Full text of the official published hearing record. Extracted from the source document — verify against the official record for citation.
- 2020 CENSUS: OUTCOMES OF THE 2016 SITE TESTS [House Hearing, 114 Congress] [From the U.S. Government Publishing Office] 2020 CENSUS: OUTCOMES OF THE 2016 SITE TESTS ======================================================================= HEARING BEFORE THE SUBCOMMITTEE ON GOVERNMENT OPERATIONS OF THE COMMITTEE ON OVERSIGHT AND GOVERNMENT REFORM HOUSE OF REPRESENTATIVES ONE HUNDRED FOURTEENTH CONGRESS SECOND SESSION __________ NOVEMBER 16, 2016 __________ Serial No. 114-169 __________ Printed for the use of the Committee on Oversight and Government Reform [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Available via the World Wide Web: http://www.fdsys.gov http://www.house.gov/reform ______ U.S. GOVERNMENT PUBLISHING OFFICE 26-171 PDF WASHINGTON : 2017 ----------------------------------------------------------------------- For sale by the Superintendent of Documents, U.S. Government Publishing Office Internet: bookstore.gpo.gov Phone: toll free (866) 512-1800; DC area (202) 512-1800 Fax: (202) 512-2104 Mail: Stop IDCC, Washington, DC 20402-0001 COMMITTEE ON OVERSIGHT AND GOVERNMENT REFORM JASON CHAFFETZ, Utah, Chairman JOHN L. MICA, Florida ELIJAH E. CUMMINGS, Maryland, MICHAEL R. TURNER, Ohio Ranking Minority Member JOHN J. DUNCAN, Jr., Tennessee CAROLYN B. MALONEY, New York JIM JORDAN, Ohio ELEANOR HOLMES NORTON, District of TIM WALBERG, Michigan Columbia JUSTIN AMASH, Michigan WM. LACY CLAY, Missouri PAUL A. GOSAR, Arizona STEPHEN F. LYNCH, Massachusetts SCOTT DesJARLAIS, Tennessee JIM COOPER, Tennessee TREY GOWDY, South Carolina GERALD E. CONNOLLY, Virginia BLAKE FARENTHOLD, Texas TAMMY DUCKWORTH, Illinois CYNTHIA M. LUMMIS, Wyoming ROBIN L. KELLY, Illinois THOMAS MASSIE, Kentucky BRENDA L. LAWRENCE, Michigan MARK MEADOWS, North Carolina TED LIEU, California RON DeSANTIS, Florida BONNIE WATSON COLEMAN, New Jersey MICK MULVANEY, South Carolina STACEY E. PLASKETT, Virgin Islands KEN BUCK, Colorado MARK DeSAULNIER, California MARK WALKER, North Carolina BRENDAN F. BOYLE, Pennsylvania ROD BLUM, Iowa PETER WELCH, Vermont JODY B. HICE, Georgia MICHELLE LUJAN GRISHAM, New Mexico STEVE RUSSELL, Oklahoma EARL L. ``BUDDY'' CARTER, Georgia GLENN GROTHMAN, Wisconsin WILL HURD, Texas GARY J. PALMER, Alabama Jennifer Hemingway, Staff Director Andrew Dockham, General Counsel Jeffrey Post, Government Operations Subcommittee Deputy Staff Director Patrick Hartobey, Counsel Sharon Casey, Deputy Chief Clerk David Rapallo, Minority Staff Director Subcommittee on Government Operations MARK MEADOWS, North Carolina, Chairman JIM JORDAN, Ohio GERALD E. CONNOLLY, Virginia, TIM WALBERG, Michigan, Vice Chair Ranking Minority Member TREY GOWDY, South Carolina CAROLYN B. MALONEY, New York THOMAS MASSIE, Kentucky ELEANOR HOLMES NORTON, District of MICK MULVANEY, South Carolina Columbia KEN BUCK, Colorado WM. LACY CLAY, Missouri EARL L. ``BUDDY'' CARTER, Georgia STACEY E. PLASKETT, Virgin Islands GLENN GROTHMAN, Wisconsin STEPHEN F. LYNCH, Massachusetts C O N T E N T S ---------- Page Hearing held on November 16, 2016................................ 1 WITNESSES The Hon. John H. Thompson, Director, U.S. Census Bureau Oral Statement............................................... 5 Written Statement............................................ 7 Mr. Kevin B. Smith, Associate Director for Information Technology and Chief Information Officer, U.S. Census Bureau Oral Statement............................................... 29 Written Statement............................................ 31 Mr. David Powner, Director of Information Technology Management Issues, U.S. Government Accountability Office Oral Statement............................................... 42 Written Statement............................................ 44 Mr. Robert Goldenkoff, Director, Strategic Issues--Census Issues, U.S. Government Accountability Office Oral Statement............................................... 61 Written Statement............................................ 63 2020 CENSUS: OUTCOMES OF THE 2016 SITE TESTS ---------- Wednesday, November 16, 2016 House of Representatives, Subcommittee on Government Operations, Committee on Oversight and Government Reform, Washington, D.C. The subcommittee met, pursuant to call, at 10:11 a.m., in Room 2247, Rayburn House Office Building, Hon. Mark Meadows [chairman of the subcommittee] presiding. Present: Representatives Meadows, Buck, Carter, Grothman, Connolly, Maloney, and Norton. Mr. Meadows. The Subcommittee on Government Operations will come to order. And, without objection, the chair is authorized to declare a recess at any time. Before I go into my opening statement, Director Thompson, I want to just say thank you to you and your team who not only made a special effort to come to my district with regards to some of the onsite testing. It was very illuminating. And I just want to say thank you for making the special effort to help me understand that process. So kudos to you and your staff in a very busy time for making the time to help us be better informed. The 2020 Census will be unlike any others in 2020-plus years since the first census happened in 1790, and for the first time the Census Bureau will be using technology ranging from mobile devices to Internet self-response to carry out one of the most important tasks for the Federal Government, counting the people of this great Nation. The use of this new technology and modernization has the potential to create tremendous cost savings for the American taxpayer. Along with this great promise, however, the Bureau's effort also can carry significant risk. For example, if the testing and delivery schedules are not met or is strictly adhered to, instead of the cost savings, there could be cost overruns. We've seen that in previous censuses, so this is not our first rodeo as we look at this. In prior hearings, and actually in prior meetings that we've had, the committee has identified a number of areas where the Bureau is at risk of cost and schedule overruns, perhaps incomplete census data, and information security breaches or the potential for that. The time is running out to address those. I think we all realize that we're on a critical time mission. And, again, I want to acknowledge the fact that as we have come together, seen those issues, it is critically important that we keep all hands on deck. I believe that even though we may have a difference of opinions in terms of the severity of the challenges that face us, that we're committed to do that. And I want to acknowledge that the last meeting that we had here, which was not a public hearing where many of you here, we had frank discussions and concerns, and yet at the same time, I think we're able to go apart to say that we're committed to making this thing happen. And I'm optimistic that we will be able to do that. But only by correcting some of the issues that we have will we be able to make sure that the new administration and the 2020 Census will be placed in a scenario for success. Decisions that we're making now, quite frankly, will have an impact for the next 4 years. And so it is critically important that we make good decisions as we start to implement a number of these areas. I have some serious concerns, as Director Thompson knows, with the Bureau's contracting schedule for the 2020 systems, as well as their commitment to making those 2020 commitments the highest priority for those programs. The Bureau's modernization effort entails a design, building, delivery, and implementation system of some 52 different systems. Any time that you have that many moving parts, it can be troubling. Each of these systems should be field tested in their final form before being rolled out in the 2020 Census. Today, obviously, we'll hear from the GAO that with 8 months left before the final testing, half of those 50 systems to be tested in 2018 will either be delivered after the start of the testing or, worse yet, their delivery date is still unknown. The fact that there is even a suggestion that IT products will go untested is unacceptable, and no system or product, not a single one, can be allowed to be used to collect and process the American public's sensitive personal information without first being tested. Despite the need to test more than 50 systems, the Bureau recently announced the cancellation of the 2017 field testing operations. And so I look forward to hearing on how we can maybe combine some of those as we start to address some of the issues, whether it be testing in a different mode, or how we make sure that there is that integrity. The Bureau has claimed that budgetary constraints are preventing them from this testing; however, when I start to look at some of the numbers, I guess my concern is that it seems like the Bureau has opted on its own to kind of reduce the 2020 budget by more than $120 million, which would have an impact now. And so I would like to get some greater clarity on why that decision was made, what we need to do there. No other program in the Bureau's portfolio is being cut anywhere near that amount. And so I guess, to give you an example, the American Community Survey, which is probably the single thing we get the most complaints about as it relates to the census because of the nature of the questions, that program is only being reduced by a mere $3 million. So I want to see where our priorities are as we start to look at it, Director. The 2020 program should be the highest priority program. It should not be shouldering roughly 75 percent of any anticipated budget reductions. I'd like to thank all of the witnesses for being here today. I believe that we will be able to make some real progress. And with that, I'll recognize the ranking member, my good friend, Gerry Connolly. And congratulations on your reelection. You did that in spite the fact that I endorsed you. Mr. Connolly. We have a mutual pact. If it helps him, I'll campaign against him or for him back home, and he's offered to do the same for me. That you, Mr. Meadows, and congratulations to you as well. I think this is an important hearing to examine the Census Bureau's 2016 test for the all-important 2020 Decennial Census. The decennial census is a cornerstone of our constitutional system. It is used to apportion seats here in the House, to define State legislative districts, and determine school district assignments. Census data helps the private sector make sound investments by identifying unsaturated or emerging growth markets and developing business plans and loan applications. The data also helps Congress make decisions in how to fund special education grants, for example, or provide adoption assistance, or create small business development centers, rural business enterprise grants, and other programs. Very important to the U.S. economy and its growth. The 2020 Decennial Census will be the first to embrace information-collection technology on a broad scale. According to the Bureau's calculations, the Bureau's information technology plans will make this census leaner and result in $5.2 billion in savings compared to the last census. But those savings can only be effectuated if the Bureau's plans are fully implemented and operate without interruption, and if the Bureau is able to accomplish its constitutional mandate and accurately count all residents. The Government Accountability Office--I'm glad to see Mr. Powner here again, thank you--is advising Congress that it has concerns. They tell us that 25 percent of the households in a recent field test could not be contacted by Bureau enumerators even after six attempts. For example, large multi-unit buildings and locked or gated communities were problematic, as one might imagine, as enumerators were unable to enter the property. The Bureau's software also made it difficult for enumerators to leave notes, which would help indicate what time of day a dwelling's resident was likely to be present or not. The Bureau also had difficulty in creating enumerated canvas lists from administrative data, and enumerators had training difficulties which resulted in diminished performance. Our goal in this hearing is to ensure that the Bureau answers the concerns of the GAO and learns the lessons from its fields tests and shares them with us and fixes problems in advance of the census itself. This subcommittee has a role to play in a successful census, and we're committed to a successful census on a bipartisan basis. We must ensure the Bureau continues to make progress in the next 3 years, improve its policies, procedures, and technology based on information gained in those field tests, and test and retest those improvements before the 2020 Census. In October, I joined Chairman Chaffetz and Ranking Member Cummings and Chairman Meadows of this subcommittee in inquiring on the status of the critical IT decisions related to the 2020 Census that had yet to be made and the Bureau's overall preparedness for the census. We were pleased to receive a quick response and will continue to work with the Bureau to ensure that the CEDCaP initiative stays on schedule. The Bureau has been making progress. It recently hired a new chief information officer, Mr. Kevin Smith, who is with us today. The presence of Mr. Smith in an office that has been without a permanent leader for nearly a year is, finally, a welcome sign. While he has only been in the position for a few months, I look forward to hearing the steps he's taken to improve the Bureau's IT infrastructure and how he proposes to implement FITARA, also known as Connolly-Issa. Mr. Meadows. I agree. Mr. Connolly. Again, I want to thank you, Chairman Meadows, for calling this hearing and keeping us focused on the importance of this subject matter. With that, I yield back. Mr. Meadows. I thank the gentleman. And I will hold the record open for 5 legislative days for any member who would like to submit a written statement. We'll now recognize our panel of witnesses. I'm pleased to welcome the Honorable John Thompson, director of the U.S. Census Bureau. Welcome, Director Thompson. Mr. Kevin Smith, the chief information officer at the U.S. Census Bureau. Welcome, Mr. Smith. Mr. David Powner, director of information technology management issues at the U.S. Government Accountability Office. Welcome. And Mr. Robert Goldenkoff, director of strategic issues and census issues at the U.S. Government Accountability Office. Welcome. And welcome to you all. Pursuant to committee rules, all witnesses will be sworn in before they testify. So if you'll please rise and raise your right hand. Do you solemnly swear or affirm that the testimony that you are about to give will be the truth, the whole truth, and nothing but the truth? Let the record reflect that the witnesses answered in the affirmative. You may take your seat. And in order to allow time for discussion, I would ask that your oral testimony be limited to 5 minutes. However, your entire written testimony will be made part of the record. And so, Director Thompson, we'll now recognize you for 5 minutes. WITNESS STATEMENTS STATEMENT OF JOHN H. THOMPSON Mr. Thompson. Good morning, Chairman Meadows, Ranking Member Connolly, and members of the subcommittee. I appreciate the opportunity to update you on the 2020 Census. In June, I testified that we were on track to execute a census that is innovative, efficient, and accurate. Since then, I've appreciated the many opportunities to engage further with you, Chairman Meadows, as well as other committee members and committee staff. We've continued to make progress in our preparations, and I am proud to report that we remain on track and on schedule. The last time I testified, we discussed the Census Bureau's Integrated Master Schedule for the 2020 Census and we provided the committee with a copy of that schedule. Our operational plan for 2020 includes 350 design decisions. As of right now, 259 of them, or 74 percent, have been made, and we are on schedule for making the rest. We update the GAO on the schedule every month, and I would be happy to provide those updates to you and your subcommittee as well to further assure you that we are ready and on time with those systems and operations. As described in my written testimony for the record, the Census Bureau is pursuing four key innovation areas that will make it easier for people to participate and save taxpayers more than $5 billion. And census tests are critical as we work toward to implement these innovation areas. We learn what works and what doesn't, and we make adjustments. Most recently, we completed our 2016 Census Test in Harris County, Texas, and in Los Angeles County, California. We tested core census operations, including how we process, store, and protect the data we receive from respondents. As I said, we learned many lessons from the 2016 Census Test, including several notable successes and insights, such as higher response rates by using a letter as the first reminder and using language services in brochures and inserts. As in past census tests, we successfully matched the large majority of respondent addresses to our address frame with non-ID processing methods. We expanded language support services to include Chinese and Korean. We used administrative records and third-party data to reduce the nonresponse follow-up work load for vacant and occupied addresses. We implemented staffing ratios that increase the number of enumerators per supervisor. These staffing ratios were successful due to increased automation. We worked with the U.S. Postal Service to reduce our nonresponse workload by gaining a better understanding of mail processing, including reasons why mail can't be delivered. And finally, we improved optimized assignments and routing for our enumerators and our use of smartphones for data collection. Just as important as the successes, we identified areas that need improvement. These areas, which are described in detail in my testimony, include better training for enumerators, better procedures for enumerators in multi-unit structures, enhancements due to proxy interview process, and continued development of closeout procedures for data- collection operations. We are already implementing what we learned from the 2016 Census Test and will continue to do so to prepare for the 2018 End-to-End Test. In addition, we began testing our address canvassing procedures and systems in parts of Buncombe County, North Carolina, and St. Louis, Missouri. And thank you to Chairman Meadows for observing our field operations earlier this month in Buncombe County. We're testing methodologies and data sources to detect new residential developments to show neighborhoods that remain unchanged since the 2010 Census. In addition to the address canvassing test, the Census Bureau had been planning field testing operations for 2017. Due to funding uncertainty, on October 18, the Census Bureau was forced to announce that we're stopping work on two planned field tests in 2017 in Puerto Rico and on two tribal reservations. Stopping these tests is not an ideal outcome for the operational risk of the 2018 End-to-End Test and the 2020 Census, but overall it's the best option given the funding uncertainty for fiscal year 2017. We will continue to prioritize our funding resources and activities that are critical to preparing for the 2018 End-to- End Census Test. I should also note, and as described in my written testimony, we've awarded four significant contracts relating to the census. I'd be happy to discuss these later. In the last three months, we've met key milestones for the 2020 Census, and we remain on track and on time. I thank the subcommittee for your continued support and interest in our work. I am confident the Census Bureau will achieve its goal of counting everyone in America once, only once, and in the right place in 2020. And I look forward to answering your questions. Thank you. [Prepared statement of Mr. Thompson follows:] [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Mr. Meadows. Thank you, Director Thompson. And thank you for keeping the dogs away from both of us when we were out there doing the field test. Mr. Smith, you are recognized for 5 minutes. STATEMENT OF KEVIN B. SMITH Mr. Smith. Chairman Meadows, Ranking Member Connolly, and members of the subcommittee, thank you for the opportunity to testify this morning. I am pleased to update you on our ongoing work to make the 2020 Census a success. I began my role as chief information officer at the Census Bureau over 4 months ago. Since then I've gained a deep appreciation of the steps the Bureau has taken to introduce innovations into their operation processes and supporting technology to make the 2020 Census more effective and efficient than ever before. The Census Bureau has taken positive steps to innovate with technology, not invent technology, by using industry available solutions wherever possible. I have been involved in developing first-of-their-kind industry solutions and large-scale technology modernization efforts in both government and in the private sector. I'm excited and honored to lead the IT team that will help deliver the most automated census ever. I have spent much of my time at the Census Bureau reviewing the IT organization and the Bureau's technology and governance. We have a talented staff of IT professionals, and the foundation for a successful census is in place. I am particularly pleased with the start of our new CEDCaP chief security engineer in July. He has a rich background in information security, a strong history of private sector experience, and a wealth of knowledge about securing federal information systems. I'd like to focus today on two key areas, our cybersecurity and the progress of the Census Enterprise Data Collection and Processing program, known as CEDCaP. We take the task of ensuring the integrity and security of our systems and data against the evolving landscape of cybersecurity threats very seriously. Our cybersecurity approach will ensure that individuals have limited, appropriate access to 2020 Census data. We'll keep the census systems running through possible cyber threats by taking informed actions in redundant systems with layers, isolation, views, and encryption whenever and wherever possible. We're committed to partnering with industry and other Federal agencies to leverage their cybersecurity services and to create as strong a cybersecurity posture as possible. In particular, we have engaged the Department of Homeland Security to review our design and security architecture for the 2020 Census systems. In addition, we have reached out to the National Security Agency and other offices within the Department of Commerce for assistance in threat identification and management. During the 2020 Census, we'll focus on phishing threats through respondents and Census Bureau employees. For respondents, we will work closely with our Integrated Communications Contract to address the threat of phishing both before and during the 2020 Census. Within the Census Bureau, we conducted regular cybersecurity training and phishing evaluations to educate our employees. We have also worked with DHS to evaluate and test our procedures for phishing. We will use an industry-leading solution to detect rogue and suspect emails and Web sites impersonating the Census Bureau. And finally, we will leverage the security protections of our cloud service providers for additional layers of security. The CEDCaP program has been underway since 2015. Much work remains. In May, the Census Bureau announced the decision to unify the functions of a number of existing systems onto a single platform-based COTS solution. The transition to the COTS platform for the 2020 Census began in June. The CEDCaP systems are built using an agile development approach that allows us to work closely with the decennial business product owners. This approach ensures that we deliver the right level of technology incrementally and make the most effective improvements for the census. We are hard at work to deliver the new version of the previously tested Internet self-response capability in the 2017 Census Test. Overall, for the 2018 End-to-End Test, half of the systems needed are already in use and going through enhancements. Over half of the needed systems will be ready for system-integration testing within the next 6 months. Looking ahead, I believe the measures we are taking will prepare us to react to any difficulty, including the threat of cyber attacks. Preserving the integrity and security of our systems and data is a top priority. First, we will protect the data we collect. And, second, we will sustain secure data-collection services so that respondents may confidently respond to the 2020 Census. To do that, we will layer our technology in ways that isolate data and systems from each other with the views that let us take immediate action when a threat is detected. This approach puts in place the majority of the technical security controls that we need to protect the data and systems from threats. This allows for systems to inherit best practice technical security within the framework while we continue to develop the systems. GAO recently expressed the importance of engaging service providers and the IT vendor community to help mitigate risks, and I completely agree with their recommendation. With the continued hard work from our team and congressional support, I am confident the Census Bureau can achieve our objectives. I look forward to discussing other aspects of our planning for the 2020 Census with you. I am grateful for this opportunity to testify before the subcommittee, and I am pleased to answer any questions you have. Thank you. [Prepared statement of Mr. Smith follows:] [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Mr. Meadows. Thank you, Mr. Smith. Mr. Powner, you are recognized for 5 minutes. STATEMENT OF DAVID POWNER Mr. Powner. Chairman Meadows, Ranking Member Connolly, and members of the subcommittee, thank you for inviting us to testify on the Bureau's plans to deliver tests and secure key technologies for the 2020 Census. Clearly, utilizing an Internet response, mobile devices for enumeration follow-up, and cloud solutions are important steps to improving our Nation's response rates and securing citizens' data. However, the Bureau has a history of poor IT delivery, and many of the broader IT initiatives this committee has focused on during this Congress are major challenges at the Bureau, including appropriate CIO authority, delivering incrementally, having the right governance and accountability over IT acquisitions, and securing systems and information. This morning, I'd like to focus on what needs to be done to deliver and secure these technologies. First, integrate IT deliverables with the 2020 schedule. My colleague, Carol Harris, made this point repeatedly before this committee at prior hearings, and this is still clearly a major issue since several of the IT schedules are still being developed. Clearly, schedule risk is a bright red risk at this point. Oversee delivery and readiness of the 50-plus systems. The Bureau needs to deliver about 50 systems to be included in the 2018 End-to-End Test starting in August of next year. I have a figure in my written statement that shows half of those systems are to be delivered after the start of the End-to-End Test or that they lack firm delivery dates. Clearly, the 12 projects associated with CEDCaP, including the Internet response capability, mobile devices, and the centralized operations component are critical systems. A key question with CEDCaP commercial products is the amount of customization that the Bureau plans. Modifying these products will have cost and schedule implications that we plan to monitor closely. Next, they need to oversee integration activities. An integration contractor recently began work in October. The contractor is responsible for delivering and implementing a decennial system of systems. The staffing plan here includes hundreds of contractors and a cost of about $900 million. The contractor acknowledges the key risks are schedule, scalability, and security. Currently, they are working on detailed schedules. It is very important for the Bureau to oversee this contractor's work and the associated costs. Currently, the Bureau is expanding its program management office by about 40 staff to do so. We think this is very important, but also very late in the game to be doing so. Next, they need to make decisions on infrastructure. The integration contractor is working on an analysis of infrastructure needs. Decisions need to be made on data centers and security operation centers. Next, they need to improve IT governance. Overseeing the delivery of the 12 CEDCaP systems, the additional 40 systems, and integration and testing activities is extremely important. The Bureau has several governing boards to do this, including the CEDCaP in 2020, executive steering committees. Most of this is the direct responsibility of the 2020 office, not the CIO. This is another instance, Mr. Chairman, where the delivery of key systems is not under the direct authority of the CIO. Mr. Smith, though, to his credit, has assured us that he can manage risks associated with IT through these governing bodies. We hope he is right, and we will be continuing to monitor this for the committee. To do so, we'll need meeting minutes of these governance bodies, and even better yet, we'd like to be observers at key governance meetings. To the Bureau's credit, we like agile development, commercial products, and using an integration contractor that does include subcontractors with prior census experience. However, Mr. Chairman, I'd like to point to a hearing you and Chairman Hurd, along with Ranking Member Kelly, held this summer on 18F and the digital service teams, and I wonder why these private sector experts are not engaged with the decennial IT deliverables. Clearly, the Bureau could use help in this area. Turning to security. The Bureau needs to continue its diligence in this area, since it has been the target of recent data breaches; although, it has been reported that no personally identifiable information was compromised. Moving forward, we have ongoing work that will focus on the Bureau minimizing the threat of phishing, securing 300 mobile devices, securing cloud services, and properly configuring the 2020 systems. A key oversight area for this committee will be the authority to test and operate the many systems associated with the decennial. In conclusion, there is a lot to do over the next 9 months leading to the End-to-End Test, and not everything is completely decided or defined. Decisions, delivery, integration, and testing needs to occur and needs to be closely monitored. In addition, the IT costs of the decennial needs much more transparency. Right now, the total IT cost for the decennial is not clear, but we will be getting you solid numbers with our upcoming review. Look forward to your questions. [Prepared statement of Mr. Powner follows:] [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Mr. Meadows. Thank you so much. Mr. Goldenkoff, you are recognized for 5 minutes. STATEMENT OF ROBERT GOLDENKOFF Mr. Goldenkoff. Chairman Meadows, Ranking Member Connolly, and members of the subcommittee, thank you for the opportunity to be here today to discuss the Census Bureau's preparations for the 2020 Census. As you know, to help control costs and maintain the accuracy of the next enumeration, the Bureau is planning to employ new procedures and technology that have not been used to a large degree in prior decennials, if at all. While the Bureau estimates that its planned innovations can save as much as $5.2 billion compared to a repeat of the approach it used in 2020, those same innovations also introduce new risks. As a result, it will be important to thoroughly test the operations plan for 2020 to help ensure that they will produce needed cost savings, function in concert with other census operations, and work at the scale needed for the national head count. The Bureau's failure to fully test some key operations prior to the 2010 Census was a key factor that led us to designate the 2010 Decennial a GAO high-risk area. In my remarks today, I will provide some preliminary observations on two such tests occurring this year: the Bureau's test of nonresponse follow-up operations in Los Angeles County, California, and Harris County, Texas, and the ongoing test of address canvassing procedures in Buncombe County, North Carolina, and in St. Louis, Missouri. I will also discuss some key lessons learned from the 2010 Census that can be applied to the Bureau's preparations for 2020. The 2016 test of nonresponse follow-up operations where enumerators visited households that did not respond to the census generally proceeded according to the Bureau's operational plans. However, preliminary data at both the Harris County and L.A. Test sites indicate that the Bureau experienced a large number of non-interviews, as much as 30 percent of the workload, where either no or insufficient data were collected. Bureau officials are not certain why there were so many non- interviews and are researching the potential causes. Going forward, it will be important for the Bureau to identify and address the factors that contributed to the non- interview rate, as they could have implications for the cost and accuracy of the final population count. Our preliminary observations also revealed that refining some key enumeration procedures could produce additional efficiencies by enabling the Bureau to be more responsive to situations enumerators encounter on the ground. These refinements include providing more flexible access to recently closed, incomplete cases, improving communication with managers of multi-unit properties to better coordinate enumerators' visits, and improving the ability to update the automated case management system with information on the best time of day to visit households. Meanwhile, in Buncombe County, North Carolina, and in St. Louis, Missouri, the Bureau is testing new procedures for building its master address list for 2020. The Bureau plans to employ aerial imagery, street imagery, and address file data from State, local, and tribal partners to update address and map information. The goal is to limit the more expensive and traditional door-to-door address canvassing only to those areas most in need of updating, such as areas with recent housing growth. The Bureau anticipates that by using this new or targeted approach, just 25 percent of housing units will require field canvassing. While this initiative shows promise for controlling cost, the Bureau has identified a series of risks that could affect its cost or quality. These risks include, for example, locating hidden housing units, such as converted garages and basements, and monitoring changes in housing stocks. In summary, while the Bureau has made considerable progress in developing more cost-effective enumeration procedures, a number of risks and uncertainties remain and key lessons learned in planning for the 2010 head count can enhance the Bureau's readiness for 2020. Based on our prior work, these lessons include ensuring key census activities are fully tested, developing and managing the enumeration on a basis of reliable cost estimates, and sustaining workforce planning efforts to ensure the Bureau has the optimal mix of skills to cost effectively conduct the enumeration. Chairman Meadows, Ranking Member Connolly, this concludes my remarks. I would be pleased to respond to any questions that you or other members of the subcommittee might have. [Prepared statement of Mr. Goldenkoff follows:] [GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT] Mr. Meadows. Thank you. Thank you all for your testimony. So the chair recognize himself for a series of questions. So, Mr. Smith, let me come directly to you, since it seems that a lot of this is following in your area. And Mr. Powner talked about your assurances with regards to a couple of areas where you don't have the direct authority but believe that you can deliver. How do you do that when things go wrong? And I guess my concern is, is when everything is going right, there's enough people around to take credit. When they go wrong, you know, then all of a sudden there is a lot of finger pointing. And so should we address that now to make sure that you have the ultimate authority as the GAO would encourage you to have? Mr. Smith. Chairman Meadows, thank you for the question. What I have in place to assure from a FATARA standpoint and also assure from an execution standpoint that the census will be successful is governance to make sure the acquisitions are done appropriately, as well as governance on all of the technology solutions being proposed from the infrastructure support to make sure they're done the right way and they are actually done with the least amount of risk from a technology standpoint. By putting these in place, I am comfortable that I am able to give the business, decennial as well as from the mission standpoint, the Census Bureau director and others, the right level of risk that potential solutions may have and we make the right decisions to move forward. I think---- Mr. Meadows. All right. Let me interrupt you. What is the right level of risk? Because that's a qualitative statement, and I want to make sure we are all on the same sheet of music. Mr. Smith. Absolutely. My first and foremost importance is to make sure that the systems are ultimately secure, the data is confidential, integral, and the systems are available. There are some potential reasons why you would want to do some of those solutions a little bit different for the 2020 based on the scale and the size of the data coming in. To me, these are not large-level risks as long as we have a large-level view of what it means from the security aspect and the ability to recover the systems. Mr. Meadows. Well, then let's go ahead and talk about that a little bit. Because I guess one of my concerns is it seems like half of the systems--and maybe I misheard what you said-- but half of the systems will not actually be implemented before we start end-to-end testing. Is that correct? Mr. Smith. That is not completely accurate. Some of the tentative dates that were not in the master--the IIP plan, those are to be delivered as well. They are all tentatively scheduled before August. Sixty percent of the systems are to be delivered before August. Mr. Meadows. So only 40 percent of them will not be implemented for end-to-end testing. So I was wrong by 10 percent. Mr. Smith. Uh-huh. But some of the ones---- Mr. Meadows. All right. So let me get to the crux of the matter. The big one, fraud prevention, according to what you've given us, it doesn't look like you plan to deliver that until March of 2020. And so I'm looking at this, and it's very troubling, because we've got a 2020 go-live estimate, and you've got fraud detection systems, the delivery is March 20 of 2020, and then we've got the other one of real-time non-ID processing, which says, basically, is this the real person that is coming in and actually filling it out, and you've got a delivery of March 20 of 2020 on that one as well. How in the world are you going to test something and plan to go live a month or 2 after that with those two critical components? I mean, I see that as the crux of why we're here today. And your delivery schedule would suggest that we're going to do all the testing and then throw in fraud prevention a month or 2 before we actually go live? Mr. Smith. Chairman Meadows, I completely agree with the concern. I will have to go look back at the tests, the system lists we delivered, and the dates to see how they would say 2020. Mr. Thompson. May I? Mr. Smith. You can. Mr. Thompson. Congressman, if I could just say something. So we are continually working on understanding the cyber risks of our fraud protection. And so we will run something in the 2018 End-to-End Test. We will continue to improve upon that using knowledge that comes in about what are potential vehicles for fraud, and we want to keep that open quite a bit so we keep updating the system. But let me assure you, we plan an extensive program where we will reinterview a sample of every response we get to the census, both non-ID and ID'ed, and use that information to determine the quality of the responses in the census. Mr. Meadows. So when do we deliver those systems? Because, we can't deliver them, Director, you know we can't deliver them in March of 2020 and expect to have any kind of integrity and confidence in those, even if we may be doing some type of testing now. Mr. Smith. And, Chairman Meadows, my answer to the delivery date that was there, there is obviously a delivery date that is going to be done for 2018. I need to get back to the group with what that date is. Our intentions are to have some of the system in place to do some testing in 2018. Mr. Meadows. All right. So do I have your commitment that the delivery of the fraud prevention and the non-ID will be done before we do end-to-end testing, you know, pure and simple? Mr. Smith. Yes, sir, you do. Mr. Meadows. Okay. And so you will get back to this committee and with Mr. Powner in terms of that? Is that something that you can live with, Mr. Powner? Mr. Powner. Yeah. I think your questions, Chairman, are right on. I mean, clearly, there are delivery dates, but there's also delivery dates, there's some integration to key systems that should occur prior to the end-to-end test. I think the thing is, if you look at this realistically, everything be done, all 52 systems by August of 2017? Likely not. Likely not. And if it's not done---- Mr. Meadows. Yeah. Of course, he modified that. He said that 40 percent of them won't be done. Mr. Powner. If it's not done, let's look at it realistically, if it's not done, then there's ways to test it by other means. I mean, this may be not ideal with the operations, but that's why it's so important that these systems, we need to really dog the delivery, the integration, and the results of the tests. So if everything doesn't go smoothly with the end-to-end test, we still do have some time to test not ideally, but we could still test leading up to the actual decennial. But I think what we need is real transparency on delivery dates, costs, and all that stuff so we can all effectively work together. Mr. Smith and I have had some really good conversations about this. I was around for the 2010 Decennial. I testified in front of this committee multiple times when the handhelds failed. There's a history of not having strong governance over these things. I do think the integration contract could help them, but also, too, the integration contractor, we're putting a lot of responsibility on them. It's close to a billion- contract, and we need to ensure that there's real delivery coming from that contractor, too. I imagine you're going to be talking to that contractor down the road. Mr. Meadows. Sooner than later, yes. And I think as we look at that, I appreciate you acknowledging the back-and-forth, because I do believe that we've had that, these key things. We're going to find a lot of stuff that is troubling, but we have a very, very short window. I mean, it's even shorter than 6 months. Some of the decisions that have to be made, as we've talked about, they need to be made, you know, in the next 60 to 90 days, and so as we look at that. So let me finish up with one, and then I'll recognize the ranking member. Director Thompson, in your opening testimony you talked about a lot of the testing and a lot of the things that are going on. So help me really take all of that testimony and, I guess, fine-tune it. With the test results that you've seen to date and where we are, how do you feel like that will compare to the actual results of the census when it comes in? I mean, what are the anomalies, I guess, is what I'm saying? Mr. Thompson. Thank you, Chairman Meadows. As GAO testified, we found in our work in the 2016 test that we needed to work on our training procedures for our enumerators more. We need to do more work on preparing to enumerate in multi-unit structures. And we need to do more work on what we refer to as closeout procedures and what Mr. Goldenkoff was discussing as more of an unresolved. But we need to work on that, we learned that, and we have processes in place to get that set for the 2018 End-to-End Test. And, again, we will have the same results coming out of that test. We'll share the results with the GAO, we'll share the results with you, so we all know where we are at that time. Mr. Meadows. So if a number of 10 says we're all the way there, where are we today based on the test results that you've seen? Mr. Thompson. I'm just smiling, because the last time I did probabilities, I got in a lot of trouble. But I would give it an eight. Mr. Meadows. Okay. Thank you, Director. And the chair recognizes the gentlewoman from New York, Mrs. Maloney. Mrs. Maloney. Thank you. And thank all the panelists. And I share the chairman's concern about security for the 2020 Census, cybersecurity. As we know, in 2015 the cyber breach of the Office of Personnel Management resulted in the theft of personal identifiable information of more than 21 million Americans. And the 2020 Census will create a repository vastly larger than what the hackers stole from OPM. So the security of the census information is of truly utmost concern by individuals, and they want to be assured that it's there before they even participate. And there is no question that encryption is a key component of information security. But, Mr. Smith, I understand that some 2020 Decennial computers for the census are too old to use encryption. Is that correct? Mr. Smith. Congresswoman, I appreciate the question. I have to go look back to see which ones are too old to use encryption. There's encryption in transit and there's encryption at rest from data. I will assure you from the standpoint from the security aspect that the approach we are taking to layer and isolate the security parameters in the network actually make these legacy systems, if they are too old for encryption, they are well behind and inside of the census data centers, that the ability to interact with these systems is limited from the outside world. From the data-collection aspect, when people are responding to the Internet site, we're taking very specific and intentional steps to make sure that that data is well protected and safeguarded and secure, which these systems you mentioned, that wouldn't apply to if they're ``too old,'' quote, to go through encryption. I'd have to look at that, though. Mrs. Maloney. Well, I think it would be important to get back to the chairman a listing of how many of these computers and their systems are too old for encryption. Mr. Smith, what concerns does that raise, and how will they be addressed? Director Powner's GAO ``Better Management of Interdependencies Between Programs Supporting the 2020 Census is Needed'' report highlighted the fact that the 2020 Census will be the first one in which respondents are heavily encouraged to respond by the Internet. And GAO has previously reported that malicious email attacks are among the growing cyber threats facing the Federal Government. So, Mr. Smith and Director Powner, your comments? Mr. Smith. So from the malicious emails aspect, what we are doing is we are going to engage industry-leading solutions to detect rogue emails, detect Web sites that are impersonating the census, and put some specific things within our system so the public can identify what is a census system they should be applying to in interacting with the Internet self-response. We're also going to be taking steps with the communications contract we have signed with the decennial to go through and educate the public on what they should be doing for phishing, what they should be doing against these attacks to get in front of it as much as possible both before the 2020 Census occurs and regularly through the 2020 Census. Mrs. Maloney. And Director Powner, adding to my question, does the 2020 Census' reliance on the Internet create an additional risk of respondents falling victim to attacks? Mr. Powner. Yeah. Clearly, we're concerned about phishing attacks. Also, too, if you look at it holistically, you have to also secure the mobile devices. We have cloud services we're likely going to be procuring, we need to ensure there is the appropriate level of security with the cloud. And then configuring all the 2020 systems, these 50 systems that we talked about, we're looking at that for you right now. We're going to continue to report back to the committee on that. But clearly, it's a concern. I do think folks at the Census Bureau, it's a high priority for them. When we talked to the contractors, there, too, it's also one of the high-risk areas. So there's a heavy focus on that right now. But we'll continue to track that for you. Mrs. Maloney. Okay. And, Mr. Smith, is the Bureau using the services of outside organizations to help implement and test cybersecurity? Mr. Smith. Yes, we are. We're using outside resources, Department of Homeland Security. We're also using industry- leading companies to help evaluate our security architecture, as well as test the vulnerabilities and test the penetration, which basically means attempt to break through the architecture we've designed for the 2020 Census systems. Mrs. Maloney. Well, Director Powner, based on what you've heard today, how confident are you that the Census Bureau and the Department of Commerce are implementing sufficient levels of cybersecurity for the 2020 Census? And, you know, your report showed that there are many intelligence challenges because of the new technologies. Mr. Powner. Yeah. Clearly, what we will be looking at, I think the penetration tests that Mr. Smith mentioned, we want to look at how complete they are, what the results are. And I mentioned in my oral statement, too, I think a key oversight area for this committee is the authority to operate signature on all these systems. You know, clearly if you look back on healthcare.gov, I testified on that, that was a key issue. There was an authority to operate, but there were vulnerabilities associated with the authority to operate that some would argue weren't acceptable. And I think that will be a key oversight item for this committee. Mrs. Maloney. And, Mr. Smith, what assurances can you give that the Bureau's IT progress is appropriate? I mean, there are new, tremendous challenges in IT. Last month, the Bureau provided the committee with information relating to its remaining IT decisions, and we were told that there are 48 IT- related decisions left. Is that correct? So how would you characterize the Bureau's progress, or lack thereof, of meeting its deadlines? Mr. Smith. Thank you for the question. There are approximately 40 IT decisions left from the list. These are decisions that we're going to make as we learn from the operational tests, things that need to be implemented. I can assure the subcommittee from the standpoint of how we're securing the technology enables us to actually develop the technology iteratively and agilely, which is recommended by the Federal Government as far as approaching large-scale systems of this nature. And I can assure the subcommittee that by developing things agilely, we are going to know how things work and how secure they are along the way as we are developing it. We are not going to wait till the end to do the testing and do the security of these systems. Mrs. Maloney. Any other comments? My time has expired. Thank you. Mr. Meadows. I thank the gentlewoman. The chair recognizes the gentleman from Wisconsin, Mr. Grothman, for 5 minutes. Mr. Grothman. Thank you. A couple of questions where, I guess, the Census Bureau gets in the news. There's been a lot of discussion in recent times in the newspapers about undocumented or illegal immigrants, and I think the number is to a certain extent based on Census Bureau data. Do you guys generate the numbers of people who are not citizens in this country? Mr. Thompson? Director Thompson? Mr. Thompson. Thank you. We don't produce a number of the undocumented people in the country. I mean, we do produce data that researchers in the field use to make estimates of that, but that's not a Census Bureau product. Mr. Grothman. Okay. So when you give us the total people in the country, you have no idea whether they are citizens or not? That is not something you concern yourself with? Mr. Thompson. Thank you, Congressman. When we produce our estimates from the decennial census, it includes everyone in the country, and we don't distinguish between those that are documented and not documented, and we don't ask. Mr. Grothman. Okay. So we have no idea whether they are a student here on a visa or a short-term visa or a person who is here illegally, you have no idea? You said you make an estimate, though. How do you make that estimate? Mr. Thompson. I'd have to get back to you on that. Mr. Grothman. You have no idea? Mr. Thompson. I'm not a demographer. But we do have a lot of good demographers at the Census Bureau. Mr. Grothman. Any other guys know? And as I understand it, in banding about this 11 million figure for 15 years or something, I'm told it's based on census data. Do you any of you guys know how they generate that number? It's used by a lot of politicians. I just wondered. Why don't you guys get back to us on that. Mr. Grothman. Okay. The next question is, you have a tendency to generate statistics by what we'd call race. Is that right? Mr. Thompson. We do. We do provide statistics by race. Mr. Grothman. And how is that determined, the race of those? They all self-identified, or how would you define? Given that in today's world we have more and more I guess what you'd call biracial people, how do you generate those numbers? First of all, give me the breakdown on the different races that you try to identify. Mr. Thompson. So we follow the guidance of the Office of Management and Budget and establish the racial classifications that are used in the United States. And this includes major categories, such as African American, such as White, such as Asian, such as Native Hawaiians and Pacific Islanders and American Indians and Alaska Natives. Within that, because we have a decennial census, we collect more detailed information. We also collect the category on the decennial census other race. Again, all this follows the guidance that the Office of Management and Budget establishes for assigning racial classifications in the United States. Mr. Grothman. We're dealing with kind of inflammatory stuff. I would think over time more and more Americans are not 100 percent something. Mr. Thompson. Congressman, let me also say that, starting in 1997, the OMB directed that respondents be given the option of responding to more than one race. So respondents can identify with any particular race or combination of races that they feel describes them. Mr. Grothman. Okay. Our President, how should he fill out that form? Mr. Thompson. That's up to our President. I mean, seriously, it's self-response. So we don't--we present the categories, we present the information to our respondents, and then they identify with what they believe describes them. Mr. Grothman. Okay. So the figures are largely no guidance. Okay. I could be one-eighth American Indian, if I put down American Indian, that's what I am, right, according to guys? Mr. Thompson. That would be your census response. Mr. Grothman. Okay. Now I'll give a question to Mr. Goldenkoff. The GAO found last time 25 percent of your cases resulted in non-interviews after six visits. So somebody went there six times, six attempts, nobody gave you the information. Is that right? Are those figures right? Mr. Goldenkoff. Yeah. And just to clarify, I believe it was 30 percent at the Texas test site and 20 percent at the Los Angeles test site. Mr. Grothman. Okay. Does this concern you? Mr. Goldenkoff. It does concern the GAO, and for several reasons. One, it concerns us because the Bureau doesn't know the factors that contributed to the high levels of non- interviews. And then also, of course, that information would need to be imputed statistically on those missing housing units. So there are a couple of factors that could have influenced those high non-interview rates. It could have been, for example, the enumerators not following certain procedures they were supposed to if they couldn't get an interview. They were supposed to go to a neighbor or somebody nearby. But then this is where Director Thompson was talking about where the training needed to be improved. Perhaps that the enumerators didn't recognize the importance of going to a proxy. There are some other factors that could have influenced that as well. But it is concerning to us, and that hopefully that the Bureau will be able to, you know, have data from this 2016 test to get a better understanding of what was behind those high non-interview rates. Mr. Grothman. Okay. I can see I've exceeded my 5 minutes, so we'll have to, when we go to conference later this week, we'll have to tell all our colleagues they missed an exciting hearing. Mr. Meadows. I thank the gentleman from Wisconsin. It's duly noted. I appreciate the gentleman's questions. We'll recognize the gentlewoman from the District of Columbia, Ms. Eleanor Holmes Norton, for 5 minutes. Ms. Norton. Thank you very much, Mr. Chairman. This is always an important hearing. It's important. It's timely. I would say it's timely. I hope it's in time. And the enigma, of course, of the census is not whether you have one drop of black blood, but of how to count people and to give them the opportunity. I indicate that because there was a time in our country when that was, if not of the census, a way in which some States decided who is African American in our country. So the only thing a diverse democracy can do is to ask you, ``Who in the hell are you?'' and to have you fill it out. I note that some people are using these apps and are finding that they are of many different ethnicities. But at the same time, they probably answer and many of them are answering ``other.'' I also want to say for the record that census is in the Constitution. At the time of the census, most people weren't even citizens, because they had arrived here and we were getting people coming in such large numbers. So it was important from the beginning just to know how many people were here. And from that, of course, scientifically, we can now--we now have a way to find who is documented and who is not. Ms. Norton. But I want to ask questions about this what I am calling the enigma of the minorities, which always vex the census. I want to turn to this Harris County field test. So that even using new technology, approximately 30 percent of the households--I think this is for you, Mr. Goldenkoff-- approximately 30 percent of the households in the test could not be contacted after six good faith attempts. My first question is, compared to how many attempts did it take for the 70 percent that could be contacted? Mr. Goldenkoff. Well, that varied. It was between one and six. Ms. Norton. Yeah, well, obviously, it was between one and six. Mr. Goldenkoff. Right. And, specifically, I don't know. Ms. Norton. I would like you to report back to this committee how many attempts does it take on the average to get a response. Because I think six is a good number. So I would like to have some explanation of these results. Mr. Goldenkoff, how is this problem to be fixed? I mean, those people are out there. Surely they can't go uncounted. Mr. Goldenkoff. Well, the first way of fixing it is getting a better understanding of what was causing it. And we were out there, I personally was out in Los Angeles, we had other GAOers out in L.A. And also in Texas. And, you know, based on our observations, as well as our conversations with Census officials, there are a couple of things that could be behind it. First, just the nature of the test itself may have high non-interview rates, and that is because it is not an actual census, and so the public may not be as aware that the census was being conducted as they would during an actual census just because during the actual census enumerators are under more time pressure; the Census Bureau also conducts a lot more outreach and promotion. So it could have been an awareness issue, people just don't want to answer their doors. Another reason could be the devices themselves, that, you know, the way those hand-held devices worked, that after an attempt was made, they couldn't reopen a case. And so if someone came home and the enumerator was still in the area--and we did see this happen on a number of cases--they couldn't reopen that case to enumerate that person. Ms. Norton. They couldn't reopen the case? Mr. Goldenkoff. Because it's just the way the hand-held devices worked, that there were certain--there are some reasons for that, you know, and I think this is just a more generic issue of being able to better balance the efficiencies that you get with the centralized control of an automated case management system with the flexibility that you get with enumerators on the ground being able to adapt procedures as they see fit. And it's a balancing act. And so maybe, and this is something that we have been discussing with the Census Bureau, is how to, through the business rules and how the devices are managed, to allow for a little bit more flexibility in the procedures. Ms. Norton. Could I just ask, some of these seem important hypotheses. Are they being acted on? They seem to be a warning sign. Mr. Thompson. Congresswoman, those rates are unacceptable for the 2020 Census. We have the information that will be available to make sure that we can go back to those households as many times as it takes to get a response. And we will do that, because we are not going to leave that much of the population---- Ms. Norton. I understand that. That is important to here. And documenting what of the hypotheses you just heard was the cause would be important for the future as well. Director Goldenkoff, is there any way to know--there should be a way to know--what percentage of those who did not use the Internet response option in the two tests were minorities? Mr. Goldenkoff. Who did not use the Internet? So basically, if I understand your question correctly, you are asking minorities who responded other than the Internet through means---- Ms. Norton. Who did not use the Internet response, who did not use the Internet response option in the two tests. I am talking about this test in this Harris County field test. Mr. Goldenkoff. Right. Ms. Norton. How many were minorities? Mr. Goldenkoff. That would be something that the Bureau could determine, I imagine. Mr. Thompson. Congresswoman, we have that information. But let me say that for 2020 we aren't offering the Internet as the only response option. So in 2020, for the first time, people can call up and give their interview over the phone. Also, we do know that there are a number of people, a number of communities where they don't want to respond by the Internet. So we will be actually mailing out a questionnaire directly to about 20 percent of the households in the United States based on information we have about what kind of method they would prefer to respond. Ms. Norton. Well, the new technology may have gotten in the way here of getting an accurate count of people, particularly undocumented people or people who are leery of responding to people who open the door, I mean. Yet you did the field test using new technology. Was it to see whether new technology would be more efficient, would work? Mr. Thompson. So our philosophy---- Ms. Norton. You didn't use enumerators in this field test. Mr. Thompson. We used enumerators, certainly. Ms. Norton. You did? Mr. Thompson. Of course. Ms. Norton. In this field test, the Harris County field test? Mr. Thompson. We certainly did. We certainly did. Now, like I said, if someone wants to respond by paper, eventually they will get a paper questionnaire, or they might get one right at the start. Ms. Norton. Do you imagine that among those that it took six attempts and still you couldn't respond, that they probably need to have some face-to-face or paper enumeration? Mr. Thompson. The ones that took six were six face-to-face attempts. We went to the door and knocked on the door and nobody was--see that nobody was home or the door wouldn't be opened. These were six attempts after they received four---- Ms. Norton. That is troubling. Mr. Thompson. --contacts by the mail to respond. Ms. Norton. Very troubling. I don't know if it's when people go. Do people go only in the daytime? People may be, I hope, away at work. Mr. Goldenkoff. Mr. Goldenkoff. Yeah. I think that is one of the points that we made in my statement, is that being able to leverage the knowledge that the enumerators see on the ground. And, for example, one of the things that we saw time and time again was that enumerators would start their assignments about the middle of the afternoon, for example, but a lot of people aren't home. But then at the end of the day they would be finishing up around 5 o'clock when people were returning home, they could not reopen those cases. We saw this in multi-unit dwellings. Ms. Norton. This is a huge error. Mr. Goldenkoff. Well, you know, this is something that, you know, this is what you learn by testing. Ms. Norton. Well, my God, we have been doing this--we have had, you know, we either have had two-parent families where both go out to work or single-parent families where we hope somebody is working for a very long time, and the notion that we don't know that people aren't home in the middle of the day. Mr. Thompson. Congresswoman, if I might. Ms. Norton. Yes, sir. Mr. Thompson. In previous censuses, what we would do is we would give an enumerator a stack of questionnaires and we would tell them to go out and get those questionnaires enumerated. We had no idea of what time they were going out or when they were going or where they were going. Now we do have the ability to know when they are going to a house, and we have the ability to tell them they can't go during the day and they have to go in the evening. So we do have that ability now that we didn't have before. Ms. Norton. This is very important. You know, I would hazard a guess--I do have one important question, if I can go before this--I would hazard a guess that most Americans are not home in the middle of the day if they are fortunate enough to have a job or if they have children and they are in school. I don't know why enumerators wouldn't begin later in the day if we want an accurate census. Mr. Goldenkoff. Well, in many cases they did. It all comes down to local knowledge, because we also saw examples where people worked the night shift, and the best time to get them was the middle of the day. But one of the things, and this is where the need for that flexibility comes in, is that even though they could not leverage in a very centralized way, you know, an enumerator could write in ``come back at 6 p.m. tomorrow,'' it was not managed centrally in a very good way. And that is what needs to be improved. I think there were procedures to do that, but it just didn't happen as often as it should have. Ms. Norton. And I see the need for a central management of the system along with the flexibility you were speaking of. I do have to ask about what I believe is really a constitutional question, and that is prison gerrymandering, a huge distortion of the count because, for example, big cities will rarely have a prison in the middle of the city, and so it will be in another area, less populated. This has gone to court. Recent cases have held that prison gerrymandering violates the Constitution. I hope we are finally going to settle this. But the Bureau recently put out a request for comment on its residence criteria, and here I am going to quote what you said: ``The usual residence for counting purposes would be the prison in which the inmate is held.'' Now, some States have moved forward to correct this huge distortion that can affect everything from where we count for Members of Congress, not to mention where resources go. So you are overcompensating where the resources may in fact not be going to schools, for example, in those schools and prisons. I don't see how we can have it both ways, and want to know are we going to settle this question or are you looking forward to the census facing really a huge constitutional question because there are a number of cases and they will be on their way up. What are you doing to settle this question before the census? Mr. Meadows. The gentlewoman's time has expired. You can answer very quickly if you have a response to that. Mr. Thompson. I do have a response. So we issued--and this was in the interests of being open, this is the first time we have done this--we issued our proposed criteria for determining residence in a Federal Register notice at the end of June. We gave a 30-day comment period. We extended the 30-day comment period based on requests from the committee. We have gotten over 78,000 comments, responses to that Federal Register notice, and we are now in the process of evaluating the very thoughtful comments that we got so that we can issue a final determination. Ms. Norton. Well, let's hope you take into account the recent litigation. You cannot have censuses where some States have corrected this, other States haven't, and then go to the American people and say we have an accurate census. You can't have two ways of counting the same population based on what State they are in. Thank you very much, Mr. Chairman. Mr. Meadows. I thank the gentlewoman for her interest in making sure that everybody is properly counted. So let me kind of close out by asking a few questions to make sure that we are all marching to the same drumbeat, because I think we are getting closer. And so, Mr. Smith, let me start with you. In October, we talked about doing, I guess it was a single operating system for deploying on mobile devices as part of that working group. And I think I read that indeed that is where you are going with that. Was that an accurate report or is it--was that correct? Mr. Smith. Chairman Meadows, I recall that we are going through to determine the mobile devices that are going to come in through a contract, and based on the operating systems the mobile device may choose to provide as a bid. The platform we are using to develop for the enumerators to produce the application can work on either mobile platform. There are very, very, very small changes beneath working on an Android or working on an iOS or working on other mobile device platforms. So the system itself can be---- Mr. Meadows. So for a non-techie, is that a ``yes'' to my question or a ``no'' to my question? I mean, is it a single operating system or not? Mr. Smith. No. Mr. Meadows. So the reports were wrong. Mr. Smith. From the mobile device aspect---- Mr. Meadows. I think the reports basically said that you were going to that. And so---- Mr. Thompson. Congressman, we have just released our request for procurement for our device as a service contract. It came out yesterday. And we are going to carefully examine the solutions that the vendors propose. And it may very well be that some vendors will propose only one operating system. Mr. Meadows. I understand what they are going to. What are you going to require? I mean, they will propose a lot of other things. I guess here was my concern, is as we get to these different devices, and I talked to you about having one operating system on one and another on another, you are going to create the potential conflict there. So I thought we had made progress, and I was about to say hurray for Mr. Smith, thank you for listening. And what I am hearing is that that may not be accurate. Mr. Smith. Yes, sir. The system itself is portable. It is like running Microsoft Word on an Apple computer or running it on a Microsoft computer. That version of the word processing software works on both. I am comfortable with picking this industry leader to help produce the application, that their application works on Android, works on iOS, and will be able to work with whatever comes back from the device as a service. Mr. Meadows. I guess it was an article by Phil Goldstein in FedTech. Were you all familiar with that? Maybe we follow this stuff closer than you do. It said Census Bureau plans to go mobile and rationalize their device portfolios. And it just talked about a device footprint. So perhaps we can clear it up. I still have great concern if we have different operating systems. So the October meeting, that concern still stays there. So if you will just let me know as the procurement, Director Thompson, gets a little bit, what you are going to anticipate from our vendors. Mr. Powner, let me come to you, because some of these timeframes, and according to Mr. Smith we are talking about 40 percent of those systems not being in place for end-to-end testing, is it possible for you to identify what systems could be descoped and would be appropriate to be descoped if the mission is too great? Mr. Powner. I think the key question on the descoping, and this is something we have talked to Mr. Smith about, is understanding what the critical path is. Not all those 50 systems are on the critical path. So there is a critical path. I mean, sure enough, the Internet response and the mobile devices and the centralized, there are some things that are very big and important, right? We know those are on the critical path. But identifying everything on the critical path and then looking at potential descoping, you don't want to descope items on the critical path. Right now, with the integration contractor and the integration test, I think that's a TBD, everything that really is on the critical path. So that's a key question about what is on that critical path from an integration point of view. Mr. Meadows. So would you agree with that, Mr. Smith? Mr. Smith. Yes, sir. Mr. Meadows. All right. So if you agree with that, then can you report back to this committee by the end of this year, and so I would say by December 31, with what I would call is a critical path analysis? Mr. Meadows. A is critical path, we have got it, these are must haves. B would be, yeah, it's critical, these are nice to haves. And C, it's not part of the critical path at all, so those would be subject to descoping, as Mr. Powner talked about. Mr. Smith. And, sir, I would like to also address in the conversation we will have with GAO, it comes down to the context of the systems, too. As Mr. Powner discussed, data collection is utmost of importance, that the systems are prioritized and being done. Some of the things that are done after August are things that can by timeframe be done after August, and some of them are enhancements to existing systems, like paper collection. So the context of the systems will also play---- Mr. Meadows. Well, and what I will ask on that A, B, and C, if you will put in parentheses a T next to it, if it is a timetable initiative that makes it perhaps less critical in terms of our end-to-end testing or where it is. Because I need to know what your thinking is on this, is, well, this is not critical because we can deal with it in November of 2019 versus this is just not critical. Does that make sense? Mr. Smith. Yes, it does. Mr. Meadows. Okay. All right. So Director Thompson, let me come back to you in terms of budgets on two different questions. So Ms. Holmes Norton was talking about going from Internet to paper and phone. If there is a disproportionate share of those that respond by phone or paper, you know, some other collection mode, how does that affect the budget and will that affect those numbers adversely in a significant way? Mr. Thompson. Right now we have estimates of getting about 63.5 percent response in directly. It would cost more if more people preferred paper and more people preferred telephone. But it would certainly cost more if people did not self-respond at all and we had to go out and do the nonresponse follow-up. So we do plan for contingencies in 2020 to allow us to be able to both fund and process if we get more paper or if we get more telephone calls. So that is in our contingencies for the 2020 Census. Mr. Meadows. So can you provide to this committee then, and I won't fine tune it too much, but for each 2 percent miss in that 63 percent response rate, financially what does that mean? All right? So we won't do 1 percent or half percent, but we will assume that you've---- Mr. Thompson. So typically what we have done, and we will do this again, is we have said for every percent of increased nonresponse follow-up this is the marginal cost increase. Mr. Meadows. All right. So you will do it for every--I was going to give you a break there. Mr. Goldenkoff is sitting there shaking his head ``yes.'' So you will submit that to the committee so that we have an expectation from a budgetary standpoint. Mr. Thompson. Yes. Mr. Meadows. All right. So let me talk about priorities. And I mentioned it in my opening statement with the reduction in the 2020 budget and what that has done in priorities. Would you not agree that the most important aspect of your job is the 2020 Census, is the highest priority out of everything that you do? Mr. Thompson. I would agree, Congressman, it's a constitutional mandate. But having said that, when we made our decisions for the 2017 budget we looked at both making sure that we could deliver a census in 2020 that would be cost- effective and innovative and on track, and at the same time maintain the important data we produce on the economy and give the Federal agencies the help---- Mr. Meadows. I was trying to give you a softball, Director Thompson, and your response is a swing and a miss. Mr. Thompson. I was anticipating where you were going. Mr. Meadows. Yeah, I am sure that you were. So let me be very clear. If we are going to cut the budget, it doesn't need to be cut from the census, it needs to be cut from other areas of responsibility, from my perspective. And then you know you have got an open door, that I am willing to go to the appropriators and I am willing to advocate on your behalf. But if we are fully funding the other operations and we are not fully funding the 2020, I am not going to get calls on whether the other aspects of your job are done or not when they don't get fulfilled as much as I would the 2020 Census. Do you understand the importance of it from my standpoint and Ranking Member Connolly's standpoint? The egg on our face will not happen with some of the other, granted, very important aspects that you have under your jurisdiction. But whether they get done or not, I get a call from someone in the bureaucracy here saying why haven't we done it, moreso than I will if we don't have the 2020 Census done properly and effectively. Would you not agree with that? Mr. Thompson. I have got to say my number one goal is to do the 2020 Census in new and innovative ways, and effectively and accurately. Mr. Meadows. I agree with that. I guess here is what I am asking you to do, is if you were making budget concerns that impact the implementation of this critical process that we have all been talking about, where it is, you know, 75 percent--or a 25 percent cut actually came out of the 2020 Census and very little cut came from some of the other responsibilities, and so I guess I question that decision in light of the priority that I deem the 2020 Census to be. And so let's have some further discussion, I will let you off a little easier on that. Mr. Thompson. Congressman, I appreciate your willingness to talk about a number of things on the census and your willingness to engage with us. It has been very refreshing to work with you. So thank you. Mr. Meadows. Well, if you will let me know--and I have told your staff this--from a budgetary standpoint, I am willing to camp out in the appropriators' lobbies until they want me to go home. How about that? And we will make sure that you get the resources, as long as you can get me those other numbers. So I want to just say thank you. Let me just tell you, to have two different missions for your agencies in terms of your primary focus, but having one mission in terms of making sure that this is successful, I am pleased with the back and forth and the tension that sometimes is there and acknowledged, but yet at the same time what I am hearing is real progress. And so, Director Thompson, I want to just acknowledge that in terms of your leadership and your willingness to do that. And so as we work going forward, I look forward to having each of you update us on a regular basis to make sure that we are making good progress. And with that being said, if there is no further business before the subcommittee, it stands adjourned. [Whereupon, at 11:35 a.m., the subcommittee was adjourned.] [all]