Interracial Marriage, Migration and Loving

Black/white interracial marriage rate for black males is defined as the ratio of black males married to white females to the ratio of married black males with either a black or a white spouse. Black female IMR is defined similarly.

There are various definitions of the “South”. In this study, I adopt the definition of the Southern region used by the U.S. Census Bureau, which includes Alabama, Arkansas, Delaware, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, West Virginia and the District of Columbia.

As of the 2000 census enumeration, 54.8 % of blacks lived in the South, while 17.6, 18.8 and 8.9 % lived in the Northeast, Midwest, and West respectively. Similarly, as of 2010 census enumeration, 56.5 % of blacks lived in the South, while 16.8, 17.9 and 8.8 % lived in the Northeast, Midwest, and West respectively. See Tolnay (2003) and Rastogi et al. (2010)..

The Loving state group does not entirely correspond to the definition of the Southern region used by the U.S. Census Bureau. The Loving states also include Missouri and exclude Maryland and D.C. Nevertheless, I loosely label the Loving states as the Southern states.

FindLaw: Loving v. Virginia, 388 U.S. 1 (1967) http://​laws.​findlaw.​com/​us/​388/​1.​html.

Marriage creates economic surplus for both parties by joint consumption, marriage specific investments, division of labor, and risk pooling. As Hamermesh (2011) argues, domestic partnership fails to create as large an economic surplus as marriage would have created. The author estimates an upper-bound value of ¡DOLLAR/¿3 billion for California to gain in marriage surplus per year if it were to allow same-sex marriages.

The samples include all married individuals regardless of the number of marriages. Foreign-born blacks (whites) make up 3.9, 4.5, and 8 % (4.9, 5, and 7.8 %) of the total black (white) population in 1980, 1990 and 2007, respectively in the U.S. (see Gibson and Jung 2006; and Grieco 2010). Following Fryer (2007), I focus on U.S.-born blacks to minimize problems that may arise from immigration and nativity issues. Batson et al. (2006) find that U.S.-born blacks are more likely to marry whites compared to non-U.S.-born blacks. Qian and Lichter (2011) confirm these findings for U.S.-born black men but not for U.S.-born black women.

Only Louisiana and Maryland had laws banning marriages between blacks and Native Americans. This study focuses exclusively on black/white marriages and all individuals of other races are dropped from the sample because black/non-white interracial marriages are not common. Calculations based on the Current Population Survey reveal that black/non-white intermarriage rates are extremely low at 0.9, 0.8 and 1.1 in 1980, 1990 and 2000 respectively (compared to black/white intermarriage rates at 4.7, 5.4, and 8.2 in 1980, 1990 and 2000 respectively). Author’s calculations are based on Table60 http://​www.​census.​gov/​compendia/​statab/​2011/​tables/​11s0060.​xls of the U.S. Bureau of the Census 2010.

In 2000, black/white IMR is only 2.6 % for black females, while it is 7.7 % for black males. The IMR for black females is still very low at 3.6 % in the 2010 census, while it is 8.1 % for black males.

The results do not change when Maryland is included in the sample.

Table 2 shows that the total number of married black males and females decreased from the 1980 to 1990 census and increased from the 1990 to 2000 census. After examining the census data sets carefully, I find that the total number of black males and females increased over the time. The total number of black males is 265,059, 274,569, and 300,759 in the 1980, 1990, and 2000 censuses respectively. The variation in the sample sizes in this study can be attributable to the fluctuations in the number of married blacks in the total sample of blacks, which may be due to changing attitudes towards marriage and/or higher incarceration rates of black males as found in Charles and Luoh (2010).

Here I focus on heterosexual marriages between blacks and whites.

The black male (female) IMRs are 5.6, 8.6, and 12.2 % (1.5, 2.9, and 4.7 %) in the 1980, 1990, and 2000 censuses respectively in the non-Loving states, while the black male (female) IMRs are only 1.4, 2.7, and 4.9 % (0.6, 1.0, and 1.6 %) for corresponding census years in the Loving states.

For instance, Loving-born and non-Loving-born black male IMR is 1.8 % and 6.1 %, respectively, while IMR among the residents of Loving and non-Loving states is lower at 1.2 % and 4.6 % respectively.

Interestingly, the first column shows that the majority of black males and females (80 and 74 % respectively) with white spouses married after 1967, while only half of the black males and females (50 and 48 % respectively) with black spouses married after 1967.

The 2010 census data were not included in the analysis because the older generation would have been 61 years of age or older as of 2010 and this would completely wipe out the older generation in the 2010 sample.

Intraracial marriages refer to those between two individuals of the same race.

This study uses a difference-in-differences estimation framework even though the unique structure of the question does not entirely correspond to a natural experiment in the classical sense. Wooldridge (2010) points out that the use of difference-in-differences is not unique to natural experiment and program evaluation settings and provides a clear explanation of the difference-in-differences estimator. In its simplest form, the outcome of interest is observed for two groups for two periods. The “treatment group” is exposed to a treatment in the second period but not in the first period, while the “control group” is not exposed to the treatment during either period. In the current framework, the treatment can be thought of as the repeal of the anti-miscegenation laws in the 16 Loving states in 1967 by the U.S. Supreme Court. The outcome of interest is the likelihood of residing in the Loving states. To be precise, in this study, there are no “treatment” and “control” groups in the classical sense. In this study the “affected group” (similar to a treatment group) consists of blacks who have white spouses. The affected group of individuals were not allowed to marry/reside in the Loving states prior to 1967, while they were free to marry/reside anywhere after 1967. The “unaffected group” (similar to a control group) consists of blacks who have black spouses. The unaffected group of individuals were not impacted by these laws at all. Because marrying and residing in Loving states was illegal for those blacks with white spouses before 1967, it is not possible to construct a typical before-after comparison depending on spousal race. A novel solution is to look at those who were 18 or younger as of 1967 (younger generation), who thus were more likely to marry after 1967 and were not legally affected by anti-miscegenation laws, and then compare them with blacks who were 19 years of age or older in 1967 (older generation). Therefore, the younger generation is similar to the “after period” or period 2, and the older generation is similar to “before period” in a natural experiment setting.

D D _{n o n − L o v i n g − b o r n} is calculated similarly to Eq. 1.

While the state group of residence in the year of the census enumeration is identified as the migration destination, some blacks may have moved to another state within their state group of birth, moved out of their state group of birth and returned, or moved after collection of the census data. Complete information on an individual’s migration history would be ideal, but using the state of residence as the destination should not invalidate the results.

Black and white unemployment rates, group size and sex ratio are calculated by using the 1970, 1980, 1990 and 2000 IPUMS census data sets. The table that shows the ratio of black male/white male unemployment rates, group size and sex ratio by states is available upon request from the author.

For instance, in a state with a lower sex ratio, black males may be more likely to intermarry due to limited availability of black females or vice versa.

Variations in these three variables arise from state of residence, and they are calculated as the 1970 and 1980 census data averages for the 1980 sample,1970, 1980 and 1990 averages for the 1990 sample,and 1970, 1980, 1990 and 2000 averages for the 2000 sample. To check the robustness of the results I repeated the estimation exercise by including these variables calculated using each of the 1970, 1980, 1990, and 2000 census data sets in all samples. The results are very similar to those reported in Table 8 and are available upon request.

Regression results are available upon request.

For the extended specification variations in the ratio of black/white unemployment rate, group size and sex ratio arise from state of residence, and they are calculated as the 1970 and 1980 census data averages for the older generation drawn from the 1980 sample; the 1970, 1980 and 1990 averages for the younger generation from the 1990 sample; and the 1970, 1980, 1990 and 2000 averages for the younger generation from the 2000 sample.

The border-states sample includes nine Loving states (Texas, Oklahoma, Missouri, Kentucky, West Virginia, Virginia, Delaware, Arkansas, and Tennessee), ten non-Loving states (New Mexico, Colorado, Kansas, Nebraska, Iowa, Illinois, Indiana, Ohio, Pennsylvania, and New Jersey), and D.C.

In 1980 the younger generation of Loving-born black males with white and black spouses are 20.5 to 11 percentage points more likely to reside in one of the Loving states compared to those of the older generation. This gap yields a double-difference estimate of 9.5 percentage points, more than twice as large as that of Table 7.

In addition, Table 5 shows that the majority of first marriages occurred after 1967 for those who turned 18 after 1967; however, the sample of married blacks used in this analysis also includes those who were married twice or more.

According to National Health Statistics Reports, life expectancy at birth is 63 for black males in the 1980s (Arias 2002).

This robustness check is not viable for the 2000 census sample, because restricting the sample to those aged 18–51 would wipe out the older generation in the 2000 sample. Unreported results show that when the samples are restricted to those aged 18–51, in most cases, the double- and triple-difference estimates are slightly larger than those in Tables 7 and 8 and they are not affected in terms of statistical significance.

I use overall divorce rate in state of residence. Recent research by Fu and Wolfinger (2011) finds no evidence that crossing a black/white racial boundary increases the divorce rates.

With the exception of 2000. The divorce rates for 2000 are calculated by taking the average divorce rates from 1990 to 1998.

The correlations between average divorce rates are available upon request.

See http://​www.​census.​gov/​hhes/​www/​housing/​resseg/​excel_​msa.​html for more information on various segregation indexes.

The data on pairwise correlations between dissimilarity indexes are available upon request.

The triple-difference estimates are highly statistically significant in five out of six cases.

Based on a survey that was conducted by Harris Interactive and Witeck-Combs Communications (see Witeck 2011).

See Gates (2011). The top ten states with the highest percentage of coupled households (married and unmarried) that are gay and lesbian are the District of Columbia, California, Vermont, Massachusetts, New York, Washington, New Mexico, Nevada, Florida, Oregon, Georgia and Rhode Island.

These states punished the offenders by whipping in puritan days or by sentencing them to life terms in antebellum days.