## Abstract

The United States compares unfavorably with other high-income countries in infant mortality, which recent literature has attributed to the poor birth outcomes among disadvantaged (i.e., unmarried and less-educated) mothers. Describing and decomposing the trend of the concentration of infant mortality among disadvantaged mothers thus provides important clues for improving birth outcomes. We develop the infant mortality disadvantage index (IMDI) to measure such concentration. Using the 1983–2013 Birth Cohort Linked Birth and Infant Death data, we show that although the IMDI—as a measure of mortality inequality—was persistently higher for Blacks than Whites, the trends were different between the two groups. The IMDI declined for Black women; for White women, however, it increased in the 1980s, then plateaued until the early 2000s, and declined thereafter. We then use Das Gupta’s decomposition method to assess the contribution of five demographic/social factors (age, education, marriage, fertility, and infant mortality) to the IMDI trend. Nonmarital fertility among women with less than 12 years of education contributed most to Whites’ changing IMDI; for Blacks, a shrinking proportion of the less-educated group and declines in infant mortality among disadvantaged mothers contributed to their declining IMDI. These findings explicate links between population-level compositional changes and infant mortality inequality.

## Introduction

Infant mortality rate (IMR) has long been the focus of much demographic, sociological, and epidemiological research. Capturing the risk of death by age 1, IMR is an important health and well-being indicator in itself. Further, as a social indicator sensitive to public investments and policy interventions (Gonzalez and Gilleskie 2017), IMR has been widely used to assess maternal health, quality of and access to medical care, socioeconomic conditions, and social well-being (Judge et al. 1998; Nersesian 1988). Therefore, much academic, public health, and policy effort has aimed to reduce infant mortality. For example, researchers have identified contributing factors, such as low birth weight (Fanaroff et al. 2007), preterm birth (Saigal and Doyle 2008), and sleep routines (Moon 2011), which have laid the groundwork for interventions or policy changes to improve the quality of prenatal care and promote infant health awareness and education.

A major challenge in reducing IMR, nevertheless, is the uneven distribution of infant deaths as a function of social and demographic factors. That is, some infants are more susceptible to death than others. In a comparative analysis of infant deaths from 2000 to 2005 in the United States, Austria, and Finland, Chen et al. (2016) showed that the less-than-desirable U.S. infant mortality performance is driven primarily by high IMR among disadvantaged groups. Indeed, infants born to White, college-educated, and married U.S. mothers have similar levels of mortality to their counterparts in Europe. In other words, the unfavorable international standing of the United States in infant mortality is largely driven by the substantial inequality in IMR between women with different socioeconomic and racial backgrounds (for an earlier discussion, see Singh and Yu 1995).

We argue that an analysis of the temporal trend of the inequality in infant mortality, as well as how population-level compositional changes and demographic processes contribute to the trend, may provide important clues for more effectively reducing IMR in the United States. Inequality can be defined in different ways, but our approach focuses on the concentration of infant mortality among disadvantaged groups. Specifically, we develop an inequality index—the infant mortality disadvantage index (IMDI), defined as the proportion of infant deaths occurring to unmarried mothers with less than 12 years of education. As we describe in detail later, this index measures the degree of infant mortality inequality by revealing the concentration of disadvantage. Thus, it is closely tied to the notion of social inequalities in health as defined by Braveman et al. (2000:233): “a disproportionate share of ill-health and premature mortality is borne by the socially disadvantaged.” Moreover, this index is decomposable, meaning that it can be separated into social and demographic sources that potentially give rise to the observed trends, thereby shedding light on the various upstream determinants of health inequalities.

Drawing on the 1983–2013 National Center for Health Statistics (NCHS) Birth Cohort Linked Birth and Infant Death data and the 1983–2013 Current Population Surveys (CPS), we examine two research questions: (1) how has the IMDI, or the concentration of infant deaths in disadvantaged women, changed over time, and (2) what are the social and demographic processes that may account for the trend in the IMDI? In answering these questions, our research contributes to the infant health literature in three ways. First, our research provides a distinct and useful angle to understand social inequality in infant mortality by, for the first time, considering the concentration of the undesirable health outcome of infant deaths in disadvantaged groups. Second, through decomposition, our findings show the extent to which changing demographic and social processes have contributed to the trends in infant mortality disadvantage concentration, thereby explicating links between structural and compositional changes and infant mortality inequality. The multiple social and demographic forces that we examine help identify the critical factor(s), as well as the ways in which they have shaped the concentration of infant deaths for Whites and Blacks from the 1980s to the 2010s in different ways, which are informative for policy-making. Third, although we do not directly examine White-Black disparities, we perform separate decomposition analyses for these two racial groups because they differ greatly in prevalence of nonmarital births, educational attainment, social support network, and so on (Cramer 1995; Frisbie et al. 1996; Frisbie et al. 2004). Our research reveals race-dependent heterogeneity in the ways through which demographic and social factors affect the concentration of infant deaths among disadvantaged groups.

## Background and Motivation

A large volume of literature has documented the relationship between social disadvantages and infant mortality at the individual level (e.g., Chen et al. 2016; Elder et al. 2016; Gage et al. 2013; Hummer et al. 1999; Partridge et al. 2012; Powers 2013; Singh and Kogan 2007; Singh and Yu 1995). By contrast, little research has been conducted at the population level to examine how aggregate-level infant mortality inequality is shaped by structural and compositional factors. To fill this research gap, we first develop a measure of inequality in infant mortality—the infant mortality disadvantage index (IMDI)—defined as the proportion of infant deaths that occurs to unmarried women with less than 12 years of education. Three reasons motivate the development of this index.

First, although inequality in infant mortality has been the subject of many investigations, most studies and public health monitoring systems have not employed measures of inequality that convey societal distributions of concentrations of deprivation (Friedman et al. 2005; Krieger et al. 2016; World Health Organization 2014). For example, often used in the infant mortality disparity literature are measures such as relative ratios or absolute differences of IMRs between more and less advantaged groups (e.g., Elder et al. 2014). These indicators are useful for many purposes, but they overlook the population-level distribution of advantaged and disadvantaged social groups. For example, the sizes of the groups being compared (advantaged vs. disadvantaged mothers) may not be the same and may change over time, leading to different population-level public health implications, which are not captured in those measures. As a complement to these absolute or relative ratio measures, we use the IMDI to reveal the extent to which an undesirable life event—infant deaths—is concentrated into the group at the extreme of disadvantage; a value of 0 means that none of the infant deaths occur to the deprived group, and a value of 1 means that 100% of the infant deaths occur to the deprived group. Because disadvantaged groups typically have higher IMR than advantaged groups, other things being equal, a higher IMDI corresponds to a higher overall IMR.

Second, our index captures the concentration of infant mortality among disadvantaged mothers: unmarried women with less than 12 years of education. We focus on this group of women because this group has been shown to be the major driving force of the poor performance in infant health in the United States relative to other countries (Chen et al. 2016). To improve the U.S. international ranking in IMR, therefore, more attention needs to be paid to these disadvantaged mothers. Our index, by focusing on these disadvantaged women, provides a useful indicator for tracking mortality disparities specific to this group, with the ultimate goal of improving the overall IMR.

We recognize that the focal group under examination can be defined in at least two alternative ways. One is to examine the concentration of infant mortality among advantaged mothers (e.g., married women with a college degree). However, we focus on disadvantaged women because it leads to clearer policy implications. By definition, a more equal IMR can be achieved by either increasing the IMR for the advantaged women or reducing the IMR for the disadvantaged women, but it is not even conceivable to suggest any policy for the former. A second alternative for defining the focal group is to add race as another dimension of disadvantage (e.g., unmarried African American women with less than 12 years of education). We choose not to include race in the definition of the index because doing so would obscure racial disparities with socioeconomic status (SES) disparities; doing so would also mask any potential racial heterogeneity in responsible factors for the IMDI. As we show later, the trends of the IMDI differ drastically between White and Black women, which would have been missed were the two groups combined.

A third motivation of this index is that, as we elaborate in the Methods section, through decomposition, our index bridges population distribution and infant mortality inequality. Specifically, proportions of infant deaths among disadvantaged mothers can be reformulated as the product of a set of sociodemographic processes, including age, education, marriage, fertility, and infant mortality. We are therefore able to demonstrate how disparities in infant health have manifested and changed as a consequence of the underlying population dynamics. Further, decomposition allows us to incorporate and pay attention to distal social factors that affects health and health inequality. For instance, when upstream social policy factors affect the nature of social stratification (e.g., reducing the number of minimally educated individuals), our measure is able to reflect that change. By incorporating the effects of demographic and social changes that occur over time, therefore, our index is helpful for thinking about upstream determinants of health inequalities.

## Trends and Contributions of Sociodemographic Factors to the IMDI

Our first research question asks how the IMDI, or the concentration of infant deaths into disadvantaged women, changed over time for White and Black women. We then conduct decomposition analysis to answer the second research question regarding how demographic and social changes give rise to the temporal trends in the IMDI for White women and Black women, respectively.

Specifically, drawing on the decomposition method developed by Das Gupta (1993), we investigate the relative contribution of 12 factors that represent trends in five sociodemographic domains: age, educational attainment, education-specific marital rates, fertility rates by education and marital status, and infant mortality rates by education and marital status. We choose these five sociodemographic dimensions because they have been frequently investigated in previous infant mortality studies and are often used to explain racial or class inequalities in infant mortality (e.g., Gage 2013; Hummer et al. 1999; Lhila and Long 2012; Partridge et al. 2012; Powers 2013; Singh and Kogan 2007). Further, the past few decades have witnessed profound changes in these sociodemographic domains, with distinct racial or class patterns. Less understood, however, are the ways in which these sociodemographic changes contribute to the changing patterns of infant mortality concentration, or any differential patterns in the IMDI between White and Black women, a research gap we begin to fill. We describe trends in the five social and demographic processes during the past three decades and discuss how the shifting patterns may have impacted the IMDI. We detail how to obtain their relative contribution to the IMDI in the Methods section.

### Age Distribution

The U.S. age distribution for women of reproductive age (aged 15–44 years) has changed over time. Specifically, an aging trend is observed through the 1990s, mostly due to the large size of the Baby Boomer population, who were aged 19–37 in 1983 and aged 27–45 in 1991. Thereafter, more women of childbearing age are from younger age brackets. In terms of racial differences, this shift in the age distribution—first toward an older population and then toward a younger population—is slightly more pronounced for White than for Black women. Because age is an important demographic factor that influences all other sociodemographic trends we examine—educational attainment, marriage, fertility, and infant mortality—all the other factors are disaggregated by age.

### Educational Attainment

Women’s educational attainment has increased over the latter half of the twentieth century (DiPrete and Buchmann 2013), especially among Blacks. As increasingly more women finish high school than ever, other things being equal, the proportion of infant deaths born to mothers with less than 12 years of education is expected to decline, especially for Black women.

### Proportions Married by Educational Attainment

It is well documented that an ever-increasing proportion of U.S. women either do not enter marriage or get married at a much later age (Cherlin 2010; Goldstein and Kenney 2001). Given that this trend may vary by educational attainment (Goldstein and Kenney 2001), we examine two factors related to marriage: age-specific marriage rates among women with less than 12 years of education (less-educated), and age-specific marriage rates among women with at least 12 years of education (more-educated). In terms of the impact on the IMDI, everything else being constant, if marriage rates declined faster for less-educated than more-educated women, an increasing proportion of infant deaths would be expected to occur to unmarried, less-educated mothers.

### Fertility by Women’s Educational Attainment and Marital Status

Another demographic trend relevant to the IMDI is the diverging fertility trends by the intersection of educational attainment and marital status. Correspondingly, we examine fertility rates among four groups: (1) married women with at least 12 years of education, (2) unmarried women with at least 12 years of education, (3) married women with less than 12 years of education, and (4) unmarried women with less than 12 years of education. Increasingly, U.S. women do not have children by the end of their childbearing years regardless of race (Dye 2010), and there is a divide in this trend of childlessness based on marital status. Fertility among married women remains largely stable and may have declined somewhat over time; nonmarital birth rates, however, have increased since the middle of the twentieth century (Ellwood and Jencks 2004; Smock and Greenland 2010; Wu and Wolfe 2001), and nonmarital births are increasingly concentrated among less-educated women (Ellwood and Jencks 2004; McLanahan and Percheski 2008). As for racial differences, nonmarital birth rates are much higher for Black than White women (Martin et al. 2012). However, whereas nonmarital childbearing appears to have stabilized for Black women since the 1990s, White women have seen increasing nonmarital fertility rates. Taken together, other things being equal, the rising fertility rates among unmarried, less-educated women—a trend that is more pronounced for White relative to Black women—is expected to increase the IMDI.

### Infant Mortality Rates by Mothers’ Educational Attainment and Marital Status

Compared with babies born to more-educated and married mothers, those born to less-educated and unmarried mothers are more likely to suffer from prematurity, low birth weight, and mortality before their first birthday (Bird et al. 2000; Singh and Kogan 2007). Although innovations in perinatal care and technology have led to declines in IMR for all groups regardless of maternal education or marital status (Gortmaker and Wise 1997; Powers 2013; Wise 2003), the extent to which the relative gap in IMR between less-educated, unmarried mothers and mothers in other groups (e.g., more-educated, married mothers) has changed over time has received scant attention. If the decline in IMR is not as fast for babies born to less-educated, unmarried mothers as for their counterparts born to more-educated, married mothers, the IMDI will increase as a consequence.

In sum, this research develops an inequality measure, the IMDI, to capture the concentration of infant deaths among disadvantaged women. We provide empirical evidence on how this index has changed over time as well as how demographic and social changes have shaped the temporal trends in the IMDI. We perform separate analysis for White and Black women. And although our focus is on overall infant mortality, we also conduct decomposition analysis for neonatal and postneonatal infant mortality, in view of the finding that postneonatal mortality contributes more to the substantially higher IMR in the United States relative to other countries (Chen et al. 2016).

## Data, Measures, and Methods

### Data

Our analyses draw on the 1983–2013 Current Population Surveys (CPS) and the 1983–2013 National Center for Health Statistics (NCHS) Birth Cohort Linked Birth and Infant Death data sets. Every year since 1962, the CPS, conducted by the Census Bureau and the Bureau of Labor Statistics, collects a variety of demographic, social, and economic information. The CPS data we use come from the Integrated Public Use Microdata Series Current Population Survey (IPUMS-CPS) (Flood et al. 2015). The IPUMS-CPS data are weighted to compute the distributions of age, education, and marriage for White and Black women that are valid at the national level.

The NCHS Birth Cohort Linked Birth and Infant Death data are a complete census of births for years 1983–2013, linked to infant deaths occurring within one year of birth. The data provide rich information on parents and infants, including infants’ conditions at birth as well as the social and demographic information of the mother. Note that the NCHS did not produce Linked Birth and Infant Death data from 1992 to 1994, so our analyses are not performed for these three years.

To maximize consistency in our definition of White and Black populations over time, we do not distinguish Hispanics and non-Hispanics because Hispanic origin was not reported in all states until 1995. For example, in 1988, items requesting Hispanic or ethnic origin were included on the birth certificates of only 30 states and the District of Columbia (National Center for Health Statistics 1994). Because an important aim of this research is to document trends over time, further distinguishing Hispanic origin would result in a loss of nine years of data. We therefore follow previous studies that used the same data (e.g., Loggins and Andrade 2014) to define White and Black populations—that is, not distinguishing Hispanics and non-Hispanics.1

### Factors Used in Decomposition

Our decomposition analysis considers 12 factors that represent trends in five sociodemographic domains. Because of our interest in examining differences between White and Black women, the following factors are calculated separately for these two racial groups.

Age distribution is percentages of women in different age groups. We examine the following age groups that taken together are typically defined as childbearing age: 15–19, 20–24, 25–29, 30–34, 35–39, and 40–44. Shares of these groups are estimated from the 1983–2013 CPS data. We do not include the 45–49 age group because of their small number of births, but adding this group yields almost identical results (see section D of the online appendix). Given the concern that the 15–19 age group coincides with the typical ages when high school is finished, we conducted another sensitivity analysis in which we excluded this age group, finding largely similar results (see section D of the online appendix).2

Percentage with less than 12 years of education by age is the percentage of women in each age group who have less than 12 years of education. For simplicity, we sometimes use “<12 years” or “less-educated” to refer to women with less than 12 years of education, and “12+ years” or “more-educated” to refer to women with at least 12 years of education. We use the 1983–2013 CPS data to estimate these percentages.

Percentage unmarried by age and education is the percentage of women, in each age group, who are unmarried at the time of the survey. We calculate this measure separately for less-educated women and more-educated women. Unmarried women are defined as those who are never married, divorced/separated, or widowed. These statistics are estimated from the 1983–2013 CPS data.

Age-specific fertility rates are four sets of fertility rates calculated for four groups of women defined based on educational attainment and marital status: married women with at least 12 years of education, unmarried women with at least 12 years of education, married women with less than 12 years of education, and unmarried women with less than 12 years of education. These four sets of fertility rates are calculated in the following way. The numerator (number of births by women’s educational attainment and marital status) is obtained from the 1983–2013 Birth Cohort Linked Birth and Infant Death data, which is a complete census of all births, and the denominator (number of women in each education/marital status group) is estimated from the 1983–2013 CPS data.

Age-specific infant mortality rates are four sets of IMRs calculated for four groups of mothers defined by educational attainment and marital status: married women with at least 12 years of education, unmarried women with at least 12 years of education, married women with less than 12 years of education, and unmarried women with less than 12 years of education. We calculate these four sets of IMR using the 1983–2013 Birth Cohort Linked Birth and Infant Death data.3 In an additional analysis, we further distinguish neonatal mortality (deaths before reaching 28 days of life) and postneonatal mortality (deaths between the ages of 28 days and 1 year).

## Methods

Previous decomposition research on infant mortality typically focused on comparing rates between two time points or two populations (e.g., Schempf, Branum, Lukacs, and Schoendorf 2007) using Kitagawa’s (1955) standardization method. Our goal is different in that we investigate changes in a measure (IMDI) across multiple years as well as the contributions to the changes in the IMDI made by the shifts in five social and demographic processes. To this end, we use the decomposition method developed by Das Gupta (1993) to identify the extent to which changes in the IMDI are attributable to changes in sociodemographic compositions, including age, education, and marriage, as well as changes in fertility and infant mortality rates among disadvantaged and advantaged women. For all analyses, we examine White and Black women separately. We use the statistical package R to perform the decomposition analysis.

The infant mortality disadvantage index (IMDI) can be expressed as
$DD+N=∑i=15−1940−44wiW×diwi×di−uidi×fidi−ui×kifi/∑i=15−1940−44wiW×diwi×di−uidi×fidi−ui×kifi+∑i=15−1940−44wiW×diwi×uidi×giui×migi+∑i=15−1940−44wiW×wi−diwi×wi−di−viwi−di×biwi−di−vi×hibi+∑i=15−1940−44wiW×wi−diwi×viwi−di×livi×jili.$
1

The symbols are denoted as follows:

 D the number of infant deaths born to disadvantaged mothers (i.e., unmarried and less than 12 years of education); N the number of infant deaths born to nondisadvantaged mothers; W the total number of reproductive-age women (ages 15–44); wi the number of women at age i (i = 15–19, 20–24, 25–29, 30–34, 35–39, and 40–44); di the number of women with less than 12 years of education (<12 years) at age i; therefore (wi − di) represents the number of women with 12 or more years of education (12+ years) at age i; ui the number of <12 years, married women at age i; therefore (di − ui) represents the number of <12 years, unmarried women at age i; vi the number of 12+ years, married women at age i; therefore (wi − di − vi) represents the number of 12+ years, unmarried women at age i; fi the number of births to <12 years, unmarried women at age i; gi the number of births to <12 years, married women at age i; bi the number of births to 12+ years, unmarried women at age i; li the number of births to 12+ years, married women at age i; ki the number of infant deaths born to <12 years, unmarried women at age i; mi the number of infant deaths born to <12 years, married women at age i; hi the number of infant deaths born to 12+ years, unmarried women at age i; ji the number of infant deaths born to 12+ years, married women at age i.
 D the number of infant deaths born to disadvantaged mothers (i.e., unmarried and less than 12 years of education); N the number of infant deaths born to nondisadvantaged mothers; W the total number of reproductive-age women (ages 15–44); wi the number of women at age i (i = 15–19, 20–24, 25–29, 30–34, 35–39, and 40–44); di the number of women with less than 12 years of education (<12 years) at age i; therefore (wi − di) represents the number of women with 12 or more years of education (12+ years) at age i; ui the number of <12 years, married women at age i; therefore (di − ui) represents the number of <12 years, unmarried women at age i; vi the number of 12+ years, married women at age i; therefore (wi − di − vi) represents the number of 12+ years, unmarried women at age i; fi the number of births to <12 years, unmarried women at age i; gi the number of births to <12 years, married women at age i; bi the number of births to 12+ years, unmarried women at age i; li the number of births to 12+ years, married women at age i; ki the number of infant deaths born to <12 years, unmarried women at age i; mi the number of infant deaths born to <12 years, married women at age i; hi the number of infant deaths born to 12+ years, unmarried women at age i; ji the number of infant deaths born to 12+ years, married women at age i.

To simplify discussion, we reexpress Eq. (1) as
$DD+N=∑i=15−1940−44Ai×Ei×1−Pi×Ni×Di/∑i=15−1940−44Ai×Ei×1−Pi×Ni×Di+∑i=15−1940−44Ai×Ei×Pi×Mi×Fi+∑i=15−1940−44Ai×1−Ei×1−Oi×Gi×Hi+∑i=15−1940−44Ai×1−Ei×Oi×Ji×Ki,$
2
where Ai is the proportion of women who are age i; Ei is the proportion of women age i who have less than 12 years of education; and Pi and Oi are the proportions of women age i who are married among those with less than 12 years of education and among those with 12 or more years of education, respectively. Ni, Mi, Gi, and Ji are, respectively, nonmarital fertility rates among <12 years, marital fertility rates among <12 years, nonmarital fertility rates among 12+ years, and marital fertility rates among 12+ years, for women age i. Finally, Di, Fi, Hi, and Ki are IMRs among the following four groups of women, respectively: <12 years and unmarried mothers age i, <12 years and married mothers age i, 12+ years and unmarried mothers age i, and 12+ years and married mothers age i.
Importantly, Eq. (2) expresses the IMDI as a function of 12 vector factors: A, E, P, O, N, M, G, J, D, F, H, K. For a given year t, Eq. (2) can be represented as
$Rt=fAt¯Et¯Pt¯Ot¯Nt¯Mt¯Gt¯Jt¯Dt¯Ft¯Ht¯Kt¯,$
where R is our outcome, the infant mortality disadvantage index; f(.) is the function as described by Eq. (2); and the bar over the various components indicates reference to their distribution over the six age groups.
Next, to compare the difference in the IMDIs between any pair of years, t = 0 and t = 1, we define a set of standardized IMDIs as described by Das Gupta (1993), $RtA¯RtE¯RtP¯RtO¯RtN¯RtM¯RtG¯RtJ¯RtD¯RtF¯RtH¯RtK¯$, such that
$R1−R0=R1A¯−R0A¯+R1E¯−R0E¯+R1P¯−R0P¯+R1O¯−R0O¯+R1N¯−R0N¯+R1M¯−R0M¯+R1G¯−R0G¯+R1J¯−R0J¯+R1D¯−R0D¯+R1F¯−R0F¯+R1H¯−R0H¯+R1K¯−R0K¯.$
3

The reason for defining the standardized IMDRs this way is that the difference in our outcome, IMDI, between any two years (i.e., the left side of Eq. (3)) can be decomposed into the sum of the effects for each of the 12 factors (i.e., the right side of Eq. (3)). The bracketed differences in Eq. (3), from left to right, are as follows: the effects of age distribution, education distribution (age-specific percentage of women with less than 12 years of education), marriage rates (age-specific marriage rates among <12 years women and among 12+ years women, respectively), fertility rates (age-specific fertility rates among <12 years unmarried women, <12 years married women, 12+ years unmarried women, and 12+ years married women, respectively), and infant mortality rates (age-specific IMRs among <12 years unmarried mothers, <12 years married mothers, 12+ years unmarried mothers, and 12+ years married mothers, respectively). These bracketed differences, calculated based on the procedure provided by Das Gupta (1993:32), are the focus of our decomposition analysis.

Last, given that we are interested in the trend of the IMDI from 1983 to 2013, not in comparison of any two specific years, we obtain a single set of standardized IMDIs for each of the N = 28 years. For example, the standardized IMDI for year t and for factor A is defined as:
$RtA¯=∑i=1,i≠tNRtiA¯N−1+∑i=1,i≠tN∑j=1,j≠t,iNRijA¯−N−2RitA¯NN−1,$
where $RtiA¯$ represents the standardized IMDI, holding age distribution (A) constant, for year i, where the standardization is based on a comparison with year t. In the Results section, we explain how to interpret the standardized IMDIs using plots.

Before turning to empirical results, we emphasize that the decomposition approach used here is essentially a descriptive accounting exercise rather than aimed at providing causal estimates. When describing results, we sometimes use terms such as “effect” to improve readability and facilitate comprehension, but we realize that our estimates do not represent unbiased estimates of causal effects.

## Results

### Trends in the Infant Mortality Disadvantage Index by Race

Drawing on the Birth Cohort Linked Birth and Infant Death data, we plot in Fig. 1 two measures over the 1983–2013 period: the proportion of disadvantaged mothers (mothers who were unmarried when giving birth and who have less than 12 years of education) and the IMDI (proportion of infant deaths born to disadvantaged mothers). Note that infant death data were not collected for years 1992–1994, so the statistics are missing for these three years.

Figure 1 is revealing in three aspects. First, for both racial groups, the two measures presented—shares of disadvantaged mothers and the IMDI—exhibit similar temporal trends. However, the IMDI is persistently higher than the share of disadvantaged mothers over the entire period, indicating that a disproportionate share of infant mortality is borne by socially disadvantaged mothers. Second, the IMDI is consistently higher for Black than White women. In any given year, the gap is 3% to 23% to Black women’s disadvantage, indicating that the concentration of infant deaths among unmarried women with less than 12 years of education is more pronounced among Black women than their White counterparts. Third, the trends in the IMDI are different for White and Black women. Black women saw an almost linear decline in the disadvantage index; for White women, the index increased rapidly in the 1980s, followed by a more or less stable trend throughout the 1990s and early 2000s, and declined thereafter at a similar rate as that for Black women. These different trends between White and Black women had led to a convergence in the IMDI over time. The racial gap in the IMDI was 23% in 1983, the first year of our analysis, but by 2013—the last year of our observational period—it had been reduced by almost seven-eighths to 3%.

Given the differential trends in the IMDI between White and Black women, our following decomposition analysis focuses on two intervals for White women—1983–2006 and 2006–2013—to reflect the nonlinear temporal trend of White women’s IMDI (i.e., increasing until 2006 and decreasing thereafter). For Black women, we consider the entire 1983–2013 period because the IMDI is decreasing monotonically throughout the period.

### Decomposition of the Infant Mortality Disadvantage Index

We present results from the decomposition procedure in Table 1 (White women) and Table 2 (Black women). In both tables, the figures are the standardized IMDIs for each of the 12 factors. Tables 1 and 2 can be used to decompose changes in the IMDI across any interval during the 1983–2013 period. To facilitate interpretation, we follow Smith et al. (1996) and use plots to present our decomposition results. These plots, as pointed out by Smith et al. (1996), are useful for reading the effects of the various factors over different intervals but cannot be used for the overall level of any particular factor or the comparative levels of two lines in a given year. In general, if a factor whose standardized IMDI declines from t0 to t1, this factor is exerting a negative effect on the IMDI; that is, all else being equal, changes in this factor alone would have forced the IMDI downward. Similarly, a factor whose standardized IMDI increases would have a positive effect (i.e., increasing the IMDI). In the following, we show how the 12 factors have evolved over the 1983–2013 period and how their changes contribute to the trend in our outcome, the IMDI.

#### Age

Throughout the 1980s and much of the 1990s, the proportions of women in the age groups of 15–19, 20–24, and 25–29 were declining, whereas the proportions of women in the age groups of 30–34, 35–39, and 40–44 were increasing (Figs. A1 and A2, online appendix). Entering the 2000s, however, an opposite pattern emerged: there were increasingly more women in their earlier childbearing years (ages 15–29) and fewer in their later childbearing years (ages 30–44). These opposing trends therefore placed first fewer and then more women in the age groups with higher incidences of marriage and childbearing. We quantify the impact of the change in the age distribution on the IMDI trend by subtracting the standardized IMDI in the age column for 1983 from that for 2006 (White women, because of the nonlinear trend, Table 1) or 2013 (Black women, Table 2). For White women, all else being equal, changes in women’s age distribution shifted the IMDI only slightly, increasing the IMDI by 0.003 (= 0.164 – 0.161) from 1983 to 2006 and decreasing the IMDI by 0.002 (= 0.162 – 0.164) from 2006 to 2013. The impact was equally small for Black women; changes in women’s age distribution alone would have resulted in a decrease in the IMDI by 0.005 (= 0.266 – 0.271) from 1983 to 2013. This can also be seen in Figs. 2 and 3, which graphically present the decomposition results of the IMDI. The lines for age in these figures are virtually flat.

#### Educational Attainment

Over time, women with less than 12 years of education accounted for a smaller proportion of the population, and the decline was especially precipitous among Black women (Fig. A3, online appendix). This change in educational attainment played a major role in the declining IMDI for Black women. All else being equal, changes in the education distribution alone would have decreased the IMDI by 0.132 for Black women, whereas the corresponding number for White women is much smaller: –0.019 in the 1983–2006 period and –0.026 in the 2006–2013 period. This racial difference is shown in Figs. 2 and 3, in which the “% <12 years” line exhibits a sharp decline for Black women (Fig. 3) but is mostly flat for White women—albeit with some declining trend in the 2006–2013 interval (Fig. 2).

#### Marriage Rates

Consistent with previous literature (e.g., Cherlin 2010), our data show that for both White and Black women, there has been a declining trend in marriage (Fig. A4, online appendix). Based on standardized ratios, we find that proportion married among women with 12 or more years of education increased the IMDI for White women (0.016 in 1983–2006 and 0.008 in 2006–2013) and Black women (0.018 in 1983–2013). The declining marriage rate among women with less than 12 years of education also increased the IMDI over the 1983–2013 period for White women (0.028 in 1983–2006 and 0.017 in 2006–2013) and Black women (0.017 in 1983–2013).

#### Fertility Rates

As shown in Figs. A5 and A6 in the online appendix, among women with less than 12 years of education, Whites saw increasing marital and nonmarital fertility rates before the 1990s, followed by largely stable rates up to around 2006 and declining rates thereafter. Fertility rates did not appear to change much for Black women with less than 12 years of education (except for an increase in the 2000s for marital fertility, possibly because of unstable estimates resulting from the small sample sizes). Among women with 12 or more years of education, the marital and nonmarital fertility rates were mostly stable for both White and Black women. Calculations for the effects of changing marital (nonmarital) fertility rates among more- and less-educated White women yield, respectively, –0.032 (–0.034) and –0.016 (0.157) during the 1983–2006 interval, and 0.005 (0.003) and 0.005 (–0.063) during the 2006–2013 interval. For Black women, the corresponding figures for 1983–2013 are –0.020 (–0.026) and –0.007 (0.033). Overall, trends in nonmarital fertility rates among less-educated women seem to have a particularly large impact on the IMDI, especially for White women (0.157 and –0.063 vs. 0.033 for Black women).

#### Infant Mortality Rates

Infant mortality rates declined over the study period for all groups—regardless of educational attainment, marital status, or race—and the decline was particularly rapid in the 1980s and the early 1990s (Figs. A7 and A8, online appendix). Correspondingly, the effects of changes in marital (and nonmarital) infant mortality rates on the IMDI among the more-educated and the less-educated White women are, respectively, 0.047 (0.017) and 0.015 (–0.093) during the 1983–2006 interval, and 0.004 (0.006) and –0.002 (–0.004) during the 2006–2013 interval. For Black women, the effects of changes in marital (and nonmarital) IMRs on the 1983–2013 IMDI among more-educated and less-educated women are, respectively, 0.040 (0.067) and 0.011 (–0.152). Notably, the decline in IMRs among less-educated, unmarried mothers played an important role in driving down the IMDI for both White women (–0.093 during the 1983–2006 interval) and Black women (–0.152).

### Contributions of Changes in Sociodemographic Factors to the IMDI Trends

To quantify the relative contribution of each of the 12 factors to the rising and declining IMDI among White women and to the monotonically declining IMDI among Black women, we divide the change in the standardized ratio of each factor by the total change over the interval. These results are summarized in Table 3, for White and Black women separately.

Table 3 shows that during the first interval, 1983–2006, the increasing nonmarital fertility rates among White women with less than 12 years of education (0.157) represent the single largest factor contributing to the increasing IMDI, accounting for 175% of the increased IMDI. By contrast, declines in IMRs among unmarried mothers with less than 12 years of education (–0.093, –103%) had been exerting strong downward, counterbalancing pressure on this index. During the second interval, 2006–2013, nonmarital fertility rates among White women with less than 12 years of education—which had been declining since around 2006—again played the largest role in pulling down the IMDI (–0.063, –131%).

For Black women, recall that their IMDI declined linearly from 1983 to 2013. Two major factors seem to underlie this decrease: declines in the proportion of women with less than 12 years of education (–0.132, –85%) and declines in infant mortality rates among unmarried mothers with less than 12 years of education (–0.152, –98%).

Taken together, we highlight two important findings. First, for White women, the temporal trend in nonmarital fertility rates among those with less than 12 years of education has the largest impact on the IMDI, first driving it up until 2006 and then pulling it down up to 2013. This is not the same for Black women, for whom the declining proportion of those with less than 12 years of education resulted in a continually declining IMDI. Second, for both White (during the 1983–2006 interval) and Black women, declines in IMRs among unmarried, less-educated mothers played an important role in forcing down the IMDI.

### Differences Between Neonatal and Postneonatal Infant Mortality

Given the previous finding that postneonatal mortality contributes more (relative to neonatal mortality) to the higher U.S. infant mortality rates compared with other countries (Chen et al. 2016), we performed separate decomposition analyses for neonatal and postneonatal infant mortality. The patterns for the two types of mortality are largely similar (see section B of the online appendix for detailed results).

## Discussion

The U.S. IMR has declined steadily over almost a century but is still consistently higher than that of other industrialized countries. As demonstrated in previous research (Chen et al. 2016), the U.S. poor performance in IMR is largely due to the concentration of infant deaths born to disadvantaged mothers. In this research, we develop an infant mortality disadvantage index (IMDI) to describe the extent to which a disproportionate share of mortality is borne by the socially disadvantaged (Braveman et al. 2000) as well as to investigate how changes in social and demographic factors may have given rise to the trend of the concentration of IMR among the socially disadvantaged.

Substantively, we highlight several key findings from analyzing trends in the IMDI and discuss potential policy implications for reducing IMR in the United States. First, we reveal differential trends in the IMDI for White and Black women. Black women saw an almost linear decline in the disadvantage index from 1983 to 2013; for White women, the index increased rapidly in the 1980s, plateaued throughout the 1990s and early 2000s, and declined since 2006. Combined, these two trends led to a convergence in the IMDI between White and Black women. This finding suggests that up until 2006, the burden of infant mortality was increasingly shouldered by the most vulnerable SES group among White women, whereas an opposite trend is observed for Black women; by comparison, between 2006 and 2013, both White and Black women saw declining inequality in infant mortality. Even though the IMDI was continually declining for Black women, we emphasize that even by the end of our observation period in 2013, the concentration of infant deaths among disadvantaged mothers was still higher for Black women (17% vs. 14% among White women). Coupled with the persistent Black-White disparities in absolute levels of infant mortality, the finding implies that infants born to disadvantaged Black mothers still suffer the highest mortality rate. A promising direction for future research, therefore, is to build on prior individual-level studies on racial disparities in infant mortality (e.g., Hamilton 2017; Schoendorf et al. 1992; Wallace et al. 2017) to develop a Black-White infant mortality disparity index. The decomposition method used in our analysis can then be applied to understand how compositional changes and demographic processes may contribute to the persistent Black-White disparities in infant mortality.

Second, using the decomposition method developed by Das Gupta (1993), we find marked racial differences in the driving forces underlying the IMDI trend. For White women, the trend in nonmarital fertility rates among those with less than 12 years of education was the leading factor for the increasing and then decreasing IMDI. The declining IMDI for Black women is mostly driven by a declining share of women with less than 12 years of education, as well as a decrease in infant mortality rates among unmarried mothers with less than 12 years of education.

Although identifying and evaluating specific policies is beyond the scope of the present study, findings from our decomposition analysis are suggestive for possible policies and programs to reduce the level of and inequality in U.S. infant mortality. One area that policy-makers and program providers can devote effort to is education. Our finding shows that as more Black women finish high school, inequality in infant mortality—as measured by the IMDI—declined. This finding implies that programs such as free community college for everyone or increased Pell grants may help further decrease the concentration of infant deaths among disadvantaged women. At first glance, these programs might not represent health programs that directly target population health or access to medical care. However, our decomposition analysis indicates that policies and programs that focus on upstream factors, such as education, can be effective for reducing infant health inequality while lowering the absolute number of infant deaths.

Another key area for policy development relates to the finding that for White women, inequality in infant mortality tracks quite well with the trend of fertility rates among less-educated, unmarried women. This finding indicates that these women may face unique challenges and difficulties compared with their more advantaged counterparts, thereby leading to particularly high infant mortality. Given that less-educated and unmarried mothers suffer higher infant mortality, it is disconcerting that the U.S. welfare system has in fact over time redistributed its spending away from single mothers and their children (Moffitt 2015). Future research is needed to better understand the exact difficulties these women face. If, for example, financial struggles or lack of social support and childcare support are the major constraints they face, policies and programs that target this disadvantaged group throughout pregnancy and during the perinatal and postpartum periods—such as financial support provision, home nurse visits, and childcare programs—may prove productive for reducing the overall level of and inequality in infant mortality. Meanwhile, because the pregnancy and childbirth of these women are sometimes unplanned, programs such as the Affordable Care Act (ACA) contraceptive mandate may help these women avoid unplanned pregnancies, whereas abortion bans without simultaneously providing sufficient social and medical support may exacerbate infant mortality inequality.

We focus on the concentration of infant mortality among women with less than 12 years of education. Given the rising educational attainment over time, the implications of not finishing high school may have changed (Bound et al. 2015; Fan and Qian 2019). To the extent that women with less than 12 years of education are now even more disadvantaged than in the past, the finding that White women’s changing IMDI is mostly driven by changes in nonmarital fertility rates among those with less education may be an overestimate. The increase in nonmarital fertility rates among these disadvantaged mothers (up to 2006) may be a marker of some other unobserved disadvantages (rather than low education or nonmarital fertility per se). By contrast, given the expansion of secondary and tertiary education, women with 12 years of education are not as advantaged as in the past. Thus, the finding that Black women’s decreasing IMDI is mostly driven by the declining share of women with less than 12 years of education may be an underestimate because Black women who were previously unlikely to finish high school would now have a high school diploma; their otherwise disadvantaged social locations may therefore limit the contribution the educational expansion could have played in decreasing the IMDI.

In addition to the trend of rising educational attainment, two other factors may complicate our findings. One is the rising immigrant population over the past few decades. We conducted a sensitivity analysis in which we included only women who were born in the United States. Results were largely similar to our main findings (see section F, online appendix). The second one is women’s labor market participation. Specifically, low-SES, unmarried mothers may not have access to affordable childcare, even as these mothers may find themselves spending more money to be in the workforce than if they were home (Edin and Lein 1997). No information on maternal employment, however, is available in the Birth Cohort Linked Birth and Infant Death data, preventing us from exploring these possibilities. We advise future research, when such measures are available, to examine the social processes linking maternal employment and infant mortality inequality.

Methodologically, the IMDI, as a population-level measure of health inequality, is related to but distinct from IMR; additionally, compared with conventional individual-level regression analysis, our decomposition analysis of the IMDI addresses different types of questions about infant mortality. Specifically, the IMDI is a distributional measure of inequality that reflects the degree to which infant mortality is unevenly distributed or concentrated among certain groups, whereas IMR is a global measure of the overall level of mortality. An increasing IMDI implies growing inequality, but it does not necessarily mean that the IMR would increase as well. Additionally, our decomposition analysis of the IMDI addresses questions that cannot be answered under the individual-level regression framework. Although individual-level regression models allow researchers to examine the associations between individual characteristics and infant health, they do not permit an understanding of how structural and compositional changes such as changes in age structures or educational expansion may affect the distribution and inequality of infant mortality. For example, even though the association between mother’s education and infant mortality might not change over time according to a regression model, infant mortality inequality as indicated by the IMDI may well change because of the shifting distribution of educational attainment (e.g., more college graduates over time).

Our research has several limitations. First, because a small percentage of infant deaths in the Birth Cohort Linked Birth and Infant Death data cannot be linked to their corresponding birth certificate, it is unclear how these unlinked records may have influenced the trend of our index or our decomposition results. Second, some of our factors are estimated indirectly; in particular, we calculate fertility rates using two sources (CPS data and Birth Cohort Linked Birth and Infant Death data). The degree to which these estimates are accurate and how they may affect our decomposition results is unknown. Given that there is no other way to obtain fertility rates by marital status and educational attainment over time, however, combining multiple sources is the only viable approach available to us. Third, a nonmarital childbirth does not necessarily imply that the mother and father have dissolved their relationship prior to the childbirth. Many unmarried biological parents are indeed living together at the births of their children (McLanahan 2004). We are unable, however, to identify these cohabiting parents.

Despite these limitations, our research is the first to show the changing patterns of the concentration of infant mortality into the most disadvantaged social group over time, as well as the social and demographic processes that have contributed to such changes. Although we focus on infant mortality, the index we develop can be easily adapted to study other health outcomes such as adult mortality. To reduce inequality in infant mortality and to improve the U.S. international ranking in infant mortality, our findings indicate that more scholarly and policy attention needs to be paid to upstream social factors, including further improvements in educational opportunities and more support for nonmarital mothers.

## Acknowledgments

An earlier version of this article was presented at the Population Association of America Conference in Chicago in 2017. The authors thank the organizer of the session, Saifuddin Ahmed, their discussant, Ndola Prata, and participants in the session for comments on the earlier version. They are also grateful for the helpful suggestions of three anonymous reviewers.

## Authors’ Contributions

Both authors conceived of the research idea, performed the analysis independently (to make sure the calculations are correct), and discussed the results. W.F. wrote the manuscript, and L.L. provided comments, revisions, and feedback that helped shape the research and manuscript.

## Data Availability

The IPUMS-CPS data are available from https://cps.ipums.org/cps/, and the National Center for Health Statistics (NCHS) Birth Cohort Linked Birth and Infant Death data are available from http://data.nber.org/data/linked-birth-infant-death-data-vital-statistics-data.html.

## Compliance With Ethical Standards

### Ethics and Consent

All research procedures were in accordance with the ethical standards of the Committee on Publication Ethics. Informed consent was obtained from all respondents for being included in the study by the relevant data collection organizations.

### Conflict of Interest

The authors declare that there is no conflict of interest regarding the publication of this article.

## Notes

1

We conducted a separate set of analysis for non-Hispanic Whites, non-Hispanic Blacks, and Hispanics, drawing on the 1995–2013 data. Results, shown in section C of the online appendix, suggest that the trends in the IMDI are similar between non-Hispanic Whites and Hispanics (i.e., more or less stable throughout the late 1990s and declining since 2006), even as the IMDI is persistently higher among Hispanic women than either non-Hispanic White women or non-Hispanic Black women. The decomposition analysis further shows that two factors played the largest role in Hispanic women’s decreasing IMDI following 2006: declines in the share of women with less than 12 years of education and decreasing fertility among unmarried Hispanic women with less than 12 years of education.

2

We note one exception. For Black women, when the 15–19 age group was removed, nonmarital fertility rates among those with less than 12 years of education played a larger role in driving up the IMDI (143%; Table D1, online appendix), compared with the result when this age group was included (21%; Table 3). This finding seems to suggest that over time, nonmarital fertility increases at a faster rate (or declines at a slower rate) for Black women older (as opposed to younger) than 19. Or perhaps nonmarital fertility poses a greater threat for Black women older (as opposed to younger) than 19; this possibility is consistent with the weathering hypothesis, which suggests that the health of Black women begins to deteriorate in early adulthood as a result of their cumulative exposure to socioeconomic disadvantage (Geronimus 1992). As age increases, therefore, nonmarital birth may lead to particularly poor birth outcomes for Black women. We advise more future research to better understand this finding.

3

One caveat is that in the Birth Cohort Linked Birth and Infant Death data, maternal education was not available in all states in early years (1983–1991). We conducted a sensitivity analysis excluding the four states with incomplete or no reporting of maternal education (California, New York, Texas, and Washington); results were substantively similar to our main findings (see section E of the online appendix).

## References

Bird, S. T., Chandra, A., Bennett, T., & Harvey, S. M. (
2000
).
Beyond marital status: Relationship type and duration and the risk of low birth weight
.
Family Planning Perspectives
,
32
,
281
287
.
Bound, J., Geronimus, A. T., Rodriguez, J. M., & Waidmann, T. A. (
2015
).
Measuring recent apparent declines in longevity: The role of increasing educational attainment
.
Health Affairs
,
34
,
2167
2173
.
Braveman, P., Krieger, N., & Lynch, J. (
2000
).
Health inequalities and social inequalities in health
.
Bulletin of the World Health Organization
,
78
,
232
235
.
Chen, A., Oster, E., & Williams, H. (
2016
).
Why is infant mortality higher in the United States than in Europe?
.
American Economic Journal: Economic Policy
,
8
(
2
),
89
124
.
Cherlin, A. (
2010
).
Demographic trends in the United States: A review of research in the 2000s
.
Journal of Marriage and Family
,
72
,
403
419
.
Cramer, J. C. (
1995
).
Racial and ethnic differences in birthweight: The role of income and financial assistance
.
Demography
,
32
,
231
247
.
Das Gupta, P. (
1993
).
Standardization and decomposition of rates: A user’s manual
(Report No. 186).
Washington, DC
:
U.S. Department of Commerce, Economics and Statistics Administration, Bureau of the Census
.
DiPrete, T. A., & Buchmann, C. (
2013
).
The rise of women: The growing gender gap in education and what it means for American schools
.
New York, NY
:
Russell Sage Foundation
.
Dye, J. L. (
2010
).
Fertility of American women: 2008
(Current Population Reports No. P20-563).
Washington, DC
:
U.S. Department of Commerce, Economics and Statistics Administration, U.S. Census Bureau
.
Edin, K., & Lein, L. (
1997
).
Making ends meet: How single mothers survive welfare and low-wage work
.
New York, NY
:
Russell Sage Foundation
.
Elder, T. E., Goddeeris, J. H., & Haider, S. J. (
2016
).
Racial and ethnic infant mortality gaps and the role of socio-economic status
.
Labour Economics
,
43
,
42
54
.
Elder, T. E., Goddeeris, J. H., Haider, S. J., & Paneth, N. (
2014
).
The changing character of the Black-White infant mortality gap, 1983–2004
.
American Journal of Public Health
,
104
(
Suppl. 1
),
S105
S111
.
Ellwood, D. T., & Jencks, C. (
2004
).
The uneven spread of single-parent families: What do we know?
. In Neckerman, K. M. (Ed.),
Social inequality
(pp.
3
78
).
New York, NY
:
Russell Sage Foundation
.
Fan, W., & Qian, Y. (
2019
).
Rising educational gradients in mortality among U.S. Whites: What are the roles of marital status and educational homogamy?
Social Science & Medicine
,
235
,
112365
. https://doi.org/10.1016/j.socscimed.2019.112365
Fanaroff, A. A., Stoll, B. J., Wright, L. L., Carlo, W. A., Ehrenkranz, R. A., Stark, A. R., . . .
NICHD Neonatal Research Network
. (
2007
).
Trends in neonatal morbidity and mortality for very low birthweight infants
.
American Journal of Obstetrics and Gynecology
,
196
,
147.e1
147.e8
. https://doi.org/10.1016/j.ajog.2006.09.014
Flood, S., King, M., Ruggles, S., & Warren, J. R. (
2015
).
Integrated Public Use Microdata Series, Current Population Survey: Version 4.0
[Data set].
Minneapolis
:
University of Minnesota
.
Friedman, D. J., Hunter, E. L., & Parrish, R. G. (
2005
).
Health statistics: Shaping policy and practice to improve the population’s health
.
New York, NY
:
Oxford University Press
.
Frisbie, W. P., Forbes, D., & Pullum, S. G. (
1996
).
Compromised birth outcomes and infant mortality among racial and ethnic groups
.
Demography
,
33
,
469
481
.
Frisbie, W. P., Song, S. E., Powers, D. A., & Street, J. A. (
2004
).
The increasing racial disparity in infant mortality: Respiratory distress syndrome and other causes
.
Demography
,
41
,
773
800
.
Gage, T. B., Fang, F., O’Neill, E., & DiRienzo, G. (
2013
).
Maternal education, birth weight, and infant mortality in the United States
.
Demography
,
50
,
615
635
.
Geronimus, A. T. (
1992
).
The weathering hypothesis and the health of African-American women and infants: Evidence and speculations
.
Ethnicity & Disease
,
2
,
207
221
.
Goldstein, J. R., & Kenney, C. T. (
2001
).
Marriage delayed or marriage forgone? New cohort forecasts of first marriage for U.S. women
.
American Sociological Review
,
66
,
506
519
.
Gonzalez, R. M., & Gilleskie, D. (
2017
).
Infant mortality rate as a measure of a country’s health: A robust method to improve reliability and comparability
.
Demography
,
54
,
701
720
.
Gortmaker, S. L., & Wise, P. H. (
1997
).
The first injustice: Socioeconomic disparities, health services technology, and infant mortality
.
Annual Review of Sociology
,
23
,
147
170
.
Hamilton, D. (
2017
).
Post-racial rhetoric, racial health disparities, and health disparity consequences of stigma, stress, and racism
(Working paper).
Washington, DC
:
Washington Center for Equitable Growth
.
Hummer, R. A., Biegler, M., De Turk, P. B., Forbes, D., Frisbie, W. P., Hong, Y., & Pullum, S. G. (
1999
).
Race/ethnicity, nativity, and infant mortality in the United States
.
Social Forces
,
77
,
1083
1117
.
Judge, K., Mulligan, J. A., & Benzeval, M. (
1998
).
Income inequality and population health
.
Social Science & Medicine
,
46
,
567
579
.
Kitagawa, E. M. (
1955
).
Components of a difference between two rates
.
Journal of the American Statistical Association
,
50
,
1168
1194
.
Krieger, N., Waterman, P. D., Spasojevic, J., Li, W., Maduro, G., & Van Wye, G. (
2016
).
Public health monitoring of privilege and deprivation with the index of concentration at the extremes
.
American Journal of Public Health
,
106
,
256
263
.
Lhila, A., & Long, S. (
2012
).
What is driving the Black–White difference in low birthweight in the US?
.
Health Economics
,
21
,
301
315
.
Loggins, S., & Andrade, F. C. D. (
2014
).
Despite an overall decline in U.S. infant mortality rates, the Black/White disparity persists: Recent trends and future projections
.
Journal of Community Health
,
39
,
118
123
.
Martin, J. A., Hamilton, B. E., Ventura, S. J., Osterman, M. J. K., Wilson, E. C., & Mathews, T. J. (
2012
).
Births: Final data for 2010 (National Vital Statistics Reports Vol. 61, No. 1)
.
Hyattsville, MD
:
National Center for Health Statistics
.
McLanahan, S. (
2004
).
Diverging destinies: How children are faring under the second demographic transition
.
Demography
,
41
,
607
627
.
McLanahan, S., & Percheski, C. (
2008
).
Family structure and the reproduction of inequalities
.
Annual Review of Sociology
,
34
,
257
276
.
Moffitt, R. A. (
2015
).
The deserving poor, the family, and the U.S. welfare system
.
Demography
,
52
,
729
749
.
Moon, R. Y. (
2011
).
SIDS and other sleep-related infant deaths: Expansion of recommendations for a safe infant sleeping environment
.
Pediatrics
,
128
,
e1341
e1367
. https://doi.org/10.1542/peds.2011-2285
National Center for Health Statistics
. (
1994
).
Public use data tape documentation: Linked birth/infant death data set: 1988 birth cohort
. (
1994
).
Hyattsville, MD
:
U.S. Department of Health and Human Services
.
Nersesian, W. S. (
1988
).
Infant mortality in socially vulnerable populations
.
Annual Review of Public Health
,
9
,
361
377
.
Partridge, S., Balayla, J., Holcroft, C. A., & Abenhaim, H. A. (
2012
).
Inadequate prenatal care utilization and risks of infant mortality and poor birth outcome: A retrospective analysis of 28,729,765 U.S. deliveries over 8 years
.
American Journal of Perinatology
,
29
,
787
793
.
Powers, D. A. (
2013
).
Black-White differences in maternal age, maternal birth cohort, and period effects on infant mortality in the U.S. (1983–2002)
.
Social Science Research
,
42
,
1033
1045
.
Saigal, S., & Doyle, L. W. (
2008
).
An overview of mortality and sequelae of preterm birth from infancy to adulthood
.
Lancet
,
371
,
261
269
.
Schempf, A. H., Branum, A. M., Lukacs, S. L., & Schoendorf, K. C. (
2007
).
The contribution of preterm birth to the Black-White infant mortality gap, 1990 and 2000
.
American Journal of Public Health
,
97
,
1255
1260
.
Schoendorf, K. C., Hogue, C. J. R., Kleinman, J. C., & Rowley, D. (
1992
).
Mortality among infants of Black as compared with White college-educated parents
.
New England Journal of Medicine
,
326
,
1522
1526
.
Singh, G. K., & Kogan, M. D. (
2007
).
Persistent socioeconomic disparities in infant, neonatal, and postneonatal mortality rates in the United States, 1969–2001
.
Pediatrics
,
119
,
e928
e939
. https://doi.org/10.1542/peds.2005-2181
Singh, G. K., & Yu, S. M. (
1995
).
Infant mortality in the United States: Trends, differentials, and projections, 1950 through 2010
.
American Journal of Public Health
,
85
,
957
964
.
Smith, H. L., Morgan, S. P., & Koropeckyj-Cox, T. (
1996
).
A decomposition of trends in the nonmarital fertility ratios of Blacks and Whites in the United States, 1960–1992
.
Demography
,
33
,
141
151
.
Smock, P. J., & Greenland, F. R. (
2010
).
Diversity in pathways to parenthood: Patterns, implications, and emerging research directions
.
Journal of Marriage and Family
,
72
,
576
593
.
Wallace, M., Crear-Perry, J., Richardson, L., Tarver, M., & Theall, K. (
2017
).
Separate and unequal: Structural racism and infant mortality in the US
.
Health & Place
,
45
,
140
144
.
Wise, P. H. (
2003
).
The anatomy of a disparity in infant mortality
.
Annual Review of Public Health
,
24
,
341
362
.
World Health Organization
. (
2014
).
Monitoring health inequality: An essential step for achieving health equity—Illustrations of fundamental concepts
. (
2014
).
Geneva, Switzerland
:
World Health Organization
.
Wu, L., & Wolfe, B. (
2001
).
Out of wedlock: Causes and consequences of nonmarital fertility
.
New York, NY
:
Russell Sage Foundation
.

## Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.