Abstract
We hypothesize that teen nonmarital birth events are influenced by adolescent girls’ perceptions of the consequences of their choices. Two such consequences are explored: (1) a teen’s expected future marriage and cohabitation relationships and (2) the present value of expected future income. We also measure the effects of the characteristics of the teen, her prior choices, her family, her neighborhood, and the social and economic environment in which she lives. The results, based on the Michigan Panel Study of Income Dynamics, suggest that teens place greater weight on the relationship consequences than the income consequences, but that both consequences influence their nonmarital birth choices.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Then-President Bill Clinton, in his 1995 State of the Union Message, referred to this problem as the nation’s “most serious social problem.” Today, the comment is most appropriately applied to the Latino and Black populations. In 2004, the Latino teen birth rate was 82.6 and the Black non-Latino teen birth rate was 62.7, compared to the overall rate of 41.2 per 1,000 females 15–19 (Hamilton et al. 2005; Ventura et al. 2000).
While teen women who have a nonmarital birth tend to have less income, more marital instability, and lower educational attainment than those who do not, some portion of these outcomes may be attributable to unmeasured adverse family background or personal characteristics. A number of the studies in Maynard (1997) attempt to account for this selection problem in studying the consequences of adolescent childbearing. Hotz, et al. (1997a,b, 1999) use a natural experiment—a comparison of teen mothers with women who became pregnant as teens but who experienced a miscarriage—to account for adverse unmeasured effects. They suggested that virtually all of the costs associated with early childbearing are a manifestation of this selection effect. Their conclusion, however, depends on the extent to which miscarriages are purely random events, and there are important reasons for believing that this is not the case. See also Geronimous and Korenman (1992, 1993), Hoffman et al. (1993), Brooks-Gunn et al. (1993), and Bronars and Grogger (1994), who also analyzed teen fertility.
There is substantial evidence that the children born to teenage mothers (especially those who are not married) are more likely to grow up in a poor and mother-only family, live in a poor or underclass neighborhood, and experience high risks to both their health status and school achievements. See Haveman et al. (1996, 2001a), Wolfe and Perozek (1996), and Wolfe and McHugh (2006). Rosenzweig and Wolpin (1995) also explore this issue.
Haveman et al. (2004) review these studies of the determinants of teenage nonmarital childbearing, providing a qualitative meta-analysis of the empirical evidence. Mother’s education, family moves, poverty, and parent’s marital status are factors that are consistently found to be important determinants of teen birth outcomes. In a related work, Kalil and Kunz (1999) focus on cumulative risk factors that go beyond parental education, poverty, and marital status to address whether the sheer number of risk factors facing an adolescent can lead to a better understanding of the determinants of nonmarital childbearing. Upchurch et al. (2002) use a life cycle model that highlights joint decision-making. Their focus, however, is not on adolescents, and they find a significant effect of earlier childbearing on subsequent childbearing.
The authors attribute these unexpected results to the small sample of Black women, to the potential for underreporting of several of events in some of the stages of the sequence, or to the existence of different racial responses to incentives.
Moffitt (1998) criticizes the specification of the Clarke–Strauss model, arguing that their instrument, state per capita income, probably belongs in their core or main equation.
Rosenzweig’s variable reflecting the “real” value of welfare benefits is plagued by missing values due to the NLSY data that he uses and, hence, may mismeasure the benefits available to women who move during their teenage years. That variable may also confound welfare generosity with time-related changes in state-specific earning opportunities for low earnings, low ability, and minority youths, because this latter variable remains unmeasured. Hence, his reported welfare effect could also be interpreted as a response to market opportunities. Hoffman and Foster (1999) reexamine the effects of AFDC benefits on nonmarital childbearing through age 22. They use an alternative data source—the Michigan Panel Study of Income Dynamics (PSID)—that allows analysis that includes more cohorts and superior information on welfare benefit levels, parental characteristics, and measures of nonmarital births. While they are able to reproduce Rosenzweig’s main finding, they fail to find a “welfare effect” on teen nonmarital births, although they find a large effect on the choices of women in their early 20s. The finding of significant welfare effects, when both cohort and state fixed effects are controlled for, is at odds with other research relying on fixed effects estimation (Moffitt 1994; Hoynes 1997).
An alternative view is that assignment of the reference sample to the two nonmarital birth outcomes is the result of a selection process that reflects all relevant determinants of this choice, and that individuals in the primary sample know this selection process and account for it in forming their economic and marriage/cohabitation expectations conditional on the nonmarital birth choice. In this view, young single women are presumed to reliably discern the effects of determinants of the nonmarital childbearing choices that are unobserved by the researcher, but which may have influenced the fertility choice of individuals in the secondary (older) sample. We judge that assigning this level of insight to the teenage girls whose outcomes we study seems unwarrantedly strong; hence, our preferred results are based on economic and marriage/cohabitation expectations that are not adjusted for such potential selectivity. However, in a related work (see Haveman et al. 2001b), we compared the results from a model in which expectations are directly estimated and an alternative model in which the estimated income variables are selectivity-adjusted. The latter model statistically controls for this selection process in estimates of expected personal income conditional on the choice that is made, using a two-stage, Heckman-type selectivity correction model (see Heckman 1979). The estimated effects of expected income in selectivity-adjusted specification are very similar to those of the preferred specification that does not reflect the assumption that youths perceive the effects of unobserved factors in forming expectations.
Only those women who remained in the survey until age 19 are included. In a few cases, observations could not be used and are excluded from the analysis. These include persons with two or more contiguous years of missing data. Those observations with but 1 year of missing data were retained and the missing information was filled in by averaging the data for the 2 years contiguous to the year of missing data. For the first and last years of the sample, this averaging of the contiguous years is not possible. In this case, the contiguous year’s value is assigned, adjusted if appropriate using other information that is reported. Studies of attrition in the PSID indicate that erosion of the sample has reduced its representativeness. See Becketti et al. (1988), Lillard and Panis (1994), and Haveman and Wolfe (1994). A recent study by Fitzgerald et al. (1998), however, finds that, while “dropouts” from the PSID panel do differ systematically from those observations retained, behavioral responses estimated from the data do not appear to be significantly affected.
For the secondary sample, we use women who remained in the survey until age 29. An older sample of 728 women is used for estimating the expected “relationship stability” variables, of whom 132 gave birth as an unmarried teenager, and 596 did not. The sample for estimated expected income is 733 (missing data account for the difference in sample size), of whom 132 gave birth as an unmarried teenager, and 601 did not.
While alternative indexes could be used, the Census Bureau describes the Consumer Price Index (CPI) as the best measure for adjusting payments to consumers when the intent is to allow them to purchase, at today’s prices, the same market basket of consumer goods and services that they could purchase in an earlier reference period. “The CPI also is the best measure to use to translate retail sales and hourly or weekly earnings into real or inflation-free dollars.” http://www.bls.gov/cpi/cpifaq.htm.
For each state, we have annual data from 1968 to 1992 on the state maximum benefits for the Aid to Families with Dependent Children (AFDC) program, the maximum Food Stamp benefit, and the average Medicaid expenditures for AFDC families. In incorporating this information into our basic data set, we match maximum benefits (the maximum amount paid by the state as of July of that year to a family of four with no other income), for the years when the child is aged 6 to 21 (deflated by the personal consumption expenditure deflator). For Food Stamps, the benefit is the amount of the allotment (or the allotment minus the purchase requirement) for a family of four with no other income, again measured as of July of that year. Finally, average Medicaid expenditures for each state equal three times the state-specific fiscal year per child Medicaid expenditures for dependent children under 21 who are in categorically needy families plus the state-specific average per person annual Medicaid payments for adults in categorically needy families. These are deflated using the CPI for medical care. We thank Robert Moffitt for providing these data.
1984 values are an average of 1983 and 1985 values for each observation; 1986 values are an average of 1985 and 1987 values.
The matching was done by combining geographic codes added to the annual PSID data over the years 1968 to 1985 by the Michigan Survey Research Center to 1970 and 1980 Census data. Using 1970 and 1980 Census data, we assign neighborhood values to the neighborhood in which each family in the PSID lived to Census data. In most cases, this link is based on a match of the location of our observations to the relevant Census tract or block numbering area (67.8% for 1970 and 71.5% for 1980). For years before 1970, we use the 1970 data; for years after 1980, we use the 1980 data, while for years 1971–1979, we used a weighted combination of 1970 and 1980 data [weights are .9 (1970) and .1 (1980) for 1971; .8 (1970) and .2 (1980) for 1972 and so on].
For example, Bumpass and Sweet (1989) report that, while most cohabiting relationships are relatively short-lived, approximately 60% end in marriage. Bumpass et al. (1991) and Bumpass and Lu (2000) also address this issue and indicate that, while the percentage of first cohabitations that result in marriage to the same person is declining, the proportion of first unions begun by cohabitation has been steadily increasing.
We include in these equations variables likely to covary with relationship stability, including race, family position (if firstborn), parental education, family structure, mother’s employment, urban residence, region, family location changes, disability status of family head, family income, family welfare recipiency, being Catholic, whether mother was divorced, number of times mother was married, whether the state has no fault divorce laws, divorce rates in the state, and ratios of males to females. Most of these variables are measured over the girls ages 12–15. This range is determined by the 25 years of observations that are available. The vectors of independent variables used in the relationship regressions are virtually the same between the two teen childbearing groups (father a college graduate was omitted in the with teen birth regression due to insufficient observations). Most of the variables have similar and expected effects on relationship stability for the two groups. The definitions, means, and SD of these variables are shown in Table 5 of the Appendix. The estimated relationships are available from the authors.
Personal income is defined as the sum of the person’s own earnings, asset income, transfer benefits (AFDC, SSI, other welfare, Social Security, veterans benefits, other retirement/pensions, unemployment insurance, and worker compensation) and unearned income from all other sources (child support, help from relatives, and “other” income). We use personal income rather than individual earnings since neither transfer income (including welfare benefits) nor child support payments are contained in earnings, hence omitting an important component of the relevant expected economic well-being concept specified in our model. These income terms are pre-tax income. It would be ideal if we could obtain estimates of disposable income by adjusting for taxes, particularly since welfare income (which is likely a larger component of personal income if the woman chooses the childbearing option than if she does not) and earned income are subject to different tax regimes. However, while we recognize this shortcoming, we are unable to reliably adjust for tax liability with the available data.
Because additional children increase the level of family needs, the family income relative to needs variable assumes that these children reduce the mother’s utility if there were no associated change in her expected income. Furthermore, using family income relative to needs to proxy for utility in those cases in which the young woman lives with her parents implies that parental income increases the young woman’s utility and that there are no other utility costs associated with living in her parents home. Similarly, if the woman would marry or cohabit, this procedure would implicitly assume that all of the benefits of this living arrangement are reflected in the partner’s income and that any costs are reflected in the increase in family needs due to the addition of another adult.
Given our desire to use neighborhood and childhood data, we are constrained from using information for ages greater than 29 years. We included in these equations variables likely to be related to the personal income dependent variable, including race, family position (if firstborn), parental education, family structure, mother’s employment, urban residence, region, family location changes, disability status of family head, family income, family poverty status, family welfare recipiency, neighborhood median income and percent of neighborhood residents in high status occupations, percent neighborhood residents with low income, neighborhood unemployment rate characteristics, and state welfare generosity, median income and unemployment rates. Most of these variables are measured over the girls ages 12–15; the range is determined by the 25 years of observations that are available. The vectors of independent variables used in the income regressions do not vary between the two teen childbearing groups, with two exceptions: (1) parental education is limited to a high school or more in the with-birth estimation rather than the high school graduate and more than high school used in the no-birth estimations, and (2) state AFDC generosity is included only in the with-birth equations. The signs on some of the coefficients in the income regressions differ between the two groups. For example, growing up in a family consistently receiving welfare benefits is associated with higher income for those women who had a teen nonmarital birth, but is negatively related to income for those who did not have a teen birth. Father’s education is positively associated with income only in the estimates for those without a teen birth. Beyond these two examples, the variables have similar and expected effects on personal income for the two groups. The definitions, means, and SD of these variables are shown in Table 5 of the Appendix. The estimated relationships are available from the authors.
We also experimented with assigning age 29 income for an additional 10 years and using the earnings growth over the 10 years with observed income to create income for subsequent years. The basic results are invariant to these alternative measures of conditional income.
See Manski (1987) and Lee (1979). There is substantial overlap in the characteristics of the teen women who do and who do not give birth out of wedlock. A reduced-form model predicting this choice fails to explain a high proportion of the choices made, suggesting but limited self-selection in terms of the economic opportunities facing young women in their choices of child birth options. Adolescents with both low and high foregone income associated with giving birth are observed to both give birth and to refrain from giving birth. This avoids a potential identification problem in the use of these income expectation variables to explain the observed childbearing choice.
The women in our sample are teens with risk of a nonmarital teen birth during 1982–1990. During this period, the nonmarital birth rate among girls aged less than 20 was about 11% (Mosher and Bachrach 1996). The rate among African-American teens was far higher than among whites. Our statistic is not a rate per year but a cumulative rate over the teen years. African-Americans are oversampled in our data; we conclude that our higher rates are consistent with the observed rates.
Alternatively, a 25% increase in the marriage/cohabitation difference is associated with a decrease of nearly 12% in the probability of a teen nonmarital birth; a 25% increase in the log income difference is associated with a decrease of about 7% in the teen nonmarital birth outcome.
The standard errors in these probit estimations were not corrected for the use of a predicted value. According to Hsiao (1986), if the income equations are estimated over a sample that is independent of the sample used for the teen birth probit conditional on the regressors, then the standard errors do not need to be corrected.
As noted, since we define our income variable to be personal income, which excludes income from a partner, this variable may also be reflecting the added income of a spouse if married or a mate if cohabiting. Also, as noted, we have also estimated the effects of using a family income variable in sensitivity tests reported below.
Alternatively, the model suggests that if state family planning expenditures were increased by 25%, the rate of nonmarital childbearing would decrease from 0.078 to about 0.063, a reduction of over 19%; for Black women, the childbearing rate would fall from 0.270 to 0.214, or over 20%.
The family variables used in the income and stability tobit equations are measured over ages 12–15 of the women in the secondary sample, while these family variables are measured over the longer ages 6–15 period in the final stage estimation. We are assuming that childhood environment during the girl’s entire childhood affects teen childbearing, but only late childhood environment (ages 12 to 15) is related to her future income and future relationship stability.
Bound et al. (1995) indicate the importance of having instruments that are not just weakly correlated with the endogenous variable. Focusing on Angrist and Krueger’s (1995) use of quarter of birth as an instrument for years of education, they show how biased estimates can result if the correlation between the instruments and the endogenous variable is weak, even thought the estimated relationship is statistically significant (because of, say, large sample size). Thus, economic significance is important, as well as statistical significance. The economic significance of our instruments in determining income is well-supported by the literature (see Datcher 1982; Corcoran and Adams 1997). In OLS estimation of our income equations at age 29 for girls who had a birth, the R-squared indicator of correlation is 0.225 when only the instruments are included in the specification. The R-squared statistic is 0.293 for the regression with the full of set of explanatory variables. Likewise, R-square is 0.107 when only the instruments are included in the income without a birth equation, compared with 0.109 with the full set of variables. In the relationship stability with a teen birth equation, the R-squared statistic is 0.04 when we regress the stability index on the instruments alone. In contrast, the R-square statistic is 0.23 when all the explanatory variables are included; the instruments account for about 17% of the full R-square indicator. Our instruments have even more explanatory power in the relationship stability without a teen birth equation. In that equation, the R-square statistic is 0.07 when we use only the instruments as explanatory variables, compared with 0.11 with the full set of variables. In this case, the instruments account for about 60% of the full R-square indicator.
References
Angrist J, Krueger A (1995) Split sample instrument variables estimates of the return to schooling. J Bus Econ Stat 13(2):225–235
Becker G (1974) A theory of marriage: part II. J Polit Econ 82(2):S11–S26 (March–April)
Becker G, Landes E, Michael R (1977) An economic analysis of marital instability. J Polit Econ 85(6):1141–1187 (December)
Becketti S, Gould W, Lillard L, Welch F (1988) The PSID after fourteen years: an evaluation. J Labor Econ 6(4):472–492
Bound J, Jaeger D, Baker R (1995) Problems with instrumental variable estimation when the correlation between the instruments and the endogenous explanatory variable is weak. J Am Stat Assoc 90(430):443–450
Bronars SG, Grogger J (1994) The economic consequences of unwed motherhood: using twin births as a natural experiment. Am Econ Rev 84(5):1141–1156
Brooks-Gunn J, Duncan G, Klebanov PK, Sealand N (1993) Do neighborhoods influence child and adolescent behavior. Am J Sociol 99(2):353–395
Bumpass L, Lu H-h (2000) Trends in cohabitation and implications for children's family contexts in the U.S. Popul Stud 54(1):29–41
Bumpass L, Sweet J (1989) National estimates of cohabitation. Demography 24(4):615–625
Bumpass LL, Sweet JA, Cherlin A (1991) The role of cohabitation in declining rates of marriage. J Marriage Fam 53(4):913–927 (November)
Clarke G, Strauss R (1998) Children as income producing assets: the case of teen illegitimacy and government transfers. South Econ J 64(4):827–856
Corcoran M, Adams T (1997) Race, sex and the intergenerational transmission of poverty. In: Duncan G, Brooks-Gunn J (ed) Consequences of growing up poor. Russell Sage, New York, pp 461–517
Datcher L (1982) Effects of community and family backgrounds on achievement. Rev Econ Stat 64:32–41
Duncan G, Hoffman S (1990) Welfare benefits, economic opportunities, and out-of-wedlock births among Black teenage girls. Demography 27(4):519–535
Fitzgerald J, Gottschalk P, Moffitt R (1998) An analysis of sample attrition in panel data: the Michigan panel study of income dynamics. J Hum Resour 33(2):251–299
Geronimous A, Korenman S (1992) The socioeconomic consequences of teen childbearing reconsidered. Q J Econ CVII:1187–1214
Geronimous A, Korenman S (1993) The socioeconomic costs of teenage childbearing: evidence and interpretation. Demography 30(2):281–290
Hamilton B, Martin J, Ventura S, Sutton P, Menacker F (2005) Births: preliminary data for 2004. National Vital Statistical Reports 54:8. National Center for Health Statistics, Hyattsville, MD
Haveman R, Wolfe B (1994) Succeeding generations: on the effects of investments in children. Russell Sage, New York
Haveman R, Wolfe B (1995) The determinants of children's attainments: a review of methods and findings. J Econ Lit 33(4):1829–1878
Haveman R, Wolfe B, Peterson E (1996) Children of young childbearers as young adults. In: Maynard R (ed) Kids having kids: economic costs and social consequences of teen pregnancy. Urban Institute Press, Washington, DC
Haveman R, Wolfe B, Pence K (2001a) Intergenerational effects of non-marital and early childbearing. In: Wu L, Wolfe B (ed) Out of wedlock: causes and consequences of nonmarital fertility. Russell Sage, New York, pp 287–316
Haveman R, Wolfe B, Wilson K (2001b) The role of economic incentives in teenage nonmarital childbearing choices. J Public Econ 81:473–511
Haveman R, Sandefur G, Voyear A, Wolfe B (2004) Trends in children’s attainments and their determinants as family income inequality has increased. In: Neckerman K (ed) Social inequality. Russell Sage, New York, NY, pp 149–188
Heckman JA (1979) Sample selection bias as a specification error. Econometrica 4(1):153–161
Hoffman SD, Foster EM (1999) AFDC benefits and nonmarital births to young women. J Hum Resour 35(2):376–391 (Spring 2000)
Hoffman SD, Foster EM, Furstenberg F (1993) Reevaluating the costs of teenage childbearing. Demography 30(1):1–13
Hotz VJ, McElroy S, Sanders S (1997a) The impacts of teenage childbearing on the mothers and the consequences of those impacts for government. In: Maynard R (ed) Kids having kids: economic costs and social consequences of teen pregnancy. Urban Institute Press, Washington, DC, pp 55–94
Hotz VJ, Mullins C, Sanders S (1997b) Bounding caused effects using data from a contaminated natural experiment analyzing the effects of teenage childbearing. Rev Econ Stud 64(4):575–603 (October)
Hotz VJ, McElroy S, Sanders S (1999) Teenage childbearing and its life cycle consequences: exploiting a natural experiment. NBER Working Paper No. 7397, NBER, Cambridge, MA
Hoynes HW (1997) Does welfare play any role in female headship decisions? J Public Econ 65:89–117
Hsiao C (1986) Analysis of panel data. Cambridge University Press, Cambridge, MA
Johnson J, DiNardo J (1997) Econometric methods, 4th edn. McGraw-Hill, New York
Kalil A, Kunz J (1999) First births among unmarried adolescent girls: risk and protective factors. Soc Work Res 23(3):197–208
Lee L-F (1979) Identification estimation in binary choice models with limited (censored) dependent variable. Econometrica 47(4):977–996
Lillard LA, Panis CWA (1994) Attrition from the PSID: household income, marital status, and mortality. Rand Corporation, Santa Monica, CA
Long JS, Freese J (2006) Regression models for categorical dependent variables using Stata, 2nd edn. Stata Press
Lundberg S, Plotnick R (1995) Adolescent premarital childbearing: do economic incentives matter? J Labor Econ 13(2):177–200
Manski C (1987) Academic ability, earnings, and the decision to become a teacher: evidence from the national longitudinal study of the high school class of 1972. In: Wise DA (ed) Public sector payrolls. University of Chicago Press, Chicago, IL
Maynard R (ed) (1997) Kids having kids: the costs and consequences of teen childbearing. Urban Institute Press, Washington, DC
McElroy M (1985) The joint determination of household membership and market work: the case of young men. J Labor Econ 3(3):293–316 (July)
Moffitt R (1994) Welfare effects on female headship with area effects. J Hum Resour 29(2):621–636
Moffitt R (1998) The effect of welfare on marriage and fertility. In: Moffitt R (ed) Welfare, the family, and reproductive behavior: research perspectives. National Academy Press, Washington, DC, pp 50–97
Mosher W, Bachrach C (1996) Understanding the U.S. fertility: continuity and change in the national survey of family growth, 1988–1995. Fam Plann Perspect 28:4–11 (January/February)
Rosenzweig MR (1999) Welfare, marital prospects and nonmarital childbearing. J Polit Econ 107(6):S3–S32
Rosenzweig M, Wolpin K (1995) Sisters, siblings, and mothers: the effect of teen-age childbearing on birth outcomes in a dynamic family context. Econometrica 63(2):303–326
Upchurch D, Lillard LA, Panis CWA (2002) Nonmarital childbearing: influences of education, marriage, and fertility. Demography 39(2):311–329
Ventura SJ, Curtin SC, Mathews TJ (2000) Variations in teenage birth rates, 1991–98: National and State Trends. National Vital Statistics Report 48:6. National Center for Health Statistics, Hyattsville, MD. http://www.cdc.gov/nchs/data/nvs48_6.pdf
Wolfe B, McHugh E (2006) Health and medical care costs to society of teenage childbearing: children from birth to age 14. Mimeograph, Madison, WI
Wolfe B, Perozek M (1996) Teen children’s health and health care use. In: Maynard R (ed) Kids having kids: economic costs and social consequences of teen pregnancy. Urban Institute Press, Washington, DC
Acknowledgements
The authors gratefully acknowledge the contributions of Scott Niemann, Kathryn Wilson, Yuichi Kitamura, Susan Lee, Elaine Peterson, Dawn Duren and two anonymous referees of this journal. Special thanks go to Guilio Zanella for his assistance with calculations done for the final version of the paper. The views expressed in the paper are those of the authors, and should not be interpreted as those of the Congressional Budget Office, the Board of Governors of the Federal Reserve System, or its staff.
Author information
Authors and Affiliations
Corresponding author
Additional information
Responsible editor: Deborah Cobb-Clark
Appendix
Appendix
Rights and permissions
About this article
Cite this article
Wolfe, B., Haveman, R., Pence, K. et al. Do youth nonmarital childbearing choices reflect income and relationship expectations?. J Popul Econ 20, 73–100 (2007). https://doi.org/10.1007/s00148-006-0109-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00148-006-0109-4