Selected Fertility and Racial Inequality

Owen Thompson

doi:10.3368/jhr.0221-11481R2

Article Figures & Data

Figures

Tables
Additional Files

Download figure
Open in new tab
Download powerpoint
Figure 1
Fertility Rates by Year, Race, and Region
Notes: Panel A shows trends in the General Fertility Rate (GFR), which is calculated as the ratio of total live births to the population of females ages 14–44, expressed in thousands. Panel B shows the simple differences in the GFRs from Panel A between the specified groups. Live birth totals are drawn from Vital Statistics print reports from 1955–1967 and from digitized Vital Statistics microdata from 1968–1975. Population counts by gender, age and race are drawn from the Decennial Census with linear interpolations for 1955–1968, and from the Survey of Epidemiology and End Results (SEER) for 1969–1975. The South consists of Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia, and the North consists of the balance of the lower 48 states. See Online Appendix A for alternative regional definitions.
Download figure
Open in new tab
Download powerpoint
Figure 2
Relative Changes in Completed Fertility
Notes: Panel A displays trends in Black–white differences in the mean number of children ever born among female census respondents ages 44–70, disaggregated by region. Dashed vertical lines indicate the 1929 and 1944 birth cohorts, which is the range of cohorts that were ages 20–35 as of 1964 and whose prime childbearing years therefore spanned 1964. Panel B displays histograms of children ever born among women from the indicated race, region and cohorts. The South consists of Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia, and the North consists of the balance of the lower 48 states. Individuals who were born in one region but resided in the other at the time of survey are excluded.
Download figure
Open in new tab
Download powerpoint
Figure 3
Heterogeneity
Notes: Each bar in Panel A indicates the change from 1960 to 1970 in the share of southern Black female census respondents ages 14–44 with an own-child under one, multiplied by 1,000 to match standard GFR units, within the indicated subgroup. Each bar in Panel B displays convergence in the share of female census respondents ages 14–44 with an own-child under one between southern Black women and southern white women, again multuiplied by 1,000 and within the indicated subgroup. The dashed horizontal lines indicate the level changes (Panel A) and convergence (Panel B) within the full sample. The utilized census samples are described in detail in Section III. The South consists of Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia, and the North consists of the balance of the lower 48 states. Individuals who were born in one region but resided in the other at the time of survey are excluded. Full tabular results of the regressions underlying this figure, with sample sizes and standard errors clustered at the state of residence level, are provided in Online Appendix C.

View popup

Table 1

General Fertility Rate from Vital Statistics versus Share of Women with Child under One from Censuses

	Vital Statistics (GFR)			Census (Share of Women with Child <1)
	1960	1970	\|Change\|	1960	1970	\|Change\|
	(1)	(2)	(3)	(4)	(5)	(6)
Southern Black	160	118	41	126	86	40
Southern white	109	85	24	102	75	27
Northern Black	142	107	35	113	82	31
Northern white	112	83	29	107	76	31

Notes: Table entries report the GFR calculated using Vital Statistics versus the share of women ages 14–44 living with an own-child under age one (multiplied by 1,000) from Decennial Censuses, within the indicated racial group, region, and year. The South consists of Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia, and the North consists of the balance of the lower 48 states. The utilized Vital Statistics and census samples are described in Sections I and II, respectively.

View popup

Table 2

Changes in Mother and Family Characteristics by Recent Birth Status and Race

	Less than 8 Years of Education	Four or More Own-Children	Tenant Farm Household
	(1)	(2)	(3)
Panel A: Southern Black
Change among women with a child <1	−0.233^***	−0.186^***	−0.104^***
	(0.015)	(0.011)	(0.021)
Change among women without a child <1	−0.179^***	0.006	−0.089^***
	(0.007)	(0.005)	(0.016)
Difference	−0.054^***	−0.192^***	−0.015^**
	(0.012)	(0.009)	(0.006)
Panel B: Southern White
Change among women with a child <1	−0.094^***	−0.085^***	−0.035^***
	(0.010)	(0.006)	(0.006)
Change among women without a child <1	−0.085^***	−0.006^**	−0.034^***
	(0.011)	(0.003)	(0.004)
Difference	−0.009^*	−0.079^***	−0.001
	(0.005)	(0.004)	(0.002)
Panel C: Black–White Difference
Southern Black difference – southern white difference	−0.045^***	−0.113^***	−0.014^**
	(0.012)	(0.008)	(0.006)
Black–white gap in 1960 census sample with child <1	0.25	0.31	0.08
Percent change	−18%	−37%	−17%

Notes: Panel A and Panel B show the average change in the indicated characteristic between the 1960 and 1970 censuses among women without a child under one and women with a child under one, as well as the difference in the changes between these two groups. Panel A reports these changes for southern Black women, while Panel B reports them for southern white women. The changes among women with and without a child under one are estimated by regressing the characteristic onto a 1970 indicator within the relevant group, while the differences between the two groups are calculated by regressing the characteristics onto a 1970 indicator, an indicator of having a child under one, and their interaction. Panel C reports the difference in the differences among southern Black women (as reported in Panel A) and southern white women (as reported in Panel B) and are calculated with the three way interaction of a 1970 indicator, an indicator of having a child under one, and a Black indicator, as well as all corresponding main effects and two-way interactions. Each table entry comes from a separate regression. All models apply sampling weights. Standard errors clustered by state are reported in parentheses. Significance: ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.

View popup

Table 3

Racial Fertility Differences and Racial Gaps in the Adult Outcomes of the Next Generation

	Baseline		State-Specific Cohort Trends		Cohort-by-Demographic Effects
	Highest Grade Completed	Log Earnings	Highest Grade Completed	Log Earnings	Highest Grade Completed	Log Earnings
	(1)	(2)	(3)	(4)	(5)	(6)
Panel A: South
Black–white GFR gap/20	−0.191^**	−0.029	−0.126^**	−0.084^***	−0.135	−0.077
	(0.061)	(0.029)	(0.056)	(0.024)	(0.125)	(0.047)
Observations (state–years)	231	231	231	231	231	231
Panel B: North
Black–white GFR gap/20	0.008	0.008	−0.015	−0.004	0.002	0.006
	(0.041)	(0.012)	(0.067)	(0.018)	(0.044)	(0.017)
Observations (state–years)	765	765	765	765	765	765
p-value for south–north difference	0.003	0.191	0.078	0.011	0.022	0.030

Notes: All models are estimated on collapsed state-cohort level census/ACS data merged with Vital Statistics fertility data. Dependent variables are the Black–white mean difference of the outcome indicated in the column headings, at the state–cohort level. Models in Columns 1 and 2 include state and cohort fixed effects, the share of each state–cohort cell that is female, and a cubic of average age at the time that the outcomes were observed. Models in Columns 3 and 4 additionally include interactions between state indicators and a continuous cohort variable, and models in Columns 5 and 6 additionally include interactions between cohort indicators and the following 1960 state characteristics: median income, median education, percent Black, total population, land area, and percent urban. Mean values for the outcomes were calculated using individual sampling weights, while each cell is given equal weight when estimating the regression models. Panel A restricts the sample to individuals born in the South, which consists of Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia, and Panel B restricts the sample to individuals born in the North, which consists of the balance of the lower 48 states. The bottom row reports p-values from chi-square tests of whether the coefficients for the South in Panel A and the coefficients for the North in Panel B are equal. Standard errors clustered by state are reported in parentheses. Significance: ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.

View popup

Table 4

Reweighting Black–White Family Background Gaps and Outcome Gaps in the NLSY-79

	Baseline	Reweighted	Change after Reweighting	Change after Reweighting × 0.25
	(1)	(2)	(3)	(4)
Panel A: Family Background
Maternal years of education	−1.34^***	−0.10
	(0.13)	(0.15)
Maternal education ≤8 years	0.10^***	0.01
	(0.02)	(0.03)
Birth order	1.08^***	0.12
	(0.10)	(0.18)
Birth order ≥4	0.22^***	0.02
	(0.02)	(0.03)
Farm residence	−0.02^**	0.00
	(0.01)	(0.01)
Panel B: Later-Life Outcomes
Highest grade completed	−0.82^***	−0.09	−0.74	−0.184
	(0.12)	(0.14)
Log earnings	−0.50^***	−0.28^***	−0.22	−0.055
	(0.06)	(0.08)
AFQT score	−1.13^***	−0.87^***	−0.27	−0.067
	(0.04)	(0.06)

Notes: Each entry in Columns 1 and 2 comes from a separate bivariate regression of the listed characteristic listed onto a Black indicator. Column 1 applies only sampling weights, while Column 2 additionally applies weights constructed using the procedure recommended by DiNardo, Fortin, and Lemieux (1996) and described in Section III.B of the text. Column 3 calculates the differences between the baseline and reweighted gaps for the outcome variables, while Column 4 multiplies these differences by 0.25, which was the approximate percent change in family background due to selective fertility decline reported in Table 2. For the regressions in Columns 1 and 2 robust standard errors are reported in parentheses. Significance: ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.