INTENDED ADMINISTRATIVE DATA USE IN THE 2020 CENSUS
I.

INTRODUCTION

Based on the results of the 2018 End-to-End Census Test, prior census tests, and other research, the
Decennial Census Programs Directorate has determined its approach to using administrative data in
many operations of the 2020 Census. In this context, we are using “administrative data” broadly to
include:



Microdata records contained in files collected and maintained by federal, state, and local
government agencies (traditionally referred to as administrative records).
Microdata records contained in files collected and maintained by commercial entities (often
referred to as third-party data).

As well as



Macro- and microdata from Census Bureau data collections for statistical purposes and address
enhancement operations (internal data).
Macrodata from publically available sources (public data).

The intended uses of administrative data and the expected sources of data have evolved throughout the
research and planning for the 2020 Census. This document reflects information as of May 1, 2020. It
does not attempt to fully reconcile information provided in previous documents or presentations, which
was current at their release. Given our changing census operations due to the corona virus pandemic,
we are exploring whether additional uses are warranted. Should the uses described in this document
change, we will be transparent about those changes. Additionally, we will provide final documentation
with complete details of the 2020 Census administrative records effort.
II.

BACKGROUND

To meet the strategic goals and objectives of the 2020 Census, the Census Bureau has made
fundamental changes to the design, implementation, and management of the decennial census. These
changes build upon the successes and address the challenges of previous censuses, while also balancing
objectives of cost containment, quality, flexibility, innovation, and disciplined and transparent
acquisition decisions and processes. The 2020 Census Operational Plan includes using administrative
records, third-party data, internal, and public sources (collectively referred to in this document as
“administrative data”) to avoid cost, maintain quality, and improve efficiency of operations.
Several core sources will be used to support this effort, including data collected from Census Bureau
operations, which is protected under 13 U.S.C. § 9, and data from the Internal Revenue Service (IRS),
which is referred to as Federal Tax Information (FTI), and is protected by 26 U.S.C. § 6103(b)(8). In
accordance with Title 26 and a 2013 Memorandum of Agreement between the Department of Treasury,
Internal Revenue Service, and the Department of Commerce, U.S. Census Bureau, the scope of work for
which FTI may be used includes frame building, enumeration, imputation, and evaluation. Other
agreements that the Census Bureau maintains with other administrative data providers similarly
delineate allowable uses of their data.

1 May 2020

1

 Below are the instances for which the Census Bureau intends to use administrative data in the 2020
Census, excluding evaluations and experiments in support of early planning and research to inform the
transition and design of the 2030 Census. Several instances will use extracts from either a “Title 26
composite” or a “Title 13 composite” rather than directly accessing original administrative datasets. The
process to create these composites is described in the next section. More detailed information on each
instance is provided in sections IV- XI. Appendix A summarizes the expected sources for each instance.


Frame Development: Develop and update the address frame, including group quarters, and
spatial data that serve as the universe for 2020 Census enumeration activities.
 Respondent Motivation:
o Support the initial contact strategy (appropriate delivery of census invitations and
questionnaires).
o Develop and execute the microtargeted advertising campaign.
 Self-Response:
o Augment respondent-provided address data to enhance matching of non-ID responses to
the updated enumeration address list: the addresses for responses that are returned
without the preassigned census identification number and do not easily match to the
Master Address File (MAF) will be compared with administrative record information in order
to obtain missing address information, or correct errors, such as misspelling. By improving
the address, another attempt can be made to associate the response data to a household in
the enumeration universe, so that no further effort is required to obtain a response, such as
follow-up mailings or fieldwork.
o Quality assurance in the Paper Data Capture operation: Optical Character Recognition (OCR)
capture of written-in names is compared to existing administrative data to ensure quality of
the OCR process.
o Enumerate or supplement field enumeration of nontraditional or unique living
arrangements, such as group quarters, military installations, and federally affiliated people
overseas.
 Nonresponse Followup (NRFU):
o Reduce contacts for cases in the NRFU workload through identification of vacant housing
units and deletes (units that do not meet the Census Bureau’s definition of a housing unit).
o Enumerate nonresponding, occupied housing units with quality, reliable information.
o Model the “best time to contact” occupied, nonresponding housing units.
 Response Data Verification:
o Self-Response Quality Assurance: Corroborate respondent-provided information to
detect potentially suspicious responses.
o Enumerator Quality Control: Validate enumerator-provided information in order to
ensure reporting accuracy and more efficiently target reinterview efforts.
 Post-Response Processing:
o Count Imputation: Determine final occupied/vacant/nonexistent status for unresolved cases
and impute the household count for those addresses determined to be occupied, but
without a specified number of occupants.
o Characteristic Imputation: Impute household and person characteristics where they are
missing from the response.
1 May 2020

2

 



III.

Publishing Data:
o Create the Citizen Voting Age Population special tabulation.
o Resolve Count Question Resolution challenges.
Coverage Evaluation:
o Improve matching and characteristic imputation in the Post-Enumeration Survey (PES).
o Create independent estimates through Demographic Analysis.
DATA SOURCES
A.
TITLE 26 COMPOSITE

Initially, administrative data will be used to create a robust repository of data and information that can
be used for activities that have been approved by data providers (e.g., the Internal Revenue Service and
the Social Security Administration). Data from several federal, state, internal, and commercial primary
sources are combined, standardized, and corroborated to create the Title 26 composite, which will
contain variables such as address, householder name, household roster and relationships, and
demographic data about the household inhabitants. FTI such as taxpayer ID and return type will be
included in this repository. These data are combined with extracts from the Master Address File
(MAF)/Topologically Integrated Geographic Encoding and Referencing (TIGER) database (MTdb). The
MTdb reflects the latest geographic information for each address. Extracts from the Title 26 composite
will be provided for additional processing as described in subsequent sections as appropriate.
The following sources are used to create the Title 26 composite and include historic and current vintages
as available to the Census Bureau:


Administrative Records
o Centers for Medicare and Medicaid Services (CMS) Medicare Enrollment Database
(MEDB)
o Housing and Urban Development (HUD) Public and Indian Housing Information Center
(PIC) and Tenant Rental Assistance Certification System (TRACS), now known as the
combined “Longitudinal” File
o HUD Federal Housing Administration (FHA) Integrated Database (IDB), which includes
data from Computerized Homes Underwriting Management System (CHUMS)
o Indian Health Service (IHS) Patient Registration
o Internal Revenue Service (IRS)
 1040 Individual Tax Returns
 IRS 1099 Information
o Selective Service System (SSS) Registration
o Social Security Administration
 Census Numident (a processed version of the SSA Numeric Identification File,
Numident, which does not contain a Social Security number) as well as the
Census Numident Alternate Name File
o State or Local Program Datasets, examples include:
 Homeless Management Information System (HMIS)
 Alaska Permanent Fund Dividend (PFD)
 Supplemental Nutrition Assistance Program (SNAP)
 Temporary Assistance for Needy Families (TANF)
 Women, Infants, and Children (WIC)

1 May 2020

3

 



o U.S. Postal Service (USPS) National Change of Address (NCOA) File
Third-Party Datasets1
o Black Knight
o DAR Partners (Data Advisory Research)
o Targus (Wireless and Federal Consumer)
o Veteran Service Group of Illinois (VSGI)
Census Bureau Data
o American Community Survey (ACS) data
o Census 2000 and 2010 Census Edited and Unedited datasets
o Census Household Composition Key File (produced using Social Security Administration
data and previous census information linking children 18 years and younger with their
parents)
o Contact Frame file (list of phone numbers) compiled from some sources listed here
(SNAP, WIC, Alaska PFD, ACS Data, DAR Partners, Targus Wireless and Federal
Consumer, and VSGI) as well as the ones below before it is delivered to 2020 Census
systems:
 Experian InSource and Experian End-Dated Records
 InfoUSA
 Melissa Data
 National Sample Survey of Registered Nurses
o “Best Race and Ethnicity” information compiled from some sources listed here (MEDB,
HUD Longitudinal, IHS, Census Numident, ACS data, prior census data, and Targus
Federal Consumer) as well as:
 CMS Medicaid and Statistical Information System (MSIS)
 National-Level Adult and Child TANF Recipient Files
 Experian In-Source and Experian End-Dated Records
 InfoUSA
B.

TITLE 13 COMPOSITE

Similarly, a robust repository of Title 13 data and information will be created from the Title 26
composite and extracts of this composite will be provided to various 2020 Census operations that do not
need to use FTI. While FTI is used in the creation of the Title 26 composite, in the creation of the Title 13
composite, the FTI is overwritten such that no FTI appears in the final Title 13 composite and FTI is not
the sole source of any information. This composite is therefore considered to be Title 13, not Title 26. An
example of how the overwriting works is in Appendix B.

1

Data from other commercial sources (e.g. CoreLogic) were used in census tests but are not considered part of the
production datasets.

1 May 2020

4

 C.

SOURCES NOT IN A COMPOSITE

Some processes use sources, either solely or in part, that are not part of the composites discussed
above. Those sources are:












ACS Contact History Information
Department of Defense
o List of deployed personnel
o Count of federally affiliated people stationed/assigned overseas, including dependents
Communications-Related
o Modeled self-response predictions (Predicted Self-Response Score and Internet
Proportion of Self-Response)
o Audience segmentation information
o Federal Communications Commission (FCC Residential Fixed Internet Access Service
Connections per 1,000 Households by Census Tract [publicly available])
Frame-Related
o USPS Delivery Sequence File (DSF)
o Additional Geographic Support Program and 2020 Census program data – address and
spatial data provided by tribal, federal, state and local governments
o Lists of group quarters and transitory locations from federal, state, local, and third-party
sources as well as ongoing operations
Planning Database (PDB) (publically available datasets at the tract and block group levels that
assemble a range of housing, demographic, socioeconomic, and census operational data;
variables have been extracted from the 2010 Census and ACS databases)
Group quarters administrator lists
USPS Undeliverable As Addressed (UAA) Information
D.

ADDITIONAL SOURCES ONLY USED FOR CITIZENSHIP

In addition to some sources mentioned in the sections above, there are several sources that are
expected to be used only to research how to and subsequently produce citizenship information in
conjunction with the census. These sources include, but are not limited to:




Census Bureau Data
o 2010 Census Coverage Measurement Files
o 2010 Census Person Identification Key (PIK) Crosswalk
o 2019 Census Test
o American Housing Survey data
o Current Population Survey data
o Survey of Income and Program Participation data
o PIK ITIN information (a list of Census Protected Identification Keys and source IDs for
people identified as having an Individual Tax Identification Number [ITIN])
Internal Revenue Service
o 1099-R information
o W-2 information

1 May 2020

5

 

















IV.

Social Security Administration
o Supplemental Security Records (SSR)
o Master Beneficiary Record (MBR)
Health and Human Services (HHS)
o CMS Transformational Medicaid and Children’s Health Insurance Program Information
System (MSIS and T-MSIS)
Veteran’s Administration Records (VA)
Department of Defense (DOD) Records
o DOD Army Post Service Panel
o DOD Army Veteran Cross Section
o DOD Army Service Member Panel with Spouse
Department of Homeland Security (DHS)
o U.S. Citizenship and Immigration Services data
o Immigration and Customs Enforcement visa information
o Customs and Border Arrival/Departure information
Department of State
o Data from Worldwide Refugee and Asylum Processing System
o Passport application data
Department of Interior
o Incident Management Analysis and Reporting System (IMARS) data
o Law Enforcement Management Information System (LEMIs)
Department of Justice U.S. Marshals data
o Facility Data
o Prisoners in Custody
o Prisoners Received
Bureau of Prisons data
Bureau of Justice Statistics National Corrections Reporting Program (NCRP)
State driver’s license data
FRAME DEVELOPMENT

Throughout the decade, the Census Bureau maintains address and spatial (e.g., roads, boundaries, and
geographic areas) data in the Master Address File (MAF) / Topologically Integrated Geographic Encoding
and Referencing (TIGER) System. The MAF/TIGER System is regularly updated with data from the United
States Postal Service; ongoing geographic partnership efforts with tribal, state, and local governments;
and fieldwork. These efforts are used to update the address frame and reflect changes to the housing
stock that occur over time.
For the 2020 Census, additional efforts were employed to finalize the frame. Focused geographic
partnership efforts helped improve the address list. State and local governments provided address
updates during the Local Update of Census Addresses (LUCA) program and New Construction (NC)
efforts. Additional data for group quarters and transitory locations was obtained from multiple sources,
including federal partners, state partners, local governments, and a third-party provider.
Frame development does not use either the T26 composite or the T13 composite. Data within the
MAF/TIGER System – extracts from the MAF/TIGER database (MTdb) – are used as inputs to several
enumeration operations.
1 May 2020

6

 V.

RESPONDENT MOTIVATION

Respondent motivation activities do not use T26 or T13 composite data. Most activities use aggregated
data at higher geographic or demographic levels; one activity uses person-level data. The activities are
outlined below.
A. INITIAL CONTACT STRATEGY
Administrative data were used to determine how the Census Bureau would send the initial invitation to
respondents. Respondents in areas more likely to respond online will receive the “Internet First” mailing
strategy, where they will receive invitations to respond online. Those who do not respond online will
receive reminders to respond and a paper questionnaire before NRFU begins. Respondents in areas least
likely to respond online (as determined by using ACS data and Federal Communications Commission
internet connectivity data), will receive the “Internet Choice” mailing strategy. The Choice strategy
consists of receiving an invitation to respond online, but with a paper questionnaire in the first mailing.
Respondents will then receive reminders to respond either online or via the questionnaire they received
earlier. Those who do not respond will receive another paper questionnaire before NRFU begins. The
“Internet First” or “Internet Choice” delineation is determined by geographic area (not by individual
household/address) and uses area-based response likelihoods, not person-level data, to identify which
contact strategy the area receives.
B. ADVERTISING CAMPAIGN
To increase the effectiveness of advertising and contact strategies, the Census Bureau, through the
communications contract, will use demographic and geographic information from various sources to
help target the advertising to specific populations.






The Census Bureau developed predictive models and used these models to estimate tract-level
self-response propensity for various mode(s). These predictions were then used to develop the
media plan and to aid in campaign optimization. Tract-level self-response rate predictions are
aggregated to larger geographic areas and help determine the media and messaging strategies
for the campaign.
Predictions were combined with geographic, demographic, housing, and sentiment information
to create audience segments. The segmentation information was used to design and execute
targeted advertising and communication strategies for geographic segments and audience
groups.
As households complete the census questionnaire, we will use campaign analytics to identify the
best messages and modes to reach various segments. We can coordinate with field teams and
partnership specialists to prioritize audiences and align messaging. We can also use internet
address and browsing history to customize user experience with our web platform.

1 May 2020

7

 In addition to the above general use, two targeted mailings also make use of administrative data.
 Native Hawaiian and Pacific Islanders (NHPI) in the continental United States will receive a direct
mailing through a partner organization. The partner organization will be using an address list
supplied to them by NHPI-related organizations. This is the only individual-level data used in the
respondent motivation activities.
 Residential addresses in targeted USPS carrier routes determined to be at risk for undercounting
young children will also receive a direct mailer. Tract-level response and demographic data from
the most recent ACS, the 2010 Census, and the planning database, along with address
information from the MAF and route information from the USPS were used to determine the
targeted routes.
VI.

SELF-RESPONSE PHASE
A.
NON-ID ADDRESS ENHANCED MATCHING

A goal in the 2020 Census is to make it easy for people to respond anytime and anywhere to increase
self-response rates. We do this by providing response options that do not require a unique Census
identifier (ID).
A “non-ID response” or a “non-ID case” refers to address information and associated person data
provided by respondents without the preassigned, unique identification number. The process of
comparing these non-ID cases to existing address information in the MTdb is referred to as “Non-ID
Processing.” The final MTdb matching and geocoding result for each non-ID case is the outcome of the
process. The Census Bureau will use administrative, third-party, and census data to augment
respondent-provided address data (referred to as “address enhancement”) to support matching to the
MAF/TIGER database (MTdb)1.
Throughout the entire self-response period, Non-ID Processing will match respondent-provided
addresses from non-ID cases to the MTdb. Census Bureau research suggests that initial address
matching efforts will pair about 87 percent of all non-ID cases with an existing record in the MTdb.
Research also suggests that an additional 2 to 3 percent could be matched to an existing record in the
MTdb as a result of address enhancement, which focuses on the non-ID cases that could not be initially
matched to the MTdb. Information derived from administrative records and third-party data sources is
used to augment respondent-provided address information, creating the most accurate addresses
possible. These “enhanced” addresses then are used in a second attempt to match non-ID addresses to
the MTdb. The remaining non-ID cases that fail the second matching attempt or were not “enhanced”
will be validated and geocoded manually.
Only data from the Title 13 composite will be used in this operation.

1

The initial census enumeration universe will have been updated through Address Canvassing and incorporated
into the MAF that is used during Non-ID processing.

1 May 2020

8

 B.

PAPER DATA CAPTURE QUALITY ASSURANCE

Administrative data are used to increase the confidence in the information that is captured from paper
questionnaires. Manual keying or Optical Character Recognition (OCR) is used to capture the
information from returned paper questionnaires, then the information is compared with existing
information as a quality check (i.e., to keep error rates low). Administrative data are not used to replace
any manual or OCR captured data, just to provide a confidence level about it. Greater confidence will
increase OCR rates and decrease keying costs.
More specifically, the 2020 Census Paper Data Capture system, Integrated Computer Assisted Data Entry
(iCADE), will use OCR to electronically capture response data provided by respondents, thereby
eliminating the majority of data needing to be captured manually. Each character within a response field
will be assigned a confidence score by the OCR software for processing.
Results from OCR are also made available to the iCADE application, where low-confidence write-in fields
are manually captured to enter into a verification process. This Quality Assurance (QA) process performs
an independent verification process manually of both OCR and keyed entries to ensure the outgoing
error rate is acceptable and meets the Service Level Agreement. Verification is done by comparing the
two data sets looking for a match or not. A Sample Stratum is defined and used to determine the
sample.
If the error rate within any sample stratum is determined to be unacceptable, those fields are then sent
to a process called Remainder Verification. Remainder Verification manually reprocesses all fields within
the targeted stratum that were not sampled and verified. This process is looking to correct any other
errors in the batch that may need corrected.
Having access to supportive demographic information (age, race, sex, date of birth) from administrative
data will allow iCADE to produce additional matches within a household. This can enhance the quality
related to the confidence levels for “First Names and Last Names” assisting the OCR and keying data
validation processes.
Names and other external data elements integrated into this process will not be used as a direct
replacement for response data. The data sources, when linked to an MTdb address, will be used to
increase or decrease the confidence levels of the data captured. External data elements are integrated
to validate existing OCR and manual keying results.
Only data from the Title 13 composite will be used in this operation.
C.

SPECIAL ENUMERATIONS

The Census Bureau focuses additional efforts on nontraditional living quarters, which may be harder to
data capture because of the unique living arrangements they present.
These enumerations will not use T26 composite or T13 composite data.
A group quarters (GQ) is a place where people live or stay in a group living arrangement that is owned or
managed by an entity or organization providing housing and/or services for the residents. This is not a
typical household-type living arrangement. These services may include custodial or medical care as well
as other types of assistance, and residency is commonly restricted to those receiving these services.
1 May 2020

9

 People living in GQs are usually not related to each other. GQs include such places as college residence
halls, residential treatment centers, skilled nursing facilities, group homes, correctional facilities,
workers’ dormitories, military barracks, and domestic violence shelters. The Census Bureau will work
with GQ administrators to enumerate the people residing at the GQ. GQ administrators can provide
2020 Census response data either electronically to a Census Bureau secured portal or can provide a
census worker with a paper listing of response data. These data may be used as the sole source of
obtaining information or as a supplemental tool to ensure data collection of an entire facility when other
enumeration methods are used. These data are not considered administrative records or third-party
data, but internally collected data.
The Census Bureau will work with Department of Defense (DOD) to acquire lists of deployed U.S.
military personnel to ensure a complete enumeration of this population. DOD will use its administrative
records to provide the Census Bureau a data file of military and civilian employees who are deployed
outside of the United States (while stationed/assigned in the United States). The Census Bureau will use
this file to enumerate these employees at their stateside address that matches to an existing address in
the MTdb.
DOD will also provide counts by home state of U.S. military and federal civilian employees stationed or
assigned overseas and their dependent living with them at their overseas duty station as part of the
Federally Affiliated Count Overseas (FACO) operation. In addition, the Census Bureau will work with
other federal departments and agencies to collect counts of their federally employed individuals
stationed or assigned overseas and their dependents living with them.
VII.

NONRESPONSE FOLLOWUP

After giving the population multiple opportunities to self-respond to the 2020 Census, addresses for
which the Census Bureau did not receive a self-response will form the initial universe of addresses for
the NRFU operation. The NRFU operation serves two purposes: 1) to determine housing unit status for
nonresponding addresses, and 2) to enumerate housing units for which a 2020 Census response was not
achieved.
For the 2020 Census, we will use administrative data to reduce the NRFU workload. Data will be used to
reduce contacts for vacant and deleted units. The NRFU workload will be reduced further by using
administrative data, where feasible, to enumerate occupied households that have failed to respond
after several contact attempts. Additionally, administrative data will be used to develop a “Best Time to
Contact” model that will reduce the NRFU workload by increasing the likelihood of finding respondents
at home.
A.

VACANT AND DELETE IDENTIFICATION

Prior to any fieldwork, an initial set of vacant and nonexistent addresses will be identified using
administrative data. Pending further information obtained for these addresses, these cases will receive
either one or multiple contact attempts.
Data from the Title 26 composite will be used in this operation.
To determine housing unit status, the Census Bureau will use data from internal sources, such as the
ACS, the MTdb (including USPS DSF status), and the 2010 Census, as well as external sources, such as the
U.S. Postal Service’s Undeliverable-As-Addressed (UAA) file and IRS records.
1 May 2020

10

 More specifically, a multinomial logistic regression model will be used to assign predicted probabilities
that a given housing unit is vacant or nonexistent. FTI from both IRS 1040 Individual Tax Payer Returns
and IRS 1099 Information Returns will be used as inputs to the model. The model also includes as
predictors administrative record household roster information (i.e., count and composition), U.S. Postal
Service UAA codes, block group characteristics from the ACS 5-year estimates, and housing unit
characteristics from the MTdb. These predicted probabilities are compared with predefined thresholds
designed to maximize the identification of vacant and nonexistent units while minimizing
misclassification error. The household status probabilities and final status designation will be
determined in a Title 26 environment and the status designation will be sent to a Title 13 environment
for use in production.
B.

ADMINISTRATIVE RECORDS ENUMERATION

Models will be used to identify units with high quality administrative data to represent occupied units.
These models use the same predictors as those used in the vacant and delete model. Occupied
households that have failed to respond after several contact attempts will be eligible for enumeration
using their own administrative data in lieu of multiple personal visits. These addresses will be matched
to administrative data that have been combined to provide the “best” person and housing unit
information found throughout the sources. Response records sent to the National Archives and Records
Administration (NARA) will not include enumerations with administrative records.
1) Count and Composition
Data from the Title 26 composite will be used in this operation.
FTI from both IRS 1040 Individual Tax Payer Returns and IRS 1099 Information Returns, as well as nonFTI from sources (including many of those listed in Section IV), will be used determine household count
and composition. Note that other data sources will be used to corroborate information from FTI and
unique FTI will be overwritten as described in Appendix B so that the final enumeration data are Title 13.
2) Characteristics
The non-FTI sources will also be used for characteristic enumeration. When possible, we will use the
2010 Census and ACS responses before using information from other sources. No FTI will be used in
characteristic enumeration.
Six characteristics will be attempted to be assigned for the 2020 Census, five at the person level and one
at the housing unit (HU) level. Only federal sources will be used to assign person-level characteristics.
Characteristics Expected to be Enumerated from Administrative Data for the 2020 Census
Person-Level Characteristics
Housing Unit Characteristic
Name
Tenure (own or rent the housing unit)
Age/Date of Birth
Sex
Race and Ethnicity
Relationship to Householder

1 May 2020

11

 C.

BEST TIME TO CONTACT MODELING

For the 2020 Census, the goal of enumerator route optimization is to increase the productivity of the
enumerators in the NRFU operation by decreasing the miles traveled and total hours spent per case.
Assignments will be based on factors including enumerator availability, workload location, and the
probability of successfully contacting cases at various hours of the day. The enumerators will receive an
optimal routing of attempts to minimize travel.
Data from the Title 26 composite will be used in this operation.
Logistic regression models will be used to develop predicted probabilities associated with contacting
occupied housing units during each hour of the workday (weekdays and weekends). The Census Bureau
will use data from internal sources, such as the ACS and the MTdb, as well as external sources, such as
IRS records, to develop input variables for the model. The predicted contact probabilities will be
determined in a Title 26 environment and then sent to a Title 13 environment for use in production to
optimize case assignments.
VIII.

RESPONSE DATA VERIFICATION

The Census Bureau will use administrative data to examine the consistency between the respondentprovided or enumerator-collected information and the administrative data sources to determine if
further analysis and/or field investigation is warranted.
Name information, in combination with address and other information, can confirm a person’s
association with an address as well as the demographic data associated with their household and
whether the person can be found at another address. A level of consistency (i.e., matched, nonmatched)
between the respondent-provided information or the enumerator-provided data and information
available from administrative data will be determined.
A.

SELF-RESPONSE QUALITY ASSURANCE

Data from the Title 26 composite will be used in this operation.
FTI and non-FTI sources will be used in matching algorithms to develop a set of Title 13 match results.
The match results are then used as inputs to scoring models that help identify suspicious cases for
further investigation. The algorithms will be run in a Title 26 environment and the match results sent to
a Title 13 environment for use in production.
B.

ENUMERATOR QUALITY CONTROL

Only data from the Title 13 composite will be used in this operation.
Like many of the other processes, quality control efforts will use the Title 13 composite described in
Section II above as the comparison data set to verify that enumerators conducted interviews
appropriately and collected accurate data. The data in the composite will never be used in lieu of
additional field verification for those housing units with consistency measures below an acceptable
level. Furthermore, the administrative data matching results will never be used as a sole indicator of
whether a response is suspicious and requires further analyst and/or field follow-up. Instead, the results
from administrative data will be used in conjunction with the results of statistical modeling on response
data, paradata, and other data elements to flag responses as suspicious.
1 May 2020

12

 IX.

POST-RESPONSE PROCESSING
A.
COUNT IMPUTATION

After data collection, count imputation 1) assigns final statuses (i.e. occupied, vacant, and nonexistent)
to those cases that were unresolved in the field and 2) assigns population counts to records with an
occupied status, but without a specified number of occupants. Response records sent to NARA will not
include imputations from administrative records.
Data from the Title 26 composite will be used in this operation.
We will use the USPS Undeliverable As Addressed (UAA) file as well as information from IRS 1040, IRS
1099, 2010 Census, Census Master Address File, 2020 Census paradata, American Community Survey,
and other sources including Medicare and Indian Health Service to place all addresses into groups with
similar status and household size (if occupied). Within each group, a hot-deck imputation process will be
used to fill in the missing status and count (if occupied) from a census address with a complete status to
the address without a complete status. FTI will never be used as a direct source for final status
assignment.





The classification of administrative records information for each address will be completed in a
Title 26 environment using the overwriting process described in Appendix B. These variables will
be transferred to a Title 13 environment to model the final status when it is unknown and the
household roster count for occupied units where the count is unknown.
UAA information will be assigned to each CensusID. The National Processing Center downloads
this information from USPS for the Census Bureau.
We are only using Protected Identification Keys (PIKs) and MAFIDs from the IRS data.
B.

CHARACTERISTIC IMPUTATION

After all responses are processed and the final population count is established, there are sometimes
missing, inconsistent, or nonvalid characteristic data. Administrative records sources will be used as part
of the process to impute missing characteristics. Imputation methods include direct substitution from
prior Title 13 responses or from other administrative record sources. They also include procedures
where missing data are imputed from donors with similar characteristics at the person- or housing-unit
level. Response records sent to NARA will not include imputations from administrative records.
Only data from the Title 13 composite above will be used in this operation. Expected sources of data
include 2010 Census data, ACS responses, the Census Numident, the Best Race and Ethnicity file, Census
Household Composition Key File, HUD PIC/TRACS, and Black Knight. When possible, we will try to impute
from the 2010 Census and ACS data before imputing from other sources. No FTI will be used in
characteristic imputation.
Six characteristics will be imputed for the 2020 Census, five at the person level and one at the housing
unit (HU) level. Only federal sources will be used to impute person-level characteristics.

1 May 2020

13

 Characteristics Expected to be Imputed for the 2020 Census
Person-Level Characteristics
Housing Unit Characteristics
Age/Date of Birth
Tenure (own or rent the housing unit)
Sex
Race and Ethnicity
Relationship to Householder
X.

PUBLISHING DATA
A.
CITIZEN VOTING AGE POPULATION (CVAP) TABLES

The Census Bureau will release a special tabulation of citizen voting age population using administrative
data1. Currently, researchers are determining the sources and methodology to accomplish this. Expected
sources will include many of the other administrative data used in 2020 Census production. However,
this data will not be first compiled into a composite, but will be accessed directly by the CVAP team.
Some additional IRS 1040 and IRS1099 variables are available for this work that are not used in the other
production activities. In addition, the CVAP team will use other sources that were acquired specifically
for this use or are unique to the citizenship effort and not a part of other production activities.
B.

COUNT QUESTION RESOLUTION

The 2020 Count Question Resolution program (CQR) allows tribal, state, and local government elected
officials to request review for corrections to their jurisdiction’s 2020 Census counts. There are
historically a small percentage of cases where an incorrect geographic boundary or coding of a housing
using was used to produce the official census population and housing count for a local area. There may
also be cases where, because of processing errors, the Census Bureau mistakenly duplicated or deleted
living quarters that were identified during the census. The Census Bureau does not collect any additional
data during the case process but will access existing information to research cases. The CQR process will
not research whether respondents or census enumerators incorrectly determined an address’s
occupancy status or household size during field operations, but only resolve boundary disputes, address
geocoding errors, and address coverage errors. Census Bureau records used in the research phase
include the MTdb, records collected or used during frame development activities (see section IV) and
2020 Census return metadata.

1

Since 2011, the ACS, which directly asks about citizenship, has been the source of the CVAP tables.

1 May 2020

14

 XI.

COVERAGE EVALUATION

The Post-Enumeration Survey (PES) is one of two primary evaluation tools used to produce estimates of
census coverage. For the PES, addresses in the selected sample blocks are listed and enumerated in
operations that are independent of the 2020 Census. The Census Bureau identifies matches and
nonmatches and discrepancies between the 2020 Census and the PES, for both housing units and people
in the sample area. Both computer and clerical components of matching are conducted. The system that
is conducting computer matching will use telephone numbers from administrative data (that is, the
Contact Frame) for census records in the sample areas when no telephone number was reported in the
census. As a result, the use of the updated telephone numbers could improve computer match rates,
thereby improving the efficiency of clerical matching and potential followup operations. Additionally,
during PES data processing, the same characteristic imputation methodology that is used for the 2020
Census will be used for the PES.
Demographic Analysis (DA) is the other approach for measuring census coverage. DA refers to a set of
methods that have historically been used to develop national-level estimates of the population for
comparison with decennial census counts. DA estimates are developed from historical vital statistics
from the National Center for Health Statistics (NCHS), data on international migration from ACS, and
Medicare records. The data are independent of the census being evaluated. The estimates are then
compared with the census counts by age-sex and limited race and/or ethnicity groups to evaluate net
coverage error in the census. In addition to the data previous used in DA, a legal permanent resident file
maintained by the Office of Immigration Statistics may also be used to assess the uncertainty of the DA
estimates in 2020. IRS tax returns and the Census Numident will also be used to develop subnational DA
estimates for young children age 0 to 4; this is not one of the official series of 2020 DA estimates.

1 May 2020

15

 APPENDIX A – EXPECTED DATA SOURCES FOR USE IN THE 2020 CENSUS
Use
Frame
Resp
Non-ID
Paper Data
Special
Source
Development Motivation Address Validation
Capture QA
Enumerations.
USPS DSF
XXX
Ongoing MAF/TIGER System updates from
XXX
partnerships and fieldwork
2020 Census partnership data (LUCA, NC)
XXX
Lists of GQs and TLs from federal sources
XXX
(Medicare, BOP, ICE, Bureau of Indian
Affairs, US Marshals, Maritime Agencies,
Military), state & local partners
STR Inc. list of hotels/motels
XXX
MTdb Extracts
XXX
XXX
Modeled self-response predictions
XXX
Audience segmentation information
XXX
FCC Internet Connectivity
XXX
GQ administrator lists
XXX
DOD Deployed data and FACO inputs#
XXX
CMS MEDB
XXX
XXX
HUD FHA IDB (CHUMS)
XXX
XXX
HUD Longitudinal (PIC/TRACS)
XXX
XXX
IHS Patient Registration
XXX
XXX
IRS 1040
IRS 1099
OIS Legal Permanent Resident
NCHS Vital Statistics (births and deaths)
SSS Registration
XXX
XXX
State/Local Program Data
XXX
USPS NCOA
XXX
XXX
ACS Data
ACS Contact History
2000 Census Data
XXX
XXX
XXX
2010 Census Data
XXX
XXX
XXX
Census Numident
XXX
XXX
Census Numident Alternate Names
XXX
Census HH Composition Key
Contact Frame
XXX
XXX
Best Race and Ethnicity
Black Knight
DAR Partners
XXX
XXX
Targus Federal Consumer and Wireless
XXX
XXX
VSGI
XXX
XXX
USPS UAA
Planning Database
XXX
Files only used for citizenship (See III.D)
* SNAP TANF and WIC only
# Other federal agencies will provide their FACO numbers directly into the system, not as a file to 2020 production systems and are
therefore not listed here.

1 May 2020

Appendix A - 1

 Use
Source

NRFU
Vacant/Delete
Identification

NRFU AR Enumeration
Occupied / AR
Count and
CharacterEnum. Eligible Composition
istics

NRFU Best
Time to
Contact

USPS DSF
Ongoing MAF/TIGER System updates from
partnerships and fieldwork
2020 Census partnership data (LUCA, NC)
Lists of GQs and TLs from federal sources
(Medicare, BOP, ICE, Bureau of Indian
Affairs, US Marshals, Maritime Agencies,
Military), state & local partners
STR Inc. list of hotels/motels
MTdb Extracts
XXX
XXX
XXX
Modeled self-response predictions
Audience segmentation information
FCC Internet Connectivity
GQ administrator lists
DOD Deployed data and FACO inputs#
CMS MEDB
XXX
XXX
XXX
HUD FHA IDB (CHUMS)
HUD Longitudinal (PIC/TRACS)
XXX
IHS Patient Registration
XXX
XXX
XXX
IRS 1040
XXX
XXX
XXX
XXX
IRS 1099
XXX
XXX
XXX
XXX
OIS Legal Permanent Resident
NCHS Vital Statistics (births and deaths)
SSS Registration
XXX
XXX
State/Local Program Data
USPS NCOA
XXX
XXX
ACS Data
XXX
XXX
XXX
XXX
ACS Contact History
XXX
2000 Census Data
2010 Census Data
XXX
XXX
XXX
Census Numident
XXX
XXX
XXX
XXX
XXX
Census Numident Alternate Names
Census HH Composition Key
XXX
XXX
XXX
XXX
XXX
Contact Frame
Best Race and Ethnicity
XXX
XXX
XXX
XXX
Black Knight
XXX
DAR Partners
XXX
XXX
Targus Federal Consumer and Wireless
VSGI
XXX
XXX
XXX
USPS UAA
XXX
XXX
Planning Database
XXX
XXX
XXX
Files only used for citizenship (See III.D)
* SNAP TANF and WIC only
# Other federal agencies will provide their FACO numbers directly into the system, not as a file to 2020 production systems
and are therefore not listed here.

1 May 2020

Appendix A - 2

 Use
SRQA Enumerator
Count
Characteristic CVAP
PES
DA
Source
QC
Imputation
Imputation
USPS DSF
Ongoing MAF/TIGER System updates from
partnerships and fieldwork
2020 Census partnership data (LUCA, NC)
Lists of GQs and TLs from federal sources
(Medicare, BOP, ICE, Bureau of Indian
Affairs, US Marshals, Maritime Agencies,
Military), state & local partners
STR Inc. list of hotels/motels
MTdb Extracts
XXX
XXX
XXX
XXX
Modeled self-response predictions
Audience segmentation information
FCC Internet Connectivity
GQ administrator lists
DOD Deployed data and FACO inputs#
CMS MEDB
XXX
XXX
XXX
XXX
XXX
HUD FHA IDB (CHUMS)
XXX
XXX
HUD Longitudinal (PIC/TRACS)
XXX
XXX
XXX
XXX
XXX
IHS Patient Registration
XXX
XXX
XXX
XXX
IRS 1040
XXX
XXX
XXX
XXX
IRS 1099
XXX
XXX
XXX
OIS Legal Permanent Resident
XXX
NCHS Vital Statistics (births and deaths)
XXX
SSS Registration
XXX
XXX
XXX
State/Local Program Data
XXX*
USPS NCOA
XXX
XXX
XXX
ACS Data
XXX
XXX
XXX
XXX XXX
ACS Contact History
2000 Census Data
XXX
XXX
XXX
2010 Census Data
XXX
XXX
XXX
XXX
XXX
XXX
Census Numident
XXX
XXX
XXX
XXX
XXX XXX
Census Numident Alternate Names
XXX
XXX
XXX
Census HH Composition Key
XXX
XXX
XXX
XXX
Contact Frame
XXX
XXX
XXX
Best Race and Ethnicity
XXX
XXX
Black Knight
XXX
XXX
DAR Partners
XXX
XXX
Targus Federal Consumer and Wireless
XXX
XXX
VSGI
XXX
XXX
USPS UAA
XXX
Planning Database
Files only used for citizenship (See III.D)
XXX
* SNAP TANF and WIC only
# Other federal agencies will provide their FACO numbers directly into the system, not as a file to 2020 production systems
and are therefore not listed here.

1 May 2020

Appendix A - 3

 APPENDIX B - OVERWRITING EXAMPLE
This is an example of how the U.S. Census Bureau can use FTI and information from multiple sources in the 2020 Census to inform census
fieldwork decisions and conduct administrative record enumeration. This approach adheres to procedures provided by IRS Office of Safeguards
regarding validation and overwriting FTI. This example was presented to IRS Statistics of Income staff in July 2016 during the review and approval
process for the 2020 Census Production Predominant Purpose Statement.
This example presents a hypothetical Mars family who filed an IRS 1040 return at 101 Main Street. The source, validation, and overwriting
involving different variables and values are presented. Data variables and values that are FTI are shown in red shaded background with white
letters. Data variables and values that have been overwritten in accordance with the IRS regulations are shown in white background with green
letters.
1.

IRS 1040 and 1099 Information

Each month, the IRS delivers the IRS 1040 Individual tax returns. Table 1 shows the two hypothetical families at 101 and 102 Main Street. The
Mars family has four people and lives at 101 Main Street. John Smith lives by himself at 102 Main Street. Since this is the 1040 information
delivered to the Census Bureau, all of the variables are FTI and thus colored in red. Among the variables, the IRS provides the Census Bureau
with the address, tax identification number (TIN) of filers and dependents, filing status and other FTI variables. For 101 Main Street, primary filer
Michael Mars had a full name, SSN present but the other people have only last name, and SSNs provided. For 102 Main Street, single primary
filer John Smith did not provide a SSN when filing. While not shown in the table, the IRS also delivers a similar file containing the 1099
Information return file as well.

Depende
nt 3 TIN

Depende
nt 3
Name

Other FTI
Variables
…

Mars

Depende
nt 2
Name

Secondar
y
Name

111-111112

Depende
nt 2 TIN

Secondar
y Filer TIN

Michael
Mars

Depende
nt 1
Name

Primary
Name

111111111
blank

Depende
nt 1 TIN

Primary
Filer TIN

101
Main
St
102
Main
St

Filing
Status

Address

Table 1: IRS 1040 FTI Record of the Mars Family at 101 Main St.

Joint
Married

111-111113

Mars

111-111114

Mars

None

None

…

John
Smith

None

None

Single

None

None

None

None

None

None

…

30 November 2019

Appendix B - 1

 2.

Census Person Validation System (PVS) Processing of IRS 1040 and 1099 Delivered File

After receipt of the IRS 1040 file, the Census Bureau processes the IRS 1040 file through the Person Validation System. This processing uses the
address provided by the IRS and links this address to an address already to the Census Master Address File (MAF). This assigns the Census
Bureau’s MAF Identification Number (MAFID) to the record. The PVS processing was able to assign the Census Bureau’s Protected Identification
Key (PIK) to replace the IRS TIN based on combinations of SSN, Name and Address fields. For John Smith, the PVS processing was unable to
assign a PIK. This is an example of a file that has a combination of FTI and non-FTI variables. The address and PIKs are colored in green since they
have been overwritten in accordance with IRS regulations. IRS provided name information has been removed. The Census Bureau does similar
processing of the 1099 Information file as well.

Census
MAFID

Address

Table 2: IRS 1040 Record for Mars Family after Census PVS Processing
Census PIK 1
Census PIK 2
Filing Status
Census PIK 3
Census PIK 4

123412341
123412342

101 Main St
102 Main St.

999-99-9991
Blank

999-99-9992
Blank

Joint Married
Single

999-99-9993
Blank

999-99-9994
Blank

Other FTI
Variables
…
…

Since 102 Main St has no PIKs, we are not able to build a roster using IRS 1040 data for this address. The remainder of this document will focus
on using the IRS data to see if we can reduce contacts for 101 Main St. For the 2020 Census, the Census Bureau will be building rosters for
households using multiple sources determined by September 2018. The remainder of this document will assume that an additional source
besides IRS information can corroborate that the Mars family lives at 101 Main St.
3.

Processing of Administrative Records to Determine Census Fieldwork

The Administrative Records Modeling team has been researching under the DMS 863 project the possibility of using administrative records and
third-party data to determine if the number of contacts can be reduced. The 2020 Production DMS 987 project will implement this usage in the
2020 Census. In this example, we will show parts of the processing. Step 2 showed that the Census Bureau had a file that had a combination of
FTI and non-FTI variables. Our research has shown that other variables created using FTI are powerful predictors in our determination. This
example highlights two of them. Table 3 shows an example of where we create additional variables based on FTI information to be used in the
30 November 2019

Appendix B - 2

 determination about how many contacts to conduct during fieldwork. There is a history of using IRS information combined with other sources to
determine fieldwork for the decennial census. For the 2010 Census, part of the workload for the Coverage Followup operation was based on the
comparison of census rosters to administrative records households that included IRS 1040, IRS 1099 and other non-FTI sources.
An example of two variables created using FTI information is:


Was any person in the Mars household found on another 1040 return filed this year? Our research has found that if any person in the
household is associated with another 1040 return then there are possible questions about the administrative record roster assembled.



Was any person in the Mars household found on the 1040 return filed at this address last year? Our research has seen that finding at
least one person on 1040 returns at the same address in consecutive years is more highly correlated with count and household
composition agreement.

Census
MAFID
123412341
123412341
123412341
123412341

Table 3: Example of Creating Two Additional FTI Variables to be used in Determination
Census PIK
Filing Status
Processing
Found on
Found on last year
Week
another 1040
1040 at same
return
address
999-99-9991
Joint Married
12
No
Yes
999-99-9992
Joint Married
12
No
Yes
999-99-9993
Joint Married
12
No
Yes
999-99-9994
Joint Married
12
No
Yes

The Administrative Record determination processing uses predictive models. These predictive models are based or “trained” on comparison of
the rosters built for 2010 NRFU addresses using persons from 2010-versions of the administrative record sources to the person enumerated at
those addresses in the 2010 Census. Based on building a roster using the most recent versions of the administrative record files, we can use
predictive models can allow us to make determinations if the roster from administrative record sources is similar enough to a census fieldwork
enumeration that we can reduce the number of contacts. In our example, we have built a roster of the Mars family.

30 November 2019

Appendix B - 3

 For every 2020 household roster built from administrative record sources, we can predict:


How likely is it that all of the Mars family should be counted at the address? From a scale of 0 to 1, we want this to be as close to 1 as
possible.



How likely is it that the household composition observed for the Mars family from IRS 1040 and other sources would match Census
fieldwork4? From a scale of 0 to 1, we want this to be as close as possible to 1 as well.

Figure 1 shows a graphical description of the number of contacts during fieldwork determination. Each axis represents one of the two
predictions listed above. The results are shown in three colors of green, yellow, and red.
For the 2020 Census, we are implementing the following three classifications. In our example, the Mars family will have high enough results for
the two predictions that we will reduce the contacts so that this address is classified as administrative record occupied and will have reduced
contacts during the NRFU operation.

4

•

Administrative Record Occupied in NRFU (Green): If the two probabilities land the address in the green area then the address will be
administrative record occupied. We will be reducing contacts for these addresses.

•

Administrative Record Occupied for Imputation (Yellow): The Census Bureau has done research on using administrative records after all
of the contacts in the NRFU as an alternative to count imputation. For the 2016 Census Test, these addresses will receive six contacts. If
the two predicted probabilities land the address in the yellow area then the address we will use the roster of administrative records as
an alternative to imputation

•

Not Use (Red) – If the predicted probabilities are in red then we would implement full fieldwork. We would not want to use these rosters
built from administrative records as an alternative to count imputation.

In this example, we would be saying how likely it would be that the census would enumerate a household with two adults and one or more children present.

30 November 2019

Appendix B - 4

 Figure 1: Graphical Representation of Hypothetical Administrative Record Occupied Determination

30 November 2019

Appendix B - 5

 4.

Overwriting for Enumeration Purposes

Based on the result for the Mars family, we have determined that we will reduce contacts during NRFU. While we used FTI in determining how
to conduct census fieldwork, the following will use the Mars family as an example to show how we will not being using FTI for enumeration.
For enumeration, the only information originally sourced from FTI that we use is the address and the PIK. Based on the IRS Office of Safeguards
response to the Census Bureau’s January 2016 overwriting inquiry, the Census MAFID and PIK have overwritten the FTI. In accordance with the
guidance from IRS, we will document the overwriting and the date that it has occurred. This file is no longer FTI. Table 4 provides an example.
Table 4: Example of Overwriting Documentation File for Future IRS Office of Safeguards Review
Census MAFID
Unique Person Number
Census PIK
Overwriting Date
123412341
1
999-99-9991
May 2, 2016
123412341
2
999-99-9992
May 2, 2016
123412341
3
999-99-9993
May 2, 2016
123412341
4
999-99-9994
May 2, 2016

30 November 2019

Appendix B - 6

 5.

Obtain Short Form Characteristics from non-FTI sources

For census enumeration purposes, we need to obtain the name, age, date of birth sex, relationship, race, Hispanic origin and tenure information
for the people at this address. For these characteristics, this information will be obtained from non-IRS Title 26 sources. This includes past
Census Bureau responses or other administrative record information sources.
Table 5 shows the administrative record enumeration for 101 Main Street. All of the fields are colored green to indicate that no FTI information
was used in these variables. Name was obtained from the Census Numident file or past Census responses. To further, emphasize in this example
that no FTI information is being used for enumeration purposes, the relationship for Mars family is set to missing. From the IRS 1040 return, we
did see that the Mars family filed a Joint Married return but we do want to point out that we are not using that information. Since IRS is the sole
source of the relationship, we are unable to use this information as a direct assignment. Our characteristic imputation will have to assign the
relationship status for this household.
Table 5: Example of Administrative Record Enumeration of Mars Family Using non-FTI information
Census Address
Unique
Name
Age
Date
Sex
Relationship
Person
Of
Number
Birth
101 Main St
1
Michael Mars
43
1/1/1973 Male
Missing
101 Main St
2
Mary Mars
40
2/2/1976 Female Missing
101 Main St
3
Hershey Mars
17
3/3/1999 Male
Missing
101 Main St
4
Lucy Mars
14
3/4/2002 Male
Missing

Race

White
White
White
White

Hispanic
Origin
Non-Hispanic
Non-Hispanic
Non-Hispanic
Non-Hispanic

Note: Census Day is April 1, 2016 for this example.

Step 3 showed that for some addresses it may be determined administrative records would only be used as an alternative to imputation. For
these addresses, they will receive the full amount of contacts during the NRFU operation. If at the end of the census fieldwork, their occupancy
and/or population status are still unresolved then administrative records will be used in a similar way as laid out in this section.

30 November 2019

Appendix B - 7

 6. Creating a Census Response File
Based on this processing, this information can then be delivered to our Title 13 server and be combined with the other Census Bureau
responses. This information can be used for all Title 13 purposes of the 2020 Census including archiving to be made available 72 years after
Census Day. Table 6 shows a simple example of variables and values that would be included.
Table 6: Census Response File
Address
Enumeration
Operation
101 Main St
101 Main St
101 Main St
101 Main St
102 Main St
102 Main St
103 Main St

Administrative Records
Administrative Records
Administrative Records
Administrative Records
Internet Self-Response
Internet Self-Response
NRFU

Unique
Person
Number
1
2
3
4
1
2
1

Name

Michael Mars
Mary Mars
Hershey Mars
Lucy Mars
Thomas Lindt
Sally Lindt
John Haribu

Age

43
40
17
14
28
26
33

Sex

Relationship

Race

Male
Female
Male
Female
Male
Female
Male

Missing
Missing
Missing
Missing
Householder
Spouse
Householder

White
White
White
White
Black
Black
SOR

Hispanic Origin

Non-Hispanic
Non-Hispanic
Non-Hispanic
Non-Hispanic
Non-Hispanic
Non-Hispanic
Hispanic

Summary
This document shows how FTI will and will not be used in relation to administrative record enumeration in the 2020 Census. While FTI will be
used during the processing to determine the number of contacts for census fieldwork purposes, the example shows that FTI will not be used in
the direct enumeration.

30 November 2019

Appendix B - 8