An Overview of the 2020 Census Design

Joint Statistical Meetings 2018Session 417:

Redesigning Federal SurveysRobin A. Pennington, Michael Bentley, Gina Walejko

Disclaimer: Any opinions stated in this presentation reflect only personal views of the authors and do not necessarily represent the views of the U.S. Census Bureau.

1

2020 Census Operational Plan

• V4.0 to be released December 2018 in 2020 Census Memo Series

• Will include updated design incorporating results from the 2018 End-to-End Census Test

2

3

Focus for Today’s Presentation

• Optimizing Self-Response• Internet self-response• Contact strategies – Internet First, Internet Choice• Integrated Partnership and Communications Program

• Main Components • Research and Communications Integration

• Minimizing Field Operations by using Administrative Records and Third-party Data

• Operational Design Integration

4

5

The 2020 Census – Optimizing Self-ResponseGenerate the largest possible self-response, reducing the number of households requiring follow-up

6

History of Self-Responses in the Census

• Since 1970, the decennial census has relied on mailout/mailback of paper questionnaires.

• In Census 2000, a reminder postcard was added, a week after the paper questionnaires.

• In the 2010 Census, a second paper questionnaire was added, with some changes to the plan when Nonresponse Followupreverted to a paper operation.

• None in high-response areas• To nonresponders only in medium response areas• Blanket mailing in low response areas

• In the 2020 Census, the primary response mode will be Internet.

7

Census Tests – Optimizing Self-Response Findings• 2012 National Census Test

• Initial contact to invite participation, followed by two reminder prompts as needed and subsequent mailing of a paper questionnaire was a promising strategy (Internet push).

• Advance letter not shown to improve response rates.• Telephone assistance needed for respondents without Internet access.

• 2014 Census Test• Low participation rate for “pre-registration,” but high questionnaire

completion rate among those who pre-registered.

• 2015 Optimizing Self-Response Test• An additional 35,249 Internet responses from housing units not

selected in mail panels as a result of advertising and promotional efforts.

• Successful real-time non-ID processing, with high match rate.

8

Census Tests (Continued)• 2015 National Content Test

• Mailing a reminder notice significantly increased response rate.• Mailing a questionnaire first in “low response areas” is effective.

• 2016 Census Test• Sending a letter instead of a postcard as the first reminder increased response

rate.• Providing more language services increased responses from non-English

speakers.• Mailing a questionnaire first in areas estimated to have lower Internet usage is

effective.

• 2017 Census TestTest of Tribal Enrollment question – no related findings

• 2018 End-to-End Census TestCurrently in production

9

Table 1: Response Rates for Decennial Census Tests Occurring between 2012 and 2018

Internet Phone Mail Total2010

Census

2012 National Census Test (50 states and D.C.) 36.5 7.4 13.0 56.9 66.5

2014 Census Test(Montgomery County and Washington D.C.)* 50.6 5.2 10.2 65.9 72.8

2015 Census Test (Maricopa County)* 39.7 6.4 8.8 54.9 63.8

2015 Optimizing Self-Response Test (Savannah) 33.4 5.6 8.4 47.5 56.5

2015 National Content Test(50 states and D.C., average of all panels) 35.6 6.1 10.1 51.9 66.5

2016 Census Test (Harris County)* 28.2 1.9 9.2 39.3 60.5

2016 Census Test (Los Angeles County)* 32.4 2.6 17.4 52.5 71.6

2017 National Census Test (50 states and D.C.) 31.7 2.4 16.2 50.3 66.5

2018 Census Test (Providence County) (Preliminary results) 32.0 3.9 16.4 52.3 64.7

*Test conducted in a portion of the selected areaSource: Response data and sample operational data for each individual test.

10

2020 Census Proposed Self-Response Mail Strategy

Dates in the table below are in-home dates.

Self-Response• Two-panel design: Internet First (invitation letter on first

contact) and Internet Choice (questionnaire on first contact)• New for 2020: four Internet First cohorts to best distribute

calls to Census Questionnaire Assistance• Internet Choice panel is in a single cohort, mailed on the

same schedule as Internet First, Cohort 2

Panel Cohort

Mailing 1Letter (Internet First)

or Letter +

Questionnaire (Internet Choice)

Mailing 2Letter

Mailing 3*Postcard

Mailing 4* Letter +

Questionnaire

Mailing 5*“It’s not too late”

Postcard

Internet First

1 March 12, 2020 March 16, 2020 March 26, 2020 April 8, 2020 April 20, 2020

2 March 13, 2020 March 17, 2020 March 27, 2020 April 9, 2020 April 20, 2020

3 March 19, 2020 March 23, 2020 April 2, 2020 April 15, 2020 April 27, 2020

4 March 20, 2020 March 24, 2020 April 3, 2020 April 16, 2020 April 27, 2020

Internet Choice N/A March 13, 2020 March 17, 2020 March 27, 2020 April 9, 2020 April 20, 2020

*Mailings targeted only to nonresponding households

Integrated Partnership and CommunicationsMain Components

The Integrated Partnership and Communications operation communicates the importance of participating in the 2020 Census to the entire population of the 50 states, the District of Columbia, Puerto Rico and the island areas to:• Engage and motivate people to self-respond,

preferably via the internet

• Raise and keep awareness high throughout the entire 2020 Census to encourage response

12

Integrated Partnership and CommunicationsResearch and Communications Integration

13

Predictive Models

STEP 1

STEP 2 Conduct the Census Barriers, Attitudes, and Motivators Study (CBAMS) 2020

STEP 3

STEP 4

STEP 5

Segmentation

Media Planning

STEP 6

STEP 7

STEP 8

Design Advertising

Deliver Advertising

Optimize Resources

Rapid Response

Use of Administrative Records in U.S. Censuses

• Administrative Records in prior censuses: • Frame-building • Imputation • Characteristic imputation

• Expanded use and improved quality of records:• Clean-up of address list prior to NRFU• Enumeration data if no response

14

15

16

AdRec: Administrative Records and Third-Party DataHU: Housing UnitNRFU: Nonresponse Followup OperationTEA: Type of Enumeration AreaUAA: Undeliverable As AddressedLE

GEND

NONR

ESPO

NSE F

OLLO

WUP

SELF

-RES

PONS

E

Shaded Box: Activity relating to AdRec Modeling

Shaded Diamond: Decision relating to AdRec Modeling

Hatched Box: NRFU Field Work

AdRec Vacant / AdRec Non-ExistentAdRec Occupied

Multiple AdRec Statuses

Case Resolved [not AdRec]

Case Resolved [with AdRec]

Case Unresolved

DATA COLLECTION POST DATA COLLECTION

Provide Opportunity

to Self-Respond:

1

TEA 1:Contact by Mail[Up to 5 Mailings]

TEA 6:Leave Packet at Door

[2 Reminder Mailingswhere Feasible]

Resolved with Self-Response

12Collect Self-Responses

Modes: Internet Response Collection, Paper Data Capture, Telephone Interview [for those who call in]

2

UAA Info

Resolved asVacant or

Non-ExistentRecord Address as

Vacant or Non-Existent

Unresolved

Resolved with AdRec Data

Enumerate Using AdRec Data

3

Run AdRec Modelson All Addresses

Create Universe of

NonrespondingAddresses

Possible Results:AdRec Vacant

AdRec Non-ExistentAdRec Occupied

AdRec No Determination

Attempt to Enumerate[First Attempt]

7

UAA?

Send AdditionalMailing toEncourage

Self-Response

Max Attempts?

Send AdRec Mailing

4

NRFUADREC VACANT

ADREC NON-EXISTENT

11

Resolved with Field Visit

[Enumerated oridentified as vacant

or not a HU]

NRFU UNIVERSE

Successful?

Appears Occupied?

AdRecOccupied?

9

NO

AdRec Vacant,AdRec Non-Existent, or AdRec Occupied?

AdRec Vacant or AdRec Non-Existent?

Successful?

End of NRFU?

NRFU Closeout

Phase Begins

Successful?

AdRec Vacant or AdRec Non-Existent?

NO

YES

YES

NO

YES

NO

YES[Unresolved]

Attempt to Enumerate

[Additional Attempts]

Attempt to Enumerate

[Final Attempts]

YES YES

NO

NO

NOUAA Info

8

6

6

6 7

9

9

6 9

10

8 10

Data Flow

Logic Flow

Reassess AdRec Models Results with

Relaxed Thresholds

YES

NO

NO[AdRec Occupied]

9

5

5

8 105

11

97

96

12

NRFU Administrative Records Contact Strategy, V2.0

NO

YESYES

NO

YESYES

7 9

NRFU Begins[6 weeks after Self-Response Begins]

17

Integration Artifacts

Operational Plan – 35 operationsDetailed Operational Plans

• Business Process Models• Activity Tree• Integrated Definition for Function Modeling• Integrated Operations Diagrams

Integration

18

19

IE

18. Nonresponse Followup Operation

(NRFU)

18-1. NRFU Planning and Preparation

18-2. NRFU Operational Workload Management

18-3. NRFU Data Collection

18-4. NRFU Quality Control Adjudication

18-5. NRFU Operation Closeout

Staff: HQ Staff, RCC Staff, ACO Staff, Field Staff, NPC

Staff

(3. SPC) Security, Privacy, and

Confidentiality Controls

(4. CFD) IE226: Content Specifications for

Automated Instruments

Sites: HQ, RCC, ACO, Field Area, NPC, Training Venues

Systems: ECaSE-ENUM, ECaSE-OCS, ECaSE-FOCS, SMaRCS, MaCS, MOJO: BLQ, MOJO:

Optimizer; MOJO: HERMES, UTS, LMSOther: Census Networks, Mobile

Networks, Mobile Devices

Mechanisms: Resources Employed by Operation

Controls: Data to Guide Operation

Inputs: Data Consumed by Operation Outputs: Data Produced by Operation

19. RPO

• IE112: NRFU Case Universe and Updates (including AdRecModeling Results)

• IE119: QC Computer Match Results

• IE120: Enumeration Hold Notification

Respond-ent

• IE183: Response Data• IE242: NRFU Contact

Information

(1. PM) Program Controls: incl. Budget, Operational

Plans and Schedule

Other: HQ Office IT Infrastructure,

Operation-Specific Kits

Version v07f – July 11, 2019

HU or Respondent

19. RPO

1. PM• IE404: NRFU Paradata

4. CFD

33. DLM• IE391: Operation-Specific Kit Requirements

32. FLDI• IE241: Draft and Final

Operation-Specific Training Materials (for NRFU)

(32. FLDI) IE367: Online and Classroom Training

Schedules

31. DSC• IE360: IT Service Outcome

Notification• IE577: Service Event Notification

31. DSC

6. GEOP

• IE066: Spatial Data (for assignment)

• IE067: Geographic Data Products (for assignment)

(5. LNG) IE206: 2020 Census Language

Support Plan

6. GEOP

• IE260: Request for Response• IE261: Notice of Visit

• IE254: Self-Response Case Hold Flagging Rules• IE255: AdRec Modeling Requirements• IE256: Contact Attempt Stopping Rules• IE257: QC Plan (for NRFU)• IE258: Areas for Early NRFU• IE104: Response and Status Data for Cases• IE259: Reinterview Anomaly Outcome• IE131: Address Data Updates from Field Ops

• IE380: IT Service Requests

32. FLDI

• IE250: Estimated Operational Workload • IE251: Operation-Specific Training Content and

Requirements • IE252: Approved Operation-Specific Training

Materials • IE615: Time and Expense Records• IE616: Administrative Actions

• IE049: Geographic Products Requirements

• IE209: NRFU Field Support Materials Content

20

(6,8) Paper Notices and/or Paper Form (i)

(7) CIResponse (i),

Status

(7) ISR Response (i,n),Status

(4) Updated MAFAfter ADC

GEOP: (MAF/ TIGER)

WorkingMAF for

Enumeration

GEOP: (MAF/ TIGER)

Final Collection and Tabulation

Universes

(11) Final Updated MAF/TIGER Data

GEOP: (MAF/ TIGER) -Updated MAF

after ADC

(3) Initial MAF from ADC and LUCA

GEOP: (MAF/ TIGER) -

Initial MAF for ADC and

LUCA

(1,2) Address and Spatial Data from

LUCA Returns

LUCA: Feedback

and Appeals

(6,8) Updates from LUCA

Appeals Outcomes

(3,4) Address and Spatial

Data for LUCA Feedback Products

ADC: (BARCA, LiMA) -

Incl. In-Office Address

Canvassing, MAFCS, In-Field

Address Canvassing

(1,2,3) Address

and Spatial Data for

ADC

(1,2) ADCAddress Adds &

Updates/ Deletes: In-Office, MAFCS

(3) ADC Address Adds &

Updates/Deletes: In-Field

RPO: Post Data Collection Phase

(CDL, DRPS) -Final Enumeration

Response Data (DRF, CUF, CEF)

RPO: Data Collection Phase

(ECaSE Survey OCS, SMaRCS, CaRDS, CDL) - Working Enumeration

Universe (incl. Supplemental)

RPO: Pre Data Collection Phase(CaRDS, ECaSE Survey OCS) -

Initial Enumeration Universe

(11) Final Collection and Tabulation

Universes

(6,8,10) Ongoing Address Adds/Updates/Deletes incl.

MAFX for Suppl. Univ.,MAFIDs for TempIDs

(5) EnumerationCases

(4) MAF Extracts (MAFX) (w/TEA Designations)

and Spatial Data

FPD

PDC

CQA (incl. CI)

NRFU:(ECaSE FLD OCS, ECaSE Enum) -

incl. Nonresponse Followup & Field Verification (FV)

NID:Post-Real Time

Phase(Async,Clerical)

NID:Real-time

Phase (RTNP)

ISR

UE

GQ

ETL

(7) Cases(n)

(7) Cases(n), Cases(i)

(6,8) Cases(i)

(8) NonIDRequests

(8) NonIDMatch/Miss

Results

(11) Final Case and Response Data

DPD

ARC

(12) FinalArchival

Response Data

RDP: Phases 3,4 & 5

CRO: Count & File

Review Phase(11) File

Review Inputs

Post-Enum. Survey Ops

(12) Final EnumerationResponse Data Inputs Ready for Tabulation

(1) Address and Spatial

Data for LUCA Outreach and

Participant Review

Materials

LUCA: Outreach,

Review and

Returns

(7) CQA AgentResponse (i,n)

(6) BCU Lists,Cases(i)

(7) GQ eResponse, Status, Address

Adds & Updates/Deletes

(10) Response,Status, Address

Adds & Updates/ Deletes

(9) NRFU Cases (i)

(9) FV Cases(n)

(7) Status, Address Adds & Updates/

Deletes

(7) PDC Response (i),Status

(12) FinalArchival Image Data

Administrative Records & 3rd

Party Data

(1,2,3) Ongoing Updates of Address and Spatial Data (incl. DSF, Geog.

Partnerships/Delineations, Other)

USPS DSFExternal Address and Spatial Data

SourcesGSS Data

(8) Occupied/ Vacant/Delete

Unit Information

Person Counts and

Characteristics

Demographic and

Geographic Data

(8,10) Address Updates/Deletes,

Address Adds with TempIDs

ID Respondent

(6) Paper Form (i) via UL

(7) Internet Response (i)(7) Telephone Response (i)

Pre Data Collection

Phase

Data Collection

Phase

Post Data Collection

Phase

CQRNonID Respondent

(7) Telephone Response (n)

(7) Internet Response (n)

IPC

(5, 6, 7, 8, 9, 10, 11, 12) IPC

Influences Public Knowledge and

Promotes Census Response

FACO

(12) Overseas Counts,Status

Demographic and

Geographic Data

(11) Administrative Records Data to

Support Tabulation

(6,8) Ongoing Updates of Address and Spatial Data (incl.DSF, New Construction, Other)

(11) Geographic Data to Support Tabulation

2020 Census Top-level End-to-End Operational Data Flow

Version v10c 11/14/2018

RDP: Phase 1

(1) Block Boundary

Suggestion Project Inputs

RDP: Phase 2

(3) Voting District Project Inputs

White House, Public

(6,8) MAF Info

cts

GQ Respondent

(7) Response ETL Respondent

(7) Response

NRFU Respondent(10) Response

Persons: Places: Things:

Office Workers Field Workers Respondents Public/Partners White House States NARA (National Archives)

Paper Materials

Internet Response

Partnerships/Communications

Automated Processing

ICON KEY

(7) Military TEAAddress/Response

(GQ+HU)

ID Cases = Cases(i)ID Responses = Response(i)

NonID Cases = Cases(n)NonID Responses = Response(n)

Frame:

Response Data:

Publish Data:

Test & Eval:

KEY

Record Sources:

GEOP Other

RPO Other

Respondent:Hi Mom!

PURPOSE: The 2020 Census Top-Level End-to-End (E2E) Operational Data Flow diagram illustrates a notional model for end-to-end data interactions between the 2020 Census

operations relating to frame management, data collection and data dissemination work. [Not depicted: Support Ops, Island Areas Censuses, Post-Enumeration Survey Ops details]

UL

(12) Final Tabulation Geography

(6,8) Status

(7) Paper ETL Response (i)

Via USPS

(7) Paper Response (i)

Via USPS

(2) LUCA Address

Validation Requests

(2) LUCA Address

Validation Responses

(11) Administrative Records Data to

Support Enumeration

UE Respondent

(7) Response

EAE

(5) Cases(i)

ATo

EAE

(6,8) Paper Notices and/or Paper Form (i)

B

A

(12) Final

Counts

(12) FinalCensus

Data Products

(8) GQ EnumerationRecords for CRO

BFrom EAE

21

22

23

Contact InformationRobin A. Pennington Robin.A.Pennington@census.gov

Michael Bentley Michael.Bentley@census.gov

Gina Walejko Gina.K.Walejko@census.gov

www.census.gov – Surveys/Programs – 2020 Census2020 Census Memo Series for 2020 Census Operational Plan, Detailed Operational Plans, and other important documents

24

Backup SlidesIntegrated Partnership and Communications

25

Step 1: Predictive models will estimate a household’s self-response propensity, the likely response mode, and the timing of response. Using Census Bureau and third-party data, we will be able to predict each Census tract’s probability to respond, enabling us to predict where low-response households are located, aggregate this information to larger geographies, and tailor outreach accordingly.

Step 2: Census Barriers, Attitudes, and Motivators Study (CBAMS) 2020 activities will help us understand the barriers, attitudes, motivators, and knowledge gaps that the communications campaign will need to address. The survey will be structured so that the findings can be integrated with small-geography level communications.

Step 3: We will group the households into segments based on their propensity to self-respond, their demographic characteristics, and our understanding of their attitudes based on responses to the CBAMS 2020 survey.

Step 4: We will plan how to reach each audience based on the segmentation data. For example, to reach a segment that is geographically concentrated, we can use local tactics like spot television, radio, events, out-of-home advertising, Spanish-language media, or earned media to reach them.

26

Step 5: We will design ads to specifically reach each audience. We will develop an overarching campaign platform, and then tailor initial designs for creative based on what we know about our audiences. Then, with our robust creative pre-testing, we will improve ads based on real feedback from a sample of people in that audience group.

Step 6: We will deliver the advertising, carefully monitoring the early results compared to predicted behaviors to identify geographic areas, audience segments, or demographic groups where response patterns are a concern.

Step 7: As households complete the census questionnaire, we will revise our models and gain efficiency by shifting resources to areas predicted to have the highest amount of remaining households who have not yet responded and limiting resources expended on geographies with high response rates. Data will inform us about what the best messages and modes to reach the segments are.

Step 8: The rapid response team will address real-time issues. We will review a daily response report among key demographic and geographic groups. We will coordinate with Field and Partnership teams to prioritize audiences and align messages.

27