Using Census and Surname Data to Oversample Racial/Ethnic Minorities in DC
1. USING CENSUS & SURNAME DATA TO
OVERSAMPLE RACIAL / ETHNIC
MINORITIES IN DC
@ jxpeugh | @ssrs_solutions
jpeugh@ssrs.com
JORDON PEUGH, SSRS | DAVID DUTWIN, SSRS | MICHAEL BADER, AMERICAN UNIVERSITY
AAPOR 72nd Annual Conference
New Orleans, LA | May 20, 2017
2. INTRODUCTION
• Study: The Washington D.C. Area Survey (DCAS) conducted in 2016 for American University. ABS mail survey.
• Research Goal: Understand factors that affect quality of life in two diverse types of neighborhoods in DC Metro:
─ Quadrivial: White, Asian, black, and Latino residents each make 10%+ and no single group comprises a majority
(12%); {Quadrivial is a Latin word meaning four roads meeting.}
─ Disproportionally Latino: Latinos make up 25%+ and not Quadrivial (14% ).
• Research Topics: Life in these communities, crime, businesses, nonprofits, local government, race relations.
• Today’s Goal: Explore the success of methods used to improve representation of ethnic minorities living within
these neighborhoods.
2@jxpeugh | @ssrs_solutions
3. RACIAL/ETHNIC DISTRIBUTION OF NEIGHBORHOODS
BASED ON CENSUS DATA
Asian
Latino /
Hispanic
African
American
White /
Other
Total
Quadrivial Neighborhoods 18% 25% 23% 35% 100%
Disproportionately Latino Neighborhoods 9% 43% 25% 24% 100%
Total 13% 34% 24% 29% 100%
3@jxpeugh | @ssrs_solutions
4. SAMPLE PLAN
Expected Incidence of Racial Groups Within Strata and Sample Allocation
% of HH Sample Allocation
Quadrivial Neighborhoods
Households (HH) with identified Asian surnames 5% 10%
HHs with identified Hispanic surnames 9% 9%
HHs in high African American neighborhoods (25%+AA) 15% 23%
All other HHs 22% 9%
Disproportionally Latino Neighborhoods
HHs with identified Asian surnames 3% 4%
HHs with identified Hispanic surnames 12% 12%
HHs in high African American neighborhoods (25%+AA) 17% 25%
All other HHs 18% 9%
Total 100% 100%
4@jxpeugh | @ssrs_solutions
5. THE RESULTS (1)
Distribution of Respondents by Race Within Strata - Expected vs. Actual (% Within Each Strata)
Self-reported race/ethnicity: Asian Latino / Hispanic
Sample Strata Expected Actual Expected Actual
Quadrivial Neighborhoods
HHs with identified Asian surnames 50% 82% 5% 2%
HHs with identified Hispanic surnames 5% 6% 75% 68%
HHs in high AA neighborhoods (25%+AA) 13% 11% 16% 6%
All other HHs 18% 17% 16% 4%
Disproportionately Latino Neighborhoods
HHs with identified Asian surnames 50% 72% 5% 1%
HHs with identified Hispanic surnames 5% 7% 75% 85%
HHs in high AA neighborhoods (25%+AA) 3% 5% 30% 14%
All other HHs 11% 14% 38% 12%
Total % 13% 22% 34% 18%
Total N Expected and Actual 156 259 408 212
5@jxpeugh | @ssrs_solutions
6. THE RESULTS (2)
Distribution of Respondents by Race Within Strata - Expected vs. Actual (% within each strata)
Self-reported race/ethnicity: African American White/ Other
Sample Strata Expected Actual Expected Actual
Quadrivial Neighborhoods
HHs with identified Asian surnames 5% 4% 40% 12%
HHs with identified Hispanic surnames 5% 0% 15% 26%
HHs in high AA neighborhoods (25%+AA) 40% 28% 31% 55%
All other HHs 22% 12% 44% 67%
Disproportionately Latino Neighborhoods
HHs with identified Asian surnames 5% 8% 40% 18%
HHs with identified Hispanic surnames 5% 1% 15% 7%
HHs in high AA neighborhoods (25%+AA) 55% 45% 12% 36%
All other HHs 15% 23% 37% 51%
Total % 24% 21% 29% 38%
Total N Expected and Actual 288 244 348 441
6@jxpeugh | @ssrs_solutions
7. RESPONSE RATES
Sample Stratum Response Rate
Quadrivial Neighborhoods 14%
Households (HH) with identified Asian surnames 16%
HHs with identified Hispanic surnames 11%
HHs in high African American neighborhoods (25%+AA) 14%
All other HHs 15%
Disproportionately Latino Neighborhoods 11%
HHs with identified Asian surnames 17%
HHs with identified Hispanic surnames 9%
HHs in high African American neighborhoods (25%+AA) 11%
All other HHs 13%
Total 13%
7@jxpeugh | @ssrs_solutions
8. DID THE DESIGN HELP?
Expected Self-Reported Race/Ethnicity Based on Actuals
Asian
Latino /
Hispanic
African
American
White/
Other
Missing/
Refused
Total
Hypothetical totals, assuming simple
random sample and actual response rates
213 213 220 515 75 1236
Compared to Actual Study Results 259 212 244 441 66 1222
8@jxpeugh | @ssrs_solutions
9. Compared to Census Data
Asian
Latino /
Hispanic
African
American
White /
Other
Quadrivial Neighborhoods
Census Data 18% 25% 23% 35%
DCAS Self-Reported 27% 13% 16% 44%
Disproportionately Latino Neighborhoods
Census Data 9% 43% 25% 24%
DCAS Self-Reported 17% 25% 28% 31%
Total
Census Data 13% 34% 24% 29%
DCAS Self-Reported 22% 18% 21% 38%
9@jxpeugh | @ssrs_solutions
10. DISCUSSION
• Asian surname sample was more productive for reaching Asians than anticipated. Also, Asians responded more
readily than other racial minorities, which was unexpected.
• Use of targeted high density AA neighborhoods was productive in increasing number of African Americans.
However, lower proportions of African Americans in these areas (or responding) than expected based on Census
data.
• Surname sample was productive in increasing number of Latinos. However, it would have helped to oversample
this group to combat lower response rates among this sample source.
• More Asians and whites than expected in these neighborhoods overall – is this due to responsiveness or
demographic shifts?
10@jxpeugh | @ssrs_solutions
11. LIMITATIONS
• Unusual character of these neighborhoods limits usefulness of conclusions to inform research in other geographic
areas.
• Fairly low response rate overall – what is the impact of responsiveness by race/ethnicity?
• Potential neighborhood transformation not being captured in census data.
11@jxpeugh | @ssrs_solutions
12. THANK YOU
JORDON PEUGH
VP, HEALTH POLICY & PUBLIC OPINION RESEARCH
JPEUGH@SSRS.COM
484-840-4337
@JXPEUGH
12