How are poverty, early teen marriage, and dropping out of high school related? I begin by presenting OLS estimates of the effect of early teen marriage and dropout status on poverty.
The top panel of Table 1 displays results for individual-level data, which include more than 3 million observations. These estimates suggest that dropping out of high school has a sizable impact on future poverty, but that teen marriage has relatively small effect.
Standard errors, adjusted for clustering by state of birth to account for arbitrary autocorrelation over time, are shown in parentheses. Data are from the , , and U. The sample is restricted to women between the ages of 20 and 60 who were born in one of the 41 states with valid marriage, compulsory schooling, and child labor laws.
The dependent variable, poor, is a dummy variable equal to 1 if the woman currently lives in a family that is at or below the poverty line. Early teen marriage is defined as marrying between the ages of 12 and 15 14 or 15 in the and censuses , and high school dropout is defined as fewer than 12 years of completed schooling.
Dummy variables for state of birth are indicators for each of the 41 states, and dummy variables for cohort of birth are single-year indicators for each birth cohort. Region of birth trends are separate linear cohort year trends for each of the four birth regions. Allocated observations refer to observations whose value for the variable age at first marriage has been logically edited or hot decked by the Census Bureau. In contrast to the individual-level estimates, the grouped data results in the bottom panel of Table 1 present a very different picture.
In contrast to OLS, the estimates in columns 1—4 are much larger, and the inclusion of controls affects both the dropout and teenage marriage coefficients.
After including controls for 1 census year, race, and current age dummy variables, 2 state of birth and birth cohort dummy variables, and 3 region of birth trends, the coefficients on early teen marriage and dropout are A key question is whether there are additional omitted variables that would drive either of these coefficients closer to zero. Column 5 expands the sample to include allocated observations. The Census Bureau allocates values for age at first marriage when data are missing or inconsistent.
First, a logical edit is performed if possible, using information from other variables and other household members. When this is not possible, the census uses a hot-deck allocation method to assign a value from an individual with similar characteristics.
Additionally, the hot-deck procedure used in and to a lesser extent in and suffers from bracketing issues for early teen marriages, with sharp spikes in marriage rates occurring for women whose current age is a multiple of five. When these allocated marriages are included in column 5, the coefficient on early teen marriages drops, particularly in the grouped OLS panel. As I show later, these allocated marriages do not have much of an impact on the IV estimates, suggesting that these allocated marriages are largely noise.
Therefore, unless otherwise noted, all allocated marriages are dropped from the data. What explains the different estimates for early teen marriage when comparing the individual versus grouped data in Table 1?
An analysis of auxiliary data later in the article indicates a large amount of measurement error in the early marriage variable. This suggests the presence of attenuation bias in the individual-level OLS estimates, whereas aggregation should minimize this type of bias. Of course, if appropriate instruments can be found, misspecification due to omitted variables or measurement error can be eliminated at both the individual and aggregate level. As I show later, the individual-level IV and aggregate IV estimates are both large and remarkably similar.
The OLS estimates presented in the previous section potentially suffer from both omitted variable bias and measurement error. One solution to these problems is to use an instrumental variables approach.
Ideally, instruments would induce exogenous variation in early teen marriage but would be uncorrelated with unobserved characteristics that affect both poverty and the decision to marry young.
Similarly, the instruments would induce exogenous variation in high school graduation but be orthogonal to the error term in the poverty equation. I use changes in state marriage, schooling, and labor laws over time as instruments for early marriage and dropping out of high school. By preventing some teens who would like to marry or drop out of high school from doing so, these legal restrictions can help identify the causal effects on poverty free of selection bias.
In the United States, wide variation has historically existed regarding the minimum age individuals are legally allowed to marry. The laws that regulate teenage marriage have appeared in the World Almanac and Book of Facts starting in the late s. Since , information has consistently been reported on the minimum marriage age with parental or court consent, separately for males and females. I collected this information annually for the years to for the 41 states with reliable information on marriage laws during this time period.
There are two sets of laws specifying minimum age requirements for marriage. The first is the minimum age with parental or court consent, while the other is the minimum age without parental consent.
In this article, I focus on the marriage age laws with parental consent, partly because there is little variation over time or across states in the laws without parental consent during the period of my data. In , men and women were granted the right to vote at age 18, which seems to have spurred most states to change their statutes for the legal age of marriage without parental consent for both men and women to age For a discussion and an interesting analysis of these laws, see Blank, Charles, and Sallee Laws with parental consent do not eliminate all early teenage marriages.
Some teens may find ways to lie about their age or may travel to states with lower age requirements to get married. These statutes imply that a judge could grant permission for an early teenage marriage if the teenage woman was pregnant. How often judges actually granted exceptions is hard to know ex post facto, but given the relatively low rate of illegitimate births and abortions during much of this period, exceptions for pregnancy were probably common.
Rather, the strength of the instrument set is that restrictive state laws make it harder to marry young, thereby preventing some fraction of teen marriages that otherwise would have occurred. I also use the compulsory schooling and labor laws originally collected by Acemoglu and Angrist and subsequently modified by Goldin and Katz These laws typically specify a minimum age or amount of schooling before a youth can drop out of school or obtain a work permit. Child labor is defined as the maximum of 1 the required years of schooling before receiving a work permit and 2 the difference between the minimum work age and the maximum enrollment age lagged 8 years.
The value of the marriage, schooling, and labor laws assigned to a woman are based on the set of laws for her birth state that are in force when she would have been age Table 2 summarizes the changes in these laws across five-year time periods in the regression analysis, year-by-year values are used.
A more detailed listing by state and year for the early marriage laws can be found in Dahl , and for the compulsory schooling and child labor laws in Acemoglu and Angrist and Goldin and Katz Summarizing the law changes another way, the average minimum marriage age across states was There have also been similar increases in the requirements governing school attendance and child labor.
Later in the article, I will also investigate the impact of divorce and use unilateral divorce laws as instruments, although the table reveals that few states enacted unilateral divorce laws prior to Entries are the fraction of states with a specified law averaged over the five-year time interval.
Sample size is the number of state-years; there are 41 states with laws available and 35 years, for a total of 1, observations. Table 2 also reports the results of chi-square tests for the pairwise independence of the marriage, schooling, labor, and divorce laws.
These tests all strongly reject the null hypothesis that the various state laws are independent. After time trends in the laws are regressed out, the state laws are still highly related. Since the marriage, schooling, and labor laws affecting youth are so highly correlated, it could be important to account for all three simultaneously when estimating instrumental variable regression models.
Past research has used the compulsory schooling and child labor laws as instruments for education in models describing human capital externalities Acemoglu and Angrist , crime Lochner and Moretti , mortality Lleras-Muney , intergenerational transmission of human capital Oreopoulos, Page, and Stevens , and fertility Black, Devereux, and Salvanes ; Leon In many of these applications, there may not be a need to instrument for early teen marriage.
However, for some outcomes, part of the observed effects might be due to changes in marriage laws and early marriage rates but mistakenly attributed to changes in compulsory schooling laws and education levels instead. In the IV regressions that follow, I use all three sets of laws in poverty regressions that instrument for early marriage and high school completion. How effective are state-specific marriage laws at restricting the age individuals marry? Other work has examined the effectiveness of compulsory schooling and child labor laws on high school graduation and is not repeated here see Acemoglu and Angrist ; Goldin and Katz ; Lleras-Muney ; Lochner and Moretti ; Margo and Finegan The combined census samples reveal that restrictive laws are associated with a smaller number of early teen marriages i.
In the IV regressions appearing in the next section, these factors will be accounted for. Are the laws actually reducing the number of teen marriages, or would states with restrictive laws naturally have lower teen marriage rates anyway? I use the and Vital Statistics Marriage Detail files, which collect data from marriage certificates, to examine the timing of teen marriages.
Data were collected from marriage certificates by the National Center for Health Statistics. Marriage rates are grouped in two-month intervals. The sample is restricted to women who married for the first time, who married between the ages of 14 and 16, and who were residents of and got married in a state that is in a marriage-reporting area MRA and has information on marriage laws. The marriage certificate data include all records for small states and a random sample for larger states; the probabilities in the figure are weighted unweighted probabilites are very similar.
The 27 states included in this figure have the following minimum marriage age with parental consent in and for women: Sharp increases in the fraction marrying occur where expected, assuming the laws are enforced. For example, in states where the legal minimum is 14 years, a fair number of women actually marry at this young age.
Moreover, there is not much of a jump in marriages once women turn age In contrast, in states where the legal minimum is 15 years, there is a sudden rise in the number of marriages immediately after women reach the minimum age of For another example, consider women marrying at age In the third graph, where the legal minimum age is 16, there is a sharp and large increase in the number of marriages occurring immediately after women turn In comparison, the rise surrounding age 16 is much less pronounced in states with minimum ages of 14 or especially Another way to test whether state laws impact the probability of marrying young is to see whether teens travel to a state with a lower age requirement to get married.
If so, this is an indication that restrictive laws impose costs on those wishing to marry before the law in their state of residence allows. Some young teens will cross state lines, while others will be deterred by these costs.
The extent to which teens cross state lines to marry in states with more permissive laws can be examined using the residence state and marriage state information in the Vital Statistics data sets.
Before looking at the entire United States, first consider the case for women residing in Tennessee. Tennessee is a long, narrow state, with population centers scattered throughout the state. Tennessee had an age requirement of 16 years for women to marry in and , the period for which Vital Statistics data are available.
Tennessee is bordered by eight states with varying age minima. Six of these states have valid marriage certificate and marriage law information. However, we should not see as many prospective teen brides traveling to Georgia, Kentucky, or Virginia, where the age requirement of 16 was the same as in Tennessee.
The pattern of out-of-state marriages strongly supports the idea that Tennessee teens traveled to bordering states with more permissive laws in order to marry young data not shown. This is not because Alabama, Mississippi, and Missouri are more convenient or attractive places to get married in general, however. Table 3 extends the Tennessee analysis of out-of-state marriages to all of the states in the sample.
I then tabulate the percentage of women who marry 1 in their state of residence, 2 in a state with a lower minimum age than their residence state, and 3 in a state with an equal or higher minimum age than their residence state. For women who married between the ages of 12 and 15, Standard errors are shown in parentheses. The sample is restricted to first marriages of women who are residents of and get married in 1 of the 32 states that are in a marriage-reporting area MRA and have information on marriage laws.
See footnote 9 in the text for a list of available MRA states. The marriage certificate data include all records for small states and a random sample for larger states; the probabilities in the table are weighted unweighted probabilities are very similar. Of course, the patterns observed in the top panel of Table 3 could be the result of the location of states with various laws or the general attractiveness of marrying in different states. To control for this possibility, in the middle panel of Table 3 , I tabulate marriage patterns for women who married at age For these women, the marriage laws should not be binding.
Indeed, fewer of the women facing an age minimum of 16 left their residence states to marry. In contrast to the top panel, women in states with laws specifying a legal minimum of 16 who chose to marry outside their states of residence were much more likely to marry in states with an equal or higher minimum age law.
A simple difference-in-differences estimate makes clear that women crossed state lines to marry young. To construct the estimate, I first compare the fraction of women who married in a state with a lower minimum versus a higher minimum.
Subtracting this difference for women who married between ages 12 and 15 from the difference for women who married at age 16 yields the estimate. For states with a marriage age requirement of 13 or 14, the difference in difference is close to 0 and not significant, as expected.
For states with an age minimum of 15, the estimated difference in difference is 4. An even greater contrast shows up for the states specifying a minimum age of 16, with a large and significant estimate of These results imply that restrictive marriage laws increase the costs to potential teen brides and likely prevent some desired early teen marriages.
As a final check on the validity of the laws as instruments, I explore the timing of law changes. One potential concern is that states that pass more restrictive laws would have experienced larger reductions in early teen marriage rates even in the absence of a law change.
However, if law changes are exogenous, then future values of the laws should not affect current early marriage rates conditional on current laws.
The results from this exercise indicate that future laws do not significantly determine current early marriage rates, while current laws do.
The F statistic for the effect of future laws is 0. Standard errors, adjusted for clustering by state of birth, are shown in parentheses. All regressions include dummy variables for census year, race, age, state of birth, and cohort of birth, and region of birth trends. See the notes to Table 1.
To investigate the effects of teenage marriage and high school completion on subsequent poverty, I use state marriage, schooling, and labor laws as instrumental variables. The bottom panel in Table 4 presents the first-stage estimates.
Since I am instrumenting for both early marriage and dropout status, there are two sets of regression estimates. Column 2 regresses a dummy variable for early teen marriage on the set of marriage, schooling, and labor laws. Additional controls mirror those used in column 4 of Table 1.
The marriage laws significantly reduce the number of teens who marry before the age of 16; ceteris paribus , states with a legislated minimum of 13 or less are between 0. In states without a legislated minimum, common law which specifies a minimum of 12 years prevails; the estimated effect of a common law is similar to a legislated minimum of 13 or less.
Interestingly, the child labor laws seem to work in the opposite direction—more restrictive child labor laws actually increase the probability of an early marriage. A woman born in a state with a child labor law age restriction of 9 or greater has a 1 percentage point higher probability of marriage at an early age. One possible explanation is that early marriage becomes more attractive to a young woman if her other options, such as working, are more limited.
The third set of laws that deal with compulsory schooling is smaller and less significant. Column 3 of Table 4 presents the same set of coefficient estimates for the first-stage dropout regressions. As expected, the compulsory schooling laws have a relatively large and jointly significant effect on whether a young woman finishes high school.
The marriage laws have nontrivial coefficient estimates but are imprecisely estimated and therefore not significant. One reason why dropout status might project onto the marriage laws is that the marriage laws are highly correlated with the compulsory schooling laws. The marriage laws are measured every year, but the schooling laws are only measured intermittently. More restrictive child labor laws seem to discourage some women from dropping out of school, but the estimates are not statistically significant.
For all of the estimates, F statistics are reported for the joint significance of the instruments. The F statistic is All of the standard errors reported in Table 4 and throughout the article are adjusted for clustering by state of birth to account for arbitrary correlation over time.
Bertrand, Duflo, and Mullainathan have shown that failure to account for such correlation can lead to severely biased confidence intervals for the estimated coefficients.
This is particularly likely to be important in IV analyses, which use laws over time as instruments, because there is typically a long time component and plausible serial correlation. The top panel of Table 4 presents the baseline results for the instrumented poverty regression.
Early teen marriage and dropping out of high school both have sizable effects on the probability a woman will end up in poverty. The estimates imply that marrying young is associated with a Dropping out of high school is associated with an I now present a series of alternative estimation approaches to assess the robustness of the baseline result. Table 1 revealed that aggregation made a large difference for OLS estimates: The first column in Table 5 repeats the baseline IV analysis, but this time with grouped data.
The grouped-data IV estimates are remarkably similar to the individual-level IV estimates 0. The similarity of the coefficient estimates is not surprising since the instruments are constant for all individuals in a state-cohort group, effectively aggregating both the individual-level and group-level estimates. Because the aggregated data produces very similar point estimates and slightly more conservative standard errors, in what follows, I present results for aggregated data unless otherwise noted.
The migration-adjusted approach is described in the text and the control function approach is described in the text and the appendix. As is well known, weak instruments can lead to biased IV estimates; under general conditions and finite samples, weak instruments bias the estimates in the same direction as OLS estimates see Bound, Jaeger, and Baker ; Staiger and Stock The first-stage F statistics appearing in Table 4 are significant but of moderate size. To help assess whether weak instruments might be biasing the results, the second column in Table 5 reports LIML estimates for the baseline model.
The consensus in the literature is that when there are many instruments or weak instruments, LIML tends to exhibit less bias compared to least squares IV, and LIML confidence intervals typically also have better coverage rates Stock This suggests that weak instruments are not a major issue for estimation.
The next task is to assess the impact migration has on the assignment of state laws for marriage, schooling, and work and the subsequent IV estimates.
Because some women have migrated out of their birth state and into a state with a different set of laws by age 15, the instruments are measured with error. I assess how this affects the IV estimates in column 3 of Table 5. To see how I examine the issue, notice that the expected value of the ideal but unobserved state laws can be calculated if migration probabilities are known. The asterisk indicates that this variable is not observed, given that she may have moved from her birth state by age However, if migration probabilities are known, the expected value of this variable can be calculated as.
The same logic applies when there are several variables for the state laws. The remaining issue is how to consistently estimate the conditional migration probabilities, p jk. Although this information is not available for all women, the migration patterns for women who were age 15 at the time of the census enumeration can be estimated because the census records both state of birth and state of current residence.
I use year-old women in the census to estimate these migration probabilities. I then calculate the expected value of the laws based on the state a woman lived in at age 15 as outlined above and use these expected laws as instruments. In the current context, a sufficient set of conditions is that the instruments are independent of 1 the individual returns to marrying young and dropping out of high school, 2 any individual-specific intercept term in the outcome equation, and 3 the reduced-form residuals in the first-stage early marriage and dropout equations see Heckman and Vytlacil To assess the impact of heterogeneous returns, I pursue a control function approach similar to the one proposed by Garen and discussed by Card The basic idea of a control function approach is to make some assumptions about the relationship between the observed variables controls and instruments and the individual-specific returns and individual-specific intercept term.
One then includes additional terms in the outcome regression to control for these relationships. The appendix details the assumptions and estimating equation. The resulting control function estimates appear in column 4 of Table 5. These results suggest that heterogeneity across individuals plays a minor role in estimation of the average treatment effect. To further investigate heterogeneity in the returns to marrying young and dropping out of school, the first two panels in Table 6 present additional IV estimates by race and region of country.
The IV estimate of the early teen marriage effect for the black sample is 0. The marriage instruments also have more power for the black sample than the white sample. The dropout coefficients are similar for whites and blacks, but statistically insignificant for blacks.
When looking at estimates by region of the country, it becomes clear that most of the identification is coming from southern states, which is not surprising given that much of the variation in laws occurs in this region of the country. Interestingly, the dropout coefficients fall for both the black and white samples in panel B. OLS estimates for these and other groupings can be found in Dahl Family income is measured in thousands of dollars.
The last three panels in Table 6 present additional robustness checks for the IV estimates. This article has focused on the laws governing marriage with parental consent for women. There are also laws specifying the minimum marriage age without parental consent for women and laws for men. As discussed earlier, there is little variation in the laws without consent for either women or men before , so I cannot effectively use these to instrument for marriages at later ages.
The laws for men with parental consent are highly correlated with the laws for women with parental consent, and are usually two years higher. Panel C uses all of the marriage laws, for men and women, with and without consent, as instruments. These additional instruments result in modest increases in the IV estimates for both the early marriage and dropout variables. In the results presented so far, the dependent variable has been poverty, a binary outcome.
I now explore the effect of early marriage and dropping out of high school on family income, a continuous outcome. The effects of early marriage and dropout status on family income are large, presenting a picture similar to the poverty regressions. An additional robustness exercise includes observations in which the age at first marriage variable was allocated by the Census Bureau. Including these observations has a large impact on the OLS estimates appearing in the bottom panel of Table 1.
In contrast, the IV estimates are robust to the inclusion or exclusion of these allocated observations. As a final exercise, Table 7 investigates the effect of divorce on poverty.
I begin by presenting estimates similar to those in column 4 of Table 1 , but with an additional variable for whether a woman is currently divorced. The estimated effect is substantial. Currently divorced is associated with a In this regression, the early teen marriage coefficient falls slightly compared with Table 1 , from The IV estimate in column 2 instruments for early teen marriage and dropout status using the same specification as Table 4 but also adds in the currently divorced variable as an additional control.
The resulting IV estimate for early teen marriage falls to Since divorce might not be exogenous, it would be useful to instrument for this variable. Previous research has analyzed the effect of changes in divorce laws on divorce rates and stocks Friedberg ; Parkman ; Peters ; Wolfers Thus, one possibility is to use these divorce laws as instruments.
I follow the approach taken by Wolfers and Gruber and use unilateral divorce laws as an instrument for the stock of divorces.
The divorce coefficient appearing in column 3 of Table 7 is negative but not statistically significant. While the results of this exercise are interesting, they should be interpreted with caution because most of the changes in divorce laws, as well as most of the rise in divorce rates, occur after my sample period see Table 2.
The IV estimates indicate that the causal effects of early teen marriage and dropout status on future poverty are substantial. The baseline estimates imply that marrying young increases the chances a young bride will end up in poverty later in life by around 31 percentage points. Dropping out of high school has a somewhat smaller, but still substantial, 11 percentage point effect on future poverty.
These results are robust to the level of aggregation, LIML estimation, corrections for migration, and a control function approach. The individual-level OLS estimates for early teen marriage are small, while aggregated OLS estimates yield an estimate that is of the same magnitude as the IV estimates.
In census data, age at first marriage is calculated from the reported date of first marriage and date of birth month and year. Usually only one person fills in the census form for the entire household.
Since the fraction of early teen marriages is so small, any mismeasurement of date of birth or date of marriage—the two variables used to construct age at first marriage—is likely to lead to a very large downward bias in the OLS estimate of the early teen marriage coefficient.
With just a small amount of measurement error, the incorrectly classified teen brides can outnumber the true teen brides, resulting in substantial attenuation bias. To better understand the prevalence of measurement error in reported dates, consider the National Fertility Survey NFS.
This was the fifth in a series of surveys conducted every five years to examine marital fertility and family planning. The interesting feature of the survey in is that the researchers chose to reinterview a selected sample of women from the survey. The reinterview sample includes 2, white women in their first marriages who were continuously married, whose age at marriage was less than 25 years, and whose husbands had also been married only once.
Both the and surveys ask date of birth and date of first marriage, with both sets of answers being recorded in the sample. Table 8 tabulates how often the responses from the survey do not match the responses from the survey. The amount of measurement error in the census is likely to be even larger because the NFS sampled only women who had never divorced and had women answer questions about themselves.
Data come from the subsample of women from the National Fertility Survey who were reinterviewed in the National Fertility Survey. Marriage age is calculated from date of marriage and date of birth. The results in Table 7 put some perspective on the size of the effect: These effects are larger than those found in much of the literature for teenage childbearing discussed earlier.
How can the current results be reconciled with that literature? There are at least two reasons why the estimated effect of early teenage marriage might not be comparable to the effects estimated for teenage childbearing in the literature.
First, the sample periods differ greatly. Most of the research on teenage childbearing focuses on births occurring in the s or later because many of these studies have used data from the National Longitudinal Survey of Youth or the Panel Study of Income Dynamics.
In contrast, I focus here on women who were 15 years old between and Comparing the two time periods, there are large differences in access to birth control and abortion, social norms, and labor market opportunities for married women and women with children.
Abortion also became legalized in the early s, first in select states and then nationwide with Roe v. To highlight one change in what was socially acceptable over time, consider illegitimacy rates, which rose from 3. In sum, there were many changes starting in the s that could make teenage motherhood after that period not comparable to my sample of early teen brides earlier in the century.
The second reason for why the estimated effect of early teen marriage is so large compared with the estimates for teen childbearing is that this article looks at a sample of particularly young teenagers—those marrying at or before age 15—while the teen child-bearing literature typically examines the effect of births to teenagers less than or equal to age There may be a large difference between marrying or having a child at or before the age of 15 versus between the ages of 16 and Some of these differences can be highlighted using the and National Fertility Surveys.
Moreover, early teen brides married men who were also relatively young and less educated. Twenty percent of women marrying at age 15 or younger married a man who was 17 or younger.
These tabulations from the NFS show that those who marry very young have substantially different divorce rates, fertility rates, schooling completion, and husbands, even in comparison with women who marry just a few years later.
One other interesting comparison can be drawn from the NFS data. In , the survey asked respondents if they would encourage a daughter to marry at a younger age, the same age, or an older age as they did. This provides some indication that early teen brides would not necessarily make the same decision to marry so young if they had it to do over again.
Based on using these laws as instruments for early marriage and high school completion, the results indicate strong negative effects on poverty status that are not due to self selection. The baseline IV estimates imply that women who marry young are 31 percentage points more likely to live in poverty when they are older. Similarly, women who drop out of school are 11 percentage points more likely to be in families below the poverty line.
The IV results are robust to a variety of alternative specifications and estimation methods, including LIML estimation and a control function approach.
In comparison, OLS estimates are sensitive to how the data are aggregated; regressions on individual-level data estimate small effects for early teen marriage, while aggregated data estimate large effects. I argue that the difference is due to a large amount of measurement error in the early marriage variable, resulting in substantial attenuation bias in the individual-level OLS regressions but not the aggregated OLS or IV regressions.
The results imply that the decisions women make early in life can have long-lasting consequences. The IV estimates suggest that legal restrictions that prevent early marriage and mandate high school completion have the potential to greatly reduce the chances of future poverty for a woman and her family. This appendix describes the control function approach taken for estimates appearing in Table 5.
Consider the following outcome equation, which allows for individual-specific coefficients and an individual-specific intercept term:. One fundamental difficulty with child marriage is that girls are financially dependent on their husbands and therefore lack the power to make demands upon them.
They cannot ask their husbands to get an HIV test; they cannot abstain from intercourse or demand condom use 20 ; they cannot insist that their husbands be monogamous; and ultimately, they cannot leave because they cannot repay their high dowry In addition, returning to their parents' home may not be an option because divorce is considered unacceptable and leaving their husbands may have serious implications on the social or tribal ties that were developed during the marriage.
Child marriage and polygamy play an important role in another deadly disease, cervical cancer. HPV infection has become endemic to sub-Saharan Africa 22 — Although many African nations do not have the capacity to adequately or effectively screen for cervical cancer or HPV, the incidence of cervical cancer in Africa is estimated to be extremely high.
Common risks for cervical cancer are child marriage, low socioeconomic status, poor access to health care, and husbands who had multiple sex partners. For example, in Mali, cervical cancer is the most common cancer in women, has an age-standardized incidence rate of Another study in Morocco had similar findings 26 , with cervical cancer risk factors identified as child marriage, high parity, long-term use of oral contraceptives, and poor genital hygiene control participants bathed more frequently, and case-participants used homemade sanitary napkins more frequently.
Other studies have also implicated hygiene as a possible factor 22 , Pregnancy poses many challenges for young girls. Because pregnancy suppresses the immune system 28 , pregnant girls are at increased risk of acquiring diseases like malaria. Approximately 25 million pregnant women are exposed to malaria per year, and pregnant women are among the most severely affected by malaria. Not only are pregnant women most susceptible to malaria during their first pregnancy 31 , but they also have higher rates of malaria-related complications predominantly pulmonary edema and hypoglycemia and death than do nonpregnant women.
However, a woman who has had malaria during pregnancy is less susceptible to malaria during subsequent pregnancies, unless the woman is also HIV infected The interaction between HIV and malaria in young married girls is devastating.
HIV-infected patients are much more susceptible to infection with Plasmodium falciparum. Pregnant women have high malaria parasitemia in the placenta and more severe clinical disease, which affects not just the first pregnancy but all subsequent pregnancies. HIV-infected patients also do not respond as well to standard antimalaria treatment. Births resulting from child marriages are said to be "too soon, too close, too many, or too late" The problem with children delivering children is that the young mothers are at a significantly higher risk than older women for debilitating illness and even death.
For example, in Mali, the maternal mortality rate for girls aged 15—19 is per , live births and for women aged 20—34, only 32 per , In Togo, for the same age groups, these rates are and 39, respectively 1. Reasons for these high death rates include eclampsia, postpartum hemorrhage, HIV infection, malaria, and obstructed labor.
Obstructed labor is the result of a girl's pelvis being too small to deliver a fetus. The fetus's head passes into the vagina, but its shoulders cannot fit through the mother's pelvic bones. Without a cesarean section, the neonate dies, and the mother is fortunate if she survives. If sepsis or hemorrhage does not occur and the girl does survive, the tissue and bones of the neonate will eventually soften and the remains will pass through the vagina.
Many times, obstructed labor leads to fistulas; the pressure of the fetal head on the vaginal wall causes tissue necrosis, and fistulas develop between the vagina and the bladder or rectum after the necrotic tissue sloughs.
Girls ages 10—15 years are especially vulnerable because their pelvic bones are not ready for childbearing and delivery. Once a fistula is formed, fecal or urinary incontinence and peroneal nerve palsy may result and may lead to humiliation, ostracism, and resultant depression. Unless the fistula is surgically repaired, these girls have limited chances of living a normal life and bearing children. Child marriage affects more than the young girls; the next generation is also at higher risk for illness and death.
These deaths may be partly because the young mothers are unhealthy, immature, and lack access to social and reproductive services.
Their babies are also at high risk of acquiring HIV at delivery and during breastfeeding. Mothers who have had malaria are at increased risk for premature delivery, anemia, and death. Untreated STDs such as gonorrhea, chlamydia, syphilis, and herpes simplex virus infection can have deleterious effects on neonates, such as premature delivery, congenital neonatal infections, and blindness. Child marriage has far-reaching health, social, economic, and political implications for the girl and her community.
It truncates a girl's childhood, creates grave physical and psychological health risks, and robs her of internationally recognized human rights. Ending child marriage requires the consent of all those involved, including fathers and religious, community, and tribal leaders. To break the cycle of poverty, programs are needed to educate and empower women. Most of these goals directly affect child marriage. Data show that improvements are being made and that sub-Saharan Africa has the most obstacles to overcome In some countries, child marriage has been declining.
Increasing mean age for marriage often results in part from overall advancement of an economy. In some countries, such as Korea, Taiwan, and Thailand, decreasing poverty effectively decreased child marriage by enabling these countries to improve education, increase employment, and provide better health care for the whole nation.
Education is a key factor for delaying first sexual activity, pregnancy, marriage, and childbearing. Programs that specifically focused on the status of girls may have directly or indirectly reduced the number of child marriages.
Successful programs have provided economic and educational opportunities to young women and their families by employing girls with the specific goal of delaying marriage 40 , giving families financial incentives to keep their daughters in school 1 , or feeding children during school to decrease families' expenses. Keeping girls in school or vocational training not only helps protect them from HIV infection, pregnancy, illness, and death but also enhances their earning potential and socioeconomic status.
Educated girls can contribute to the health and welfare of their family and marry men of their own choosing and age. Lack of enforcement renders laws against child marriage ineffective. Through media campaigns and educational outreach programs, governments need to take responsibility for stopping this practice. Local, regional, and national governments can also implement health outreach programs for girls and boys.
Learning about reproductive and sexual health, STD prevention, contraception, AIDS, and how to seek health care helps girls negotiate safer sex. Governments must incorporate preventive and treatment programs for reproductive health issues into their health services. Necessary preventive services include supplying mosquito netting and condoms; educating patients about contraceptive methods; providing diagnostic screening for HIV and HPV; and offering treatment options such as medications, cesarean sections, and postpartum care.
Ending child marriage requires a multifaceted approach focused on the girls, their families, the community, and the government. Culturally appropriate programs that provide families and communities with education and reproductive health services can help stop child marriage, early pregnancies, and illness and death in young mothers and their children.
She is committed to the eradication of female genital cutting. In , Dr Nour received a MacArthur Foundation Fellows "genius grant" for creating this country's only center that focuses on issues regarding the health, public policy, and legal needs of circumcised women. Suggested citation for this article: Health consequences of child marriage in Africa. Emerg Infect Dis [serial on the Internet].
National Center for Biotechnology Information , U. Journal List Emerg Infect Dis v. This article has been cited by other articles in PMC. United Nations Efforts and National Laws Since , the United Nations and other international agencies have attempted to stop child marriage. Incentives for Perpetuating Child Marriages Poverty plays a central role in perpetuating child marriage.
Risk for HIV and Other Sexually Transmitted Diseases A common belief is that child marriage protects girls from promiscuity and, therefore, disease; the reality is quite different. Cervical Cancer Child marriage and polygamy play an important role in another deadly disease, cervical cancer. Children Bearing Children Pregnancy poses many challenges for young girls.
Children Delivering Children Births resulting from child marriages are said to be "too soon, too close, too many, or too late" Effects on Offspring Child marriage affects more than the young girls; the next generation is also at higher risk for illness and death. Discussion Child marriage has far-reaching health, social, economic, and political implications for the girl and her community.
Footnotes Suggested citation for this article: International Center for Research on Women;Adults found simulation of sex by children to be funny. A social norms perspective on child marriage: Subtracting this difference for women who married between ages 12 and 15 from the difference for women who married at age 16 yields the estimate. What explains the different estimates for early teen marriage when comparing the individual versus grouped data in Table 1? After time trends in the laws are regressed out, the state laws are still highly related. African countries have enacted marriageable age laws to limit marriage to a minimum age of 16 to 18, depending on jurisdiction. Regardless of how birth order plays out in your marriage, both of you can choose to change how you respond to and interact with each .