2020 Census Evaluation: Evaluating Privacy and Confidentiality Concerns Study Plan

2019.23.i_2020 Census Privacy and Confidentiality Evaluation Study Plan.pdf

2020 Census

2020 Census Evaluation: Evaluating Privacy and Confidentiality Concerns Study Plan

OMB: 0607-1006

Document [pdf]
Download: pdf | pdf
The memorandum and attached document(s) was prepared for Census Bureau internal use. If
you have any questions regarding the use or dissemination of the information, please contact
the Stakeholder Relations Staff at dcco.stakeholder.relations.staff@census.gov.

2020 CENSUS PROGRAM INTERNAL MEMORANDUM SERIES: 2019.23.i
Date:

May 31, 2019

MEMORANDUM FOR: The Record
From:

Deborah M. Stempowski (Signed May 31, 2019)
Chief, Decennial Census Management Division

Subject:

2020 Census Evaluation: Evaluating Privacy and Confidentiality Concerns Study
Plan

Contact:

Jennifer Reichert
Decennial Census Management Division
301-763-4298
jennifer.w.reichert@census.gov

This memorandum releases the final version of the 2020 Census Evaluation: Evaluating Privacy and
Confidentiality Concerns Study Plan, which is part of the 2020 Census Program for Evaluations and
Experiments (CPEX). For specific content related questions, you may also contact the authors:
Aleia Clark Fobia
Center for Behavioral Science Methods
301-763-4075
aleia.yvonne.clark.fobia@census.gov

Jennifer Hunter Childs
Center for Behavioral Science Methods
301-763-4927
jennifer.hunter.childs@census.gov

Jessica Holzberg
Center for Behavioral Science Methods
301-763-2298
jessica.holzberg@census.gov

Kelly Mathews
Decennial Statistical Studies Division
301-763-5639
kelly.m.mathews@census.gov

Matthew Virgile
Center for Behavioral Science Methods
301-763-4745
matthew.virgile@census.gov

census.gov

2020 Census Evaluation
Evaluating Privacy and Confidentiality
Concerns Study Plan

Aleia Clark Fobia, CBSM
Jessica Holzberg, CBSM
Jennifer Hunter Childs, CBSM
Kelly Mathews, DSSD
Matthew Virgile, CBSM

5/31/19
Version 3

Page intentionally left blank.

<>

Table of Contents
I.

Introduction ......................................................................................................................... 3

II.

Background ......................................................................................................................... 4

III.

Assumptions........................................................................................................................ 7

IV.

Research Questions ............................................................................................................. 8

V.

Methodology ....................................................................................................................... 8

VI.

Data Requirements ............................................................................................................ 16

VII.

Risks.................................................................................................................................. 16

VIII.

Limitations ........................................................................................................................ 16

IX.

Issues That Need to be Resolved ...................................................................................... 16

X.

Division Responsibilities .................................................................................................. 17

XI.

Milestone Schedule ........................................................................................................... 17

XII.

Review/Approval Table .................................................................................................... 18

XIII.

Document Revision and Version Control History ............................................................ 18

XIV. Glossary of Acronyms ...................................................................................................... 18
XV.

References ......................................................................................................................... 19

1

<>

Page intentionally left blank.

2

<>

I.

Introduction

Privacy and confidentiality has been at the forefront of concerns as the census moves online and
increases reliance on administrative records. The Census Bureau has been conducting research
on respondents’ privacy and confidentiality concerns with online response and administrative
records use as the focus of one of the teams from the Research and Testing phase leading up to
the 2020 Census. Thus far, much of this work has been hypothetical, with respondents asked how
they would feel if x strategy were to be employed in the census. The 2020 Census provides an
opportunity to evaluate respondent privacy and confidentiality concerns and their relationship to
response mode, item nonresponse, and mismatches between administrative records and selfreported data in a decennial census environment.
Privacy and confidentiality research addresses key elements in the 2020 Census Operational Plan
and the guiding principles for the 2030 Census. As the 2020 Census is the first time that the
majority of respondents will be encouraged to respond to the census on the internet, a key
element of the Optimizing Self-Response innovation area is assuring respondents that their data
is secure and treated as confidential (US Census Bureau, 2017; p. 19). Government and private
sector data breaches are salient public events that have potentially weakened respondents’ trust
in the Census Bureau’s ability to maintain privacy and keep data confidential. Respondents need
to be sure that their personal information is protected, particularly when responding online. In
fact, the public’s perception of the Census Bureau’s ability to safeguard response data has been
identified as a high-level risk to the 2020 Census Program (Blumerman & Fontenot, 2017).
This evaluation is a telephone and in-person survey of decennial census respondents focused on
their privacy and confidentiality concerns. The evaluation is an opportunity to measure how the
web response option affects privacy and confidentiality concerns of respondents who have had
the chance to use this option. Based on previous research, we expect that respondents will have
particular privacy and confidentiality concerns associated with responding online (Holzberg &
Fobia, 2016; Morales, Holzberg, & Eggleston, 2017). The public perception of how the Census
Bureau handles privacy and confidentiality in 2020 will shape how the Census Bureau prepares
for and executes a 2030 Census, especially one that might be all-electronic.
Expanded use of administrative records in 2020 and the principle of a primarily records-based
census in 2030 are also areas where research about respondent privacy and confidentiality
attitudes is crucial. The 2020 Census plans to use administrative records and third-party data to
target advertising, validate respondent submissions, and reduce nonresponse follow-up
workloads (US Census Bureau, 2017; p. 22). Administrative record use for these purposes has
been identified as a high-level risk to the 2020 Census (Blumerman & Fontenot, 2017). Unlike
online response, administrative record use may not be a census strategy of which many
respondents are aware. However, previous research has shown that how administrative record
use is framed has an impact on its favorability (Singer et al. 2011; Childs 2015; Childs, Walejko,
and Eggleston 2015). People are also more skeptical of the Census Bureau’s ability to keep data
secure and confidential when sharing between agencies, which will occur when administrative
records are used (Childs et al. 2015a). While some groups support the use of administrative
records to replace or prepopulate census forms, a misstep in this area as we move toward a
records-based 2030 Census could jeopardize the trust that the public has in the Census Bureau
3

<>

(Mitre, 2017; JASON 2016). This evaluation allows us to collect up-to-date feedback from
respondents on administrative record use in the census environment.
This evaluation also provides an opportunity to investigate any potential link between privacy
and confidentiality concerns and mismatches between self-reported data and administrative
records. It is an open question whether people with more privacy and confidentiality concerns
are more likely to have mismatches between survey responses and administrative records, or
missing administrative records. By linking responses to this evaluation with administrative
records we could begin to address the relationship between privacy concerns and consistency of
administrative records with self-reported data.
Concerns about privacy and confidentiality will continue to shape how the Census Bureau
interacts with the public and how we address these concerns is of critical importance as we
execute the 2020 Census and begin to prepare for 2030. There is growing, recent evidence that
these types of concerns are increasingly salient and if unaddressed could contribute to the
undercount of certain populations and item nonresponse (CSM 2017).
It is critical to conduct this research within the 2020 Census environment for three related
reasons. First, the decennial census environment is unique in that the salience of government data
collection will likely be quite high for most Americans. Assessing respondents’ concerns with
government data collection shortly after having made a decision about whether and how to share
data with the government is an opportunity to gauge attitudes about privacy and confidentiality
more accurately than at other times. Second, public discourse such as news media might also
discuss matters of privacy and confidentiality during a decennial census that people might not
often think about, helping to shape opinions and attitudes. Finally, the privacy and confidentiality
concerns with the amount and types of data collected in a decennial census might also be
different than those associated with a survey that has a smaller sample size but more in-depth
data collection, such as the American Community Survey (ACS).
To fulfill our constitutional mandate, the 2020 and 2030 censuses will be used to apportion
districts for representation in Congress. Public trust in the accuracy and reliability of the census
will be important to support the fulfilling of that mandate, and this evaluation will provide us
with the tools to craft messaging and approaches that will ensure that trust.

II.

Background

This work continues studies conducted as part of previous decennial census evaluations as well
as more recent work that has been ongoing throughout this decade. Surveys of privacy and
confidentiality concerns were undertaken as part of the 1990 and 2000 decennial census
evaluation programs and this evaluation continues that work. This decade, the Gallup Daily
Tracking Survey and Census Test Focus Groups provide background for this evaluation.
In a follow-up to the 1990 Census, the Census Bureau contracted with NORC to conduct a
nationwide in-person survey focusing on issues related to census participation. One of the
concerns that the survey was designed to address was privacy and confidentiality concerns with
the data. The questionnaire included items about general privacy concerns and items specific to
4

<>

the census. The sample was nationally representative and also included nonrespondents to the
census. Data from this survey was linked to actual census response by asking respondents to
provide an address for the purpose of matching back to census response records. This design
allowed Singer et al. to analyze the role of privacy and confidentiality in census participation.
The authors found that privacy and confidentiality concerns explained around 1.5 percent of the
variance in return rates after controlling for demographics (Singer et al. 1993).
Similarly, after the 2000 Census, the University of Michigan, under contract with the Census
Bureau, collected data with the Gallup organization to examine trends in beliefs about
confidentiality and privacy. This study also investigated trends in attitudes toward data sharing.
Again this evaluation matched back to actual census responses by asking respondents for their
address for matching purposes. As in the 1990 Census evaluation, this study also found that
privacy and confidentiality concerns explained about 1.5 percent of the variance in the mail
return rate after controlling for demographics (Singer et al. 2003). This study also found
increasing concerns about the sharing of confidential data among federal agencies.
In 2011, the Census Bureau’s Communications Directorate conducted the second iteration of the
Census Barriers, Attitudes, and Motivators Survey (CBAMS II) as a follow-up to the original
CBAMS conducted prior to the 2010 Census in 2008. CBAMS was conducted to gain an indepth understanding of the public’s opinions about the 2010 Census, with the specific intention
to understand those who have negative attitudes toward the census and the government more
generally or those who are unaware/lack extensive knowledge of the census. CBAMS II
provided a post-2010 Census measurement of the same issues as well as information on the use
of administrative records for the decennial census. In CBAMS II, respondents were
experimentally divided into three groups in order to test their views of administrative records use
as a means of (1) reducing census (government) costs, (2) reducing respondent burden or (3) as
simply an alternative option to a self-response (the control group). From this research, the study
found that both arguments of reducing cost (when citing a $10 billion census price tag) and of
alleviating respondent burden increased public support of administrative records usage, though
the cost reduction frame was more powerful (Wroblewski, Bates and Pascale, 2012; Conrey,
ZuWallack, and Locke, 2011).
Additionally, the CBAMS II found that some administrative records are less sensitive than
others. People were more comfortable with obtaining one’s name, date of birth, gender, and race
from tax returns (50 percent), or other government records such as unemployment or social
security (45 percent); whereas they were much less in favor of the census obtaining credit bureau
data (25 percent) or medical records (22 percent) for use in a decennial census. Further, in the
study, most people (65 percent) expressed unwillingness to allow the Census Bureau to use
social security numbers to obtain sex, age, date of birth, and race information from other
government agencies. Other research has suggested the importance of providing a context for
answering such questions, and CBAMS II, like many telephone surveys, afforded limited
opportunity to provide such context.
Beginning in February 2012, the Census Bureau has asked a random sample of approximately
200 respondents nightly questions on trust, confidentiality, credibility, transparency, and data use
5

<>

as part of the Gallup Daily Tracking Survey. The data from this survey provides us with a timeseries on trust in federal statistics that can be extended by this proposed research on privacy and
confidentiality.
Between 2014 and 2016, the Center for Behavioral Science Methods has also conducted focus
groups as follow-up research to annual census tests. The groups focused on privacy and
confidentiality concerns of different segments of the populations in scope for each test. Groups
included both respondents from different modes and nonrespondents and were separated
demographically (e.g., by age, race, and language) when feasible. Findings from these focus
groups suggest that demographics are indicative of important differences in terms of the types of
privacy and confidentiality concerns that people have (Morales et al. 2017; Fobia et al.
forthcoming). Related research also indicates that Spanish-language speakers have particular
concerns about privacy and confidentiality as well (CSM 2017; Sha et. al 2018).
The Gallup Daily Tracking Survey and focus groups inform this work by suggesting the types of
privacy and confidentiality concerns that might be prominent for census respondents. For
example, in terms of responding to the census online, respondents in the 2014 Census Test focus
groups were concerned about individuals posing as the Census Bureau via malicious links or
contact attempts and stealing information (Holzberg & Fobia, 2016). In the Gallup survey,
respondents who had concerns about answering the census online also report being concerned
about hacking and data security (Childs et al., 2017). In terms of administrative record use, some
respondents in 2014 and 2015 Census Test focus groups thought that government agencies
should not share information with each other (Holzberg & Fobia, 2016; Morales, Holzberg, &
Eggleston, 2017). Lack of trust in the government and concerns about hacking were reasons why
some Gallup respondents did not support administrative records use (Childs et al. 2015a).
Meanwhile, other research has shown that rates of both consent to link data and overall survey
participation have declined, raising concerns about the accuracy of results drawn from linked
data and survey responses (Fulton 2012; Sakshaug and Kreuter 2012; Curtain, Presser, & Singer
2005; National Research Council 2013). A study by Singer and Presser (1996) demonstrated that
individuals’ reactions to data-sharing arrangements (to facilitate mandatory census activities)
were influenced by demographics, especially gender and education. Research by Huang, Shih,
Chang, and Chou (2007) in Taiwan revealed that the elderly, lower income, less educated, and
minorities were less likely to consent to sharing and linking their information for research
purposes, but that gender was not a factor.
In addition to demographics, some of respondents’ opinions and knowledge are also related to
one’s openness to data linkage. For example, research by Singer and Presser (1996) established
that people’s propensity to share or link their data was swayed by their understanding of the
statistical agencies involved in those endeavors, their belief that the information is already being
shared, and the importance they attach to the use of shared information. Similarly, negative
attitudes toward the use of administrative records have also attributed to respondents’ lack of
understanding of what administrative records are, how statistical agencies make use of that
information, the authority of the statistical agencies, and their ability to protect confidentiality
(Bates and Pan, 2009; Gerber and Landreth, 2007, Holzberg & Fobia, 2016; Morales, Holzberg,
& Eggleston, 2017).
6

<>

Respondents’ reactions may also not always be in the direction we might expect, and therefore
they should continue to be studied. For example, previous research has shown respondents to be
less favorable to the use of administrative records to determine the occupancy of housing units
than they are for administrative record use to fill in basic census demographic information
(Childs et al. 2015a).
Privacy and confidentiality concerns have been cited as a potential reason for nonresponse in
web surveys in particular (Couper 2000; Cho and LaRose 1999). In the 2020 Census there will
be three modes of self-response available and an in-person Nonresponse Followup (NRFU). In
this evaluation, we will investigate how privacy and confidentiality concerns might affect selfresponse choices between online, mail and telephone response as well as a respondent being
enumerated in-person in the 2020 Census. Another behavior we will study is the relationship
between distinct privacy and confidentiality concerns and item nonresponse. Literature suggests
that item nonresponse is connected to respondent concerns with confidentiality of disclosure
(Booth-Kewley et al.,2007; Joinson et al., 2004). Concerns about privacy and confidentiality
might be related to specific items asked on the decennial census questionnaire that might be seen
as sensitive.
The relationship between respondents’ privacy and confidentiality concerns and the likelihood
that their self-response data does not match administrative records is a gap in the literature on
discrepancies between records and survey data that we plan to address. We plan to link responses
to our evaluation survey with administrative record data. If increased privacy and confidentiality
concerns are related to a higher likelihood of a mismatch between self-reported and
administrative records data it could indicate bias in either the records or self-reported data. This
evaluation could provide a starting point for further research into record mismatches as part of
the 2030 Census research program.
This evaluation, in line with evaluations in earlier decades, will connect privacy and
confidentiality concerns with respondent behavior in a decennial census environment. What our
research from this decade has shown is that respondents have particular concerns about
responding to surveys online. However, much of that data collection has been hypothetical and
qualitative. We will investigate the relationship of privacy and confidentiality concerns to
response mode, item nonresponse, and mismatches between administrative records and selfreported data.

III.

Assumptions

1. The project team will obtain adequate funding to implement the evaluation as it is
designed in this study plan.
2. The 2020 Census will have an online response option.
3. The 2020 Census will use administrative records for operations as planned.
4. The Census Data Lake will contain 2020 Census response and operational data required
for analysis.
7

<>

5. The project team assumes that the Census Bureau will be able to obtain the services of a
contractor to support the design and implementation of this evaluation.

IV.

Research Questions

Three research questions are central to this project:
1. Are privacy and confidentiality concerns related to response mode?
a. How do these concerns vary by demographic group?
In the 2020 Census there will be three modes of self-response available and an in-person followup. In this evaluation, we will investigate how privacy and confidentiality concerns might affect
self-response choices between online, mail, and telephone response. Privacy and confidentiality
concerns might also be related to a respondent being enumerated in-person rather than selfresponding.
2. Are privacy and confidentiality concerns related to partial responses?
a. How does this relationship vary by demographic group?
Based on data completeness measures from the 2010 Census, we expect 89 percent of selfresponse forms to include all five person-level variables while item nonresponse rates for
household-level items range from 1.8 percent (household count) to 7.8 percent (telephone
number) (Rothhaas et al. 2012). For this project, we define a partial response as missing one or
more items. We expect nonresponse rates to different items to be related to privacy and
confidentiality concerns.
3. Are privacy and confidentiality concerns related to mismatches between administrative
records and self-reported data?
a. How does this relationship vary by demographic group?
We plan to link survey responses from this evaluation with Internal Revenue Service (IRS)
and/or Social Security Administration (SSA) records. We are using decennial response data to
select our sample so decennial responses will be available. A mismatch occurs when the selfresponse data from the decennial responses or from the evaluation survey response do not match
data from IRS or SSA records. We expect that households with mismatches will have different or
increased privacy and confidentiality concerns when compared with households that do not have
mismatches.

V.

Methodology

In this section, we detail the methodology for a survey of decennial census respondents’ privacy
and confidentiality concerns. Data collection will begin shortly after April 1, 2020, and include
both telephone and in-person modes. The survey will be administered in both English and
Spanish. The sample will focus on detecting differences in demographic groups. The instrument
will take between 20 and 35 minutes to administer and include question items in four topic areas:
8

<>

1) privacy and confidentiality concerns, 2) opinions on administrative records, 3) concerns about
decennial census items, and 4) related constructs (see page 11 for details). Analysis plans include
logistic regression models, multinomial logistic regression, and t-tests for differences between
demographic groups. Our analysis plan also includes linking survey responses with
administrative records from the IRS and/or SSA.
This study plan also includes a qualitative component. Some respondents with privacy and
confidentiality concerns will likely not complete the census or allow entrance to enumerators and
observers. To capture these respondents we will conduct a qualitative study that leverages the
2020 Census Partnership Program in addition to the survey. The qualitative study will include
three components: 1) observations of events, 2) interviews with national partners, and 3) focus
groups with community partners.
A. Evaluation Design
Data Collection
Quantitative Component
Data will be collected in two modes: telephone and in-person. Self-respondents to the decennial
census will be interviewed by telephone and NRFU census respondents will be interviewed inperson and may be offered a small incentive. The survey will be conducted in both English and
Spanish.
Telephone Survey. Telephone data collection will use an instrument programmed by a
contractor. Telephone interviewers would make the contact attempts to ask respondents to
participate and administer the instrument. 2020 Census respondent-provided phone numbers
would be used to contact respondents. Contact frame telephone numbers may be used if no
respondent-provided phone numbers have been collected, or if the reuse of respondent-provided
phone numbers is prohibited for this purpose. Then, a sample for the telephone survey would be
drawn from census response data on a flow basis and sent to telephone interviewers for followup. We would like the follow-up survey to be conducted in close proximity to when a respondent
fills out their census form. Ideally, data collection would begin in April.
In-Person Survey. Census respondents who fill out their forms with enumerators during NRFU
will be sampled for an in-person follow-up survey. A sample of addresses that do not respond to
the telephone survey will also be selected for in-person follow-up. The interviewers that would
be used for this task will be employed by the contractor.
Sample

9

<>

The universe for this evaluation encompasses households that responded to the decennial census
(omitting households selected for other decennial census experiments and evaluations when
necessary), the ACS, and the census Post-Enumeration Survey. We will draw sample using 2020
Census response data for Person 1. Based on findings from previous research, we are primarily
interested in three race/ethnicity groups: Hispanic (any race), White (alone, non-Hispanic), and
Black (alone, non-Hispanic). 1 We plan to sample so that we will be able to cross these race
groups with age groups (18-24, 25-44, 45-64, 65+). We will include both self-respondents and
NRFU respondents in our sample.
Using an alpha of 0.10, beta of 0.20, and a detectable difference of 8 percentage points in privacy
concerns, the national sample size necessary for this evaluation is 103,340 housing units.
We will draw sample during the 2020 Census data collection using characteristics of the 2020
Census return (e.g., response mode, complete or partial response, race data, age data).
Self-responses to the 2020 Census with available phone numbers, either provided by the
respondent or from the Census Bureau’s contact frame, are first sorted by geography, partial
response, 2020 Census response mode, contact strategy, and language. Next, these are stratified
by the race and age stratums of interest: non-Hispanic White ages 18-29, non-Hispanic White
ages 30-44, non-Hispanic White ages 45-59, non-Hispanic White ages 60+, non-Hispanic Black
ages 18-29, non-Hispanic Black ages 30-44, non-Hispanic Black ages 45-59, non-Hispanic Black
ages 60+, Hispanic ages 18-29, Hispanic ages 30-44, and Hispanic ages 45-59. Finally, a
systematic random sample is taken to obtain a sample of 9,059 housing units from each of the
race and age stratums of interest resulting in a total of 99,653 housing units selected from selfresponses.
The NRFU responses are sorted by geography, contact strategy, by the race and age stratums of
interest: non-Hispanic White ages 18-29, non-Hispanic White ages 30-44, non-Hispanic White
ages 45-59, non-Hispanic White ages 60+, non-Hispanic Black ages 18-29, non-Hispanic Black
ages 30-44, non-Hispanic Black ages 45-59, non-Hispanic Black ages 60+, Hispanic ages 18-29,
Hispanic ages 30-44, and Hispanic ages 45-59, and language before taking a systematic random
sample to obtain the 3,687 NRFU housing units.
Instrument
The instrument assesses respondents’ privacy and confidentiality concerns. We will ask
questions that surround four themes: 1) privacy and confidentiality concerns, 2) opinions on the
use of administrative records, 3) concerns about particular census questions, and 4) other related
constructs.
1

Previous research has shown differences in privacy and confidentiality attitudes by race, Hispanic origin, sex, and
age. See Morales et al. 2017; Fobia et al. forthcoming; Sha et al. 2018; CSM 2017.
10

<>

The instrument will be written by Center for Behavioral Science Methods (CBSM) staff and
cognitively tested in English and Spanish. We anticipate that the instrument will take between 20
and 35 minutes to administer.
Privacy and Confidentiality Concerns. Respondents will be asked if they have any privacy or
confidentiality concerns about their census responses and data; interviewers will code the type of
concern and ask follow up questions about the level of respondent concern. Additionally,
respondents will be asked to choose a level of concern for different types of privacy and
confidentiality concerns that have been found in previous research or identified as emerging
issues. Some of these may include hacking, misuse of data, the government having too much
information, re-identification, data sharing, and computer scams. Respondents will be asked
about their perceptions of each of the four Census Bureau privacy principles: necessity,
openness, respect for respondents, and confidentiality (U.S. Census Bureau, 2006). The privacy
and confidentiality practices asked about will also include the differential privacy methods that
the Census Bureau plans to implement. This project is an opportunity to evaluate respondent
confidence and beliefs about our practices in this arena.
We will replicate questions from privacy and confidentiality studies in past decades. In
particular, we plan to replicate questions about privacy beliefs, confidentiality concerns, and
opinions about administrative records (Singer et al. 1993; Singer et al. 2003).
Opinions on Administrative Records. Results from attitude questions about administrative
records will inform the research and testing phase of a records-based 2030 Census. The items for
this topic will be replicated from the Gallup Nightly Survey as well as previous decennial
evaluations of privacy and confidentiality (See Singer et al. 1993; Singer et al. 2003; CBAMS II
Final Report). We plan to ask respondents for their income information. Past research has shown
that failure to report income on a public opinion survey is highly correlated with reported privacy
and confidentiality concerns.
Concerns about decennial census items. We will also ask respondents about their level of
privacy and confidentiality concern for each of the census items. This data will help us
understand whether certain items are more sensitive than others and whether the sensitivity of
particular items are associated with different demographic groups. This evaluation will provide
data about privacy and confidentiality concerns about the citizenship question that is planned for
the 2020 Census questionnaire. Data on individual census items can also help inform decision
making around privacy budgets for data releases that use differential privacy methods.
Related constructs. Other questions will replicate those asked in the Gallup Nightly Survey that
have been shown in previous studies to be related to privacy and confidentiality concerns. For

11

<>

example, respondents may be asked questions about their knowledge of federal statistics as well
as trust in the federal government, the Census Bureau, and other institutions. In previous work,
knowledge about the federal statistical system, data use, and belief in the relevance of statistics
are important correlates of trust in federal statistics and self-reported response (Childs et al.
2015; Childs et al. 2017; Conrey et. al. 2012). Trust in government has also been identified as an
important challenge for the 2030 Census (Mitre, 2016).
Analysis
We will begin by running correlations between our predictor and outcome variables (See
Appendix A, Table x1). We may create indices for privacy and confidentiality concerns
depending on the final items selected for the questionnaire. We will run descriptive statistics on
attitude items and outcome variables as well as other exploratory analyses in addition to what is
described below.
1. Are privacy and confidentiality concerns related to response mode?
a. How do these concerns vary by demographic group?
For the first research question, we will use a multinomial logistic regression model to test the
relationship between response mode and our predictor variables. For this model, the predictor
variables include items about privacy and confidentiality concerns, related constructs, concerns
about census items, demographics, and whether or not the respondent reported income (See
Appendix A, Table RQ1). Base models will not include demographic controls.
2. Are privacy and confidentiality concerns related to partial responses?
a. How does this relationship vary by demographic group?
For the second question, we will use logistic regression models to test the relationship between
partial response and our predictor variables. We will have a binary outcome variable for
complete versus partial response. We will also run models for binary outcome variables for each
missing data item (e.g. missing citizenship versus not missing citizenship, missing birthdate
versus not missing birthdate, etc.). The predictor variables include items about privacy and
confidentiality concerns, related constructs, concerns about census items, demographics and
whether or not the respondent reported income (See Appendix A, Table RQ2). Base models will
not include demographic controls.

3. Are privacy and confidentiality concerns related to mismatches between administrative
records and self-reported data?
a. How does this relationship vary by demographic group?
For the third research question, we will link responses from this evaluation survey to IRS and/or
SSA administrative records. We will use logistic regression models to test the relationship
between administrative records and self-reported data mismatches. Decennial response data will
provide the self-reported items to be compared to IRS and SSA data on the same items. We will
ask for respondent income in the evaluation survey. We will create binary outcome variables for
12

<>

discrepancies for each data item (e.g. number of people in household reported in the 2020
Census does not match most current administrative records versus data items match correctly).
Predictor variables include items about privacy and confidentiality concerns, opinions on
administrative records, related constructs, concerns about census items, and demographics (See
Appendix A, Table RQ3).
For all three subquestions, we will use a chi-square test to compare the distributions of responses
by demographic groups. If significant differences are found (p < 0.10), we will run t-tests
adjusted for multiple comparisons using a Bonferroni adjustment to further examine the pattern
of these differences.
Qualitative Component
Since it is likely that some respondents with privacy and confidentiality concerns will not
complete the census or allow entrance to enumerators and observers, we will conduct a
qualitative study that leverages the 2020 Census Partnership Program in addition to the survey.
The qualitative study will include four main components: 1) Qualitative interviews with cultural
experts recommended by the partnership program, 2) focus groups with trusted messengers, 3)
observations of partnership events, and 4) focus groups with respondents.
This component complements research on the effects of the citizenship question on respondent
participation as well as on the survey of privacy and confidentiality concerns. This qualitative
component is not representative research and the findings will be limited in their generalizability
to larger populations. However, since people who do not respond to the 2020 Census are not
likely to be captured using other methods, this aspect of the research will fill this gap.
Research Questions
1. What can community partners tell us about reasons people in their communities do not
complete census forms?
2. What effect did including the citizenship question on the 2020 Census questionnaire have
on participation? What are the reasons respondents and community partners give for this
effect?
3. What other privacy and confidentiality concerns are expressed, if any?
Methodology
The study will include four main components: 1) Qualitative interviews with cultural experts
recommended by the partnership program, 2) focus groups with trusted messengers, 3)
observations of partnership events, and 4) focus groups with respondents.
We will leverage the partnership program for this study since people who might have concerns
about the citizenship question will likely not speak directly with government employees or
contractors. The partnership program seeks to partner with people in hard-to-count (HTC)
communities who are already trusted in those communities.

13

<>

Qualitative interviews with cultural experts. Cultural experts are individuals who will be
recommended by the 2020 National Partnership Program (NPP). These individuals will come
from organizations that have experience working with HTC communities and knowledge of and
access to networks of trusted messengers. In depth, qualitative interviews with cultural experts
will allow us to draw on the experience of those that have been able to successfully reach HTC
groups and have chosen to be part of the Partnership Program. Partnership specialists will play a
key role by connecting researchers to partners. These interviews will identify concerns they have
encountered about respondent participation in the 2020 Census because of the citizenship
question and will collect information about strategies they used to increase response and their
effectiveness. Interviews will be both in-person and by telephone when necessary. We plan to
conduct 15 interviews in 2020.
Focus groups with trusted messengers. Trusted messengers are individuals who are influential
in their local communities and may affect others’ decisions about when, how, and whether to
respond to the census or other surveys. These are also individuals who are targeted through the
2020 Community Partnership and Engagement Program (CPEP). In the context of the decennial
census, CPEP aims to engage community partners to increase decennial participation of those
who are less likely to respond or are often missed. While the National Partnership Program
partners with larger organizations, the 2020 CPEP will engage at the grassroots level to reach out
to those who are less likely to respond to the national campaign (Hall 2017). CPEP plans to
leverage trusted messengers (also called “trusted voices”) to increase response rates in hard-tocount populations. Our research goal with this group is to learn more about concerns trusted
messengers may have about respondent participation in the 2020 Census, how they addressed
these concerns with respondents, and whether their strategies were effective. We plan to conduct
six focus groups in 2020.
Observations of partnership events. Researchers will observe 2020 Partnership Program
events at both the national level and community program level. One example of an event is the
Census Solutions Workshops hosted by members of the NPP. By observing these types of
events, we can see how communities are engaging with the 2020 Census, the questions and
concerns that respondents are mentioning, and how the partners respond to those questions and
concerns. From these preliminary interviews, we will ask respondents about planned events or
examples of grassroots efforts that we can observe in the next phase of the research. We plan to
observe 10 events: five at the NPP level and five at the CPEP level.
Focus groups with respondents. Researchers will conduct focus groups with members of hardto-count populations who might have been impacted by the inclusion of the citizenship
questions. Focus groups will be conducted in English and other languages. This set of focus
groups will allow us to assess the impact of privacy and confidentiality concerns for respondents
who might not be captured in the larger quantitative survey because of the sampling strategy and
sample size constraints. We plan to conduct about 12 focus groups with respondents across the
United States.
Researchers could be Census Bureau staff or contractors. Some researchers for certain
populations will need to be bilingual English-Spanish speakers. Other languages might also be
necessary.
14

<>

Analysis
Researchers will complete individual summaries for each qualitative interview and observation
event. Focus groups will be transcribed and translated into English when necessary. Summaries
and transcripts will be reviewed for evidence of recurring themes and patterns. Each analyst will
code the interview summaries, observation summaries, and focus group transcripts separately
before meeting as a group. The research team will then discuss the recurring themes and patterns
in the data and work to reach a consensus on the themes and codes for reporting and conduct
further analysis if needed.
B. Interventions with the 2020 Census
This proposal does not require direct interventions with the 2020 Census systems or processes.
However, it does propose to use 2020 systems independently from production. We will need
response data to draw sample and to make sure that all cases are unduplicated with other
decennial census experiments and evaluations cases when necessary and ACS sampled
households.
The response data to the follow-up should probably ultimately reside in the Census Data Lake
(CDL). That means there needs to be a connection somewhere into the CDL, and CDL has to
expect the file.

C. Implications for 2030 Census Design Decisions and Future Research and Testing
As we begin the research and testing phase for the 2030 Census, this evaluation will provide a
starting point for research on respondents’ perceptions and understanding of how the census uses
administrative records. Future research into how to communicate around a primarily recordsbased census could build on the results of this study about the relationship of privacy and
confidentiality concerns and mismatches between administrative records and self-reported data.
This evaluation is also designed to detect differences in privacy and confidentiality concerns for
different demographic groups. This data can inform the design of qualitative studies to take place
throughout the next decade to analyze the meanings that respondents attach to the census and
other surveys and how that will change with a records-based census. It will also provide a
baseline for other quantitative studies to assess changes in privacy and confidentiality concerns
as the coming decade progresses. The data from this evaluation will also inform the way that the
Census Bureau communicates privacy protections with its respondents in the coming decade. We
will gather nationally representative data on the relationship between item nonresponse and
concerns about privacy and confidentiality. This can inform messaging as well as decisions
around privacy budgets for specific data points for this and other surveys.

The results will potentially lead to cost savings by informing our messaging on privacy and
confidentiality, which can possibly lead to increased unit response in the 2030 Census. The
15

<>

proposed research will also provide some insight into how to reduce item nonresponse, as some
respondents may choose to skip questions because of privacy and confidentiality concerns. This
evaluation will collect information that can help inform messaging for the Census Bureau to use
to reassure members of the public and stakeholders who are concerned about privacy and
confidentiality. Privacy and confidentiality concerns have a longstanding history in the census
and Census Bureau surveys. For example, the contact centers and respondent advocate field
phone calls and emails from respondents who find the ACS to be “too intrusive” or who are
unsure whether it is a legitimate request or a scam. These concerns will likely increase as
technology evolves throughout the next decade. Research in this area and related areas such as
differential privacy and data disclosure avoidance is of critical importance.

VI.

Data Requirements

Data File/Report

Source

Purpose

Expected
Delivery Date

Census response file
Decennial response event file
Contact frame phone numbers

Census Data Lake
CDL
CARRA/PEARSIS

Universe for sample
Universe for sample
Append phone numbers to
sample cases

mm/dd/yyyy
mm/dd/yyyy
mm/dd/yyyy

VII. Risks
There are no risks that impact the completion of this evaluation.

VIII. Limitations
1. We do not currently plan to survey people who did not respond to the decennial census.
This would significantly increase the costs of this evaluation, but would also increase the
value of the results.

IX.

Issues That Need to be Resolved

1. It is possible that there may be some respondents who find issues of privacy and
confidentiality sensitive in this evaluation. We will carefully script a survey introduction
to reassure these respondents to the maximum extent possible.
2. It is an open question whether we can link the data from this survey with administrative
records from IRS and/or SSA.

16

<>

X.

Division Responsibilities

Division or Office
CBSM

Responsibilities
• Project design and coordination
• Drafting and pretesting of instrument
• Analysis
• Design of qualitative component
• Qualitative data collection and analysis

DSSD

•
•

Project design
Sample design and specifications

Contractor

•

Telephone and in-person instrument and case
management
Telephone and in-person data collection
In-person incentive management
Qualitative data collection and analysis

•
•
•

XI.

Milestone Schedule

Privacy and Confidentiality Evaluation Milestone

Date

Begin qualitative data collection

03/2020

Draw universe from Census Data Lake

04/2020-09/2020

Select sample for follow ups

04/2020-09/2020

Begin telephone data collection

04/2020

Begin in-person data collection

06/2020

Wrap telephone data collection; send sample to in-person follow-up

08/2020

Wrap in-person data collection

09/2020

Wrap qualitative data collection

09/2020

Receive, Verify, and Validate Data For Privacy and Confidentiality Evaluation

mm/dd/yyyy

Distribute Initial Draft Privacy and Confidentiality Evaluation Report to the Decennial
Research Objectives and Methods (DROM) Working Group for Pre-Briefing Review

mm/dd/yyyy

Decennial Census Communications Office (DCCO) Staff Formally Release the
FINAL Privacy and Confidentiality Evaluation Report in the 2020 Memorandum
Series

mm/dd/yyyy

17

<>

XII. Review/Approval Table
Role

Approval Date

Primary Author’s Division Chief (or designee)

08/21/2018

Decennial Census Management Division (DCMD) ADC for Nonresponse,
Evaluations, and Experiments

02/19/2019

Decennial Research Objectives and Methods (DROM) Working Group

02/19/2019

Decennial Census Communications Office (DCCO)

mm/dd/yyyy

XIII. Document Revision and Version Control History
Version/Editor
1.0/Fobia
2.0/Fobia
3.0/Fobia

Date
8/10/2018
2/07/2019
3/15/2019

Revision Description
Original
Revised after DROM 10/2/2018
Revised after DROM review 2/19/2019

XIV. Glossary of Acronyms
Acronym
ACS
ADC
CBAMS
CBSM
CDL
CPEP
CSM
DCCO
DCMD
DROM
DSSD
EXC
HTC
IPT
IRS
NPP
NRFU
R&M
SSA

Definition
American Community Survey
Assistant Division Chief
Census Barriers, Attitudes, and Motivators Survey
Center for Behavioral Science Methods
Census Data Lake
Community Partnership and Engagement Program
Center for Survey Measurement
Decennial Census Communications Office
Decennial Content Management Division
Decennial Research Objectives and Methods
Working Group
Decennial Statistical Studies Division
Evaluations & Experiments Coordination Branch
Hard-to-Count
Integrated Project Team
Internal Revenue Service
National Partnership Program
Nonresponse Followup
Research & Methodology Directorate
Social Security Administration

18

<>

XV. References
Bates, N.A. and Yuling, P. (2009). "Motivating Non-English Speaking Populations for Census
and Survey Participation" Paper presented at the Federal Committee on Statistical Methodology
2009 Research Conference, Washington DC, November 3, 2009.
Bates, N., Wroblewski, M., & Pascale, J. (2012). Public attitudes toward the use of
administrative records in the U.S. Census: Does question frame matter? Technical Report,
Survey Methodology Series. #2012-04, United States Bureau. Available at:
https://www.census.gov/srd/CBSMreports/byrsm.html#y2012.
Blumerman, L., & Fontenot, A. (2017). 2020 Census Program Management Review: Welcome
and High-Level Program Updates. Washington, DC: U.S. Census Bureau. Retrieved from:
https://www2.census.gov/programs-surveys/decennial/2020/program-management/pmrmaterials/07-11-2017/pmr-welcome-high-level-updates-07-11-2017.pdf
Booth-Kewley, S. Larson, G. & Miyoshi, D. ( 2007). Social desirability effects on computerized
and paper-and-pencil questionnaires. Computers in Human Behavior. Vol 23:1.
https://doi.org/10.1016/j.chb.2004.10.020
Center for Behavioral Science Methods (CBSM). (2017). Respondent Confidentiality Concerns,
Memorandum for ADRM.
Childs, J.H., Eggleston, C., Morales, G., & Fobia, A.C. (2017). Attitudes Towards Relevance of
Statistics in Public Policy-Making. Paper presented at the meeting of the World Association for
Public Opinion Research, Lisbon, Portugal.
Childs, J.H., King, R. & Fobia A.C. (2015). Confidence in US Federal Statistical Agencies.
Survey Practice. Volume 8. Issue 5.
Childs, J.H., King, R., Eggleston, C., & Holzberg, J. (2015a). “Public Attitudes towards Use of
Administrative Records to Supplement the U.S. Census.” Paper presented at the 68th Annual
Conference of the World Association for Public Opinion Research, Buenos Aires, Argentina.
Cho, H., & Larose, R. (1999). “Privacy Issues in Internet Surveys” Social Science Computer
Review. Vol. 17: 4. https://doi.org/10.1177/089443939901700402
Conrey, F. R., ZuWallack R. & Locke, R. (2012). Census Barriers, Attitudes, and Motivators
Survey II Final Report. Washington, DC: U.S. Census Bureau. Retrieved from:
https://www.census.gov/2010census/pdf/2010_Census_CBAMS_II.pdf
Couper, M. (2000). Review: Web Surveys: A Review of Issues and Approaches. The Public
Opinion Quarterly, 64(4), 464-494. Retrieved from http://www.jstor.org/stable/3078739

19

<>

Curtin, Richard, Stanley Presser, and Eleanor Singer. 2005. Changes in Telephone Survey
Nonresponse over the Past Quarter Century, Public Opinion Quarterly, Volume 69, Issue 1, 1
January 2005, Pages 87–98, https://doi.org/10.1093/poq/nfi002.
Das, Marcel, and Mick P. Couper. 2014. “Optimizing Opt-Out Consent for Record Linkage.”
Journal of Official Statistics 30(3):479-497.
Gerber, E. and Landreth, A. (2007). “Respondents’ Understandings of Confidentiality in a
Changing Privacy Environment,” U.S. Census Bureau, Statistical Research Division.
Huang, N., Shih, S., Chang, H., and Chou, Y. (2007). “Record Linkage Research and Informed
Consent: Who Consents?” BMC Health Services Research, 7:18 (February 2007).
Fobia, A. C., Morales, G. and Childs, J.C. (forthcoming). 2020 Census Research and Testing:
Final Report on the 2016 Census Test Focus Groups. Washington, DC: U.S. Census Bureau.
Fulton, Jenna. 2012. “Respondent Consent to Use Administrative Data.” Doctoral dissertation,
University of Maryland.
Holzberg, J.L., & Fobia, A.C. (2016). 2020 Census Research and Testing: Final Report on the
2014 Census Test Focus Groups (2020 Census Program Internal Memorandum Series:
2016.43.i). Washington, DC: U.S. Census Bureau.
JASON Program Office, Mitre Corporation (2016). Alternative Futures for the Conduct of the
2030 Census.
Joinson, A., Woodley, A., & Reips, U. (2004). Personalization, authentication, and
self-disclosure in self-administered Internet surveys. Computers in Human
Behavior, 23, 275–285. doi:10.1016/j.chb.2004.10.012.
Mitre Corporation. (2017). US Census and American Enterprise Institute: Census 2030 Blue Sky
Event.
Mitre Corporation. (2016). 2030 and the Census Environment. A Discussion Sponsored by the
Brookings Institution Hutchins Center on Fiscal and Monetary Policy.
Morales, G., Holzberg, J., & Eggleston, C. (2017). 2015 Census Test Focus Groups: Final Report
(2020 Census Program Internal Memorandum Series: 2017.4.i). Washington, DC: U.S. Census
Bureau.
National Research Council. 2013. Nonresponse in Social Science Surveys: A Research Agenda.
R. Tourangeau and T.J. Plewes (Eds.) Panel on a Research Agenda for the future of Social
Science Data Collection, Committee on National Statistics, Division of Behavioral and Social
Sciences and Education. Washington, DC: The National Academies Press.
Pascale, Joanne. 2011. “Requesting Consent to Link Survey Data to Administrative Records:
Results from a Split-Ballot Experiment in the Survey of Health Insurance and Program Benefit
20

<>

Wording and Data-Linkage Consent.” Technical Report, Survey Methodology Series #2011-03,
United States Census Bureau.
Rothaas, C., F. Lestina and J. Hill. (2012). 2010 Decennial Census: Item Nonresponse and
Imputation Assessment Report. Decennial Statistical Studies Division. Washington, DC: U.S.
Census Bureau.
Sakshaug, Joseph W., and Frauke Kreuter. 2014. “The Effect of Benefit Wording on Consent to
Link Survey and Administrative Records in a Web Survey.” Public Opinion Quarterly
78(1):166-176.
Sakshaug, Joseph W., Mick P. Couper, Mary Beth Ofstedal, and David. R. Weir. 2012. “Linking
Survey and Administrative Records: Mechanisms of Consent.” Sociological Methods &
Research 41:535–69.
Sakshaug, Joseph, Valerie Tutz, and Frauke Kreuter. 2013. “Placement, Wording, and
Interviewers: Identifying Correlates of Consent to Link Survey and Administrative Data.” Survey
Research Methods 7:133–144.
Sha, M., Son, J., Pan, Y., Park, H., Schoua-Glusberg, A., Tasfaye, C., Sandoval Giron, A.,
García Trejo, A., Terry, R., Goerman, P., Meyers, M., and L. Lykke. 2018. “Multilingual
Research for Interviewer Doorstep Messages Final Report.” Research Report Series. 2018.
Singer, Eleanor, Nancy Bates, and John Van Hoewyk. 2011. “Concerns about Privacy, Trust in
Government, and Willingness to Use Administrative Records to Improve the Decennial Census.”
Paper presented at the 66th Annual Conference of the American Association for Public Opinion
Research, Phoenix, AZ, USA. Available at:
http://ww2.amstat.org/sections/SRMS/Proceedings/y2011/Files/400168.pdf.
Singer, E. and Presser, S. (1996). “Public Attitudes toward Data Sharing by Federal Agencies.”
Paper presented at the Census Bureau Annual Research Conference, Washington DC.
Singer, E. and Van Hoewyk, J. (2001). “Trends in Attitudes Toward Privacy, Confidentiality,
Data Sharing, 1995-2000.” Paper presented at the Proceedings of the Annual Meeting of the
American Statistical Association, August 5-9, 2001.
Singer, E., Van Hoewyk, J., and Neugebauer, R.J. (2003). “Attitudes and Behavior: The Impact
of Privacy and Confidentiality Concerns on Participation in the 2000 Census.” Public Opinion
Quarterly, 67, 368-384.
Singer, E, Van Thurn, D.R., and Miller, E.R. (1995) “Confidentiality Assurances and Response:
A Quantitative Review of the Experimental Literature.” The Public Opinion Quarterly, 59:1,
66.77.

21

<>

Singer, E., Van Hoewyk, J., and Presser, S. (1997). “Public Attitudes Toward Data Sharing by
Federal Agencies.” In Record Linkage Techniques-1997: Proceedings of an International
Workshop and Exposition (pp. 237-247). Washington, DC: National Academy Press.
Smirnova, M., & Scanlon, P. (2017). Experience and Cultural–Repertoire Based Avenues of
Trust: An Analysis of Public Trust in Statistical Agencies and their Data. Social Policy and
Society 16(2), pp. 219-236. doi: 10.1017/S147474641500069X.
U.S. Census Bureau. (2017). 2020 Census Operational Plan: A New Design for the 21st Century.
U.S. Census Bureau, (2016). American Community Survey, July 2016 Panel.
U.S. Census Bureau. (2006). U.S. Census Bureau Privacy Principles. Retrieved from:
https://intranet.ecm.census.gov/apps/policyportal/Pages/PolicyViewer.aspx?docID=375.
Wroblewski, M., Bates, N., and Pascale, J. (2012). Public Attitudes toward the Use of
Administrative Records in the U.S. Census: Does Question Frame Matter? Federal Committee on
Statistical Methodology Research Conference. Washington, DC.

22

<>
Appendix A: Planned Tables for 2020 Privacy and Confidentiality Evaluation

Table x1. Bivariate Correlations of Predictors and Outcome Variables
Mode of
response
Concerns about privacy
Attitudes towards administrative records
Concerns about census items
Demographics
Race
Age
Race*Age
Education
Region
Sex
Marital status
Employment status
Income not reported

23

Partial
Response

Mismatch
with
admin recs

<>

Table RQ1. Demographic and attitudinal predictors of web response mode v. all other modes
Regression
Odds Ratios
coefficients
Privacy and confidentiality concerns
Related constructs
Concerns about census items
Demographics
Race
Age
Race*Age
Education
Region
Sex
Marital status
Employment status
Income not reported

Table RQ2. Demographic and attitudinal predictors of complete vs. partial response
Missing
Any missing
citizenship
data
status
Privacy and confidentiality concerns
Concerns about census items
Related constructs
Demographics
Race
Age
Race*Age
Education
Region
Sex
Marital status
Employment status
Income not reported

24

Missing
birthdate

<>

Table RQ3. Demographic and attitudinal predictors of mismatches between administrative
records and self-reported data items
Household
Citizenship
Any
count
Income
mismatch
mismatch
mismatch mismatch
Concerns about privacy
Opinions on administrative records
Related constructs
Concerns about census items
Demographics
Race
Age
Race*Age
Education
Region
Sex
Marital status
Employment status
Income not reported

25

<>
Table x2. Privacy and confidentiality attitudes by demographic group
Concerns
about
privacy
Demographics
Age
18-29
30-44
45-59
60+
Race

White (alone), non-Hispanic
Black (alone), non-Hispanic
Hispanic
More than one race

Race*Age

White (alone), nonHispanic

18-29
30-44

Black (alone), nonHispanic

45-59
60+
18-29
30-44
45-59
60+

Hispanic

More than one race

Sex
Education

18-29
30-44
45-59
60+
18-29
30-44
45-59
60+

Male
Female
Less than HS and
HS/GED
Some
college/associates
Bachelors
Post-Bachelor’s

Region

Marital
Status
Employment Status
Income not reported

1

Attitudes
towards
admin
records

Concerns
about census
items

Related
constructs

Mismatch
admin
records


File Typeapplication/pdf
Authordouglass Abramson
File Modified2019-05-31
File Created2019-05-31

© 2024 OMB.report | Privacy Policy