Breaking the Cycle: Unraveling Intergenerational Teen Childbearing in Latin America
Based on research by Matilde P. Machado, Ricardo Mora, and Karen Olivo.
We are filled with dismay at images of pregnant girls. They are kids having kids, and the vast majority lack the maturity to engage in such a lifetime endeavor. These pregnancies will affect not only their health (Karataşlı et al. 2019) but also their education and, ultimately, their chances in the labor market and elsewhere (Hoffman and Scher 2008). Although teen fertility has been decreasing worldwide for decades, the pace of such decrease has been modest in Latin America and the Caribbean (LAC). As a result, the region now has the third highest teen fertility rates in the world, only surpassed by sub-Saharan Africa and Oceania (UNFPA 2022). The disparity with developed nations is striking, as evidenced by the statistics from 2020: LAC recorded a rate of 60.6 births per 1000 girls aged 15-19, in contrast to 15.9 in the United States and 7.3 in Canada (United Nations Population Division, World Population Prospects).
Now, there are many potential reasons why the rates of teen childbearing are high and persistent in Latin America. Keep in mind that this is a region with low average income and low social mobility, where access to contraception is limited, and abortion is either very restricted or banned altogether (Guttmacher Institute 2018). To look for all possible causes would be a daunting task. Instead, in this post and in our research related to it, we ask to what extent the source of the persistently high teen childbearing rates in Latin America lies within the family, concretely in the mother-daughter link. The answer is critical to policy design: if daughters of teen mothers are more at risk of a teen motherhood than other adolescents, effective preventive policies should target them.
Researchers using data from developed countries have shown that daughters of teen mothers are indeed more likely to become teen mothers themselves. Recent examples include Haveman, Wolfe, and Peterson (2008) for the US, Francesconi (2008) for the UK, and Aizer, Devereaux, and Salvanes (2020) for Norway. Their results constitute evidence of “intergenerational transmission of teen childbearing”. Studies from Latin America reveal statistical associations between teen childbearing and, among others, poverty, low educational inputs, low aspirational objectives, low sexual literacy, and high levels of violence (Drewry and Garcés-Palacio 2020; Tsaneva and Gunes 2020; Dongarwar and Salihu 2019; Alzate 2014). Yet little is known about intergenerational transmission in those regions of the world where teen motherhood seems endemic, including Latin America.
This research gap may be attributed to the challenges associated with data availability. To effectively study the mother-daughter link in teen childbearing, it is crucial to have access to data that captures the teen childbearing status of both mothers and daughters. In many countries, particularly in the Latin American region, this information is primarily sourced from household surveys or census data. Matching mothers and daughters in such data, however, is impossible for those mothers and daughters who do not live in the same household. Consider the typical household survey where information is gathered only for individuals living in the household. Daughters who no longer live with their parents usually appear as household heads or their spouses and information on their mothers is missing. Similarly, the daughter information is also missing in households of interviewed mothers whose daughters have already left their parents’ home. These unmatched women are not present when data are restricted to matched mother-daughter pairs. This feature of the data raises two questions. Would this be a problem? If yes, can we fix it?
The answer to the first question is “yes” and has to do with a well-known problem that plagues many empirical analyses in Economics: sample selection bias. There are many examples of sample selection bias. Whereas the market wage for working women is known, the potential wages of women who do not work are not. Likewise, the wages of migrants are not, in general, representative of what non-migrants would have earned had they migrated. The lifetime earnings of college graduates are a biased estimate of the earnings of dropouts had they finished their studies. In short, self-selected samples are not representative and lead to difficulties in the identification of useful counterfactuals. In our setting, teen childbearing increases the probability that the daughter leaves the parental home and, hence, the probability that she self-selects out of the matched mother-daughter sample. This implies that using only matched pairs likely bias the sample in favor of teenagers who are not teenage mothers. One could argue that the problem of sample selection can be reduced if we focus on young daughters, say those aged between 15 and 18 because young daughters are more likely to live with their parents. This is true, but the fundamental problem, which is that teen childbearing may trigger abandoning the parental home, persists.
To illustrate the importance of these issues, consider the population of women aged 15 to 18 years old and their mothers in the following Latin American countries: Bolivia, Colombia, Dominican Republic, Guatemala, Haiti, and Peru. In all these countries, the Demographic and Health Survey (DHS) is available. This survey is ideal for reconstructing the birth history of all women and making international comparisons about teen motherhood rates. Unfortunately, it is a household survey and does not follow daughters who abandon the parental home.
Figure 1: Daughter’s Teen Childbearing Status by her Mother’s Teen Childbearing Status. Notes: Own elaboration using birth histories from sample Standard Demographic Health Surveys (DHS). These are: (i) Bolivia: 2008; (ii) Colombia: 2004-05, 2009-10 and 2014-15; (iii) Dominican Republic: 2002, 2007, and 2013; (iv) Guatemala: 2014-15; (v) Haiti: 2012, 2016-17; and Peru: 2005-08, 2009, 2010, 2011, and 2012.
A first look at mother-daughter pairs in the DHS data, as shown in Figure 1, reveals that matched daughters of teen mothers have a two- to six-percentage point higher probability of teen childbearing compared to the other matched daughters. In the case of the Dominican Republic, for example, 12 percent of daughters of mothers who were teen mothers are teen mothers themselves. The proportion decreases to 7 percent among matched teenagers whose mothers were not teen mothers. We would then say that the unconditional effect of having a teen mother would be 5 percentage points, a value in the range of the estimates for developed countries. However, the figure also shows that teen daughters who are not matched with their mothers are close to three times more likely to be teen mothers (31 percent in the Dominican Republic, 27 percent in Colombia). Assuming the existence of mother-to-daughter transmission, we expect that a more than proportional percentage of the missing mothers were teen mothers as well. Hence, Figure 1 suggests that estimation with only matched data creates a negative sample selection bias.
At this point, it might look like we have reached a dead end, and, answering the second question above, the problem cannot be fixed. In a recent working paper, we argue otherwise (Machado, Mora, and Olivo 2021). We propose an econometric methodology based on the maximum likelihood (ML) estimation method that gives estimates that are not subject to sample selection bias. The fundamental idea behind the proposed methodology is that an unmatched observation still gives valuable information and should not be ignored. Concretely, even for unmatched pairs, there is always information on either the daughter or the mother, and these data helps identify the probability of being unmatched as well as the magnitude of the intergenerational transmission. To be clear, our ML estimator does not construct any type of imputation. Instead, it fits the individual data to the model while taking statistically into account all possibilities in the missing observations.
Using our methodology with the six Latin American countries shown in Figure 1, we estimate very large effects of having a teen mother on the probability of teen childbearing, between 9.1 and 23.7 percentage points (75 and 123% relative to the mean incidence of teen childbearing). They imply that inertia within the family is an important driver of the persistence of teen childbearing in these Latin American countries. Moreover, we show that restricting the estimation sample to matched mother-daughter pairs results in a sizable downward sample selection bias, between 79 and 93%, of the transmission of teenage childbearing from mothers to their daughters. Not surprisingly, we also find that having a teen mother is associated to a higher probability of having sex before the age of 16, marrying before 19, and having no more than primary education.
The magnitude of intergenerational transmission of teen motherhood in our six Latin American countries is somewhat comparable to previous estimates for the US and is clearly larger than estimates obtained for other developed countries. Such international differences (and those we report among the six Latin American countries in our data) suggest that the intergenerational transmission of teen childbearing is sensitive to the availability of public programs, the use of contraceptive methods, the legality of abortion, and, more generally, cultural values and educational achievements. More research on the role of these factors would be desirable.
Machado, Matilde P, Ricardo Mora, and Karen Olivo. 2021. “Intergenerational Transmission of Teen Childbearing in Latin America.” CAF Working papers; https://scioteca.caf.com/handle/123456789/1840
About the authors:
Matilde Machado is Associate Professor at the Economics Department of Universidad Carlos III de Madrid. She works on Health Economics, Gender Economics, and Applied Microeconometrics.
Ricardo Mora is Associate Professor at the Economics Department of Universidad Carlos III de Madrid. He works on Labor Economics, Development Economics, and Applied Microeconometrics.
Aizer, Anna, Paul Devereaux, and Kjell Salvanes. 2020. “Grandparents, Moms, or Dads? Why Children of Teen Mothers Do Worse in Life.” Journal of Human Resources, 1019–10524R2.
Alzate, Monica M. 2014. “Adolescent Pregnancy in Colombia: The Price of Inequality and Political Conflict.” In International Handbook of Adolescent Pregnancy, 241–55. Springer.
Dongarwar, Deepa, and Hamisu Mohammed Salihu. 2019. “Influence of Sexual and Reproductive Health Literacy on Single and Recurrent Adolescent Pregnancy in Latin America.” Journal of Pediatric and Adolescent Gynecology 32 (5): 506–13.
Drewry, Jonathan, and Isabel C Garcés-Palacio. 2020. “Socio-Demographic Characteristics and Female Empowerment as Determinants of Adolescent Pregnancy in Colombia.” International Journal of Sexual Health 32 (3): 293–305.
Francesconi, Marco. 2008. “Adult Outcomes for Children of Teenage Mothers.” Scandinavian Journal of Economics 110 (1): 93–117.
Guttmacher Institute. 2018. “Abortion in Latin America and the Caribbean.”
Haveman, Robert, Barbara Wolfe, and Elaine Peterson. 2008. “Children of Teen Mothers as Young Adults: Consequences of Teen Childbearing for the Life Chances of Children, 1968–88.” Kids Having Kids: Economic Costs and Consequences of Teen Pregnancy, 324–41.
Hoffman, Saul D, and Lauren S Scher. 2008. “Consequences of Teen Childbearing for the Life Chances of Children, 1979-2002.” Kids Having Kids: Economic Costs and Social Consequences of Teen Pregnancy, 342–58.
Karataşlı, Volkan, Ahkam Göksel Kanmaz, Abdurrahman Hamdi İnan, Adnan Budak, and Emrah Beyan. 2019. “Maternal and Neonatal Outcomes of Adolescent Pregnancy.” Journal of Gynecology Obstetrics and Human Reproduction 48 (5): 347–50.
Machado, Matilde P, Ricardo Mora, and Karen Olivo. 2021. “Intergenerational Transmission of Teen Childbearing in Latin America.” CAF Working papers; http://cafscioteca.azurewebsites.net/handle/123456789/1840.
Tsaneva, Magda, and Pinar Mine Gunes. 2020. “The Effect of Violent Crime on Teenage Pregnancy in Mexico.” Review of Economics of the Household 18 (1): 141–64.
UNFPA. 2022. “Motherhood in Childhood: The Untold Story.” United Nations Population Fund, New York.