Empirical Evidence to Understand the 2020 Census Citizenship - - PowerPoint PPT Presentation

▶
empirical evidence to
SMART_READER_LITE
LIVE PREVIEW

Empirical Evidence to Understand the 2020 Census Citizenship - - PowerPoint PPT Presentation

Jun 24, 2023 180 likes •502 views

Empirical Evidence to Understand the 2020 Census Citizenship Controversy Gina Walejko, Brian Kriz, Sarah Evans, Anna Sandoval Girn, Yazmn A. Garca Trejo, Kyley McGeeney, & Nancy Bates AAPOR Annual Conference May 18, 2019 Toronto,

slide-1
SLIDE 1

2020CENSUS.GOV

Empirical Evidence to Understand the 2020 Census Citizenship Controversy

Gina Walejko, Brian Kriz, Sarah Evans, Anna Sandoval Girón, Yazmín A. García Trejo, Kyley McGeeney, & Nancy Bates AAPOR Annual Conference May 18, 2019 Toronto, Canada

Disclaimer: Any views expressed are those of the authors and not those of the U.S. Census Bureau.

slide-2
SLIDE 2

2020CENSUS.GOV

Outline

Background Research Question Data & Methods Results Discussion & Conclusion

2

slide-3
SLIDE 3

2020CENSUS.GOV

Background

3

slide-4
SLIDE 4

2020CENSUS.GOV

Effect of citizenship question?

4

  • March 26, 2018 – Secretary of Commerce requested Census Bureau add citizenship

question to 2020 Census

  • One week into 2018 Census Test in Providence, RI
  • Too late to test effects of question during this 2020 Census dress rehearsal
  • Secondary data analysis attempted to estimate the effect of question on response rate
  • Brown and colleagues (2018) estimated a 5.8 percentage point greater drop in self-response

rates for households with any noncitizens compared to households without noncitizens; 1.7 percentage point drop in self-response in 2020 Census

  • 2019 Census Test will examine effect of question on response (Velkoff, 2019), but results

will come after Supreme Court decision expected in late June and printing of 2020 Census forms

slide-5
SLIDE 5

2020CENSUS.GOV

Research Question

5

slide-6
SLIDE 6

2020CENSUS.GOV 6

How does the addition of a citizenship question affect intention to respond?

slide-7
SLIDE 7

2020CENSUS.GOV

Data & Methods

7

slide-8
SLIDE 8

2020CENSUS.GOV

2020 CBAMS

8

5 10 15 20 25 30 35 40 45 10 20 30 40 50 60 70 80 90 100 2/20/2018 2/22/2018 2/24/2018 2/26/2018 2/28/2018 3/2/2018 3/4/2018 3/6/2018 3/8/2018 3/10/2018 3/12/2018 3/14/2018 3/16/2018 3/18/2018 3/20/2018 3/22/2018 3/24/2018 3/26/2018 3/28/2018 3/30/2018 4/1/2018 4/3/2018 4/5/2018 4/7/2018 4/9/2018 4/11/2018 4/13/2018 4/15/2018 4/17/2018 4/19/2018 Number Focus Groups Survey Response Rate Dates of CBAMS Data Collection Google Trends Interest Survey Response Rate

  • No. Focus Groups
slide-9
SLIDE 9

2020CENSUS.GOV

Dependent Variable

9

Variable Question Wording Values=1 Intent If the census were held today, how likely would you be to fill out the census form? Mark (X) ONE box.

□ Extremely likely □ Very likely □ Somewhat likely □ Not too likely □ Not at all likely

Extremely likely; very likely

slide-10
SLIDE 10

2020CENSUS.GOV

2020 CBAMS Matching & Modeling

10

Variables Categories Matched? Modeled? Contact Strategy Internet “First,” Internet “Choice” Yes Yes Response Mode Mail, Internet Yes Yes Civic Engagement Index Low (0-3 items), High (4-10 items) Yes Yes Race and Hispanic Origin Hispanic, White, Black, Asian, Small sample races Yes Yes Age 18-34, 35-44, 45-64, 65+ Yes Yes Sex Male, Female Yes Yes Education No college, At least some college Yes Yes Nativity Born in US, Born outside US (including Puerto Rico) Yes Yes Knowledge Index Low (0-4 correct), High (5-9 correct) Yes Yes Treatment Control (responded before announcement), Treatment (responded after announcement) N/A Yes

slide-11
SLIDE 11

2020CENSUS.GOV

2020 CBAMS Subgroup Analysis

11

Population

  • f Interest

Control Weight Count Treatment Weight Count Control Treatment Difference p-value Difference in difference Difference in difference SE Difference in difference p-value Hispanic 7665 2558 0.6973 0.5736

  • 0.12

0.0009

  • 0.09

0.022494 0.0001 Everyone else 3490 1428 0.6855 0.6524

  • 0.03

0.0275 Source: 2020 CBAMS Survey, Public Use Dataset, CBDRB-FY18-422

slide-12
SLIDE 12

2020CENSUS.GOV

Results

12

slide-13
SLIDE 13

2020CENSUS.GOV

Treatment Effect

13

Coefficient Internet Choice contact strategy +* Internet response mode + Low civic engagement

  • ***

Non-Hispanic, White

  • Non-Hispanic, Black

+ Non-Hispanic, Asian

  • **

Non-Hispanic, Other races

  • Ages 35-44

+ Ages 45-64 +*** Ages over 65 +*** Female +* No college

  • ***

Born outside U.S. + Knowledge Index – High Knowledge +*** Treatment

  • ***

Constant +* Observations 15,141 Note: *=<0.05; **=<0.01; ***=<0.001 Source: 2020 CBAMS Survey, Public Use Dataset, CBDRB-FY18-422

If the census were held today, how likely would you be to fill out the census form? Extremely likely & very likely = 1

slide-14
SLIDE 14

2020CENSUS.GOV

Evidence from Focus Groups

  • Citizenship question as barrier to

participation was highest among those who believed:

  • 1. the purpose of the question is to find

undocumented immigrants,

  • 2. their information will be shared across

agencies, potentially leading to deportation, and

  • 3. their ethnic group is facing an

inhospitable political environment.

14

“I wouldn’t answer it because ICE is working with a lot of different groups on deportation sweeps and stuff, and I guess it would make me feel like I’m aiding in

  • that. They’re doing a lot of illegal

stuff, and so I feel like I wouldn’t fill out any of the questions.”

  • Focus group participant

Source: 2020 CBAMS Focus Groups, CBDRB-FY19-043

slide-15
SLIDE 15

2020CENSUS.GOV

Subgroup Identification

15

1. In what country or territory were you born?  Outside of the United States (print name of country, or Puerto Rico, Guam, etc.) 1. Are you of Hispanic, Latino, or Spanish origin?  Yes 2. How well do you speak English  Not well, not at all

  • Some racial and ethnic minorities: Focus group research suggests participation barrier related to how vulnerable

participants believe they and their ethnic community are to retaliatory actions based on ethnicity; focus groups with non-English speakers

  • U.S. born & second-generation Hispanics: Research suggests U.S. born Hispanics and second-generation

Hispanics also react negatively to addition of citizenship question (Escudero & Becerra, 2018; Kissam et al. 2019)

  • Non-citizens: Secondary data analysis suggests households with non-citizens less likely to complete a form with

citizenship question (Brown et al. 2018) 1. What is your race?  Asian Indian, Chinese, Filipino, Japanese, Korean, Vietnamese, or Other Asian 2. How well do you speak English  Not well, not at all Foreign born Hispanic, Spanish speaker, U.S. born Hispanic, not speak English Asian NH, not speak English 1. Are you of Hispanic, Latino, or Spanish origin?  Yes 2. Do you speak a language

  • ther than English at home?

 Yes 3. What is the language?  Spanish 4. In what country or territory were you born?  In the United States.

slide-16
SLIDE 16

2020CENSUS.GOV

Subgroup Analysis - Intent

16

  • 0.09***
  • 0.14***
  • 0.15**
  • 0.02
  • 0.21**
  • 0.09***
  • 0.3
  • 0.25
  • 0.2
  • 0.15
  • 0.1
  • 0.05

Hispanic Hispanic, not speak English well Hispanic, Spanish speaker, U.S. born Asian, NH Asian, not speak English well Foreign born Source: 2020 CBAMS Survey, Public Use Dataset, CBDRB-FY18-422 *=<0.05; **=<0.01; ***=<0.001

slide-17
SLIDE 17

2020CENSUS.GOV

Discussion & Conclusion

17

slide-18
SLIDE 18

2020CENSUS.GOV

Discussion

18

  • Effect of citizenship question addition on intended response greater among some

subgroups.

  • Effect of citizenship question addition on intended response may not be limited to

those worried about direct repercussions  effect also seen in Hispanic Spanish speakers who are born in the U.S.

  • Supports other research:
  • U.S. born Hispanics react as negatively to addition of citizenship question as foreign born in

Providence County, Rhode Island (Escudero & Becerra, 2018)

  • 34 percent of second-generation Hispanic survey respondents in San Joaquin Valley said they

would no longer respond with addition of citizenship question (Kissam et al. 2019)

slide-19
SLIDE 19

2020CENSUS.GOV

In conclusion…

19

  • After attempting to control for differences between responders

before and after citizenship question announcement…

  • …those who responded to CBAMS after addition of citizenship

question less likely to intend to respond to census.

  • Effect more drastic with certain subgroups.
  • Effect supported by focus group participants.
slide-20
SLIDE 20

2020CENSUS.GOV 20

Questions? gina.k.walejko@census.gov

slide-21
SLIDE 21

2020CENSUS.GOV

Appendix

21

slide-22
SLIDE 22

2020CENSUS.GOV

Limitations

22

  • Cannot be certain respondents heard news about citizenship and – if so –

how much or via what channel.

  • Analysis represents “intention-to-treat” rather than actual treatment.
  • Pre-treatment covariates may not fully explain treatment assignment.
  • Focus group data may not be representative of subgroups.
slide-23
SLIDE 23

2020CENSUS.GOV

Bibliography

23

  • Brown, D., Higgeness, M., Dorinski, S., Warren, L., & Vi, M. (2018). “Understanding the Quality of

Alternative Citizenship Data Sources for the 2020 Census.” Center for Economic Studies Memorandum 18-38. U.S. Census Bureau. Available at https://www2.census.gov/ces/wp/2018/CES-WP-18-38.pdf

  • Escudero, K. A. & Becerra, M. (2018). “The Last Change to Get it Right: Implications of the 2018

Test of the Census for Latinos and the General Public.” NALEO Educational Fund.

  • Kissam, E., Mines, R., Quezada, C., Intili, J. A., & Wadsworth, G. (2019). “San Joaquin Valley

Latino Immigrants: Implications of Survey Findings for Census 2020.” Available at: https://www.shfcenter.org/assets/SJVHF/SJVCRP_Survey_Findings_Report_011819_Web.pdf

  • Velkoff, V. (2019). “2019 Census Test.” Presentation during the 2020 Census Program

Management Review, February 1, 2019. Available at https://www2.census.gov/programs- surveys/decennial/2020/program-management/pmr-materials/02-01-2019/pmr-2019-test-2019- 02-01.pdf

  • Wadsworth, G., Kissam, E., Quezada, C., & Intili, J. A. (2019). “Troubled Reflections: Latino

Immigrants’ Thinking About Census 2020.” San Joaquin Valley Census Research Project. Available at: https://www.shfcenter.org/assets/SJVHF/SJVCRP_Troubled_Reflections_022719.pdf

slide-24
SLIDE 24

2020CENSUS.GOV

Next steps

24

  • Model additional dependent variables – for example –

willingness to encourage others to fill out the census.

  • Perform a content analysis of news surrounding the

addition of a citizenship question to better understand the media environment.

slide-25
SLIDE 25

2020CENSUS.GOV

Dependent Variables

25

Variable Question Wording Values=1 Intent If the census were held today, how likely would you be to fill out the census form? Extremely likely; very likely Encourage How likely are you to encourage someone you know to fill out the 2020 Census form? Extremely likely; very likely Locate Is the census used to locate people living in the country without documentation? Yes, used for this Who Counts Does the census count both citizens and non-citizens, or only citizens? Yes, used for counting both citizens and non-citizens Share Information How concerned are you, if at all, that the Census Bureau will share answers to the 2020 Census with other government agencies? Extremely concerned; very concerned Use Against How concerned are you, if at all, that the answers you provide to the 2020 Census will be used against you? Extremely concerned; very concerned Confidentiality How concerned are you, if at all, that the Census Bureau will not keep answers to the 2020 Census confidential? Extremely concerned; very concerned Trust How much of the time do you think you can trust the FEDERAL government to do what is right? Just about always; most of the time Benefit Do you believe that answering your 2020 Census form could personally benefit or harm YOU in any way?. Harm; both benefit and harm

slide-26
SLIDE 26

2020CENSUS.GOV

2020 CBAMS Survey

  • 2020 Census Barriers, Attitudes, and Motivators Study (CBAMS) Survey
  • February 20, 2018 – April 17, 2018
  • 50,000 housing units
  • Oversampled populations of interest in order to make inferences to these groups
  • 50 states and DC; excludes Puerto Rico
  • Mail or internet; English or Spanish
  • Pre-paid incentive in first of five mailings
  • 17,283 respondents; 39.4 percent response rate (AAPOR III)

26

slide-27
SLIDE 27

2020CENSUS.GOV

2020 CBAMS Focus Groups

27

Audience Group Recruitment Criteria

  • No. of

Focus Groups

  • No. of Focus Groups

after Citizenship Announcement American Indian and Alaska Native (AIAN) Identifies as AIAN; speaks English at home. 6 groups 4 groups Black or African American Identifies as Black or African American; has a high school education or less, or has a household income under $30,000 per year. 4 groups 4 groups Chinese—Cantonese and Mandarin Speaks Chinese at home. 4 groups 2 groups Low Internet Proficiency Uses the internet a few times a week or less; speaks English at home. 4 groups 0 groups Middle Eastern or North African (MENA) Identifies as MENA; speaks English at home. 4 groups 4 groups Native Hawaiian and Pacific Islander (NHPI) Identifies as Native Hawaiian or from Micronesian islands or Polynesian/Melanesian islands; speaks English at home. 4 groups 4 groups Rural Identifies as White, non-Hispanic; lives in a ZIP Code Tabulation Area identified as rural in the 2010 Census. 2 groups 0 groups Spanish (Puerto Rico) Resident of Puerto Rico; speaks Spanish at home. 4 groups 4 groups Spanish (U.S. Mainland) Speaks Spanish at home; is a resident of mainland U.S. 4 groups 4 groups Vietnamese Speaks Vietnamese at home. 4 groups 2 groups Young and Mobile 18- to 24-year-old who is not married; rents their home. 2 groups 2 groups

slide-28
SLIDE 28

2020CENSUS.GOV

Treatment Effect

28

Intent Encourage Share info. Use against Coefficient Coefficient Coefficient Coefficient Internet Choice contact strategy +* +

  • **
  • **

Internet response mode + +***

  • ***
  • ***

Low civic engagement

  • ***
  • ***

+ + Non-Hispanic, White

  • ***

Non-Hispanic, Black + +*** + + Non-Hispanic, Asian

  • **

+ + +*** Non-Hispanic, Other races

  • +

+ Ages 35-44 + +

  • Ages 45-64

+*** +*** +** + Ages over 65 +*** +*** +** + Female +* +**

  • No college
  • ***

+ +** +* Born outside U.S. + + +*** +*** Knowledge Index – High Knowledge +*** +***

  • ***

Treatment

  • ***
  • ***
  • **
  • *

Constant +*

  • ***
  • ***
  • ***

Observations 15,141 15,141 15,141 15,141 Note: *=<0.05; **=<0.01; ***=<0.001 Source: 2020 CBAMS Survey

slide-29
SLIDE 29

2020CENSUS.GOV

Subgroup Analysis – Share Info.

29

  • 0.01
  • 0.06
  • 0.05
  • 0.04
  • 0.05
  • 0.05*
  • 1
  • 0.8
  • 0.6
  • 0.4
  • 0.2

0.2 0.4 Hispanic Hispanic, not speak English well Second gen. Hispanic proxy Asian, NH Asian not speak English well Foreign born

slide-30
SLIDE 30

2020CENSUS.GOV

Subgroup Analysis – Use Against

30

  • 0.06
  • 0.06
  • 0.12
  • 0.04
  • 1
  • 0.8
  • 0.6
  • 0.4
  • 0.2

0.2 0.4 Hispanic Hispanic, not speak English well Second gen. Hispanic proxy Asian, NH Asian, not speak English well Foreign born

slide-31
SLIDE 31

2020CENSUS.GOV

Subgroup Analysis - Encourage

31

0.06

  • 0.12
  • 0.05*
  • 0.14*
  • 1
  • 0.8
  • 0.6
  • 0.4
  • 0.2

0.2 0.4 Hispanic Hispanic, not speak English well Second gen. Hispanic proxy Asian, NH Asian, not speak English well Foreign born

slide-32
SLIDE 32

2020CENSUS.GOV

Analysis Approach

1. Tested effect of intervention with weighted logistic regression model:

32

𝑚𝑝𝑕𝑗𝑢 𝑞(𝑦) = 𝛾0 + 𝛾1Internet Choice + 𝛾2Internet Response Mode + 𝛾3Low Civic Engagement + 𝛾4Non−Hispanic, White + 𝛾5Non−Hispanic, Black + 𝛾6Non−Hispanic, Asian + 𝛾7Non−Hispanic, Other + 𝛾8Age 35−44 + 𝛾9Age 45−64 + 𝛾10Age 65 or older + 𝛾11Female + 𝛾12No College + 𝛾13Born Outside U.S. + 𝛾14High Knowledge + 𝜸𝟐𝟔Treatment

Where 𝜸𝟐𝟔 is the coefficient for the treatment fixed effect variable

2. Tested effect of treatment on hard-to-count subpopulations using weighted conditional difference in means (i.e. conditional average treatment effect) for subpopulation and compliment