The Unique Impact of Abolition of Jim Crow Laws on Reducing Inequities in Infant Death Rates and Implications for Choice of Comparison Groups in Analyzing Societal Determinants of Health

Abstract

Objectives. We explored associations between the abolition of Jim Crow laws (i.e., state laws legalizing racial discrimination overturned by the 1964 US Civil Rights Act) and birth cohort trends in infant death rates.

Methods. We analyzed 1959 to 2006 US Black and White infant death rates within and across sets of states (polities) with and without Jim Crow laws.

Results. Between 1965 and 1969, a unique convergence of Black infant death rates occurred across polities; in 1960 to 1964, the Black infant death rate was 1.19 times higher (95% confidence interval [CI] = 1.18, 1.20) in the Jim Crow polity than in the non–Jim Crow polity, whereas in 1970 to 1974 the rate ratio shrank to and remained at approximately 1 (with the 95% CI including 1) until 2000, when it rose to 1.10 (95% CI = 1.08, 1.12). No such convergence occurred for Black–White differences in infant death rates or for White infants.

Conclusions. Our results suggest that abolition of Jim Crow laws affected US Black infant death rates and that valid analysis of societal determinants of health requires appropriate comparison groups.

Surprisingly little research has quantified the health impact of the abolition of Jim Crow legislation, that is, state laws legalizing racial discrimination overturned by the 1964 US Civil Rights Act (78 Stat 241).^1–6 The 4 extant empirical population-based investigations, however, provide suggestive evidence of improvements in health among Black Americans and decreases in health inequities between Blacks and Whites.^7–10 Far from a matter of historical interest only, the contemporary significance of whether abolition of Jim Crow laws had any health impact is both substantive and methodological.

At issue is understanding not only determinants and consequences of trends in US racial/ethnic health inequities within and across generations^11,12 but also the assumptions underlying such analyses. In particular, conceptualizing Jim Crow legislation as a political determinant of health shifts the focus from “race/ethnicity” to race relations as a causal exposure, thereby raising an important question from a counterfactual standpoint: should contrasts be within or across the racial/ethnic groups defined by these race relations?^13–15 This question of who should be contrasted adds another dimension to current methodological discussions on quantifying health inequities, which chiefly emphasize which effect measures should be used (e.g., relative vs absolute difference).^16–18

We chose to focus on the abolition of Jim Crow laws on a priori theoretical grounds, because it is precisely the type of epochal event^2–6 that, according to leading social epidemiological and political sociology theories, should produce clear-cut period and cohort effects.^19–23 The abolition of Jim Crow legislation was a critical turning point in the process of formal political incorporation of Black Americans into the US society,^3–6 whereby the political incorporation of socially defined groups refers not only to their political rights and political participation as citizen members of an electorate but also to their politically gaining the capacity and agency to advance fulfillment of their human rights.^{4–6,23–27}

To date, however, scant research has explicitly examined the health impact of political incorporation²²; 2 notable exceptions—a study addressing post-Apartheid policies in South Africa²⁸ and a study focusing on the extension versus denial of political rights to women in 61 countries²⁹—showed that political incorporation reduced health inequities. The case of Jim Crow is thus not only important in its own right but potentially can provide useful substantive and methodological insights relevant to assessing the health impacts of current efforts to promote political incorporation (e.g., the extension of voting rights to migrants^26,30).

We selected infant death as the “outcome” because it is a widely used indicator of population health that is highly sensitive to living conditions and access to medical technology during pregnancy and the first year of life and is also reflective of mothers’ cumulative health status before and after conception.^31–33 For our comparisons, we took the novel step of contrasting outcomes between what we term Jim Crow and non–Jim Crow polities. These entities, defined by their politics as opposed to their geography,^22,23 respectively comprised sets of states that did and did not have laws legalizing racial discrimination overturned by the 1964 US Civil Rights Act, regardless of geographic location (i.e., these states were not restricted solely to the US South).¹ We assessed birth cohort effects and trends in Black and White infant death rates and differences in these rates, taking into account state, county, and income effects, thereby enabling a direct test of the role of political incorporation in reducing health inequities.

METHODS

We used 3 sources of preexisting, deidentified data to construct our study database: US mortality data, US census population data, and US census county income data. We obtained a computerized file of US county mortality data for 1960 to 1967 from the US National Center for Health Statistics (NCHS)³⁴; the file contained individual records with data on calendar year of death, age, gender, race/ethnicity, and cause of death, which we aggregated to the county level on the basis of county identifiers. Because these data did not incorporate the post-1968 Federal Information Processing Standard county codes, we manually located and identified the correct county code for each of the 3073 counties and assigned the corresponding standard code.³⁵ We extracted corresponding data for 1968 to 2006, already aggregated to the county level, from the NCHS compressed mortality files.³⁶

In addition, we obtained published US vital statistics data³⁷ on annual average state-level infant death rates for 1939 to 1941, 1949 to 1951, and 1959 to 1961, stratified by race/ethnicity; the 1939 to 1941 data did not include Alaska or Hawaii (both then US territories and whose populations in 1940 respectively represented 0.055% and 0.322% of the US total population). We used these data in calculating the descriptive results presented in Figure 1a but not in the analytic models.

FIGURE 1—

Infant death rates among US Black and White infants in the Jim Crow (JC) and non–Jim Crow polities, by (a) rates per 1000 (rolling 3-year average) and (b) absolute difference in rates by birth cohort: 1940–2006.

Note. Gray lines are based on published state vital statistics data; black lines are based on 1960–2010 compressed mortality file.

In US census population data, the denominator for infant death rates comprises children younger than 1 year in the same calendar year in which the infant deaths occurred.³⁶ We obtained the relevant denominator data, by calendar year and race/ethnicity and aggregated to the county level, from the 1960 US census³⁵ and 1970 US census³⁸; we estimated the counts for 1961 to 1967 via linear interpolation. Data on yearly population counts for 1968 to 2006 were obtained from the NCHS compressed mortality files.³⁶ Given data limitations, we used the NCHS algorithm to calculate the infant death denominator for the years 1960 to 1987 (i.e., we multiplied “the population in the 1–4 age category by 0.25”).³⁶

Finally, we included decennial 1960 to 2000 county-level data on median family income from the US census^39,40 that we adjusted for inflation and regional cost of living.^35,41 We used linear interpolation for intercensal years and extrapolated for 2001 to 2006 on the basis of the slope for 1990 to 2000. In the case of both denominators and numerators, missing information due to counties lacking income data was minimal (< 1%).

Exposure Data and Exposed Groups

Exposure data: Jim Crow versus non–Jim Crow polity.

The Jim Crow polity consisted of the District of Columbia and the 21 states whose state laws, prior to enactment of the 1964 US Civil Rights Act, legalized racial discrimination in one or more of the following domains: education, transportation, hospital and penal institutions, welfare institutions (pauper homes), employment, marriage, public accommodation, other public venues, and voting. These states were Alabama, Arizona, Arkansas, Delaware, Florida, Georgia, Kansas, Kentucky, Louisiana, Maryland, Mississippi, Missouri, New Mexico, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, West Virginia, and Wyoming.¹ Within these states, additional laws sanctioned racial discrimination at the county and city levels.¹

Exposed groups: classification of race/ethnicity.

For our analyses, exposed groups comprised individuals designated “Black” and “White” in the mortality and population data. The only exceptions were the 1960 to 1967 compressed mortality file data³⁴ and the 1940 to 1960 US vital statistics,³⁷ which, as is well documented,^34,42,43 designated people only as White and non-White; for these data, we followed standard practice by reclassifying non-White individuals as Black.⁴³ Suggesting that this approach was reasonable, 92% of US “non-White” individuals were Black in 1960, and the mortality rates of these 2 groups were almost identical.⁴³ New Jersey death certificates did not identify race/ethnicity in 1962 or 1963, precluding use of these 2 years’ data (which represented less than 3% of the US population).

Statistical Analysis

We initially conducted descriptive analyses whereby we plotted infant death rates (Black, White, and Black–White difference), stratified by Jim Crow polity, and generated smoothed figures based on 3-year rolling rate averages. We then used Poisson log-linear models to analyze and test for differences in birth cohort effects in 2 ways: by Jim Crow polity within racial/ethnic groups and by race/ethnicity within each polity. In our analyses, we categorized birth year in 5-year groupings (1960–1964 through 2000–2004) and also included data for 2005–2006. We refer to these groups as birth cohorts (despite the simultaneity of birth cohort and period effects for analyses of an age group restricted to children younger than 1 year) to emphasize intragenerational and intergenerational implications for health status.^9,11,12

Recognizing that Jim Crow laws and their implementation varied by state and county,^1–6 as did resources and efforts to ameliorate their impacts once the laws were abolished,^{6–8,44–46} we took a multilevel approach to modeling state- and county-level variation, including state-level variation in cohort effects, by incorporating normally distributed random state, state by cohort, and county effects. Penalized quasi-likelihood in SAS PROC GLIMMIX (SAS Institute, Cary, NC) was used in fitting all models. We first fit models with no random effects and then added, sequentially, state-level random effects, state by cohort interactions, county-level random intercepts, and county median family income as a fixed effect.

Finally, we used joinpoint regression techniques to test for changes in temporal trends in Black and White infant death rates and their differences.^47,48 In these models, also called segmented line regression models, line segments are joined at points called “joinpoints,” whose location is estimated from the data and which denote statistically significant changes (P < .05) in the time trend; the slope of these line segments is interpretable as the change in the mortality rate or the rate difference per year.^47,48

RESULTS

Figure 1 presents descriptive data on 1940 to 2006 trends in infant death rates in the Jim Crow and non–Jim Crow polities among Black and White infants (Figure 1a) as well as the absolute difference in rates across these polities within each racial/ethnic group (Figure 1b). The percentages of infants born in the Jim Crow polity underscore the large and disproportionate concentration of Black infants in that polity. These percentages ranged from a high of 83.9% in 1940 to a low of 51.8% in 1990 among Black infants and from a high of 43.5% in 1940 to a low of 35.2% in 1960 among White infants.

As can be seen in Figure 1a, in the mid-1960s there was a unique convergence of Black infant death rates in the Jim Crow and non–Jim Crow polities. This convergence occurred against a background of generally declining infant death rates and a death rate among Black infants that was persistently twice as high as the rate among White infants in both polities. In 1960 to 1964, the Black infant death rate was 1.19 times higher (95% confidence interval [CI] = 1.18, 1.20) in the Jim Crow polity than in the non–Jim Crow polity (Table 1 ), corresponding to an absolute difference of approximately 8 per 1000 (Figure 1b) at a time when Black infant death rates were 49.1 per 1000 (95% CI = 48.7, 49.4) in the Jim Crow polity and 41.3 per 1000 (95% CI = 41.0, 41.7) in the non–Jim Crow polity.

TABLE 1—

US Black and White Infant Death Rates and Jim Crow Cohort Effects, by Jim Crow Polity: 1960–2006

	Jim Crow Polity			Non–Jim Crow Polity			Cohort-Specific Jim Crow Effect
Type of Effect	Deaths	Person-Years	Rate per 1000 (95% CI)	Deaths	Person-Years	Rate per 1000 (95% CI)	No State or County Effects, IRR (95% CI)	Random State Intercepts, IRR (95% CI)	Random State Intercepts + Random State by Cohort Interactions, IRR (95% CI)	Random State Intercepts and Random State by Cohort Interactions + County Random Effects, IRR (95% CI)	Random State Intercepts and Random State by Cohort Interactions + County Random Effects + County Income, IRR (95% CI)
Black
Cohort effects
1960–1964	86 234	1 757 858	49.1 (48.7, 49.4)	49 436	1 196 569	41.3 (41.0, 41.7)	1.19 (1.18, 1.20)	1.21 (1.08, 1.36)	1.11 (0.98, 1.26)	1.13 (1.00, 1.27)	1.08 (0.97, 1.22)
1965–1969	63 464	1 509 532	42.0 (41.7, 42.4)	45 567	1 154 192	39.5 (39.1, 39.8)	1.07 (1.05, 1.08)	1.09 (0.97, 1.22)	0.90 (0.80, 1.03)	0.92 (0.82, 1.04)	0.89 (0.79, 1.00)
1970–1974	44 501	1 353 443	32.9 (32.6, 33.2)	35 050	1 078 782	32.5 (32.2, 32.8)	1.01 (1.00, 1.03)	1.03 (0.92, 1.16)	1.09 (0.96, 1.25)	1.09 (0.96, 1.23)	1.05 (0.93, 1.19)
1975–1979	36 463	1 319 503	27.6 (27.4, 27.9)	28 284	1 028 877	27.5 (27.2, 27.8)	1.01 (0.99, 1.02)	1.03 (0.91, 1.16)	1.13 (0.99, 1.30)	1.13 (0.99, 1.27)	1.09 (0.97, 1.23)
1980–1984	32 107	1 456 421	22.0 (21.8, 22.3)	26 018	1 115 331	23.3 (23.0, 23.6)	0.95 (0.93, 0.96)	0.97 (0.86, 1.08)	1.02 (0.89, 1.17)	1.02 (0.90, 1.15)	0.99 (0.88, 1.12)
1985–1989	30 752	1 526 045	20.2 (19.9, 20.4)	27 343	1 212 897	22.5 (22.3, 22.8)	0.89 (0.88, 0.91)	0.91 (0.81, 1.02)	1.04 (0.91, 1.19)	1.03 (0.91, 1.16)	1.00 (0.89, 1.13)
1990–1994	29 592	1 717 586	17.2 (17.0, 17.4)	26 998	1 443 682	18.7 (18.5, 18.9)	0.92 (0.91, 0.94)	0.94 (0.83, 1.05)	1.04 (0.91, 1.18)	1.01 (0.89, 1.15)	0.99 (0.87, 1.12)
1995–1999	24 388	1 640 890	14.9 (14.7, 15.0)	19 504	1 303 910	15.0 (14.7, 15.2)	0.99 (0.98, 1.01)	1.01 (0.90, 1.14)	1.13 (0.99, 1.29)	1.10 (0.97, 1.25)	1.08 (0.95, 1.22)
2000–2004	24 996	1 830 446	13.7 (13.5, 13.8)	17 693	1 420 818	12.5 (12.3, 12.6)	1.10 (1.08, 1.12)	1.11 (0.99, 1.25)	1.27 (1.11, 1.45)	1.23 (1.09, 1.39)	1.21 (1.07, 1.36)
2005–2006	10 482	802 087	13.1 (12.8, 13.3)	7 071	584 501	12.1 (11.8, 12.4)	1.08 (1.05, 1.11)	1.10 (0.97, 1.23)	1.26 (1.09, 1.45)	1.23 (1.08, 1.40)	1.21 (1.06, 1.37)
P for Jim Crow by cohort effect							< .001	< .001	.003	.003	.001
State-level random effect								0.0426 (0.0097)	0.0103 (0.0037)	0.0083 (0.0038)	0.0075 (0.0035)
State by cohort interaction random effect									0.0379 (0.0034)	0.0290 (0.0027)	0.0287 (0.0027)
County random effect										0.0663 (0.0033)	0.0645 (0.0033)
County income quintile fixed effect
Q1 (lowest)											1.13 (1.11, 1.16)
Q2											1.15 (1.13, 1.18)
Q3											1.10 (1.08, 1.12)
Q4											1.07 (1.06, 1.09)
Q5 (highest; Ref)											1.00
White
Cohort effects
1960–1964	137 655	5 549 933	24.8 (24.7, 24.9)	247 872	11 397 912	21.7 (21.7, 21.8)	1.14 (1.13, 1.15)	1.18 (1.09, 1.29)	1.15 (1.07, 1.24)	1.13 (1.07, 1.21)	1.15 (1.08, 1.22)
1965–1969	103 755	5 050 307	20.5 (20.4, 20.7)	193 735	10 466 009	18.5 (18.4, 18.6)	1.11 (1.10, 1.12)	1.16 (1.06, 1.26)	1.15 (1.07, 1.25)	1.14 (1.08, 1.22)	1.16 (1.09, 1.23)
1970–1974	85 102	4 890 928	17.4 (17.3, 17.5)	141 564	9 330 582	15.2 (15.1, 15.3)	1.15 (1.14, 1.16)	1.20 (1.10, 1.30)	1.13 (1.05, 1.23)	1.13 (1.06, 1.20)	1.14 (1.07, 1.21)
1975–1979	63 672	4 764 425	13.4 (13.3, 13.5)	104 154	8 318 181	12.5 (12.4, 12.6)	1.07 (1.06, 1.08)	1.11 (1.02, 1.21)	1.10 (1.02, 1.19)	1.09 (1.03, 1.16)	1.11 (1.04, 1.18)
1980–1984	56 974	5 219 053	10.9 (10.8, 11.0)	90 952	8 775 090	10.4 (10.3, 10.4)	1.05 (1.04, 1.06)	1.10 (1.01, 1.20)	1.08 (1.00, 1.17)	1.08 (1.01, 1.15)	1.09 (1.02, 1.16)
1985–1989	50 667	5 523 983	9.2 (9.1, 9.3)	81 290	9 090 874	8.9 (8.9, 9.0)	1.03 (1.01, 1.04)	1.07 (0.98, 1.17)	1.10 (1.02, 1.19)	1.10 (1.03, 1.17)	1.11 (1.04, 1.18)
1990–1994	43 175	5 809 313	7.4 (7.4, 7.5)	69 530	9 718 433	7.2 (7.1, 7.2)	1.04 (1.03, 1.05)	1.09 (0.99, 1.18)	1.11 (1.03, 1.20)	1.11 (1.04, 1.18)	1.12 (1.05, 1.19)
1995–1999	38 304	5 866 374	6.5 (6.5, 6.6)	55 114	9 019 432	6.1 (6.1, 6.2)	1.07 (1.05, 1.08)	1.12 (1.02, 1.22)	1.13 (1.04, 1.22)	1.13 (1.06, 1.20)	1.14 (1.07, 1.21)
2000–2004	39 635	6 371 746	6.2 (6.2, 6.3)	51 504	9 169 004	5.6 (5.6, 5.7)	1.11 (1.09, 1.12)	1.16 (1.06, 1.26)	1.17 (1.08, 1.27)	1.17 (1.10, 1.25)	1.19 (1.11, 1.27)
2005–2006	16 716	2 683 092	6.2 (6.1, 6.3)	20 201	3 697 683	5.5 (5.4, 5.5)	1.14 (1.12, 1.16)	1.19 (1.09, 1.30)	1.21 (1.11, 1.31)	1.21 (1.13, 1.29)	1.22 (1.14, 1.31)
P for Jim Crow by cohort effect							< .001	< .001	.143	.114	.111
State-level random effect								0.0244 (0.0050)	0.0114 (0.0026)	0.0028 (0.0011)	0.0028 (0.0011)
State by cohort interaction random effect									0.0078 (0.0006)	0.0079 (0.0006)	0.0080 (0.0007)
County random effect										0.0368 (0.0012)	0.0370 (0.0012)
County income quintile fixed effect
Q1 (lowest)											0.98 (0.97, 0.99)
Q2											1.02 (1.01, 1.03)
Q3											0.99 (0.98, 1.00)
Q4											1.02 (1.02, 1.03)
Q5 (highest; Ref)											1.00

Note. CI = confidence interval; IRR = incidence rate ratio.

The Black Jim Crow versus non–Jim Crow polity relative risk declined to 1.07 (95% CI = 1.05, 1.08) in the 1965 to 1969 birth cohort and then dropped to approximately 1 in subsequent birth cohorts until the 2000 to 2004 birth cohort, when it rose to 1.10 (95% CI = 1.08, 1.12). However, this translates to an absolute difference of only about 1 in 1000 (Figure 1b) given overall declines in rates in both the Jim Crow polity (13.7 per 1000; 95% CI = 13.5, 13.8) and the non–Jim Crow polity (12.5 per 1000; 95% CI = 12.3, 12.6).

The P value for the Jim Crow by cohort effect was less than .004 in all models, and the temporal pattern, by birth cohort, in the Jim Crow effect was robust to inclusion of random state effects, random county effects, and fixed county income level. Inclusion of random state by cohort effects, however, slightly attenuated the 1960 to 1964 Jim Crow excess, strengthened the convergence in 1965 to 1969, and magnified the post-2000 Jim Crow excess. Thus, in the fully specified model the relative risk for the 1965 to 1969 birth cohort dropped below 1 (relative risk [RR] = 0.89; 95% CI = 0.79, 1.00), and the relative risk for the 2000 to 2004 birth cohort was 1.21 (95% CI = 1.07, 1.36).

No such temporal changes in the magnitude of the Jim Crow effect by birth cohort were evident for White infants before or after random state by cohort interactions had been taken into account (Table 1, Figure 1b), and the P value for the Jim Crow by cohort effect exceeded .10 in all models containing this interaction term. Instead, White infant death rates in the fully specified model consistently were about 1.1 to 1.2 times higher (with the 95% CI excluding 1) in the Jim Crow than in the non–Jim Crow polity in all birth cohorts, despite overall declining White infant death rates in both polities. In the Jim Crow polity, the White infant death rate decreased from 24.8 per 1000 (95% CI = 24.7, 24.9) in 1960 to 1965 to 6.2 per 1000 (95% CI = 6.1, 6.3) in 2005–2006, and the corresponding decrease in the non–Jim Crow polity was from 21.7 per 1000 (95% CI = 21.7, 21.8) in 1960 to 1965 to 5.5 per 1000 (95% CI = 5.4, 5.5) in 2005–2006.

The Black versus White relative risk for infant death, in the basic model with no state or county effects, ranged across birth cohorts from 1.9 to 2.3 in the Jim Crow polity and from 1.9 to 2.6 in the non–Jim Crow polity (Table 2; all 95% confidence intervals excluded 1). In both polities, the P value for the Jim Crow by cohort effect was below .001 in all models, and in both the basic and fully specified models the Black versus White relative risk was greatest in the 1985 to 1989 through 1995 to 1999 birth cohorts (Table 2).

TABLE 2—

US Black vs White Differences in Infant Death Rates and Jim Crow Birth Cohort Effects, by Jim Crow Polity: 1960–2006

			Cohort-Specific Race Effect
Type of Effect	White, Rate per 1000 (95% CI)	Black, Rate per 1000 (95% CI)	No State Effects, IRR (95% CI)	Random State Intercepts, IRR (95% CI)	Random State Intercepts + Random State by Cohort Interactions, IRR (95% CI)	Random State Intercepts and Random State by Cohort Interactions + County Random Effects, IRR (95% CI)	Random State Intercepts and Random State by Cohort Interactions + County Random Effects + County Income, IRR (95% CI)
Jim Crow polity
Cohort effects
1960–1964	24.8 (24.7, 24.9)	49.1 (48.7, 49.4)	1.98 (1.96, 1.99)	1.96 (1.94, 1.97)	1.96 (1.94, 1.97)	1.89 (1.88, 1.91)	1.89 (1.87, 1.91)
1965–1969	20.5 (20.4, 20.7)	42.0 (41.7, 42.4)	2.05 (2.03, 2.07)	2.02 (2.00, 2.04)	2.04 (2.02, 2.06)	1.97 (1.95, 1.99)	1.96 (1.94, 1.98)
1970–1974	17.4 (17.3, 17.5)	32.9 (32.6, 33.2)	1.89 (1.87, 1.91)	1.88 (1.85, 1.90)	1.89 (1.87, 1.91)	1.83 (1.81, 1.85)	1.83 (1.81, 1.85)
1975–1979	13.4 (13.3, 13.5)	27.6 (27.4, 27.9)	2.07 (2.04, 2.09)	2.06 (2.03, 2.09)	2.08 (2.05, 2.11)	2.01 (1.98, 2.04)	2.01 (1.99, 2.04)
1980–1984	10.9 (10.8, 11.0)	22.0 (21.8, 22.3)	2.02 (1.99, 2.05)	2.02 (1.99, 2.05)	2.01 (1.98, 2.04)	1.94 (1.91, 1.97)	1.94 (1.91, 1.97)
1985–1989	9.2 (9.1, 9.3)	20.2 (19.9, 20.4)	2.20 (2.17, 2.23)	2.20 (2.16, 2.23)	2.16 (2.13, 2.20)	2.08 (2.05, 2.11)	2.08 (2.05, 2.11)
1990–1994	7.4 (7.4, 7.5)	17.2 (17.0, 17.4)	2.32 (2.28, 2.35)	2.32 (2.28, 2.35)	2.29 (2.25, 2.32)	2.21 (2.17, 2.24)	2.20 (2.17, 2.24)
1995–1999	6.5 (6.5, 6.6)	14.9 (14.7, 15.0)	2.27 (2.24, 2.31)	2.28 (2.25, 2.32)	2.25 (2.21, 2.29)	2.18 (2.14, 2.22)	2.18 (2.14, 2.21)
2000–2004	6.2 (6.2, 6.3)	13.7 (13.5, 13.8)	2.19 (2.16, 2.23)	2.20 (2.17, 2.24)	2.16 (2.13, 2.20)	2.10 (2.07, 2.14)	2.10 (2.07, 2.14)
2005–2006	6.2 (6.1, 6.3)	13.1 (12.8, 13.3)	2.10 (2.05, 2.15)	2.11 (2.06, 2.16)	2.07 (2.02, 2.13)	2.03 (1.98, 2.08)	2.03 (1.98, 2.08)
P for Jim Crow by cohort effect			< .001	< .001	< .001	< .001	< .001
State-level random effect				0.0064 (0.0020)	0.0049 (0.0017)	0.0035 (0.0016)	0.0030 (0.0015)
State by cohort interaction random effect					0.0057 (0.0007)	0.0056 (0.0006)	0.0063 (0.0007)
County random effect						0.0388 (0.0002)	0.0379 (0.0016)
County income quintile fixed effect
Q1 (lowest)							1.13 (1.11, 1.15)
Q2							1.19 (1.16, 1.21)
Q3							1.10 (1.08, 1.12)
Q4							1.09 (1.08, 1.11)
Q5 (highest; Ref)							1.00
Non–Jim Crow polity
Cohort effects
1960–1964	21.7 (21.7, 21.8)	41.3 (41.0, 41.7)	1.90 (1.88, 1.92)	1.90 (1.88, 1.91)	1.93 (1.91, 1.95)	1.79 (1.77, 1.81)	1.79 (1.77, 1.81)
1965–1969	18.5 (18.4, 18.6)	39.5 (39.1, 39.8)	2.13 (2.11, 2.15)	2.12 (2.10, 2.15)	2.13 (2.11, 2.15)	1.96 (1.94, 1.98)	1.95 (1.93, 1.97)
1970–1974	15.2 (15.1, 15.3)	32.5 (32.2, 32.8)	2.14 (2.12, 2.17)	2.13 (2.11, 2.16)	2.14 (2.11, 2.16)	1.97 (1.94, 1.99)	1.96 (1.94, 1.99)
1975–1979	12.5 (12.4, 12.6)	27.5 (27.2, 27.8)	2.20 (2.17, 2.22)	2.19 (2.16, 2.22)	2.17 (2.15, 2.20)	2.01 (1.98, 2.04)	2.01 (1.98, 2.04)
1980–1984	10.4 (10.3, 10.4)	23.3 (23.0, 23.6)	2.25 (2.22, 2.28)	2.24 (2.21, 2.28)	2.22 (2.19, 2.25)	2.05 (2.02, 2.08)	2.05 (2.02, 2.08)
1985–1989	8.9 (8.9, 9.0)	22.5 (22.3, 22.8)	2.52 (2.49, 2.56)	2.52 (2.48, 2.55)	2.47 (2.44, 2.51)	2.29 (2.26, 2.33)	2.29 (2.26, 2.33)
1990–1994	7.2 (7.1, 7.2)	18.7 (18.5, 18.9)	2.61 (2.57, 2.65)	2.61 (2.57, 2.64)	2.57 (2.54, 2.61)	2.39 (2.36, 2.43)	2.40 (2.36, 2.43)
1995–1999	6.1 (6.1, 6.2)	15.0 (14.7, 15.2)	2.44 (2.40, 2.48)	2.44 (2.40, 2.48)	2.41 (2.37, 2.45)	2.24 (2.20, 2.28)	2.25 (2.21, 2.29)
2000–2004	5.6 (5.6, 5.7)	12.5 (12.3, 12.6)	2.22 (2.18, 2.25)	2.22 (2.18, 2.25)	2.18 (2.14, 2.22)	2.04 (2.00, 2.07)	2.04 (2.01, 2.08)
2005–2006	5.5 (5.4, 5.5)	12.1 (11.8, 12.4)	2.21 (2.15, 2.27)	2.22 (2.16, 2.28)	2.20 (2.13, 2.26)	2.06 (2.00, 2.12)	2.07 (2.01, 2.13)
P for Jim Crow by cohort effect			< .001	< .001	< .001	< .001	< .001
State-level random effect				0.0060 (0.0016)	0.0055 (0.0017)	0.0029 (0.0013)	0.0029 (0.0014)
State by cohort interaction random effect					0.0074 (0.0007)	0.0071 (0.0007)	0.0071 (0.0007)
County random effect						0.0392 (0.0016)	0.0324 (0.0016)
County income quintile fixed effect
Q1 (lowest)							0.96 (0.95, 0.98)
Q2							1.00 (0.98, 1.01)
Q3							1.00 (0.99, 1.01)
Q4							1.02 (1.01, 1.03)
Q5 (highest; Ref)							1.00

Note. CI = confidence interval; IRR = incidence rate ratio.

Joinpoint analyses (Table 3 ) of the crude rates (in analyses not constrained to the temporal bounds of the 5-year birth cohorts) likewise showed that, during the mid-1960s, significant slope changes (P < .05) occurred for Black but not White infant death rates. Among Black infants, the slope declined more steeply in the Jim Crow than in the non–Jim Crow polity (−0.0768 vs −0.0413), and these declines both exceeded the small, significant mid-1960s change in slope estimate for the Black–White difference in infant death rates (−0.0028) observed only in the Jim Crow polity. Additional small, significant changes in slope occurred. Among Black infants, there was a steeper decline in both polities in the late 1980s and increasing rates in the late 1990s and early 2000s. Among White infants, there was a decline in the early 1970s in the Jim Crow polity and a rise in the late 1990s in both polities. Finally, with respect to the Black–White difference, there were increases in the post-1980 Jim Crow polity and the post-2000 non–Jim Crow polity.

TABLE 3—

Joinpoint Regression Analyses: US Black and White Infant Death Rates, and Absolute Difference in Black–White Infant Death Rates, in the Jim Crow and Non–Jim Crow Polities, 1960–2006

		Change in Slope Estimates
Rate and Joinpoint Segment	Joinpoint (95% CI)	Slope Estimate	Slope SE	Slope Probability > t
Jim Crow polity
Black
1	…	…	…	…
2	1964 (1962, 1967)	−0.0768	0.0253	0.0049
3	1968 (1965, 1986)	0.0245	0.0222	0.2791
4	1984 (1977, 1990)	0.0349	0.0145	0.0229
5	1989 (1980, 1997)	−0.0394	0.0170	0.0273
6	1995 (1991, 2004)	0.0310	0.0094	0.0026
White
1	…	…	…	…
2	1964 (1962, 1965)	−0.0632	0.0369	0.0966
3	1967 (1965, 1968)	0.0936	0.0517	0.0803
4	1970 (1969, 1972)	−0.1123	0.0531	0.0429
5	1973 (1972, 1977)	0.0537	0.0375	0.1622
6	1995 (1993, 1998)	0.0318	0.0022	0.0000
Black–White difference
1	…	…	…	…
2	1965 (1962, 1967)	−0.0028	0.0010	0.0093
3	1969 (1966, 1974)	0.0018	0.0009	0.0574
4	1982 (1971, 2004)	0.0002	0.0001	0.0226
Non–Jim Crow polity
Black
1	…	…	…	…
2	1965 (1962, 1970)	−0.0413	0.0135	0.0043
3	1984 (1974, 1987)	0.0514	0.0302	0.0988
4	1988 (1986, 1991)	−0.0626	0.0303	0.0469
5	2001 (1992, 2004)	0.0403	0.0114	0.0012
6
White
1	…	…	…	…
2	1970 (1965, 1972)	−0.0553	0.0375	0.1498
3	1973 (1969, 1976)	0.0684	0.0428	0.1197
4	1977 (1972, 1994)	−0.0233	0.0209	0.2728
5	1996 (1991, 2000)	0.0234	0.0033	0.0000
6
Black–White difference
1	…	…	…	…
2	1985 (1979, 1987)	0.0011	0.0013	0.4287
3	1988 (1986, 1991)	−0.0012	0.0013	0.3578
4	2001 (1995, 2004)	0.0006	0.0002	0.0005

Note. CI = confidence interval.

DISCUSSION

Our analysis of US Black and White infant death rates from 1959 to 2006 provides novel evidence of a unique convergence in Black infant death rates in the Jim Crow versus non-Jim Crow polity that occurred in 1965 to 1969, following passage of the 1964 US Civil Rights Act, which overturned Jim Crow laws legalizing racial discrimination.^2,6 It also delineates the 21st-century reemergence of higher rates in the Jim Crow polity. Both trends notably coexisted with a persistent 2-fold excess Black infant death rate. Providing perspective on the magnitude of the mid-1960s absolute decline (of approximately 8 per 1000) in the Jim Crow versus non–Jim Crow difference in Black infant death rates is the fact that this precipitous drop sharply contrasts with the prior decade of reduced declines in these rates (Figure 1), which included notable increases in some US southern states.^31,37 In addition, the size of the decline exceeds the rate observed for White infant death rates in 2005–2006, is equal in magnitude to nearly two thirds of the Black infant death rate during these same years, and is 8 times larger than the subsequent 21st-century Jim Crow effect (approximately 1 per 1000).

These temporal patterns of birth cohort effects (big effect, no effect, reemergence of smaller effect) were robust to inclusion of random state and county effects and county income level; the magnitude of these effects, moreover, was increased by inclusion of random state by cohort interactions. Together, our findings suggest that these temporal patterns in effects occurred above and beyond county differences in income level (e.g., as influenced by the mid-1960s “war on poverty”^49,50) and that state variations in the enforcement and abolition of Jim Crow laws^{1–6,44–46} probably mattered. Further attesting to the specificity of the effects observed, no comparable temporal patterns in Jim Crow birth cohort effects occurred among White infants or were evident in Black versus White comparisons within these polities.

Study Limitations

One potential limitation of our study concerns flaws in the case (numerator) and population (denominator) data. Suggesting, however, that our findings are not artifacts arising from inaccurate data, federal reports indicate that 99% of all US deaths and births since 1960 have been registered^37,51 and that the US census undercount, disproportionately affecting people who are poor or of color, decreased substantially between 1960 and 2000, from 3.1% to 0.1% for the total population and from 6.6% to 2.8% for the Black population.⁵² Although some local studies suggest there may have been a degree of underreporting of infant births and deaths in rural and impoverished counties up to the mid-1970s, especially among African Americans,^51,53,54 such selective underreporting would lead to a conservative bias (because of the disproportionate deflating of infant death rates in the Jim Crow polity during the Jim Crow era).

In addition, misclassification of “White” and “Black” deaths in US mortality data has been shown to be minor.⁵⁵ The effect of our having to equate the non-White with the Black population for 1960 to 1967 was also likely small⁴³ and, if extant, would have produced a conservative bias against obtaining a mid-1960s birth cohort effect (given evidence suggesting that infant death rates were higher in the Black vs non-Black “non-White” population⁵⁶).

Our findings are also unlikely due to bias arising from the well-established “healthy immigrant” effect, whereby immigrants typically have better health status and better birth outcomes than their nonimmigrant counterparts in both the sending and receiving country or region.^57,58 In the case of US internal migration, the probable impact of the great migration (referring to the exodus of Black Americans from the Jim Crow to non–Jim Crow states that extended from the early 1900s to the 1970s^2–6,59) would be conservative, that is, increasing rather than decreasing the Jim Crow effect. The post-1970 increase in the US foreign-born Black population, a consequence of changes in US immigration law, in turn would likely have had little impact, given evidence indicating that the proportions of the US foreign-born population residing in the states composing the Jim Crow and non–Jim Crow polities are similar.⁶⁰

Study Findings in Context

Further lending credence to our findings are the results of the 4 extant US empirical investigations explicitly focused on the health impacts of the abolition of Jim Crow laws.^7–10 Three analyzed infant and maternal health,^7–9 with data spanning the period 1955 to 2000. These studies showed greater gains in maternal health status among Black than White women who gave birth in the early versus late 1960s, especially in the US South. They also showed sharper declines in postneonatal mortality rates among Blacks than among Whites between 1965 and 1971, especially in southern US states, and county variations in declines in these rates in Mississippi were highly correlated with desegregation of hospital facilities.⁹

The fourth study, based on data from 1950 to 1980, compared gender-stratified Black–White differences in life expectancy at ages 35 and 65 years and mortality rates (both all cause and cause specific) among US residents 35 to 64 years of age and within 4 US regions (West, Southwest, Midwest, and South). The study authors concluded that “the Civil Rights Act of 1964 did have a salutary impact on the health of black women in the United States, especially in the South.”^10(p166) Three of the studies just described also discussed the likely decades-long implications of the mid-1960s changes,^8–10 given the salience of early life conditions with respect to adult health status as well as the health of children born to the next generation.^11,12,19,20

Our study builds on and adds to these earlier investigations in 4 ways. First, we analyzed all states that had Jim Crow laws as a single polity defined by policy regime, as opposed to the focus in prior research^7–10 on single states or geographic regions (e.g., the US South, which does not include states outside the South that also had Jim Crow laws). Second, our novel approach to modeling birth cohort effects allowed us to take into account intrapolity and interpolity state and county heterogeneity as well as county income level, thereby newly providing indirect evidence of likely variations in the extent and enforcement of Jim Crow legislation^1–6,59 and its abolition^{6,8,44–46,59} (otherwise difficult to document empirically within any given locality, let alone across all US states and counties¹) and also related policies affecting economic conditions and access to medical care (e.g., the mid-1960s war on poverty^49,50 and the concurrent establishment of Medicaid and Medicare^44–46). Third, using an extended time frame, we were able to observe both the disappearance of the initial large Jim Crow effect and its later reduced reappearance.

Finally, by extending beyond the primarily Black–White comparisons of the earlier studies^7–10 and much of the contemporary research on racial/ethnic health inequities,^11,12,16,42 our results underscore the critical need for careful specification of relevant causal contrasts.^13–15 As our findings demonstrate, it is not the same to assess the impact of the abolition of Jim Crow laws on the health of people whose legal status and political incorporation were directly affected by the policy change (e.g., comparing trends in Black infant death rates in the Jim Crow polity before and after the abolition of Jim Crow laws and comparing birth cohort effects among Black infants born in the Jim Crow vs non–Jim Crow polity) as it is to assess the magnitude of health inequities in different groups created and defined by the inequitable social relations^{1–6,11,14,22–27,42} underlying the existence of inequitable policies (e.g., comparing Black vs White Americans). Both contrasts are important but do not necessarily yield the same results.

Directions for Future Research

Several promising lines of research are accordingly suggested by our findings. The first pertains to elucidating the diverse pathways by which political incorporation can potentially reduce health inequities,^19–30 a research agenda that will require far richer, as well as longer term, institutional data and microdata than are typically available in nationally representative data sets.²² In the case of Jim Crow legislation, for example, it will be important to develop rigorous ways to quantify the pre-1965 social, medical, and economic hardships (well recognized at the time^{1–6,46,59,61–63}) imposed by both Jim Crow and the entrenched nonlegally mandated racial discrimination extant throughout the United States. It is also important to document how these hardships began to be altered by the myriad changes in social conditions and health care access enabled by the abolition of Jim Crow laws and the concurrent war on poverty, along with the social movements that led to the enactment of these policies.^{2–6,44–46,49,50,59,61–63} Finally, it is important to quantify subsequent policy changes (including backlash against Great Society programs)^{6,19–22,35,45} that may have contributed to the 21st-century reemergence of a small Jim Crow effect.

In addition, research is needed to account for why, despite political incorporation, a still unexplained 2-fold Black excess risk in infant death rates has persisted. This discrepancy points not only to gaps in knowledge about and action regarding relevant societal determinants of probable biological pathways (e.g., involving preterm birth)^{7–9,11,28,31–33,35,56–58,64–67} but also to limitations in theories that assume political incorporation will, in and of itself, resolve societal inequalities.^23–27

A second potentially useful avenue of research, relevant to valid analysis of societal determinants of health, involves the theoretical work needed to demarcate appropriate comparison groups, carefully conceptualized in relation to the underlying generative causal societal relationships that give rise to health inequities.¹⁴ For example, in our study the political incorporation effect observed among the Black population did not automatically translate into corresponding birth cohort effects for Black versus White infant death rate inequities. The most likely explanation, warranting empirical testing, is that comparisons of trends in the magnitude of health inequities can be affected by additional changes that differentially affect the societal groups being compared, above and beyond the impact of a given policy change.

Conclusions

Our results add to the scant empirical data indicating that abolition of Jim Crow laws has contributed to shaping US infant death rates. These findings are not of historical interest only. Instead, if our causal interpretation of the observed trends is correct, the enduring embodied consequences are potentially large, given not only likely life-course and intergenerational effects^8–12 but also the numbers of people affected. As we have documented, nearly two thirds of all Black infants born in 1960 to 1964 were born in Jim Crow states.

Moreover, in 2013, all US-born individuals 67 years or older (i.e., the age group in which the bulk of mortality occurs) were born, came of age, and had already lived the first 20 years of their lives—and perhaps had their first child—when Jim Crow was legal in 21 of the 50 states in addition to the District of Columbia, alongside de facto discrimination in the remaining 29 states.¹¹ Engaging with embodied histories^11,14,19 is thus essential if efforts to understand, let alone alter, current health inequities are to succeed.