June 11, 2015

Multiracial in America

Appendix A: Survey Methodology

The analysis in this report is based on a Pew Research Center survey conducted online Feb. 6-April 6, 2015 among a probability based sample of 1,555 multiracial Americans ages 18 and older. The sample of multiracial adults was identified after contacting and collecting basic demographic information from 21,224 adults nationwide. For comparative purposes, an additional 1,495 adults from the general public were surveyed, including oversamples of 154 adults who are single-race black and non-Hispanic and 208 adults who are as single-race Asian and non-Hispanic.

The survey was conducted by the GfK Group for the Pew Research Center using KnowledgePanel, its nationally representative online research panel. KnowledgePanel members are recruited through probability sampling methods and include individuals both with and without internet access. KnowledgePanel provides internet access for those who do not have it and, if needed, a device to access the internet when they join the panel. A combination of random digit dialing (RDD) and address-based sampling (ABS) methodologies have been used to recruit panel members (in 2009 KnowledgePanel switched its sampling methodology for recruiting panel members from RDD to ABS). The panel includes households with landlines and cellular phones, including those only with cell phones, and those without a phone. Both the RDD and ABS samples were provided by Marketing Systems Group (MSG). KnowledgePanel continually recruits new panel members throughout the year to offset panel attrition as people leave the panel. The survey was conducted in English and Spanish.

All active members of the GfK panel were eligible for inclusion in this study. This project was divided into two stages. For Stage 1 a sample of general population adults age 18 and over was selected along with oversamples of non-Hispanic single-race blacks and Asians that were identified using GfK’s panelist profile data. Stage 2 consisted of a general population sample that was split randomly into four replicates of approximately 8,000 panel members each.

Those selected for each replicate were asked their race or origin, the race or origin of their biological mother and father as well as the race or origin of their grandparents, great-grandparents and earlier ancestors. At each stage respondents could select more than one race or origin. In addition, those with a Hispanic background were asked if they considered Hispanic to be their race, their ethnicity, or both their race and ethnicity. Only those who were identified as having a mixed-race background based on these initial screening questions qualified to continue on to take the main multiracial survey. Slightly different qualifying criteria were used for some of the replicates, as described in the next section.

I. Defining Multiracial Adults

To determine their racial background, a series of five screening questions were asked of all 21,224 GfK panelists selected to participate in the online survey.

The first question asked individuals to select their own race or origin and is reproduced below as it appeared on the questionnaire. Respondents were asked about each of their biological parent’s race or origin, their grandparents’ race or origin, and the race or origin of great-grandparents or any earlier ancestors, if different than any of the races indicated in the previous questions. The same choice options, including the option to mark ‘Hispanic,” were used in all five questions; to avoid repetition, the option list only appears with the first question. While respondents had the option to mark “Some other race or origin” in each question, the option to list any other race(s) or origin(s) was only offered on the first question. Respondents could mark as many races or origins as applied and were instructed to consider only the background of their biological parents, grandparents or earlier ancestors.

The screening questions were:

Those with a Hispanic background as indicated in their responses to any of the previous five questions were asked:

To qualify as multiracial for the purposes of taking the survey (though the definitions used in the report are somewhat narrower), a panel member in Replicate 1 must have met one of the following criteria:

In addition, Hispanics could continue on to take the full multiracial survey in Replicate 1 if they did not fit the definition of multiracial based on criteria 1-4 but considered being Hispanic a race and met one of the following criteria:

In Replicates 2 and 4, a panel member would qualify to take the multiracial survey if they fit the definition of multiracial based on criteria 1-3 or 5-7. Those who only met the great grandparent or earlier ancestor criteria (4 and 8) did not qualify.

In Replicate 3, only those who met criteria 1-3 qualified for the multiracial survey.

For purposes of analysis, those who qualified as multiracial on the basis of the racial makeup of great-grandparents or earlier ancestors (criteria 4 and 8) were not part of the multiracial sample.

All sampled members received an initial email to notify them of the survey and provide a link to the survey questionnaire. Follow-up reminders were sent after three and seven days to those who had not yet responded.

II. Weighting

Individuals included in the Stage 1, nationally representative sample were first assigned base weights that account for their probability of selection from the overall panel. This includes weighting down single-race blacks and Asians to account for the fact that they were oversampled. Next, cases that completed the survey were weighted to match the March 2014 Current Population Survey (CPS) in order to correct for potential bias due to nonresponse.

The overall sample of respondents was weighted to match CPS estimates of the following characteristics:

Additionally, individual racial and ethnic groups were weighted to be internally representative on the following variables:

The groups were: non-Hispanic whites, non-Hispanic blacks, non-Hispanic other or mixed races, and Hispanic.

For the Stage 2 sample, the weighting process was similar. The full, sample of 21,224 respondents was weighted to be nationally representative using the characteristics listed above.

For the Stage 2 samples, five race/ethnicity groups were weighted to be internally representative: non-Hispanic whites, non-Hispanic blacks, non-Hispanic other, Hispanic, and non-Hispanic mixed races.

After the full Stage 2 sample was weighted to be nationally representative, respondents who did not qualify for inclusion in their replicate were removed. At this point, a final adjustment was made to account for the fact that the criteria for inclusion differed across replicates, ensuring individuals who qualified under different criteria were represented proportionally.

Details about the GfK panel-level weights can be found at: http://www.gfk.com/Documents/GfK-KnowledgePanel-Design-Summary.pdf

III. Sampling Error and Design Effects

Weighting to adjust for disproportionate sampling and nonresponse reduces the precision of estimates beyond what would be achieved under simple random sampling. In this report, all measures of sampling error and statistical tests of significance take into account the design effect of weighting.

The margin of sampling error at the 95% confidence level for results based on the total Stage 2 sample (n=21,224) is plus or minus 1 percentage point.

The margin of sampling error for results based on the general population sample and single-race black and white oversamples (n=1,495) is plus or minus 3 percentage points.

The margin of sampling error for results based on the multiracial sample (n= 1,555) is plus or minus 3.8 percentage points.

The sample sizes and margins of sampling error for the six largest multiracial groups referenced in this report are:

In addition to sampling error, question wording and the practical difficulties in conducting surveys can introduce error or bias into the findings of opinion polls.