap measure is calculated as one minus the ratio of the number of births estimated from the first census to those estimated from the second census 1−birthst birthst+10  , over the age of the mother at birth. The interval between censuses averages 10 years, but it includes spans ranging from 8 to 12 years. Samples exclude children of foreign citizenship. We find substantial under-reporting of child-mother births, with baseline censuses missing more than 60% births to mothers aged 13 or younger that are recorded in subsequent censuses. For mothers aged 14, this figure is around 50%, and for those aged 15, it is close to 20%. Based on census and survey data that identify the share of young children without birth certificates, previous research has shown very low levels of birth registration in the developing world, with only about one-third of children under five registered in South Asia (Kisambira and Schmid,2022), around 45% in Sub-Saharan Africa (United Nations DESA,2020a), and between 8% and 25% in Latin America (Duryea et al., 2006). However, our findings indicate that census and survey samples themselves are missing many children, particularly those born to very young mothers. Therefore, the level of under-registration in birth registries is 19 likely significantly higher than previously estimated for this population. This is because a significant fraction of the offspring of young mothers is likely absent from these censuses and surveys, and those missing are also the most likely not to have a birth certificate. 4.3 Instrumental Payoffs and the Propensity to Report The evidence in Figure 2also offers insights into the differences in instrumental payoffs across reporting methods and how these payoffs compare to social and judicial costs as mothers age. According to Proposition 3, if the mother’s age at the time of childbirth and reporting are held constant, reporting should be higher in the method that provides the greater net instrumental payoff. Our data can approximate these conditions when the census occurs close to the child’s birth date. In Figure 2, we observe that for births occurring in the years prior but close to the census years, and for both age groups and across all three countries, the number of births recorded in the registries exceeds the census counts. As per Proposition 3, this suggests that the net payoffs of reporting are higher for administrative records than for the census. This is unsurprising, as birth registries typically offer more direct benefits, such as access to services and public subsidies, while the benefits of reporting to the census are less immediate and direct. Thus, if a birth to a young mother occurs close to the time of the census, households that report to the census are likely to also report to the registry, but those reporting in the registry may not report in the census. If social and judicial costs are absent or minimal, the pattern described above should persist: reporting will always be higher with the method that provides a higher net instrumental payoff. This is what we observe in the three countries for mothers aged 15 to 24 (Figure 2, left column), where the number of births in the registries consistently exceeds those estimated by the censuses. 13 However, in the presence of social and/or judicial costs that decline with the mother’s age at reporting, the model predicts that for very young mothers, reporting may become higher in a later census when the mother is older, and these costs have diminished. This is what we observe in all three countries for mothers aged 10 to 14 (Figure 2, right column). For births occurring close to and prior to the time of a census, birth counts based on censuses are lower than in the registries when using the contemporaneous census but higher than the registries when using data from the subsequent census, collected 10 years later. This phenomenon leads to the single-crossing prediction described in Proposition 4. 13 In the data, this gap also reflects the fact that census estimates only include births where the mother and child reside in the same household at the time of the census. 20 These results reveal that even administrative registries–—the "gold standard" for fertility measurement– —significantly undercount child-mother births. Current United Nations guidelines regard high-coverage birth registries as the most reliable data source for estimating adolescent birth rates, favoring them over survey and census data (Kisambira and Schmid,2022). However, we find that retrospective census birth counts, despite their tendency to underestimate the actual number of births for older age groups, can provide a more accurate estimate of child-mother births than registries. 5 Mechanisms In this section, we investigate the underlying mechanisms contributing to the under-reporting of child-mother births. 5.1 Attributing Children to Different Mothers Under-reporting a child-mother’s birth does not always imply that the child’s existence is entirely unreported. The household could acknowledge the child but attribute their parentage to an older woman in the household, such as the child’s grandmother or aunt. This strategy could help families avoid the social and legal costs associated with reporting a birth for a young mother. Moreover, it offers a plausible explanation for the child’s presence when noticed by census enumerators. Figure 6: Gap in Reported Childbirths between Consecutive Censuses by Mother’s Age, Including Women beyond Fertile Age Notes: The sample comprises 80 pairs of national censuses from 59 countries, with one observation per census pair, age of the mother at birth and the year of birth. We estimate the number of births by identifying the age at birth of the reported mother. The bias measure is calculated as one minus the ratio of the number of births estimated from the first census to those estimated from the second census 1−birthst birthst+10  , over the age of the mother at birth. The interval between censuses averages 10 years, but it includes spans ranging from 8 to 12 years. Samples exclude children of foreign citizenship. 21 Duryea, S., Olgiati, A., and Stone, L. (2006). The Under-Registration of Births in Latin America. SSRN Electronic Journal, (January). Ebbers, A. L. and Smits, J. (2022). Household and context-level determinants of birth registration in SubSaharan Africa. PLoS ONE, 17(4 April):1–17. Hurst, E. and Pugsley, B. (2010). Are Household Surveys Like Tax Forms: Evidence from the Self Employed. Review of Economics and Statistics, 96(1):19–33. Integrated Public Use Microdata Series, IPUMS [dataset, accessed March, . (2024). Minnesota Population Center. Jayachandran, S. (2015). The Roots of Gender Inequality in Developing Countries. Annual Review of Economics, 7(1):63–88. Johnson, M. S. (2020). Regulation by shaming: Deterrence effects of publicizing violations of workplace safety and health laws. American Economic Review, 110(6):1866–1904. Karing, A. (2024). Social Signaling and Childhood Immunization: A Field Experiment in Sierra Leone. The Quarterly Journal of Economics, pages 2083–2133. Kearney, M. S. and Levine, P. B. (2012). Why is the teen birth rate in the United States so high and why does it matter? Journal of Economic Perspectives, 26(2):141–166. Kearney, M. S. and Levine, P. B. (2015). Media influences on social outcomes: The impact of mtv’s 16 and pregnant on teen childbearing. American Economic Review, 105(12):3597–3632. Keskin, F. and Çavlin, A. (2020). Unregistered births of adolescent mothers in Turkey: Invisible children. Comparative Population Studies, 45:179–200. Kisambira, S. and Schmid, K. (2022). Selecting adolescent birth rates (10-14 and 15-19 years) for monitoring and reporting on sustainable development goals. https://www.un.org/ development/desa/pd/sites/www.un.org.development.desa.pd/files/undesa_pd_2022_tp_ selecting-abr-data-for-sdg-reporting.pdf. [Accessed: 2023-03-17]. Klepinger, D., Lundberg, S., and Plotnick, R. (1999). How does adolescent fertility affect the human capital and wages of young women? Journal of Human Resources, 34(3):421–448. Lang, K. and Weinstein, R. (2015). The Consequences of Teenage Childbearing before Roe v. Wade. American Economic Journal: Applied Economics, 7(4):169–197. Meneses, E. and Ramírez, M. (2018). Niveles y tendencias de la fecundidad en niñas y adolescentes de 10 a 14 años en méxico y características de las menores y de los padres de sus hijos e hijas, a partir de las estadísticas vitales de nacimientos de 1990 a 2016. (106):117–152. Pullum, T. W. and Becker, S. (2014). Evidence of omission and displacement in DHS birth histories . DHS Methodological Reports No. 11, (September). Pullum, T. W., Becker, S., and Charlton, C. (2018). Demographic and health surveys methodology - household questionnaire. Technical report, The DHS Program. Sandefur, J. and Glassman, A. (2015). The Political Economy of Bad Data: Evidence from African Survey and Administrative Statistics. Journal of Development Studies, 51(2):116–132. Schoumaker, B. and Sánchez-Páez, D. A. (2022). Under-15 fertility around the world. Population & Societies, 601(6):4. 28 Singh, S. and Darroch, J. E. (2000). Adolescent pregnancy and childbearing: Levels and trends in developed countries. Family Planning Perspectives, 32(1):14–23. Spoorenberg, T. (2014). Reverse survival method of fertility estimation: An evaluation. Demographic Research, 31(1):217–246. United Nations (1983). Manual X. Indirect techniques for Demographic estimations. United Nations. United Nations DESA (2017). Principles and recommendations for population and housing censuses, volume 3. United Nations DESA (2020a). Fertility among Young Adolescents at Ages 10-14 Years - A global assessment Fertility among Young Adolescents. Technical report. United Nations DESA (2020b). Fertility among young adolescents at ages 10-14 years – a global assessment. https://www.un.org/development/desa/pd/sites/www.un.org.development.desa.pd/files/ desa_pd_2017_fertility_among_young_adolescents.pdf. [Accessed: 2023-03-17]. Wendt, A., Hellwig, F., Saad, G. E., Faye, C., Boerma, T., Barros, A. J., and Victora, C. G. (2022). Birth registration coverage according to the sex of the head of household: an analysis of national surveys from 93 lowand middle-income countries. BMC Public Health, 22(1):1–11. WHO, W. H. O. (2024). Adolescent birth rate (per 1000 women) – global health observatory data. Accessed: 2024-09-24. 29 Appendix (For Online Publication) A Additional Figures and Tables A1 Additional Figures (a) Births in Mexico by Mother Age Group (1985-2018) Figure A1: Total Births by Mother’s Age Group According to Census and Registry Data Note: The figure presents annual birth estimates derived from national censuses and birth registries for Mexico. The left panels represent mothers aged 15 to 25, while the right panels depict those aged 10 to 14. The graph portrays Mexican birth estimates, employing data from the 2010 and 2020 Censuses supplemented by the birth registry with only registered births within one year from birth. Information from 2020, potentially affected by COVID-19 pandemic-related reporting delays, is excluded. 1 Figure A2: Average of Births Reported by Mother’s Age 2 Mothers aged 10-14 Mothers aged 15-19 Mothers aged 20-24 Mothers aged 25-29 Figure A3: Exhibit 2: Mexico by Age Groups Notes: Sample does not include registry information from 2020 onwards due to the COVID-19 emergency. 3 Figure A4: Judicial Costs and Minimum Age of Consent Note: This graph depicts a RD plot of our bias measure alongside the difference between a mother’s age at birth and the minimum age of consent. The sample excludes children born in the year of the first census. MAoC ranges between 12 and 18 years old. 4 A2 Additional Tables Table A1: Estimated Unrecorded Births, Relative to Subsequent Census (1) (2) (3) (4) (5) (6) (7) Mother’s age at child birth Total (ages 10 - 14) Baseline Census year 10 11 12 13 14 Count (in 1000s) Share of counted High income countries (2023) Austria 1991 0.050 . 0.030 0.020 0.100 0.200 95.24% Chile 1992 0.290 0.440 0.320 0.190 0.240 1.480 42.53% France 1990 0.360 0.636 0.800 0.924 1.484 4.204 95.11% France 1999 0.101 0.181 0.072 0.123 0.107 0.585 70.90% Greece 1991 . . 0.020 0.130 0.230 0.380 55.07% Hungary 1990 0.140 0.260 0.020 0.180 0.200 0.800 80.00% Hungary 2001 0.140 0.160 0.080 0.060 0.040 0.480 63.16% Panama 1990 0.270 0.350 0.210 0.320 0.430 1.580 65.02% Panama 2000 0.150 0.210 0.340 0.380 0.250 1.330 58.59% Portugal 1991 0.020 0.240 0.080 0.180 0.140 0.660 56.90% Portugal 2001 0.080 0.040 0.080 0.060 0.220 0.480 77.42% Puerto Rico 1990 0.052 0.127 0.214 0.493 0.157 1.043 69.03% Puerto Rico 2000 . . 0.414 . 0.216 0.630 71.92% Romania 1992 0.190 0.140 0.070 0.120 0.770 1.290 52.02% Romania 2002 0.040 0.030 0.240 0.210 0.130 0.650 46.76% Russia 2002 1.380 2.200 1.420 2.220 2.980 10.200 88.24% Spain 1991 0.200 0.140 0.120 0.240 0.037 0.737 67.03% Spain 2001 0.717 0.396 0.345 0.347 0.317 2.122 94.65% Switzerland 1990 . 0.020 0.020 . 0.020 0.060 75.00% Trinidad and Tobago 2000 0.011 0.011 0.034 0.034 0.026 0.117 85.38% United States 1990 4.935 7.949 12.000 19.397 30.122 74.403 91.15% United States 2000 5.858 4.623 11.623 13.770 18.333 54.207 80.53% Uruguay 1996 0.082 0.018 0.076 . 0.269 0.445 56.54% Upper middle income countries (2023) Argentina 1991 1.435 1.811 2.045 2.131 3.886 11.308 68.66% Armenia 2001 0.050 0.050 0.080 0.100 0.160 0.440 91.67% Belarus 1999 0.030 0.030 0.080 0.070 0.180 0.390 90.70% Botswana 1991 0.030 0.120 0.080 0.100 0.140 0.470 88.68% Botswana 2001 0.050 0.020 0.140 0.100 0.180 0.490 85.96% Brazil 1991 11.422 13.527 14.807 15.389 14.839 69.982 54.07% Brazil 2000 11.931 13.012 12.578 14.396 13.935 65.853 46.82% China 1990 6.100 11.600 10.300 14.700 25.500 68.200 87.44% Costa Rica 2000 0.190 0.180 0.160 0.240 0.110 0.880 53.99% Cuba 2002 0.310 0.440 0.350 0.300 0.430 1.830 57.55% Dominican Republic 2002 0.460 0.920 1.110 1.450 2.180 6.120 56.25% Ecuador 1990 1.250 0.970 . 1.070 . 3.290 34.41% Ecuador 2001 0.580 0.680 0.800 1.100 1.370 4.530 47.73% Fiji 1996 0.060 0.110 0.080 0.130 0.030 0.410 91.11% Guatemala 1994 0.810 1.430 2.060 3.040 4.090 11.430 69.48% Indonesia 1990 33.105 41.298 51.900 75.853 99.034 301.190 90.75% Indonesia 2000 1.310 1.000 26.210 34.750 44.840 108.110 75.26% Jamaica 1991 0.149 0.168 0.111 0.239 0.455 1.122 78.90% Malaysia 1991 0.800 1.550 2.600 2.350 2.700 10.000 76.34% Mauritius 1990 0.040 . 0.030 0.020 0.040 0.130 81.25% Mauritius 2000 0.020 0.040 . 0.040 . 0.100 90.91% Mexico 1990 5.553 7.694 8.866 12.756 20.231 55.100 74.18% Mexico 2000 4.846 6.631 7.233 9.261 11.227 39.198 64.54% Paraguay 1992 0.332 0.332 0.474 0.729 0.917 2.785 81.08% South Africa 2001 1.650 1.863 3.400 3.544 5.304 15.760 86.67% Thailand 1990 2.090 2.848 3.095 3.746 2.818 14.596 56.11% Turkey 1990 4.280 5.420 7.420 9.420 12.920 39.460 76.71% Lower middle income countries (2023) Bangladesh 1991 123.530 116.580 109.680 215.780 156.660 722.230 81.76% Bangladesh 2001 33.360 50.050 86.830 167.520 195.530 533.290 80.63% Benin 1992 2.600 2.560 4.020 5.060 3.910 18.150 71.12% Benin 2002 4.890 4.440 7.450 11.000 5.760 33.540 84.27% Bolivia 1992 0.590 0.650 0.930 1.280 1.970 5.420 76.23% Bolivia 2001 1.423 1.815 1.454 1.595 1.948 8.234 84.42% Cambodia 2004 0.315 0.019 0.185 0.019 0.199 0.737 100.00% Egypt 1996 9.772 11.660 18.968 30.029 41.137 111.566 91.83% Ghana 2000 5.390 5.790 6.150 9.530 11.000 37.860 94.37% Kenya 1999 9.060 10.890 19.300 27.730 31.090 98.070 92.30% Kyrgyzstan 1999 0.530 0.490 0.320 0.490 0.620 2.450 97.61% Nepal 2001 3.242 4.452 5.521 11.888 14.349 39.452 83.03% Nicaragua 1995 0.710 0.800 0.910 1.120 1.140 4.680 59.62% Papua New Guinea 1990 0.500 0.480 1.870 3.250 3.520 9.620 83.72% Philippines 1990 5.042 6.250 7.003 8.279 9.458 36.032 79.89% Philippines 2000 0.475 0.996 1.056 1.217 0.174 3.917 22.01% Senegal 2002 5.609 4.941 6.074 7.088 2.352 26.064 68.44% Tanzania 2002 13.459 15.042 26.083 27.007 30.652 112.242 87.79% Vietnam 1999 1.218 1.640 1.363 2.553 3.350 10.125 82.66% Zambia 1990 3.870 5.010 4.340 6.470 6.960 26.650 90.80% Zambia 2000 3.540 4.080 5.470 7.870 9.770 30.730 83.80% Low income countries (2023) Burkina Faso 1996 4.450 5.290 4.180 5.960 3.760 23.640 63.91% Malawi 1998 4.290 5.400 5.860 7.640 7.680 30.870 79.66% Mali 1998 4.240 4.560 4.220 4.720 . 17.740 54.02% Mozambique 1997 9.780 11.650 11.390 11.420 11.760 56.000 71.46% Rwanda 1991 0.070 0.280 0.400 0.430 0.770 1.950 50.52% Rwanda 2002 0.410 0.680 0.570 0.760 1.050 3.470 79.95% Sierra Leone 2004 6.090 4.560 5.530 9.690 5.900 31.770 94.78% Uganda 1991 9.286 11.325 14.480 19.492 22.164 76.747 92.68% Uganda 2002 9.930 11.640 12.820 14.930 19.270 68.590 93.96% Notes: Columns 1-5 present the total number of unrecorded births by mother’s age at birth, calculated as the difference between the estimated total births in census t+10 and census t. Column 6 reports the aggregate number of unrecorded births, while Column 7 reports the share of unrecorded births relative to the total recorded births in census t+10. 5 C Data Appendix C1 Sample Table C1: Country-Census Pairs in Sample Country T0T1Country T0T1 Argentina 1991 2001 Mauritius 1990 2000 Austria 1991 2001 Mauritius 2000 2011 Bangladesh 1991 2001 Mexico 1990 2000 Bangladesh 2001 2011 Mexico 2000 2010 Armenia 2001 2011 Morocco 2004 2014 Bolivia 1992 2001 Mozambique 1997 2007 Bolivia 2001 2012 Nepal 2001 2011 Botswana 1991 2001 Nicaragua 1995 2005 Botswana 2001 2011 Panama 1990 2000 Brazil 1991 2000 Panama 2000 2010 Brazil 2000 2010 Papua New Guinea 1990 2000 Belarus 1999 2009 Paraguay 1992 2002 Cambodia 2004 2013 Philippines 1990 2000 Chile 1992 2002 Philippines 2000 2010 China 1990 2000 Portugal 1991 2001 Costa Rica 2000 2011 Portugal 2001 2011 Benin 1992 2002 Puerto Rico 1990 2000 Benin 2002 2013 Puerto Rico 2000 2010 Dominican Republic 2002 2010 Romania 1992 2002 Ecuador 1990 2001 Romania 2002 2011 Ecuador 2001 2010 Russia 2002 2010 Fiji 1996 2007 Rwanda 1991 2002 France 1990 1999 Rwanda 2002 2012 France 1999 2011 Senegal 2002 2013 Ghana 2000 2010 Sierra Leone 2004 2015 Greece 1991 2001 Vietnam 1999 2009 Guatemala 1994 2002 Spain 1991 2001 Hungary 1990 2001 Spain 2001 2011 Hungary 2001 2011 Switzerland 1990 2000 Indonesia 1990 2000 Thailand 1990 2000 Indonesia 2000 2010 Trinidad and Tobago 2000 2011 Ireland 1991 2002 Turkey 1990 2000 Ireland 2002 2011 Uganda 1991 2002 Italy 2001 2011 Uganda 2002 2014 Jamaica 1991 2001 Tanzania 2002 2012 Kenya 1999 2009 United States 1990 2000 Kyrgyz Republic 1999 2009 United States 2000 2010 Malawi 1998 2008 Burkina Faso 1996 2006 Malaysia 1991 2000 Uruguay 1996 2006 Mali 1998 2009 Zambia 1990 2000 Zambia 2000 2010 Notes: . 6 C2 Definitions and Sources of Variables Table C2: Definitions and Sources of Variables Used in the Analysis Measure Description Source Bias measure This variable is calculated by one minus the ratio between the number of births reported in the first census over the number of births reported in the second census. To estimate the yearly number of births, we count the total number of individuals born in a specific year from the census microdata obtained from IPUMS international. IPUMS International % women making informed sexual decisions This indicator measures the percentage of women aged 15-49 who are married or in a union and have autonomy in three key areas: making decisions about sexual relations, contraceptive use, and their own reproductive health care. Specifically, only women who have an active decision in the three components are considered to have autonomy in these areas. UN Demographic and Health Surveys Log income per capita This metric is derived by computing the natural logarithm of the GDP at purchaser’s prices, which is measured in constant 2017 international dollars. World Bank Difference between age and MAoC This variable is calculated by subtracting the minimum age of consent in a given country from the age of the mother at the time of giving birth. ageofconsent.net C3 Further Details on Measures Used C3.1 Census Data The data used in this project were obtained directly through IPUMS. IPUMS International harmonizes and integrates census microdata from around the world into a consistent format over time and space, which makes it useful for cross-temporal and cross-national comparative research. 7 C3.2 Administrative Birth Registries Sistema de Informações sobre Nascidos Vivos, SINASC (Brazil). This information system collects data on live births through the issuance of a Live Birth Certificate (Declaração de Nascido Vivo - DNV) at the time of birth, primarily in hospitals and healthcare facilities. When a child is born, healthcare professionals fill out the DNV, which includes comprehensive details about the birth. The hospital is responsible for entering this information into the SINASC system. In cases of home births, midwives or attending healthcare providers are responsible for issuing the DNV. The DNV is mandatory and serves as the primary document for recording a live birth. However, efficiency and effectiveness of data collection and reporting can vary across states due to differences in local healthcare infrastructure and administrative practices. Parents must take the DNV to a civil registry office (cartório) to officially register the birth. This process links the healthcare data collection with the civil registration system. Despite its comprehensive coverage, a significant share of births still go unregistered. This can happen when births do not occur in hospitals and fail to be registered afterwards, because parents of hospital-born children report false or erroneous information, or because hospital births are not registered at a civil registry office by the parents, and the cross-validation system flags those births as requiring further checks. Estadística de Nacimientos del Registro Civil, INEGI (Mexico). This dataset is part of Mexico’s official vita statistics (Estadística Vitales), and is compiled by INEGI, the Mexican National Institute of Geography and Statistics (Instituto Nacional de Estadística y Geografía), for each year since 1985. Each year contains detailed microdata of all births registered in Mexico in that year, including date 15 and place of birth, and socio-demographic information on parents (including age at birth). In Mexico, when a birth occurs, healthcare providers are required to issue a Certificate of Birth (Certificado de Nacimiento) within the first 24 hours. This certificate is mandatory and serves both legal and statistical purposes. The parents are then required to take the Certificate of Birth to the local Civil Registry office (registro civil) to obtain the official Acta de Nacimiento (birth certificate). The Civil Registry offices spread around the country then send these Acta de Nacimiento documents directly to INEGI, whether in printed form or electronically. After receiving the data, INEGI processes, validates and diffuses the final dataset. As in the case of Brazil, births still can go unregistered in Mexico for similar reasons: unregistered home births, non-registration of births in the civil registry, the purposeful or accidental registration of incorrect information, and lack of access to civil registry units. There is still another potential source of data on live 15 The date of birth is different from the registration date, such that a given year’s dataset can include births that took place several years before the registration date. 8