Assessing the Impact of Birth Spacing on Child Health Trajectories

Global Development Policy Center HCI WORKING PAPER 001 • SEPT 2018

HUMAN CAPITAL INITIATIVE

MAHESH KARRA

ABSTRACT

We investigate the effect of birth spacing between siblings on child health trajectories and investments using longitudinal data from the Young Lives Study. We analyzed a birth cohort of over 7,000 children with data collected between 2002 and 2013 from four low- and middle-income countries. We found increased height among children who were more widely spaced relative to children who were narrowly spaced (within two years from an older sibling). However, we also found evidence of Mahesh Karra is an Assistant compensatory growth (converging height and HAZ scores) for closely spaced children as they aged. Professor of Global Development There was a positive association between birth spacing and prenatal investments, which suggests Policy at the Frederick S. that the emergence of height gaps could have been driven by parental behavior in addition to bio- Pardee School of Global Studies logical factors. However, we found little evidence that compensatory growth was driven by remedial at Boston University and health investments in closely spaced children after birth, suggesting catch-up growth as a biological Associate Director of the Human phenomenon. Available data suggests financial constraints and caregiver perceptions of child health Capital Initiative at the Global Development Policy Center. as possible explanations for lack of remedial investments. Keywords: birth spacing, health, nutrition, child investments, compensatory growth, height

HCI@GDPCenter www.bu.edu/gdp Pardee School of Global Studies/Boston University 1 Introduction

The importance of birth spacing for maternal and child health has been a long-standing source of interest to researchers and policymakers (World Health Organization, 2005). Evidence from systematic reviews and other empirical studies have suggested that short birth intervals (less than two years between births) may be associated with increased risk of maternal and child mortality and morbidity (Conde-Agudelo et al., 2006). These increased morbidity risks in childhood may, in turn, lead to lower educational attain- ment (Oreopoulos et al., 2008), poor labor market outcomes (Smith, 2009; Case et al., 2005), lower human capital and social status (Case et al., 2001), and lower earnings in adulthood (Case et al., 2005; Schultz, 2002). Examined morbidity risks have primarily concerned birth and early life outcomes including pregnancy-related complications (high blood pressure, pre-eclampsia), preterm birth, low birthweight, and small for gestational age. However, the evidence of birth spacing eﬀects on downstream child morbidity and health outcomes is scant and either weak or mixed (Dewey and Cohen, 2007). To our knowledge, no studies have investigated whether adverse early life health outcomes associated with short intrapartum spacing persist in older children, especially as they transition into adolescence. In this study, we investigate the eﬀect of birth spacing between siblings on child health outcomes (height and height-for-age) using longitudinal data that was collected on a cohort of children and their siblings in four low- and middle-income countries. As part of our analysis, we assessed whether the estimated impact of short intrapartum spacing changed for the sample as children aged. We also investigated potential mechanisms behind observed patterns by examining the relationship between birth spacing and parental investments in child health from conception to early adolescence.

2 Background

The relationship between short proceeding birth interval and high infant and child mortality is well established in a wide range of populations (DaVanzo et al., 2004). Conversely, there is little empirical evidence that directly assesses the links between birth intervals and child morbidity, which is surprising considering that the mechanisms through which birth intervals may inﬂuence child health and well-being have been extensively discussed in the literature (DaVanzo et al., 1983; Miller, 1991). In particular, the consequences of a short birth interval for child health outcomes, partic-

HCI@GDPCenter 2 www.bu.edu/gdp Pardee School of Global Studies/Boston University ularly at younger ages, have often been attributed to the physiological effects related to the “maternal depletion syndrome,” which postulates that the woman may not have fully recuperated from one pregnancy before supporting the next one (Conde-Agudelo et al., 2012; Dewey and Cohen, 2007). Other mechanisms that have been hypothe- sized to contribute to a detrimental effect of a short preceding interval include: 1) behavioral effects that are associated with competition between siblings, which may include competition for parental time or resources; 2) depleted parental investments or household resources that were used for the earlier birth, which may include a lack of physical resources or even a psychological or emotional inability to provide the later child with adequate attention if its birth came sooner than desired; and 3) larger mor- bidities through higher disease transmission among closely spaced siblings, particularly at younger ages (DaVanzo et al., 2004; Conde-Agudelo et al., 2012). The closest approximation of child morbidity effects from birth spacing is provided by studies that examine the relationship between indicators of childhood malnutrition (stunting, wasting, underweight) and family formation patterns. A systematic review by Dewey and Cohen (2007) assessed the evidence from 52 studies and noted that approximately half found that a previous birth interval of at least 36 months was associated with a 10 to 50 percent reduction in childhood stunting (similar for wasting), whereas the remaining studies found no association or were inconclusive. A more recent study by Fink et al. (2014), which pooled 153 Demographic and Health Surveys (DHS) surveys across 61 countries conducted between 1990 and 2011, found that birth intervals of less than 12 months and between 12 and 23 months were associated with higher risks for stunting (relative risks of 1.09 and 1.06) as compared to a 24 to 35 month inter-pregnancy interval. Due to the cross-sectional nature of the data, however, Fink et al. (2014) were limited in their ability to make inferences on the persistence of these associations in children over time. Our study aims to address the shortcomings in the literature in two ways. First, we improve on prior methodologies that have almost exclusively relied on cross-sectional data by exploiting a longitudinal dataset that allowed us to observe changes in birth spacing effects over stages of child development. Importantly, the cross-sectional structure of the DHS does not allow one to adequately control for birth cohort effects when examining health trajectories over childhood. Moreover, the DHS does not include height measures for children after age 5. Second, we explore the biological and behavioral mechanisms driving results by examining parental investment patterns on the

HCI@GDPCenter www.bu.edu/gdp 3 Pardee School of Global Studies/Boston University basis of birth spacing.

3 Data and Methods

3.1 Data

For our analysis, we used longitudinal data from the Young Lives Study (YLS), which investigates the determinants of childhood poverty and well-being (Oxford Depart- ment of International Development, 2017). As part of the YLS, data was collected on a cohort of children born between 2001 and 2002 from four low- and middle-income countries—Ethiopia, India, Peru, and Vietnam. The sampling design included selecting 20 communities in each country and randomly selecting 100 children from each. We used data available on approximately 8,000 children (2,000 from each country) over four survey waves that were conducted in 2002, 2006, 2009, and 2013, when children were approximately 1, 5, 8, and 12 years old. The study also collected information on household and child characteristics in each survey wave, including the anthropometric markers height and weight. Beginning in the third survey wave, anthropometric markers were also collected for a sibling of the primary cohort of children. We derived birth spacing by combining the household rosters collected during each wave of the survey. Date of birth was collected directly for the primary cohort and for siblings with anthropometric data. For the remaining siblings who were reported on the household rosters, reported age in years was combined with the date of interview to calculate an approximate date of birth. For those with an older sibling, we grouped spacing to next oldest sibling into four categories: spaced under 2 years, spaced 2 to 3 years apart, spaced 3 to 7 years apart, and spaced 7 years or more apart. A number of household and child characteristics were also used in analyses to help control for demographic and socioeconomic eﬀects on child health outcomes (see Appendix B for outcome and control variable construction details). The YLS consists of a total of 31,366 observations for the primary cohort of children summed across the four survey waves and four countries in the study. Of this sample, we dropped 0.3 percent of observations due to missing data on birth spacing and another 5.6 percent due to missing household or child characteristics (including height). This left us with a panel sample of 29,505 observations from 7,480 children born between 2001 and 2002.

HCI@GDPCenter 4 www.bu.edu/gdp Pardee School of Global Studies/Boston University 3.2 Empirical Speciﬁcation

We are primarily interested in examining the eﬀects of birth spacing on longitudinal health trajectories. In our main empirical speciﬁcation, we exploited the panel structure of the YLS by estimating the following model:

2 Yis = δsSpacei + βsXi + γsais + κsais + ηs + λsζi + εis (1)

where Yis is an outcome for child i measured in survey round s; Spacei is a categorical variable for the number of years to next oldest sibling (< 2 years is the reference group);

Xi is a vector of child-speciﬁc characteristics; ai is age in days at time of measurement;

ηs is a survey round intercept; ζi is an unobserved child-level random eﬀect; and εi is a random error term. This approach allows for comparison of eﬀects at ages 1, 5, 8, and 12, estimated longitudinally for a single birth cohort. Included in the time

invariant characteristics Xi are mother’s age and age squared at birth, wealth index, total number of siblings, caregiver’s education and dummies for sex, number of older siblings, season of birth1, and community of residence. It should also be noted that being a first-born child was included as a category in the birth spacing variable (Space). Note that coefficients are allowed to vary by survey wave to capture heterogeneity in 2 effects over childhood. Identification of coefficients on child random effects λs requires

a normalization, so we set λ1 =1. We also assume the error term is independent and identically distributed across individuals and independent across survey waves. As we want to examine effect changes over time, we include only children with complete data over all four survey waves for each outcome. This ensures sample composition changes are not influencing results. The coefficient of interest is that on birth spacing, δ. Interpretation of the coefficient of interest requires careful consideration. Effects estimated from this model can only be interpreted as causal if birth spacing is uncorrelated with any unobserved determinants of examined outcomes. It is clearly the case that geographic residence is likely to be correlated with both health outcomes and birth spacing, as access to family planning and other health services vary considerably across countries and locales. However, effects associated with geographic area were controlled for with the inclusion of community fixed effects. An additional concern is the existence of seasonal patterns

1Season of birth was controlled for with a month of birth by country dummy. 2We used Stata’s gsem command to estimate the system via maximum likelihood.

HCI@GDPCenter www.bu.edu/gdp 5 Pardee School of Global Studies/Boston University of fertility that correlate with our independent variables of interest. If, for example, pregnancies that are associated with shorter birth intervals are correlated with times of the year when food is relatively scarce, then results could be attributed to season of birth as opposed to birth spacing (e.g. Rayco-Solon et al. 2005; McEniry 2011; Miller 2017). Moreover, studies have documented seasonal patterns of fertility across a variety of countries (e.g. Rajagopalan et al. 1981; Panter-Brick 1996; Buckles and Munnich 2012). However, the inclusion of month by country of birth dummies controlled for seasonal eﬀects that occurred at the country level and that were independent of birth spacing.

3.3 Outcomes

We focus on three primary child health outcomes—height (cm), height-for-age, and stunting. Height based measures capture a child’s restricted growth potential associated with the chronic or long-term effects of health shocks and/or undernourishment and are an important predictor of later-life well-being and productivity (Schultz, 2002; Case et al., 2005; Heckman, 2007). Height-for-age is measured by a child’s WHO height-for-age z-score (HAZ), which is standardized against an international reference population sample.3 Stunting is a binary indicator that is defined by a child’s HAZ falling below two standard deviations from the WHO MGRS reference median height. Including stunting as an outcome allows for more direct comparison of our findings with the existing estimates from the literature. In addition to our primary health outcomes, we are also interested in understanding the underlying mechanisms driving the results of our analysis. To this end, we examined the association between birth spacing and additional measures related to parental investments in children. These include outcomes on prenatal and birth investments including level of prenatal care (Prenatal care) and indicators for place of delivery (Home birth), presence of a medical professional at birth (Pro at birth), and whether the pregnancy was reportedly wanted (Wanted). In order to understand parental investment responses to health after birth and over childhood, we also examined child weight-for-height (WFH) and the variety and frequency of meals. Weight-for-height—weight in grams divided by height in centime-

3As HAZ is standardized based on the age-speciﬁc standard deviation of height in the reference population—and this standard deviation increases with age—it is possible for two children to diverge in height over time but the gap in their respective HAZ scores to decrease. This is why we use both absolute height and HAZ as outcomes.

HCI@GDPCenter 6 www.bu.edu/gdp Pardee School of Global Studies/Boston University ters—is a measure associated with acute nutrition as it is more sensitive to short-term health inputs and environment. These analyses provide some suggestive evidence on the extent to which the observed eﬀects of birth spacing may have been operating through biological channels (e.g. “maternal depletion syndrome”) relative to behavioral mechanisms (e.g. competition for resources).

4 Results

4.1 Descriptive Statistics

Table 1: Descriptive statistics Mean SD Min Max Height (cm) 109.52 26.33 54.70 174.20 HAZ -1.33 1.17 -6.49 3.00 Stunted 0.27 0.44 0.00 1.00 Preceding birth interval <2 years 0.09 0.29 0.00 1.00 2-3 years 0.15 0.36 0.00 1.00 3-7 years 0.27 0.44 0.00 1.00 7+ years 0.08 0.27 0.00 1.00 First-born child 0.41 0.49 0.00 1.00 Older siblings 1 0.30 0.46 0.00 1.00 2 0.14 0.34 0.00 1.00 3+ 0.15 0.36 0.00 1.00 Male 0.52 0.50 0.00 1.00 Mom’s age at birth 25.52 6.09 11.00 53.00 Age (months) 79.38 48.27 3.72 163.00 Wealth index 0.53 0.21 0.00 1.01 Total siblings 2.08 1.76 0.00 11.00 Caregiver’s edu. 4.97 4.54 0.00 17.00 Ethiopia 0.24 0.43 0.00 1.00 India 0.25 0.44 0.00 1.00 Vietnam 0.26 0.44 0.00 1.00 Peru 0.24 0.43 0.00 1.00 Observations 29505 Individuals 7480 Source: Young Lives Study, young cohort. Sample of children with non-missing outcomes or covariates.

Table 1 presents descriptive statistics for the primary estimation sample. Average height was quite low, with 27 percent of the sample classiﬁed as stunted. Nine percent of the sample were spaced within 2 years of an older sibling, while 8 percent were

HCI@GDPCenter www.bu.edu/gdp 7 Pardee School of Global Studies/Boston University spaced 7 or more years apart. Children in the sample had an average of 2.08 total siblings (by the ﬁnal survey wave) and 41 percent were ﬁrst-born children. The sample was spread evenly over the four YLS countries. 4 .6 3 .4 2 HAZ gap HAZ .2 Height (cm) gap Height 1 0 0

1 5 8 12 1 5 8 12 Age Age Years to next oldest sibling Years to next oldest sibling 2-3 3-7 7+ 2-3 3-7 7+ .05 0 -.05 -.1 Share stunted gap stunted Share -.15 -.2 1 5 8 12 Age Years to next oldest sibling 2-3 3-7 7+

Figure 1: Birth spacing and average child height by age

Notes: Figure plots gap for each birth spacing category in average absolute height in centimeters (panel one) and height-for-age z-score (panel two). Gaps are relative to being spaced < 2 years from a older sibling.

Average height by spacing category is plotted by age in Figure 1. Raw average height gaps (both absolute centimeters and HAZ) at age 1 are small between those spaced less than 2 years and those space 2-3 years from older sibling. This gaps decline with age, even becoming negative by age 12. The gap for children spaced 3-7 years is substantially larger at age 1, but also declines with age. Finally, those spaced greater than 7 years from an older sibling have the largest average height gap. Moreover, the absolute gaps rises steadily over ages, though there is still some decline in the HAZ

HCI@GDPCenter 8 www.bu.edu/gdp Pardee School of Global Studies/Boston University gap. The ﬁnal panel shows that similar patterns hold when examining the share of children stunted in each spacing category. In the next section, we use the empirical model and plot how these patterns change when adjusted for confounding variables.

4.2 Main Results

The association between birth spacing and child outcomes at each age for the primary YLS birth cohort are presented in Figure 2 (point estimates in Appendix Table A1). The first panel plots point estimates and significance levels for the effect of birth spacing on absolute height measured in centimeters. At age 1, wider spacing was significantly associated with increased height. For example, relative to being spaced less than two years apart from one’s next oldest sibling, being spaced 3 to 7 years apart was associated with an increase in height of 0.84 cm at age 1—or about 17 percent of the standard deviation of age 1 height in the sample. There was a similar positive association between birth spacing and HAZ at age 1 (see panel two). However, the magnitude of the associations between birth spacing and child height—both absolute height and HAZ—declined over time. The only exception was absolute height for children who were very widely spaced (at least 7 years apart). Moreover, the effect of birth spacing only statistically persisted to age 12 for the most widely spaced group.4 We also formally tested if coefficients differed as children aged. For those spaced 2-3 from an older sibling, we cannot statistically reject that coefficients are equal at each age at conventional significance levels. However, for those spaced 3-7 years from an older sibling, we can reject the null hypothesis—that effects are equal at ages 1 and 12—at the 5% level for height (χ2 =4.58) and 1% level for HAZ (χ2 = 23.97).5 For children spaced >7 years, we can only reject the null (at the 5% level) for effects on HAZ (χ2 =6.38). The third panel of Figure 2 (and Appendix Table A3) plots results of logit regressions of model (1) with child stunting as the outcome. For children who were spaced within two years of an older sibling, the odds of stunting at age 1 were 23 percent higher than children spaced 2 to 3 years apart and 79 percent higher than children spaced 7 or more years apart. We again found that the spacing point estimates attenuated and become statistically insignificant as children became older. The effects also statistically differed at ages 1 and 12 for each of the three spacing groups.6 Taken together, 4Results are also robust to the exclusion of first-born children (see Appendix Table A2). 5All χ2 results from Wald test with one degree of freedom. 6Spaced 2-3 years at 5% level (χ2 =4.80), 3-7 years at 1% level (χ2 = 10.20), and >7 years at

HCI@GDPCenter www.bu.edu/gdp 9 Pardee School of Global Studies/Boston University .4 *** 1.5 *** *** *** *** .3

1 *** *** *** ***

.2 *** HAZ *** Height (cm) Height

.5 * * *** * .1 * 0 0 1 5 8 12 1 5 8 12 Age Age Years to next oldest sibling Years to next oldest sibling 2-3 3-7 7+ 2-3 3-7 7+ 1.2 1

.8 ** .6 *** *** Stunted odds ratio Stunted

.4 *** .2 1 5 8 12 Age Years to next oldest sibling 2-3 3-7 7+

Figure 2: Association between birth spacing and child height by age

Notes: Figure plots the estimated eﬀects of birth spacing category on absolute height in centimeters (panel one), height-for-age z-score (panel two), and odds of stunting (panel three). Eﬀects are relative to being spaced < 2 years from a older sibling. Stars indicate p-values—*** p<0.01, ** p<0.05, * p<0.1.

the observed attenuation of birth spacing eﬀects provides evidence of catch-up growth among more narrowly spaced children.

4.3 Prenatal and Childhood Investments

Table 2 presents the association between birth spacing and prenatal investments. Chil- dren who were more widely spaced received more prenatal care, were less likely to be born at home, were more likely to have a medical professional present at birth,

10% level (χ2 =3.06).

HCI@GDPCenter 10 www.bu.edu/gdp Pardee School of Global Studies/Boston University and were more likely to be from a reportedly wanted pregnancy. These differences in prenatal care suggest that the health benefits of increased birth spacing observed by age 1 could be partially driven by differential parental investment behavior. To explore this possibility further, we ran our benchmark specification with and without the inclusion of the prenatal investment variables (results in Appendix Table A4). The available investment variables had a mediating influence on the estimated effects of birth spacing, supporting a role for the parental investment mechanism. However, the mediation effect was generally mild, which suggests that biological factors may still play a significant role as a primary channel of influence.

Table 2: Prenatal investments and birth outcomes

Prenatal care Wanted Pro at birth Home birth

Space 2-3 1.125 1.257∗ 1.028 0.822 (0.106) (0.152) (0.135) (0.110) Space 3-7 1.255∗∗∗ 2.077∗∗∗ 1.325∗∗ 0.666∗∗∗ (0.110) (0.282) (0.187) (0.083) Space 7+ 1.466∗∗∗ 2.606∗∗∗ 1.850∗∗∗ 0.488∗∗∗ (0.166) (0.427) (0.360) (0.084) Obs 7029 7076 6521 7023 Odds ratios reported (except for birthweight). Robust standard errors (clustered at the community level) in parentheses, p-values—*** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, and dummies for ﬁrst-born, older siblings, sex, survey round, year and season of birth, and community.

In order to examine the association between birth spacing and nutritional investments after birth and over childhood, Figure 3 presents results from our empirical model with weight-for-height as the outcome (point estimates in Appendix Table A5). Wider birth spacing was associated with increased weight-for-height across all spacing groups suggesting wider spaced children continued to receive increased nutritional investments over childhood. Moreover, unlike primary height outcomes, this association did not attenuate with age. Gaps were also largest for the widest spacing group (7+ years), potentially explaining why absolute height gaps increased for this group over ages. In contrast to weight-for-height, more direct measures of nutritional inputs over childhood—frequency and variety of meals in past 24 hours—did not reveal a strong statistical relationship with birth spacing (save meal variety for very widely spaced children—results in Appendix Table A5). However, point estimates were almost all positive suggesting, if anything, wider spaced children received more nutritional inputs around the time of each survey. Overall, weight-for-height and meal based measures

HCI@GDPCenter www.bu.edu/gdp 11 Pardee School of Global Studies/Boston University 8 *** 6

*** 4

Weight-for-height *** ** *** *** *** ***

2 *** *** *** ** 1 5 8 12 Age Years to next oldest sibling 2-3 3-7 7+

Figure 3: Association between birth spacing and child weight-for-height by age

Notes: Figure plots the estimated effects of birth spacing category on weight-for-age. Effects are relative to being spaced < 2 years from a older sibling. Spikes indicate 95 percent confidence intervals.

provide no evidence that closely spaced children received more nutritional investments over childhood that allowed them to catch up in height. This supports the observed compensatory growth after age 1 as a biological phenomenon.

4.4 Heterogeneity

Appendix Tables A6 and A7 provide results separately for each of the four countries in the YLS. Overall, effects of birth spacing on outcomes were strongest for Ethiopia, followed by India and Vietnam; birth spacing effects were found to be the smallest in Peru. Appendix Tables A8 and A9 present results when analyses were conducted on sex-specific sub-samples. Overall, birth spacing effects were stronger for females, while males exhibited faster catch-up growth. As expected, the confidence intervals around most of the country- and sex-specific coefficient estimates were wider, particularly among those children who were most widely spaced.

5 Discussion

We used data on a cohort of over 7,000 children to assess the impact of birth spacing on health outcomes and investments over childhood. We found increased absolute height and relative HAZ score and lower stunting odds among children who were more widely

HCI@GDPCenter 12 www.bu.edu/gdp Pardee School of Global Studies/Boston University spaced compared to children who were more narrowly spaced (< 2 years). However, we also found evidence of compensatory growth (estimated gaps in height that converge to the null) for closely spaced children relative to children spaced 3 to 7 years from an older sibling. Moreover, our HAZ results indicate that closely spaced children outgrew even their most widely spaced counterparts (7+ years) relative to their projected height-for- age trajectory based on an international reference population. However, wider spaced children started taller (at age 1), and it is unclear whether closely spaced children will be able to completely close the height gap over time. The association between birth spacing and child growth varied across our four countries, and our stratified results provided evidence of slower catch-up growth for girls when compared to boys. Moreover, we found a positive association between birth spacing and prenatal care-seeking. Together, these results suggest that the effects of birth spacing on child development may be partially driven by parental investment behavior. However, our mediation analysis of prenatal investments suggests that biological factors continue to play an important role in explaining the emergence of height gaps by age 1. After age 1, our weight-for-height results suggest that closely spaced children did not receive significant additional nutritional investments over their childhood that would allow them to catch up in height to children who were more widely spaced. This empirical finding—catch-up growth without remedial investments—provides evidence of substitutability in the health production function. However, economic theory emphasizes that substitutability should be accompanied by compensatory investments (see Appendix C), which we did not observe in our data. There are several possible explanations that may serve to reconcile these two seemingly contradictory observations. We discuss four such possible explanations in the Appendix D.

5.1 Study Limitations

There are several important limitations to our study that warrant discussion. First, we found considerable attenuation over time in both height and height-for-age gaps across birth spacing groups, and the trajectory indicates a potential convergences of gaps to the null. However, given the relatively short (12-year) period over which our sample was observed, we are unable to say whether convergence is assured in the long-run, particularly as children enter into periods of rapid growth and development during adolescence.

HCI@GDPCenter www.bu.edu/gdp 13 Pardee School of Global Studies/Boston University Second, while our main panel analysis controlled for many child- and household- level characteristics, it is still conceivable that fertility patterns could correlate with additional unobserved characteristics of children or their families. We examined the robustness of our results by running a family fixed effects model specification using a pooled sample of YLS children and their siblings with available anthropometric data (see Appendix E). Overall, results from the pooled analysis provides no evidence that unobserved family characteristics are substantially biasing our panel model results. Moreover, it is important to recall that our main results are estimated longitudinally on a single birth cohort of children. Therefore, even if some residual confounding remains, it does not invalidate the fact that there was catch-up growth among children who were more narrowly spaced in our sample—nor does it invalidate evidence that supports compensatory growth as a biological as opposed to behavioral phenomenon. Lastly, while we observed possible convergence in height in our sample across birth spacing groups, disparities in other outcomes may persist. For example, several studies have found longer intrapartum spacing to be associated with improved school test scores in older siblings, though the effects were found to be minimal for younger siblings (Broman et al., 1975; Buckles and Munnich, 2012). Further investigation along this line is warranted in order to determine the extent to which gaps in other key outcomes of health and development may persist over time for children who are more closely spaced.

6 Conclusions

Our results indicate that short proceeding birth intervals are associated with dimin- ished height by early childhood. However, it may be possible for closely spaced children to catch up to those who are more widely spaced over time. Our ﬁndings cannot speak to the extent to which catch-up is possible across other child health and development outcomes; however our results suggest that interventions that aim to increase birth intervals, including family planning and reproductive health services, may still be important in improving stunting in children (particularly at early ages) as well as positively contributing to child health more generally. Finally, it is important that we continue to investigate the biological, social, and behavioral mechanisms by which adequate birth spacing might contribute to child health. A more thorough understanding of these pathways is essential for the development of eﬀective policies, programs, and

HCI@GDPCenter 14 www.bu.edu/gdp Pardee School of Global Studies/Boston University evidence-based interventions that seek to promote healthy growth and development from conception through adolescents.

References

Broman, S. H., P. L. Nichols, and W. A. Kennedy (1975). Preschool IQ: Prenatal and early developmental correlates. Hillsdale, NJ: Erlbaum Associates.

Buckles, K. S. and E. L. Munnich (2012). Birth Spacing and Sibling Outcomes. Journal of Human Resources 47 (3), 613–642.

Case, A., A. Fertig, and C. Paxson (2005). The lasting impact of childhood health and circumstance. Journal of Health Economics 24 (2), 365–389.

Case, A., D. Lubotsky, and C. Paxson (2001). Economic status and health in childhood: The origins of the gradient.

Conde-Agudelo, A., A. Rosas-Bermudez, F. Castano, and M. H. Norton (2012). Eﬀects of Birth Spacing on Maternal, Perinatal, Infant, and Child Health: A Systematic Review of Causal Mechanisms. Studies in Family Planning 43 (2), 93–114.

Conde-Agudelo, A., A. Rosas-Bermúdez, and A. C. Kafury-Goeta (2006). Birth spacing and risk of adverse perinatal outcomes: a meta-analysis. JAMA: The Journal of the American Medical Association 295 (15), 1809–1823.

Currie, J. and D. Almond (2011, January). Chapter 15 - Human capital development before age ﬁve. In D. Card and O. Ashenfelter (Eds.), Handbook of Labor Economics, Volume 4, pp. 1315–1486. Elsevier.

DaVanzo, J., W. P. Butz, and J.-P. Habicht (1983). How Biological and Behavioural Inﬂuences on Mortality in Malaysia Vary during the First Year of Life. Population Studies 37 (3), 381–402.

DaVanzo, J., A. Razzaque, M. Rahman, L. Hale, K. Ahmed, M. A. Khan, G. Mustafa, and K. Gausia (2004). The eﬀects of birth spacing on infant and child mortality, pregnancy outcomes, and maternal morbidity and mortality in Matlab, Bangladesh. Technical Consultation and Review of the Scientiﬁc Evidence for Birth Spacing.

HCI@GDPCenter www.bu.edu/gdp 15 Pardee School of Global Studies/Boston University Dewey, K. G. and R. J. Cohen (2007). Does birth spacing aﬀect maternal or child nutritional status? A systematic literature review. Maternal and Child Nutrition 3 (3), 151–173.

Fink, G., C. R. Sudfeld, G. Danaei, M. Ezzati, and W. W. Fawzi (2014, July). Scaling- Up Access to Family Planning May Improve Linear Growth and Child Development in Low and Middle Income Countries. PLOS ONE 9 (7), e102391.

Grossman, M. (1972). On the concept of health capital and the demand for health. The Journal of Political Economy 80 (2), 223–255.

Heckman, J. J. (2007, August). The economics, technology, and neuroscience of human capability formation. Proceedings of the National Academy of Sciences 104 (33), 13250–13255.

McEniry, M. (2011, March). Infant mortality, season of birth and the health of older Puerto Rican adults. Social Science & Medicine 72 (6), 1004–1015.

Miller, J. E. (1991). Birth Intervals and Perinatal Health: An Investigation of Three Hypotheses. Family Planning Perspectives 23 (2), 62–70.

Miller, R. (2017). Childhood health and prenatal exposure to seasonal food scarcity in Ethiopia. World Development 99, 350–376.

Oreopoulos, P., M. Stabile, R. Walld, and L. L. Roos (2008, January). Short-, Medium- , and Long-Term Consequences of Poor Infant Health An Analysis Using Siblings and Twins. Journal of Human Resources 43 (1), 88–138.

Organization, W. H. (1995). Physical Status: The Use and Interpretation of Anthro- pometry. WHO Technical Report Series 854, World Health Organization, Geneva, Switzerland.

Oxford Department of International Development (2017). The Young Lives Study.

Panter-Brick, C. (1996, July). Proximate Determinants of Birth Seasonality and Con- ception Failure in Nepal. Population Studies 50 (2), 203–220.

Rajagopalan, S., P. K. Kymal, and P.-a. Pei (1981). Births, Work, and Nutrition in Tamil Nadu, India. In Seasonal Dimensions to Rural Poverty (Chamber R., Longhurst R., Pacey A. ed.)., pp. 156–62. London: Frances Pinter.

HCI@GDPCenter 16 www.bu.edu/gdp Pardee School of Global Studies/Boston University Rayco-Solon, P., A. J. Fulford, and A. M. Prentice (2005, January). Diﬀerential eﬀects of seasonality on preterm birth and intrauterine growth restriction in rural Africans. The American Journal of Clinical Nutrition 81 (1), 134–139.

Rosenzweig, M. R. (1986). Birth Spacing and Sibling Inequality: Asymmetric Infor- mation within the Family. International Economic Review 27 (1), 55–76.

Rosenzweig, M. R. and K. I. Wolpin (1988). Heterogeneity, Intrafamily Distribution, and Child Health. The Journal of Human Resources 23 (4), 437–461.

Schultz, T. P. (2002). Wage gains associated with height as a form of health human capital. Yale Economic Growth Center Discussion Paper (841).

Smith, J. P. (2009). The Impact of Childhood Health on Adult Labor Market Out- comes. The Review of Economics and Statistics 91 (3), 478–489.

Svirydzenka, K. (2016). Introducing a New Broad-based Index of Financial Develop- ment. IMF Working Papers 16 (5).

World Health Organization (2005). Report of a WHO Technical Consultation on Birth Spacing. Technical Report, World Health Organization, Geneva, Switzerland.

HCI@GDPCenter www.bu.edu/gdp 17 Pardee School of Global Studies/Boston University Appendix A: Additional Tables

Table A1: Birth spacing and child height

Height HAZ Age 1 Age 5 Age 8 Age 12 Age 1 Age 5 Age 8 Age 12

Space 2-3 0.268∗ 0.185 0.316 0.172 0.107∗ 0.037 0.057 0.024 (0.160) (0.206) (0.250) (0.320) (0.064) (0.043) (0.043) (0.046) Space 3-7 0.840∗∗∗ 0.568∗∗∗ 0.459∗ 0.274 0.338∗∗∗ 0.122∗∗∗ 0.080∗ 0.040 (0.169) (0.213) (0.254) (0.291) (0.066) (0.045) (0.044) (0.042) Space 7+ 0.984∗∗∗ 1.142∗∗∗ 1.257∗∗∗ 1.373∗∗∗ 0.396∗∗∗ 0.242∗∗∗ 0.219∗∗∗ 0.198∗∗∗ (0.217) (0.289) (0.359) (0.403) (0.086) (0.061) (0.063) (0.058) Ind 7101 7101 Robust standard errors (clustered at the community level) in parentheses, p-values—*** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, wealth index, number of siblings, caregiver’s education, and dummies for ﬁrst-born, older siblings, sex, survey round, year and season of birth, and community.

Table A2: Birth spacing and child height: excluding ﬁrst-born children

Height HAZ Age 1 Age 5 Age 8 Age 12 Age 1 Age 5 Age 8 Age 12

Space 2-3 0.279∗ 0.229 0.312 0.185 0.111∗ 0.048 0.055 0.025 (0.162) (0.204) (0.246) (0.315) (0.064) (0.043) (0.043) (0.045) Space 3-7 0.886∗∗∗ 0.690∗∗∗ 0.543∗∗ 0.418 0.356∗∗∗ 0.147∗∗∗ 0.094∗∗ 0.061 (0.175) (0.217) (0.256) (0.287) (0.069) (0.046) (0.044) (0.041) Space 7+ 1.055∗∗∗ 1.343∗∗∗ 1.512∗∗∗ 1.705∗∗∗ 0.424∗∗∗ 0.284∗∗∗ 0.264∗∗∗ 0.246∗∗∗ (0.238) (0.308) (0.372) (0.428) (0.094) (0.065) (0.065) (0.061) Ind 4181 4181 Robust standard errors (clustered at the community level) in parentheses, p-values—*** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, wealth index, number of siblings, caregiver’s education, and dummies for ﬁrst-born, older siblings, sex, survey round, year and season of birth, and community.

HCI@GDPCenter 18 www.bu.edu/gdp Pardee School of Global Studies/Boston University Table A3: Birth spacing and stunting

Age 1 Age 5 Age 8 Age 12

Space 2-3 0.775∗∗ 0.863 1.008 1.246 (0.094) (0.176) (0.369) (0.259) Space 3-7 0.520∗∗∗ 0.505∗∗∗ 0.652 0.949 (0.064) (0.112) (0.304) (0.174) Space 7+ 0.501∗∗∗ 0.344∗∗∗ 0.220 0.863 (0.103) (0.116) (0.214) (0.224) Ind 7101 Odds ratios reported. Robust standard errors (clustered at the community level) in parentheses, p-values— *** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, wealth index, number of siblings, caregiver’s education, and dummies for ﬁrst-born, older siblings, sex, survey round, year and season of birth, and community.

Table A4: Mitigation eﬀect of prenatal investment variables

Age 1 Age 5 Age 8 Age 12 Age 1 Age 5 Age 8 Age 12 Space 2-3 0.235 0.153 0.294 0.132 0.227 0.129 0.264 0.111 (0.164) (0.208) (0.259) (0.344) (0.164) (0.210) (0.260) (0.350) Space 3-7 0.804∗∗∗ 0.546∗∗∗ 0.410 0.222 0.771∗∗∗ 0.454∗∗ 0.308 0.131 (0.161) (0.204) (0.254) (0.302) (0.164) (0.206) (0.258) (0.309) Space 7+ 1.010∗∗∗ 1.257∗∗∗ 1.420∗∗∗ 1.600∗∗∗ 0.968∗∗∗ 1.131∗∗∗ 1.286∗∗∗ 1.469∗∗∗ (0.213) (0.293) (0.358) (0.421) (0.216) (0.296) (0.367) (0.426) Low prenatal care 0.138 0.402∗ 0.146 0.526 (0.133) (0.213) (0.258) (0.331) Mid prenatal care 0.124 0.399∗ 0.221 0.263 (0.126) (0.219) (0.255) (0.312) High prenatal care 0.327∗∗ 0.760∗∗∗ 0.498 0.514 (0.161) (0.270) (0.338) (0.357) Wanted 0.098 0.509∗∗∗ 0.554∗∗∗ 0.436∗∗ (0.088) (0.143) (0.165) (0.176) Pro at birth 0.236 0.311∗ 0.375∗ 0.445∗ (0.145) (0.178) (0.227) (0.267) Home birth 0.064 0.142 -0.010 -0.075 (0.120) (0.167) (0.219) (0.310) Ind 6567 6567 Robust standard errors (clustered at the community level) in parentheses, p-values—*** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, wealth index, number of siblings, caregiver’s education, and dummies for ﬁrst-born, older siblings, sex, survey round, year and season of birth, and community. No prenatal care is the reference group for level of prenatal care.

HCI@GDPCenter www.bu.edu/gdp 19 Pardee School of Global Studies/Boston University Table A5: Childhood nutritional investments

Weight-for-height Meal frequency Meal variety Age 1 Age 5 Age 8 Age 12 Age 5 Age 8 Age 12 Age 5 Age 8 Age 12

Space 2-3 1.720∗∗∗ 1.896∗∗∗ 2.700∗∗∗ 3.520∗∗ 0.034 0.029 -0.043 0.134∗ 0.056 -0.099 (0.583) (0.629) (0.840) (1.435) (0.054) (0.038) (0.039) (0.082) (0.077) (0.111) Space 3-7 2.180∗∗∗ 1.628∗∗ 2.999∗∗∗ 5.226∗∗∗ 0.045 0.018 0.009 0.052 0.082 -0.005 (0.496) (0.632) (0.737) (1.503) (0.052) (0.033) (0.037) (0.072) (0.070) (0.113) Space 7+ 3.074∗∗∗ 2.454∗∗∗ 3.623∗∗∗ 7.469∗∗∗ 0.095 0.075 0.025 0.231∗∗∗ 0.209∗∗ 0.005 (0.698) (0.796) (1.223) (2.349) (0.060) (0.052) (0.060) (0.084) (0.093) (0.146) Ind 7001 6694 7233 Robust standard errors (clustered at the community level) in parentheses, p-values—*** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, wealth index, number of siblings, caregiver’s education, and dummies for ﬁrst-born, older siblings, sex, survey round, year and season of birth, and community.

Table A6: Birth spacing and child height: India and Ethiopia

India: Height India: HAZ Ethiopia: Height Ethiopia: HAZ Age 1 Age 12 Age 1 Age 12 Age 1 Age 12 Age 1 Age 12

Space 2-3 0.268 -0.112 0.115 -0.017 0.644∗ 0.485 0.252∗ 0.070 (0.276) (0.492) (0.110) (0.071) (0.383) (0.599) (0.147) (0.086) Space 3-7 0.835∗∗∗ 0.282 0.342∗∗∗ 0.043 1.681∗∗∗ 0.685 0.671∗∗∗ 0.104 (0.260) (0.498) (0.103) (0.071) (0.415) (0.610) (0.161) (0.088) Space 7+ 0.671 -0.176 0.287 -0.023 2.185∗∗∗ 1.276 0.874∗∗∗ 0.188 (0.466) (0.823) (0.185) (0.118) (0.599) (1.059) (0.233) (0.153) Ind 1823 1823 1706 1706 Robust standard errors (clustered at the community level) in parentheses, p-values—*** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, wealth index, number of siblings, caregiver’s education, and dummies for ﬁrst-born, older siblings, sex, survey round, year and season of birth, and community.

Table A7: Birth spacing and child height: Vietnam and Peru

Vietnam: Height Vietnam: HAZ Peru: Height Peru: HAZ Age 1 Age 12 Age 1 Age 12 Age 1 Age 12 Age 1 Age 12 Space 2-3 0.128 0.476 0.052 0.056 0.064 0.160 0.024 0.025 (0.318) (1.040) (0.127) (0.145) (0.284) (0.690) (0.117) (0.100) Space 3-7 0.709∗∗∗ 0.593 0.284∗∗∗ 0.076 -0.030 -0.481 -0.003 -0.069 (0.269) (0.817) (0.102) (0.115) (0.234) (0.548) (0.094) (0.079) Space 7+ 0.836∗∗∗ 1.904∗∗ 0.344∗∗∗ 0.269∗∗ 0.354 2.053∗∗ 0.130 0.294∗∗ (0.311) (0.843) (0.119) (0.120) (0.429) (0.822) (0.178) (0.119) Ind 1826 1826 1746 1746 Robust standard errors (clustered at the community level) in parentheses, p-values—*** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, wealth index, number of siblings, caregiver’s education, and dummies for ﬁrst-born, older siblings, sex, survey round, year and season of birth, and community.

HCI@GDPCenter 20 www.bu.edu/gdp Pardee School of Global Studies/Boston University Table A8: Birth spacing and child height, females only

Height HAZ Age 1 Age 5 Age 8 Age 12 Age 1 Age 5 Age 8 Age 12

Space 2-3 0.478∗ 0.167 0.443 0.186 0.201∗ 0.034 0.079 0.029 (0.268) (0.299) (0.368) (0.469) (0.103) (0.060) (0.063) (0.068) Space 3-7 0.938∗∗∗ 0.658∗∗ 0.804∗∗ 0.719 0.365∗∗∗ 0.137∗∗ 0.139∗∗ 0.105 (0.264) (0.302) (0.355) (0.448) (0.100) (0.062) (0.061) (0.065) Space 7+ 1.185∗∗∗ 1.345∗∗∗ 1.819∗∗∗ 1.909∗∗∗ 0.453∗∗∗ 0.282∗∗∗ 0.315∗∗∗ 0.278∗∗∗ (0.374) (0.437) (0.521) (0.648) (0.143) (0.090) (0.090) (0.095) Ind 3394 3394 Robust standard errors (clustered at the community level) in parentheses, p-values—*** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, wealth index, number of siblings, caregiver’s education, and dummies for ﬁrst-born, older siblings, survey round, year and season of birth, and community.

Table A9: Birth spacing and child height, males only

Height HAZ Age 1 Age 5 Age 8 Age 12 Age 1 Age 5 Age 8 Age 12 Space 2-3 0.128 0.215 0.190 0.249 0.044 0.046 0.032 0.034 (0.187) (0.259) (0.332) (0.481) (0.076) (0.056) (0.058) (0.067) Space 3-7 0.808∗∗∗ 0.458∗ 0.103 -0.114 0.336∗∗∗ 0.101∗ 0.018 -0.015 (0.183) (0.271) (0.356) (0.435) (0.074) (0.058) (0.063) (0.061) Space 7+ 0.883∗∗∗ 0.981∗∗ 0.815∗ 1.112∗ 0.378∗∗∗ 0.214∗∗∗ 0.145∗ 0.156∗ (0.239) (0.381) (0.487) (0.572) (0.099) (0.081) (0.086) (0.080) Ind 3707 3707 Robust standard errors (clustered at the community level) in parentheses, p-values—*** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, wealth index, number of siblings, caregiver’s education, and dummies for ﬁrst-born, older siblings, survey round, year and season of birth, and community.

Table A10: Caregiver perceptions of child size

Birthsize Weight Height Age 1 Age 5 Age 1 Age 5 Space 2-3 1.154 1.134 1.158 1.044 0.890 (0.115) (0.097) (0.105) (0.089) (0.071) Space 3-7 1.151 1.156∗ 1.059 1.167∗∗ 0.962 (0.104) (0.098) (0.088) (0.083) (0.081) Space 7+ 1.327∗∗ 1.336∗∗ 1.122 0.931 1.040 (0.147) (0.167) (0.132) (0.105) (0.115) Obs 7191 7259 7387 7249 7396 Odds ratios reported. Robust standard errors (clustered at the community level) in parentheses, p-values—*** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, wealth index, number of siblings, caregiver’s education, and dummies for ﬁrst-born, older siblings, sex, survey round, year and season of birth, and community.

HCI@GDPCenter www.bu.edu/gdp 21 Pardee School of Global Studies/Boston University Appendix B: Variable Construction

Height/HAZ/Stunted: We used height (cm) provided directly in the survey data sets. We used height-for-age z-scores (HAZ) directly from the data for survey rounds 1/2. For rounds 3/4, we used raw height to calculate HAZ using the Stata program provided on the WHO website. Included in analysis were all observations within WHO’s recommended HAZ flexible exclusion range (Organization, 1995). Specifically, we dropped observations whose z-scores were more than plus or minus five from the mean of the sample for each age, or whose HAZ was greater than three. Such extreme values are believed by the WHO to be measurement error. Community: Binary indicator for reported community of residence in the first round of the survey. Male: Binary indicator for reported sex in latest available survey round. Mother’s Age at Birth: Derived from the combined houshold rosters. Wealth Index: The study provided a constructed wealth index based on sub-indices of housing quality, access to services, and consumer durables. We used the YLS constructed household wealth index from the fourth survey round. Caregiver’s Education: Years of education of the primary caregiver reported in the first round of the survey. Total Siblings: Derived from the combined houshold rosters. Older Siblings: Derived from the combined houshold rosters, top coded at having three or more older siblings. Meal Frequency: In the last three survey rounds data was collected on the frequency of eating in the past 24 hours (or a “normal” day). Seven yes/no questions were asked on if the child ate any food before breakfast, breakfast, food between breakfast and midday meal, midday meal, food between midday meal and evening meal, evening meal, and food after the evening meal. Our variable is a simple sum of these frequencies. Meal Variety: In the last three survey rounds data was collected on the variety of food eaten in the past 24 hours (or a “normal” day). Yes/no questions were asked for eating each of up to 20 food groups depending on the survey round and country (i.e. eggs, cheese/milk, cactus). Our variable is a simple sum of the total number of categories reported eaten. Prenatal Care: We used the level of antenatal care variable provided in the first round data set. The YLS study team constructed the variable as follows: a mother that reported no antenatal care was given a zero. For those who had antenatal visits, one was added if the first visit was when they were four months pregnant or before, one was added if the mother had five or more visits in total,

HCI@GDPCenter 22 www.bu.edu/gdp Pardee School of Global Studies/Boston University and one was added if the mother was given tetanus injections. This gave a value between zero and three for all mothers. Wanted: This variable is an indicator that takes a value of one if the parent responded yes to following: “At the time you became pregnant with ‘NAME’ did you want to become pregnant?” Professional at Birth: This variable is an indicator that takes a value of one if the parent reported a doctor, nurse, or midwife was present during the child’s delivery. Home Birth: This variable is an indicator that takes a value of one if the parent reported the child’s delivery was at home. Perceived Birthsize: Caregiver’s perception of birth size was collected in round one: “When ’Name’ was born, was he/she very large, large, average, small, or very small?” This variable takes values between one (very small) and ﬁve (very large). Perceived Height: Caregiver’s perception of comparative height was collected in survey rounds 2-4: “Compared to other child of this age would you say ’Name’s’ is the same height, taller, or shorter?” This variable takes values between one (shorter) and three (taller). Perceived Weight: Caregiver’s perception of comparative weight was collected in survey rounds 2-4: “Compared to other child of this age would you say ’Name’s’ is the same weight, heavier, or lighter?” This variable takes values between one (lighter) and three (heavier).

HCI@GDPCenter www.bu.edu/gdp 23 Pardee School of Global Studies/Boston University Appendix C: Theoretical Framework for Parental Invest- ment

One of our key aims is to understand the extent to which parental investment behavior may alter any biological eﬀects of short birth spacing over childhood. The general model of human capital formation proposed by Heckman (2007) provides a useful framework for analyzing the parental investment response to birth spacing eﬀects on child health. Consider a stock of child health capital that evolves over time in response to parental health investments:

ht+1 = ft (ht,It) ,

where ht is health capital and It are general investments made in the child when they are t years old. Health investments are a perfect substitute for the existing health stock when ∂ft(ht,It) =0. With perfect substitution and standard preferences over own ∂ht∂It consumption and child’s health, Currie and Almond (2011) show that parents have an incentive to compensate low health stocks with additional investments ∂It < 0 . In ∂ht contrast, there is “dynamic complementarity” when the return to health investments are increasing in the stock of existing health capital: ∂ft(ht,It) > 0. The stronger the ∂ht∂It complementarity, the more incentive parents have to reinforce existing health stocks ∂It > 0 . For example, if a closely spaced child is of poor health at age 1, parents ∂ht may shift resources to other siblings where marginal returns to health investments are higher. The model’s predicted evolution of health over time depends on the strength of dynamic complementarity. Consider a cohort of children with differing initial health stocks due to differential birth spacing. If complementarity is strong, the theory predicts reinforcing investments and a divergence in health within the cohort over time. If there is substitutability, the theory predicts compensatory investments with the potential for converging health over time. It is also possible that parental investments do not respond to the initial health stock and are instead equal for all children across the cohort ∂It =0 . We may ∂ht expect this to be the case if differences in health stocks are unobservable or if parent’s directly value equitable investments, for example across peers or siblings. With equal investments and strong dynamic complementarity, the model predicts a divergence in health within the cohort. In contrast, if there is adequate substitutability, the initial

HCI@GDPCenter 24 www.bu.edu/gdp Pardee School of Global Studies/Boston University diﬀerences in health will persist but will not grow over time, and may fade away as in the widely used Grossman (1972) model of health capital.

HCI@GDPCenter www.bu.edu/gdp 25 Pardee School of Global Studies/Boston University Appendix D: Subgroup Findings and Mechanisms

Economic theory emphasizes that substitutability should be accompanied by compensatory investments, which we did not observe in our data. We propose four possible explanations that may serve to reconcile these two seemingly contradictory observations. Unobserved correlates Firstly, it is possible that unobservable differences between closely and widely spaced children were driving differential investments in child health. In order to examine the robustness of our results to additional confounding variables, we used data collected on siblings of the primary birth cohort in the YLS to compare birth spacing effects across sibling pairs in the same family (see details in Appendix E). Even in our family fixed effects model, there continued to be a significant positive association between birth spacing and parental investment (weight-for-height).

Financial constraints Secondly, it is possible that some families were unable to op- timally compensate closely spaced children due to constraints on available resources. This seems a viable potential explanation given that birth spacing effects were more pronounced in India and Ethiopia, where there are less developed ﬁnancial institutions (Svirydzenka, 2016).

Parental perceptions Thirdly, it may be that parents were not able to easily observe the adverse effects of short spacing and, as a result, did not see a need for improving their child’s growth through compensatory investments. In order to explore this possibility, we examined the association between birth spacing and caregiver perceptions of child size (see Appendix Table A10). Caregivers perceived wider spaced children to be larger at birth, though the relationship was only statistically significant for those children who are spaced more than 7 years apart. Similarly, there was some evidence that caregivers believed wider spaced children to be heavier and (to a lesser extent) taller than other children at age 1. However, these relationships became weaker by age 5, particularly for height. While these findings indicate that parents may have perceived closely spaced children as smaller at birth and at age 1, we found less evidence that parents perceived closely spaced children to be shorter at older ages. It therefore may be that the bulk of parental investments to compensate for poorer growth among

HCI@GDPCenter 26 www.bu.edu/gdp Pardee School of Global Studies/Boston University closely spaced children are provided between ages 1 and 5 and that we simply do not have the necessary data within this time frame to observe these behaviors.

Noisy investment measure Finally, it may be that weight-for-height is too blunt of a short-term measure of nutritional and other remedial investments. However, we found that closely spaced children received no more investments than widely spaced children across other possible measures of investment in our data—meal frequency and variety—which does not support this explanation. Nonetheless, it is possible that a more precise measure of investment or speciﬁc types of investments may exhibit negative associations with birth spacing. In particular, we lack data on parental time and emotional investments.

While here we have proposed several possible explanations for limited evidence of compensatory parental investments in our analyses, it is clear that additional research is needed to convincingly disentangle the biological and behavioral channels through which birth spacing may alter childhood growth and development.

HCI@GDPCenter www.bu.edu/gdp 27 Pardee School of Global Studies/Boston University Appendix E: Family Fixed Eﬀects Model

We employ a secondary statistical model that relies on within-family sibling com- parisons of birth spacing for identification. This approach serves to minimize residual confounding by adjusting for all time-invariant factors that remain constant within the family. Of the 7,480 children included in our primary panel sample, 915 (12%) were excluded from the sibling sample because they were an only child and 1,906 (25%) were excluded because no sibling anthropometric data was available.7 This left a sibling sample of 27,180 observations from 4,659 unique sibling pairs. Compared to the panel sample, a somewhat higher 14 percent of the sibling sample were spaced within two years of an older sibling (see Table A11). Given that this sample excluded single child families by construction, the average number of siblings was also higher than the panel sample at 2.58. The sample was also somewhat skewed towards countries with higher overall fertility rates, namely Ethiopia and India. We estimated the following family fixed effect model:

2 Yifs = δSpaceif + βXif + γaifs + κaifs + ηs + ζi + θf + εifs

where Yifs is an outcome for child i from family f measured in survey round s; θf is a family fixed effect; and other independent variables are as previously defined. Due to collinearity with the family fixed effect, we dropped the household wealth index, total number of siblings, and caregiver’s education from the vector of child-level

characteristics, Xif . This approach controls for any remaining permanent unobserved correlation between a child’s family and the spacing measures by comparing children within the same family. As data was only collected for siblings of the primary cohort after round two (and only if they were over 2-3 years old) we do estimate round specific coefficients. The limitation of this model is that we cannot observe trends in effects over ages, but can check robustness of the overall height and investments gradients in birth spacing. Results from the model are presented in Table A12. The first column shows birth spacing effects on height without the inclusion of a family fixed effect. Relative to being spaced within 2 years of an older sibling, being spaced 2 to 3 years apart was associated with a 0.56 cm increase in a child’s height, while being spaced at least 7

7Markers were collected for closest aged younger sibling provided they were over two years old for Peru and India or over three years old for Ethiopia and Vietnam. If no such younger sibling was present, data was collected on closest aged older sibling (except for Peru).

HCI@GDPCenter 28 www.bu.edu/gdp Pardee School of Global Studies/Boston University Table A11: Descriptive statistics: sibling sample Mean SD Min Max Height (cm) 114.34 25.57 54.70 182.00 HAZ -1.41 1.19 -6.49 3.00 Stunted 0.29 0.46 0.00 1.00 Preceding birth interval <2 years 0.14 0.35 0.00 1.00 2-3 years 0.21 0.41 0.00 1.00 3-7 years 0.33 0.47 0.00 1.00 7+ years 0.04 0.19 0.00 1.00 First-born child 0.28 0.45 0.00 1.00 Older siblings 1 0.36 0.48 0.00 1.00 2 0.16 0.37 0.00 1.00 3+ 0.20 0.40 0.00 1.00 Male 0.51 0.50 0.00 1.00 Mom’s age at birth 25.40 5.74 12.00 52.00 Age (months) 90.02 49.94 5.06 228.83 Wealth index 0.51 0.21 0.00 1.00 Total siblings 2.58 1.80 1.00 11.00 Caregiver’s edu. 4.11 4.36 0.00 17.00 Ethiopia 0.30 0.46 0.00 1.00 India 0.34 0.47 0.00 1.00 Vietnam 0.20 0.40 0.00 1.00 Peru 0.16 0.37 0.00 1.00 Observations 27180 Individuals 9318 Sibling Pairs 4659 Source: Young Lives Study, young cohort. Sample of children with non-missing outcomes or covariates.

years apart was associated with a 1.79 cm increase.8 The second column shows results when the family fixed effect was added to the previous model specification. There was a moderate increase in coefficient estimates when moving from the simple OLS to the family fixed effect specification for all but the most widely spaced group. For example, the relative effect on child height of being spaced 3 to 7 years from an older sibling increased from 1.05 to 1.28 cm with the inclusion of family fixed effects. However, the confidence intervals around the latter estimates were also wider than those from simple OLS, particularly for larger birth intervals where there were generally fewer observations. Columns 3 and 4 show a similar pattern holds for HAZ scores.9

8The higher point estimates compared to the panel models is due to sample selection. For example, the pooled sibling sample had more children from Ethiopia and India where birth spacing effects were stronger. 9There was not enough within family variation to estimate the sibling fixed effect model for

HCI@GDPCenter www.bu.edu/gdp 29 Pardee School of Global Studies/Boston University We also ran the family fixed effect model with weight-for-height to check for robustness of results. Birth spacing continued to have a strong positive association with weight-for-height using the pooled sibling sample. However, in contrast to height measures, spacing effects on weight-for-height declined with the inclusion of the family fixed effect.

Table A12: Family ﬁxed eﬀects model

Height Height HAZ HAZ WFH WFH

Space 2-3 0.565∗∗∗ 0.730∗∗∗ 0.104∗∗∗ 0.120∗∗∗ 2.551∗∗∗ 2.186∗∗ (0.204) (0.239) (0.035) (0.044) (0.751) (1.105) Space 3-7 1.059∗∗∗ 1.287∗∗∗ 0.206∗∗∗ 0.249∗∗∗ 3.801∗∗∗ 3.146∗∗ (0.230) (0.315) (0.037) (0.056) (0.761) (1.379) Space 7+ 1.795∗∗∗ 1.634∗∗ 0.339∗∗∗ 0.331∗∗∗ 5.196∗∗∗ 1.570 (0.315) (0.715) (0.068) (0.112) (1.391) (2.520) Sibling FE No Yes No Yes No Yes Obs 27180 27180 27180 27180 27089 27089 Robust standard errors (clusted at the community level) in parentheses, p-values— *** p<0.01, ** p<0.05, * p<0.1. Additional independent variables in all regressions: mother’s age at birth, mother’s age at birth squared, age (months), age squared, and dummies for ﬁrst-born, older siblings, sex, survey round, and year and season of birth. OLS regressions also include wealth index, number of siblings, caregiver’s education, and community dummies.

Overall, results from the pooled sibling sample provides no evidence that unobserved family characteristics are substantially biasing our panel model results. How- ever, a common methodological criticism in literature relying on family fixed effects is the inability to adequately account for within-family heterogeneity in unobservable characteristics (Rosenzweig and Wolpin, 1988; Rosenzweig, 1986). More specifically, the impact of birth spacing on child health outcomes is likely to be driven by a wide number of individual, social, and contextual determinants, and these determinants may vary from birth to birth within the same family as parents choose to differentially invest in children who may be differentially spaced (Fink et al., 2014). For these reasons, our use of a family fixed effect would not be sufficient in adjusting for any residual confounding that is driven by differential birth timing decisions.

stunting.

HCI@GDPCenter 30 www.bu.edu/gdp Pardee School of Global Studies/Boston University Boston University [email protected] Global Development Policy Center 53 Bay State Road www.twitter.com/gdpc_bu Boston, MA 02215 www.bu.edu/gdp

HUMAN CAPITAL INITIATIVE

The Human Capital Initiative (HCI) is a research inititiative at Boston University’s Global Development Policy Center. The GDP Center is a University wide center in partnership with the Frederick S. Pardee School for Global Studies. The Center’s mission is to advance policy- oriented research for financial stability, human wellbeing, and environmental sustainability.

www.bu.edu/gdp

The views expressed in this Working Paper are strictly those of the author(s) and do not represent the position of Boston University, or the Global Development Policy Center.

HCI@GDPCenter Pardee School of Global Studies/Boston University