Birth Spacing and Sibling Outcomes - PowerPoint Presentation

Birth Spacing and Sibling Outcomes Kasey S. Buckles, University of Notre Dame Elizabeth L. Munnich, University of Notre Dame

Birth SpacingQuestion: How does the spacing (in time) between siblings affect their outcomes? In our study, we: Investigate the effects of spacing on children’s test scores Use OLS and an instrumental variables strategy that uses miscarriages between siblings as an IV for spacingEstimate effects separately for older and younger children 2

Birth SpacingEvidence on the effects of spacing would: Add to our understanding of the effects of family structure --“Child spacing patterns have received relatively less attention in the literature, even though differences in the timing of births may well be the most important aspect of fertility differentials in low-fertility societies.” (Wineberg and McCarthy, 1989)Inform parents making decisions about spacing—over which they have some controlInform public policy 3

4

Source: http://cchealth.org/topics/birth_spacing/ 5

Source: USAID Issue Brief6

Potential MechanismsPhysiological/Development Effects Nutritional depletion (Smits and Essed 2001, Eijsden et al. 2008)Autism (Cheslack et al. 2011)Parental Investments Time (Price 2008) Financial resources ( Steelman et al. 2002, Kidwell 1981) Complementarities/Economies of Scale Between closely-spaced (Jones 2009) Between further-spaced ( Zajonc 1976) Effect on Parents Post-partum depression more common with short pregnancy intervals ( Gurel and Gurel 2000) Mothers’ career interruptionsSibling Rivalry 7

DataNational Longitudinal Survey of Youth, 1979 Survey of 12,686 youth, ages 14-22, in 1979 Detailed fertility histories for women Link to child outcomes from the NLSY Child and Young Adult surveyAge-adjusted math and reading score from Peabody Individual Achievement Test (PIAT)SampleWomen with at least two live births before 2006Gaps of < 10 Years Use up to six live births (five sibling pairs) per mother Omit those with more than one pregnancy loss, or with abortions after the first pregnancy 5,067 observations, where each observation is a sibling pair 8

Distribution of Gap Between Child 1 & 2,NLSY79 and Natality Detail Files, 1986 NLSY79 Natality, 1986 Mean gap: 40.8 monthsMedian: 34 monthsPercent <2 Years: 24% Mean gap: 40.8 months Median: 34 months Percent <2 Years: 28% 9

Table 2: Summary Statistics10

Estimation: OLSEstimated model: Score is = β0 + β1*gapi + X s β 2 + Z i β 3 + u is where s indexes the sibling (older or younger) and i indexes the sibling pair. Score is = age-adjusted, normalized PIAT score in Math, Reading (Cognitive)gapi = spacing between siblings in months (days/30), ln(spacing in months) or a dummy = 1 if gap is < Y years. X s = a vector of child-specific characteristics, including gender, race, birth order, and year-of-birth dummies Z i = a vector of pair-specific characteristics, including mother’s age and marital status at first birth, number of total children (by 2006), education, and AFQT score. Estimates are weighted by NLSY child sampling weights. Standard errors are clustered at the family level. 11

Table 3: OLS Estimates of Effect of Spacing on Test Scores for OLDER SiblingsPanel A: Reading Scores12 Std . errors in parenthesis, clustered at family level. All regressions also include controls for child gender and birth order, mother’s education, race, and AFQT score, total fertility, age and marital status at first birth, and child year-of-birth dummies. NLSY child weights are used .

Table 4: OLS Estimates of Effect of Spacing on Scores for YOUNGER SiblingsPanel A: Reading Scores 13 Std . errors in parenthesis, clustered at family level. All regressions also include controls for child gender and birth order, mother’s education, race, and AFQT score, total fertility, age and marital status at first birth, and child year-of-birth dummies. NLSY child weights are used.

OLSResults: Greater spacing associated with slightly better test scores for olders and little or no effect for youngers. Issues with OLS: Potential bias if unobserved characteristics of mother or child are correlated with spacing (Rosenzweig 1986, Rosenzweig & Wolpin 1988). 14

Miscarriages as an Instrumental VariableIdea: Use miscarriages (and stillbirths) that occur between two live births as an instrument for sibling spacing. A miscarriage is a pregnancy that is lost before the 20 th week of pregnancy. 10-20% of confirmed pregnancies end in miscarriage.Most due to chromosomal abnormalities of the fetus.Mechanically increases the time to the next birth (weeks at miscarriage + time to conceive).Previous research has used miscarriage as an instrument for fertility timing: - Hotz , McElroy, and Sanders (1996, 2005), Miller (2009) - Hotz , Mullin, and Sanders (1997) show that miscarriage is an appropriate instrumental variable for women who experience random miscarriages, and use this instrument to explore the effect of teenage childbearing on teen mothers’ outcomes. 15

Distribution of Gap in NLSY79, by Miscarriage6.4% of gaps contain a miscarriage (recall that we omit women with >1 pregnancy loss). 16

Miscarriages as IV: IssuesMiscarriages might actually be abortions, or be “latent abortions” (Lang and Ashcraft 2006). - Less likely for our sample, with live births bracketing the miscarriage. - As in Hotz, Mullin, and Sanders (1997), we assume that underreporting of miscarriages is random with respect to child outcomes; to the extent that women underreport miscarriages randomly, this would downward bias our estimates . Miscarriages might be correlated with unobservables that affect test scores, such as community-level factors, risky behaviors, or maternal health (Fletcher & Wolfe, 2008). - Again, possibly less likely for our sample - Can control for known risk factors 17

Table 5: Coefficients from Probit Regression of Miscarriage on Pre-existing Characteristics 18 . . . Pseudo R-squared: 0.0206Wald Chi-Squared: 31.11 (p-value 0.1505)

Miscarriages as IV: IssuesMiscarriages might have a direct effect on the well-being of live-born children. - Medical literature suggests that miscarriage is associated with depression, anxiety, but that these symptoms decrease over time (Thapar and Thapar, 1992; Janssen, et al., 1996; Hughes, Turton, and Evans, 1999). - Additionally, women who have a healthy pregnancy following a miscarriage or stillbirth may be at decreased risk for depressive symptoms (Swanson, 2000; Theut, et al., 1989). - Miscarrying appears to have no effect on investment or early life outcomes in subsequent children (Armstrong, 2002; Theut, et al., 1992). Importantly, the evidence seems to suggest that a miscarriage would have (if anything) a negative effect on child outcomes, which would bias our estimates of the effect of spacing downward.19

Table 6: First Stage Estimates20 Clustered standard errors in parenthesis

Table 7: IV Estimates of Effect of Spacing on Reading Scores*, ** indicate significance at 10% and 5%. Std. errors in parenthesis, clustered at family level. All regressions also include controls for child gender and birth order, mother’s education, race, and AFQT score, total fertility, age and marital status at first birth, and child year-of-birth dummies. NLSY child weights are used . Effect of birth order on IQ is 0.2-0.25 SD (BDS 2007 & Bjerkedal et al. 2007) 21

22 Table 8: IV Estimates of Effect of Spacing in Years on Reading Scores, for Selected Subsamples

23 Table 9: IV Estimates of Effect of Spacing on Inputs into Child Quality for Older Child

ConclusionsWe use miscarriages as an instrument for birth spacing to consider the effects of spacing on achievement test scores. IV specifications find that spacing benefits older children A one-year increase in spacing increases reading scores for older child by 0.17 SD Spacing of less than two years decreases scores by 0.65 SD. There is no evidence of a negative effect for younger children. Spacing could be an important channel through which family structure influence child outcomes 24

Future Work Consider additional outcomes: Early behavioral measures Health limitations, height and weightAccidents, illnesses, injuriesRisky behaviors25