Arona L. PistinerInternational Relations Officer, U.S. Census BureauUnited Nations Economic Commission for EuropeConference of European StatisticiansGroup of Experts on Population and Housing Censuses Palais des Nations, Geneva30 September – 2 October 2015
Changing Times, Changing Methods, Changing Technologies:
Innovation and Testing for the United States 2020 Population and Housing Census
1
One Chance to get the 2020 Census Right
2
3
Four Major Innovation Areas to Reduce the Census Cost
4
2020 Census Innovation Areas and Testing
Reengineering Address Canvassing
Optimizing Self‐Response
Utilizing Administrative Records
Reengineering Field Operations
2012 National Census Test
X
2013 Census Test X 2014 Census Test X X 2015 Address Validation Test
X
2015 Optimizing Self Response Test
X
2015 Census Test X X
2015 National Content and Self Response Test
X
Geographic Support System Initiative
X
Decennial Modeling and Analysis
X X
ROCkIT Simulations X 2012 Gallup Poll X American Community Survey
X X X X
Optimizing Self-Response
What are we talking about? Giving people more options to respond to the Census, so we do not
have to knock on their doors. Internet is cheaper than paper, paper is cheaper than a field
operation.
What are we researching and developing? Testing ideas to measure how much we can increase self-
response. Researching new techniques such as email, text, social media, and
leveraging the Internet to give the respondent options. Testing multi-language support.
5
6
2020 Census Innovation Areas and Testing Optimizing Self-Response
7
Assumptions:
Online response rate will be about 50-55 percent.
We will include paper questionnaires using a mail-out/mail-back approach.
Will result in a reduction of paper data capture operations workload and field infrastructure as compared to the 2010 Census.
Approach has potential savings of US$548 Million.
8
Key Design Components:
Employ the use of a pre-registration phase (with a tool called “Notify Me”).
Allow respondents to answer the 2020 Census Questionnaire without an identification code (“Non-ID” processing).
• Option 1 is to provide on-line tools that allow real-time matching and geocoding at the time of questionnaire completion.
• Option 2 is to conduct batch processing of matching and geocoding at specific periods during the day.
9
Key Research Questions What are the best methods for communicating the
importance of responding to the 2020 Census as measured by public response?
What is the estimated self-response rate? What are the response rates for different modes?
What infrastructure is necessary to support the Internet as the primary mechanism for self-response?
Is there value in asking households to pre-register for the census?
Is it necessary to provide households with an identification code to respond via the Internet?
Reengineering Field Operations
What are we talking about? Using technology, data, and GPS to collect
interviews efficiently.
What are we researching and developing? Streamlining data collection using smart phones
and tablets to lower costs. Managing through technology not brick-and-mortar
field offices. Automating logistics to increase
workforce efficiency.
10
11
2020 Census TestingReengineering Field Operations
12
Assumptions Increase Non-response Follow-up (NRFU) productivity by 20 percent
using automation. Remove late responses from the NRFU workload. Reduce the total number of local census offices (LCOs) by 5 percent. Reduce the total square footage of LCOs by 70 percent. Reduce the number of clerical staff by 20 percent. Redesign the training strategy to reduce enumerator-training hours by 35
percent. Establish a lower training pay rate than the production pay rate. About 75 percent of enumerators will use bring their own device (“Bring
Your Own Devise”). Reduce the phone/personal visit contact cycle (relative to the 2010
Census) from a maximum of six contacts to a maximum of three contacts. Use adaptive design (routing and dynamic case management) to allocate
resources efficiently. Approach has the potential savings of US$2.3 Billion.
13
Key Research Questions How can the Census Bureau reduce the cost of field
operations with automation? What is in the model for reengineered field operations? How will the field staff structure change? How will we manage interaction with employees? What is the optimal workload assignment for field staff? Where will workload management occur? What type of systems and devices will be used for:
• Case assignment • Routing and navigation • Data collection• Administrative activities
Will the Census Bureau build or buy these systems? How will the employees obtain the handheld devices for
enumeration?
14
Key Design Components Reengineer the field work:
• Provide enumerators with daily, optimal, and sequence contact attempt assignments.
• Provide supervisors with electronic access to enumerator workload and status information.
• Automate training.
Reengineer the field staff structure:• Redefine field staff roles.• Change how we think about work schedules.• Change staffing ratios.
15
2014 Census Test
We tested:
Notify Me” – the use of pre-registration of emails and phone numbers in order to enable email and automated voice invitations to participate.
The use of the Internet and Telephone as response options.
The use of mobile devices and a “data collection app” for conducting field interviews.
The use of alternative contact strategies for non-responders, such as telephone calls or personal visits.
The use of administrative data to remove cases form the field visit workload (both vacant and occupied housing units).
16
2014 Census Test - Continued
Preliminary findings include:
A self-response rate of 70.9 percent.
Internet self-response rate of 58.3 percent.
Email invitations or email reminders were not an effective replacement for postal mail.
Enumerators were successful in using the smartphone and the “data collection app” for nonresponse follow-up.
There were potential benefits of using administrative data for enumeration.
17
2014 Census Test - Continued
We also learned some valuable lessons:
Using email as an initial invitation and reminder contact strategy was not an effective replacement for traditional mail pieces.
Also, the short time-frame we allowed between mailings limited our ability to
exclude households who responded from subsequent mailings.
We also experienced a much higher call volume to our Telephone Questionnaire Assistance Center than we projected.
18
Overview of the 2015 Tests
The Address Validation Test enabled us to assess the performance of the methods and models that will help us develop the 2020 Census address list and develop workload models.
The 2015 Census Tests conducted in Maricopa County, Arizona and Savannah, Georgia allowed us to employ a variety of new methods and advanced technologies that are under consideration for the 2020 Census.
The 2015 National Content Test will aid us in obtaining a nationwide measure of response rates for Internet and self-response.
19
2015 Census Tests – Maricopa County, Arizona
2015 Census Test
We are testing the reengineering of the field roles, responsibilities, and infrastructure for conducting field data collection.This includes: Reengineer the roles, responsibilities, and infrastructure for the field. Automate tasks traditionally performed by people. Test and implement routing and navigation. Test a Bring Your Own Device component. Determine how administrative data and adaptive design can reduce the
NRFU workload and determine contact strategies. Three panels: Control, Hybrid Administrative Records Panel, Full
Administrative Records Panel. Using focus groups to explore reactions to the contact methods, use of
administrative data, privacy and confidentiality concerns, and how the Census Bureau might address those concerns.
20
Maricopa County, Arizona - Continued
Preliminary findings include:
COMPASS:• Application was easy to use.• Experienced crashes and freezes of the application; further investigation into
root causes is needed.• Coverage questions added to respondent burden.
Field Test Procedures:• Work needed to define a coordinated approach to enumeration within multi-
units and gated-communities.• Refinement to data collection application “pathing” needed to better assist
enumerators in cases on proxy responses and non-interviews.
Bring Your Own Device:• Training was fairly labor intensive.• Based on observations, no adverse respondent reactions to the device being
used for data collection.
21
2015 Census Tests – Savannah, Georgia
Optimizing Self-Response Test
The scope of the test includes: Conduct early research on the use of advertising an outreach to
engage and motivate respondents. Determine the utility of an early engagement, pre-registration
option. Understand the operational feasibility of real-time Non-ID
processing as well as its impact on self and internet response rate, and the system workloads generated by it.
Research how advertising, outreach, and promotion can engage and motivate respondents for action.
Three initial mail panels designed to determine if we can optimize self and internet response rates: “Internet Push,” “Notify Me,” and “Non-ID.”
22
Savannah, Georgia- Continued
Preliminary findings include:
The weighted total response rates of the three panels were: Internet Push – 47.5 percent Non-ID Internet Push – 44.4 percent Notify Me Postcard – 46.3 percent
Low Participation in “Notify Me:” 1,925 participants pre-registered
Self-Response among Non-Sampled Housing Units: An additional 35,249 Internet responses from housing units not selected in mail
panels Response attributed to advertising and promotional efforts
Successful implementation of real-time Non-ID processing: Matched 98.5 percent of cases
New postcard panel generated response of approximately 8 percent.
23
2016 Tests:Harris County, Texas and Los Angeles, California
Refine technologies and methods for assigning cases to field staff conducting nonresponse follow-up and continue to refine the use of mobile technology for data collection and case management.
Continue to research how best to use existing government and commercial data to reduce the follow-up workload (using data to identify vacant units or to reduce additional visits for household we are unable to contact).
Study partnership and outreach efforts for communicating with historically hard-to-count populations, including providing assistance in multiple languages.
Continue to explore self-response options, including how people respond using the Internet, mobile devices, the telephone, and paper questionnaires.
Continued refinement of methods to process responses in real time submitted without a unique Census-provided address identification code (non-ID processing).
24
2017 Testing and Beyond
The 2017 Test will focus on address canvassing.
In 2018, we will conduct operational readiness testing, including an end to end test of census operations and systems.
In 2019, our early 2020 operations begin.
25
Summary
The better part of this past year we have spent narrowing our efforts to lead us to our design decisions.
26
The 2020 U. S. Census will be fundamentally different from prior censuses to meet our goals and the changing needs of the American public.
Major design decisions for the 2020 Census will be announced October 6, 2015.
26
Questions
27
Top Related