Secondary data

Analyses are based primarily upon the American Community Survey (ACS), as well as the 1980 decennial census, both of which were obtained from IPUMS-USA. Since 2008, the ACS, which is an annual, nationally representative survey, has included a question asking if the respondent married within the past 12 months, which is used to classify people as newlyweds for those years.

The 1980 census, which was the first to collect reliable data on Hispanic origin, also collected data allowing for the identification of first-time newlyweds. The questionnaire asked people to list the age at which they first married. For this analysis, anyone whose age at first marriage was the same as their age at the time of the survey was identified as a newlywed, as was their spouse. As a result, only those newlyweds who are part of a couple where either the bride or groom (or both) recently married for the first time are identified as newlyweds in 1980. About 90% of the married population in 1980 included people who were in a first marriage, or who were married to someone in a first marriage, and the intermarriage rates for those in “first marriage” couples differed little from the rates among all married couples.

For the estimates of intermarriage in years other than 1980 and 2008 to 2015, a retrospective or “look-back” method was used. The 1980 census data were used to estimate intermarriage among newlyweds from 1967 to 1979. Using the 1980 census data regarding respondent age at the time of survey and age at first marriage, the year of first marriage among couples who were still in their first marriage was established. Then, annual estimates of newlywed intermarriage were calculated. For instance, all couples who were first married in 1967 were identified as newlyweds in that year and were classified as either being intermarried or not intermarried. This same approach was used for subsequent years through 1979.

The same general approach was used to estimate intermarriage rates for the years 1981 to 2007 using the 2008 to 2015 ACS data. However, for these years, all married couples were included, regardless of whether they were in their first marriage or a subsequent marriage. To establish intermarriage rates among newlyweds in 1981, for instance, a combined file of 2008 to 2015 data was used to identify all people who had wed in that year. These 1981 newlyweds were then classified as either being in an intermarriage, or not. This same procedure was used to calculate intermarriage rates for newlyweds in subsequent years.

While using census and ACS data to create estimates for prior years would be problematic if intermarriages break up more often than other types of marriages, a number of additional analyses suggest that using this retrospective approach produces reliable estimates of intermarriage rates. See Chapter 3 of “Marrying Out: One-in-Seven New U.S. Marriages Is Interracial or Interethnic” for more details.

While statistics regarding overall intermarriage rates are based on single year estimates, more detailed analyses using ACS data combine multiple years of data in order to increase sample size. Analyses examining age patterns or patterns by metro status are based on a combined sample of 2011-2015 ACS data. All other detailed analyses are based on a combined sample of 2014 and 2015 ACS data.

Estimates regarding the total share of presently married people who are intermarried are based on data from the 1990 and 2000 decennial census, as well as the 1980 decennial census and 2008-2015 ACS data.

In analyses that are based on presently married people, only those who are married and living with a spouse are included, since data regarding the racial and ethnic profile of spouses living apart are not available through the ACS or census. The vast majority (95%) of people who state that they are married in the ACS are married and living with their spouse.

Since 2013, it has been possible to identify most same-sex married couples in the ACS. For almost all analyses regarding 2013 and later, individuals in a same-sex marriage are included. The only exception occurs for the couple-level analysis, which is limited to other-sex couples in order to highlight the interaction of gender and race.

Beginning with the 2000 census, individuals could choose to identify with more than one group in response to the race question. In this analysis, these multiracial people are treated as a separate race category, different from those who identify as a single race, including those who identify as “some other race.” (As with single-race individuals, a multiracial person who also identifies as Hispanic would be classified as Hispanic.) Since the introduction of the multiracial option on the census, the share of individuals who identify as such has grown substantially, and this has likely contributed to the increases in the share of married couples who are classified as intermarried.¹⁴

Survey data

The survey data in this report come from two sources. The question on whether more people of different races marrying each other is a good thing or bad thing for society comes from Pew Research Center telephone surveys conducted between 2010 and 2017. Data reported for 2017 are drawn from a mode experiment conducted Feb. 28-March 12, 2017, on the American Trends Panel (ATP). In order to avoid any potential mode effects, only data from the telephone portion of the mode experiment are used in this report. A total of 1,778 panelists were interviewed by phone and the margin of error is plus or minus 4.0 percentage points. Interviews are conducted in both English and Spanish, but the Hispanic sample in the ATP is predominantly native born and English speaking. For more information, see the Methodology for that survey.

The series of questions on favorability of a close relative marrying someone of a specified race or ethnicity is drawn from NORC’s General Social Survey (GSS).

Sampling errors and statistical tests of significance take into account the effect of weighting. In addition to sampling error, one should bear in mind that question wording and practical difficulties in conducting surveys can introduce error or bias into the findings of opinion polls.