americans if we got that compliance number up. it's going to save money. it's also more accurate. i'm not talking about an ad campaign where you're on an ad at the super bowl. just use the bully pulpit of being a cabinet level official to kind of reassure our citizens that you want to do it the right way and you want it to be accurate and you could use their help. >> i think that's very good advice and i will try very hard to use whatever bully pulpit comes with commerce to try to achieve that. >> on behalf of everyone, i want to thank you not just for today but also the briefing you gave us yesterday. we have another panel. i know that you have other things to do. so we're going to recess and then allow the second panel to come forward. with that, thanks on behalf of all of us for your time this

morning. >> thank you, mr. chair. >> yes, sir. i want to thank you for your patience. i will recognize you for your five-minute openings. mr. cummings is coming. he graciously told me to go ahead and start. from my left to right, mr. robert goldencough.

mr. david pounder, director of information technology management issues at the government accountability office. ms. carol rice, department of commerce office of inspector general. ms. gupta, leader on civil rights. i think all of you have testified before. green means go, yellow means run through the light. yellow is speed up, get through it as quick as you can. and red, you're going to get stopped. i won't stop you right at five minutes because it's really important. but your opening statement is part of the record. so with that, mr. goldenkoff. >> chairman gowdy, ranking member cummings and the rest of the committee, in recent years

we have identified a number of operational, management and other challenges that raise serious concerns about the bureau's ability to conduct a cost effective enumeration. and in february 2017 we added the 2020 census to our list of high risk government programs. my remarks today will focus on two such challenges, implementing design innovations aimed at controlling costs and developing reliable cost estimates that better account for risks and inform annual budget requests. my colleague will then discuss the challenges the bureau faces in implementing and securing critical i.t. systems. as secretary ross mentioned, it is now projected to cost 15.6 bltds.

management and oversight will be needed in the short time remaining until census day to ensure systems are fully tested. to help control costs while maintaining ak rast ras si, the bureau will use new procedures for 2020 including greater use of collection methods, administrative records in place of collected by enumerators. they have not been used to a great extent in prior di -- at

all. the bureau is currently conducting the 2018 end to end test. the tests began in august 2017 and is scheduled to run through april of 2019. essentially a dress rehearsal for the actual enumeration, the 2018 test is the bureau's final chance to ensure it will perform under operational conditions. in may 2017 because of budgetary constraints the bureau descoped the test and only plans to conduct a complete test in rhode island. the washington and west virginia sites will only address address canvassing. this is not the first time the bureau has curtailed its

testing. the bureau cancelled field components of a 2017 test. without sufficient testing across a range of groups, operational problems can go undiscovered and the opportunity to refine procedures and systems will be lost. another risk factor is the quality of the bureau's estimate of the cost of the 2020 census which does not conform to best practices. cost estimates can help an agency manage complex assignments and make funding decisions and provide oversight. however, the bureau's 2015 cost estimate only partially met the characteristics of two best programs and minimally met the other two. additionally the bureau has not yet publicly released an update to its 2015 cost estimate. yet several events since then

eindicate that the life cycle cost will be substantially higher than the original estimate of $12.5 billion. the bureau's preparations for 2020 have been further complicated by ongoing vacancies in the positions of census director and deputy director. although acting leadership has since been named, turnover in the top ranks makes it difficult to ensure accountability and continuity as well as to mitigate risks and control costs over the long-term. this concludes my prepared remarks. i will now turn it over to my colleague. thank you for inviting us to testify on the bureau's plans to deliver and secure key technologies for the 2020 census. these technologies can improve our nation's response rates and secure citizens' data. this technology has not been effective lly planne ened or ma

and much work remains to deliver it. schedule, security and cost growth,ly expand on each of these. the bureau needs to deliver 43 systems for 2020. some are new systems in infrastructure while others are changes to existing systems. clearly the internet response capability, mobile devices and the centralized operations component are critical systems. these systems need to be ready for the end to end test that started in august to ensure they are integrated appropriately. the bureau has missed many milestones and currently only 4 of the 43 have completed development and testing. many of these systems have key delivery dates in early 2018 so they can be included in the end to end test.

the mobile device application has been running behind schedule for some time. and we have serious concerns about the timely delivery of the fraud detention system. i'm not sure i'd agree with the secretary's comment this morning that the systems are on schedule. the appendix in our statement lays out each of these 43 systems' delivery dates. we are especially concerned that schedule pressure and late deliveries will lead to inadequate testing. the bureau needs to minimize the threat of phishing, secure about 400,000 mobile devices, ensure security of cloud services and properly configure all systems. to ensure all systems are as secure as possible, the bureau needs to assess security controls, fix known deficiencies and have the proper sign-off to

ensure that each system is authorized to operate. this process is critical since 43 systems contain personally identifiable information. the bureau has such a process, but we have concerned about whether there will be ample time to complete all the security work given the late delivery dates of the system and whether shortcuts will be taken due to pressure to authorize pro mat e prematurely. all 43 systems need to be reauthorized. finally, we have concerns about cost growth associated with these technologies and have had these concerns for some time. the sedcap program has overr overrun about $400 million. the latest from the bureau is that i.t. costs were about 3.4

billion but that total did not include about 1.4 billion, bringing the total cost closer to 5, not 3.5. the new life cycle estimate includes this additional billion and a half that was previously unaccounted for. we will be tracking these costs closely because given the remaining work and past mismanagement we remain concerned about cost growth in this area even with built-in contingencies. schedule security and cost growth are major concerns we will continue to monitor for the congress. thank you for holding this hearing. >> ms. rice. >> good afternoon, chairman, go gowdy. i appreciate the opportunity to testify about the bureau's 2020 life cycle cost estimate. we all support the bureau's ultimate goal. but over the last several years we have produced a number of

reports that have found weaknesses in the how the bureau allocates costs, collects cost data and estimates cost for the 2020 census. my testimony today provides additional found that the bureau's 2020 census life cycle cost estimate cannot be validated. second, much of our work throughout this decade reveals cost overruns, unaccounted for cost, and decreased cost avoidance. to the first point, the bureau's cost estimate is not auditable. it takes into account more than 100 inputs, but when we tried to verify their accuracy during the 2014 census test, there was no supporting documentation. additionally, we evaluated the interactive review portion of the inoffice canvassing operation and found that the estimate was derived from undocumented verbal communications. this lack of documentation prevented us from oauditing the

estimated cost avoidance the bureau reported to stakeholders. similarly, we identified three instances where the census test failed to capture cost data that could validate and update the cost estimate. the 2014 census test and the 2015 address validation test included to collect data in the field, but it didn't. finally, during the 2015 census test, we found that the bureau could not differentiate between the cost performing versus administrative activity. it could not determine whether the new methods were more effective and more efficient than achieved by just reducing administrative burden. we also reported weaknesses with how the bureau accounts for cost. in auditing the process for implementing the fy '14 budget cuts, we found the bureau recorded employee costs based on predetermined budget allocations instead of the actual hours worked. in addition, we found multiple instances where contractor costs were not charged to the correct

activity. all of these accounting and estimation weaknesses prevent the bureau from making informed decisions. second, we're concerned about cost overruns, unaccounted for costs, and decreased cost avoidance. for example, the bureau initially projected spending $650 million for the cap program, and now it's estimated at $965 million, and i think based on what the secretary just said, it will be higher. one audit -- our audit of the canvassing test also found substantial overruns. the bureau estimated that spending for in-office center canvassing would be $44 million between fy' 16 to '19, but our review found it would cost nearly three times that amount at $125 million. finally, our evaluation of the 2016 census test identified unaccounted for costs, redesigned operations. we found that the operational control system allowed enumeraters to make more contact attempts than the cost estimate

accounted for. so the expectation was they would go up to six times, and we found them going many, many more times. we're also concerned about the increased unresolved rate. as observed in the test, the potential need to expend additional resources that are not currently accounted for in the bureau's cost estimate. in conclusion, we applaud the census bureau for undertaking several major initiatives to modernize the decennial operation. we recognize that not all the innovations will perform as expected and that some of the 2020 cost savings may not materialize. however, for stakeholders to have confidence in the cost estimates, the bureau must strengthen its cost accounting processes by documenting inputs and tracking all project costs correctly. i'm encouraged by the recent efforts to improve the life cycle cost estimates. the bureau started linking documentation to the variables in its estimate and as we have just heard, the secretary's task

force worked closely with the bureau to improve that estimate. we are looking forward to reviewing that final product. so thank you very much for inviting me. i'm pleased to answer any questions. >> thank you, ms. rice. >> chairman gowdy, ranking member cummings and members of the committee, thank you for inviting me to be here. it's a coalition of more than 200 international organizations that has been deeply engaged in the last several cycles of the decennial census. we share this committee's concerns and interested in a modern, secure, and cost effective 2020 census. technology can facilitate easy and lest costly participation in the census however, to uphold its constitutional duty and insure an accurate and fully inclusive count, congress must allocate resources to count underresourced communities. it must also provide funding for comprehensive risk management in preparations for realtime backup

methods and operations, and we're pleased to submit for the record two reports about the risk posed by technology and the records in the 2020 census. the primary and overarching goal of the census is a fair and accurate enumeration of all people living in the united states on census day, and the goal of a census that is equally successful in all communities is really nonnegotiable. these aren't as the chairman said, republican goals or democratic goals. they're shared american goals and achieving them is going to be very key to our representative system of government. but insufficient, uncertain, and frequently late annual funding has delayed and derailed testing and preparations as you have heard already. most at risk are our operations that are specifically designed to enumerate historically hard to count communities. the census bureau needs a ramp-up in spending for a dress

rehearsal, ark checkter and feel infrastructure, and a campaign that is going to encourage people to participate, and therefore, at the back end, help keep census costs in check. we support the proposal and congresswoman maloney's new bill. you have that before you, but the senses has political consequences, in fact, the constitution says as much by basing congressional apportionment and equal representation on the population count, but the conduct must be strictly nonpartisan and must strive to achieve an accurate account in all communities. unfortunately, even with careful planning, we may be looking at an environment that could yield significant undercounts in poor communities, rural areas, communities that are recovering from natural disasters, communities of color as well as undercounts of children, immigrants, and others who have a palpable mistrust of government, who have deep concerns about cybersecurity and the security of their information as well as in rural communities where the digital

divide is particularly pronounced. there are hard to count communities in every state, and hard to count populations in communities of all sizes, from large urban areas to smaller cities, to rural and remote communities, including american indian tribal lands and reservations, and just two points to note for both members here, nearly 10% of census tracks in south carolina are hard to count per census bureau research, and 27% of baltimore's population live in hard to count census tracks. we fear the strict budget constraints congress has imposed add to these formidable barriers. the census bureau is going to attempt to minimize underaccounting but is going to be hampered by a smaller footprint in the field. when your constituents aren't counted in the census, they remain invisible for the next ten years. there aren't any doovers and no question that the census bureau must get it right for the first time because all of our members of congress, county officials, mayors, school principals,

veterans, advocates, businesses large and small, and indeed all of us in the united states, have to live with the results of it for the next ten years. so the leadership conference and our member organizations look forward to working with all of you to insure a cost effective, secure, and above all, an accurate and inclusive census in every single one of our nation's communities. thank you. >> i want to thank all four of your for your opening statements. i also want to thank you for doing it in less than five minutes, which is something that members of congress cannot do. so i hope you consider running some time. mr. cummings, my friend from maryland. >> certain populations, groups, have historically been undercounted in dissenial census. specifically minorities such as african-americans, hispanics, and american indians on reservations are counted at rates much lower than others.

ms. gupta, what are some of the reasons these groups are undercounted? >> well, i think that's one significant reason and it's particularly of concern right now is that mistrust of the federal government and whether communities will be -- whether their information when filling out the census will be safe and not used for enforcement purposes or and the like. it's particularly i would say right now in immigrant communities. there's also a real concern around the digital divide and what that might do, and i would say even more acute in rural communities. as well as, you know, real concerns around cybersecurity and whether their information is going to be breached and the like. a lot of these harder to count communities, communities of color, there's a lot of mobility in these communities. they need to be -- have trusted partners in the field from the census, whether it's through organizations like the ones that

make up the leadership conference or through trusted partnerships with the census bureau to have messengers that are explaining the purpose of the census, why it's important to be counted, and that's why the partnership program and the communications aspect of the census need to be adequately funded, because without there, there's no chance of getting an accurate account, and that will fail the entire census and the country. >> one thing that secretary ross said to the chairman the other day is that the cost of labor, he thinks, has to go up because you have to get a certain kind of person, the kind that you just talked about. and to get -- i was trying as you were talking, i was trying to figure out, how do you penetrate that? how do you attack that problem? because if people are not trusting of government, if they're worried about immigration issues, and maybe not for themselves but for family members, and then there's

some people that just don't even know what the census is. all they know is that the government is knocking on their door. and any time the government is knocking on their door, they assume it's something bad. i know because i live in a neighborhood like that. so how do you attack that? >> i really think it's critical that the census have, the bureau have the infrastructure around a communications issuing and a partnership program. we were encouraged, i was encouraged today when i herds the secretary talk about a projected communications budget of $500 million, but i think it's really important to insure that there's enough money to actually promote census participation in our many diverse communities, and to address some of the unprecedented challenges that the 2020 census is going to be facing around, as i said, the mistrust of government, cybersecurity fears, fear of federal government agencies, and i think it's deeply important that the commerce department and the census bureau consult

closely with stakeholders who know these communities well, who are in these communities, who can be trusted messengers and know best how to convey the information and the value of what it means to be counted. and that's the only way i think that we're going to be able to insure that there's a fair and accurate count. and of course, there's already deep concerns about the canceled testing and the like that could have other implications for the accuracy of the census in the next -- in 2020. >> talking about the canceled testing, i understand that one test the bureau canceled in 2017 due to insufficient funds was to be conducted on tribal land, including on the standing rock reservation. is that what you're referring to? considering that native americans were the most undercounted group in the 2010 census, this is very significant. ms. gupta, what are the implications of canceling this and other critical tests,

particularly for rural areas and american indian reservations? >> well, you know, there's a statistic that we have, 87% of the hardest to count counties in the 2010 census were rural counties. and given that this is the first high-tech census, i think there's a lot of questions that you have heard, a lot of concerns both about cybersecurity, about whether there's going to -- the height that technology is going to work. while obviously we want a modernized census, you have to be able to test it, and test it in rural communities as well as hard to count communities is going to be really important to be able to actually get it donent 2020. the fact they canceled because of insufficient funding three out of four of the census test runs and that's of great concern, and it will hamper this lack of preparation, you know, could very well hamper the accuracy and fairness of the census count in 2020. >> would the chairman indulge

me? i have one last question. the alleged cost savings of such cancellations are penny wise but pound foolish. mr. goldenkoff. >> yes. >> gao has reported that higher response rates save money. can you try to explain that for us? >> sure. the most cost effective way to count people is when they fill out the questionnaire that's either mailed to them or when they self-respond via the internet. every time -- and if they don't self-respond through either of those methods. then the census bureau has to send out enumerators to knock on their doors, and as many as six times. and so you can see how costs add up very quickly. it's both in terms of the time that the enumerators are paid for, the wages they earn, as well as for the mileage they're reimbursed. so the most cost effective way of conducting the census is to

get up to increase that initial count. >> all right. gentleman yields back. mr. powner, i'm sitting here trying to figure out whether anything could be more inherently predictable than the decennial census. i'm struggling to think of something you see that far in advance that's coming. so not to ask this in too sophisticated of a way, but how the hell did we get here? if we know it's coming every ten years. how do you get on a critical watch list? >> well, the thing that's a bit frustrating is the same thing happened in 2010, and we were involved in that in 2010. so that's when we cancelled the handheld project. this committee held multiple hearings on the overruns there and the eventual cancellation. the interesting comment that secretary ross talked about 80% of the money going from beyond 2018. from an i.t. perspective, that

should not be the case. we should start early, plan more appropriately, get the right contracts in place. and deliver. what happens with the decennial historically and it's going to happen again, can we get this done late and secure it late? yes. but you know what it's going to be? it's going to be a higher cost. you pay a premium for doing everything late. i would have major concerns about the 15.6 and still hitting that even with the 10% when you look at the i.t. management of this. the thing that's good is having the commerce department's governance and oversight, i think will greatly help. but we still haven't changed the basic management how we're doing things at the census bureau, so we would still have concerns about additional cost overruns going forward. but it is -- it's not acceptable that we repeat the same sins of the past that we do every ten years with the decennial. >> rueally hard to explain to te same people you're trying to

convince to voluntarily comply. so, which leads to my next question. what are the impediments, mr. goldenkoff, what are the impediments to what i call voluntary compliance, although i'm sure that's not the right technical term? >> sure, there are a number of barriers. one that was already mentioned was just declining civic engagement. people are increasingly disconnected from the government. you know, so that requires trust. building trust between the census bureau and the public. there's also a cultural divide. you know, it was already mentioned, too, when the government comes knocking on your door, you know, the census is sometimes seen as not something that's done for you or on your behalf. it's sometimes seen as something that's done to you. it's not always seen in a positive light. then there's structural barriers. people living in makeshift housing, nonconventional housing, people doubling up.

people living in basements and converted attics, so they're just physically hard to find. there are language barriers out there, so just reaching those folks is difficult as well. and this is why all this up front work is so important and why the census bureau is at the point that it is right now. it's very difficult to make up for either funding that was not provided or time that has already elapse sad. you can get behind the curve very quickly. >> accepting the fact we can't do all of the preplanning for 2020 that needed to be done in 2012, 2013, so the best we can do is the 2030 census there, with specific reference to the 2020 that's coming up, if i heard the testimony right, there's an opening in the census director's office? so i'm going to start with ms. gupta. if you were the census director,

you had to do one thing, you got to do it today, on the chance that mr. ross or his folks may be listening, what is the single most important thing you would do today for the 2020 census? >> it's hard to say a single thing. i think a couple things need to happen all at once. the i.t. infrastructure needs to be tested, clearly, for all of the reasons that have been previously stated. and i think that the partnership come communications campaigns need to be out there starting in the next couple of months. they've got to ramp up this whole infrastructure in order to really educate folks about why it's important to be counted, and to overcome some of the real concerns around mistrust in government and cybersecurity. and so -- you know, i want to add that when the secretary was testifying, he said that he anticipated 800 partnership specialists, which is well below the 2010 number of 3,800. so those plans are still

inadequate, and i think that right now, there is a lot of concern that the hardest to count communities are not going to be adequately counted, and that will result in a failed census even with the $1.6 billion allocated. there has to be an immediate investment in that infrastructure and kind of real planning around it just as there is with the i.t. >> ms. rice, what's the single most or two if you want. what needs to be done today given the fact that we can't change the lack of preplanning? >> i really think the i.t. infrastructure has to be in place. we don't want to be in a position where we were in 20 tent where they had to revert to a paper based operation at the last minute. and they didn't have the -- they had some infrastructure there, and they had to build on it. it was a problem. there were backlogs and the office sizes weren't compatible for a paper operation. it was just trickled down all the problems. they got it done, but it wasn't pretty.

so getting that i.t. infrastructure in place. >> mr. powner, you can't use the word infrastructure. what would you do? >> so, with the i.t., and that's my world, mr. chairman, i would say those 43 systems theed to be delivered and secured. we have this data in our testimony that looks all neat, and it's very difficult to get a handle on where all those systems are from a delivery point of view integration and security. that needs to be managed weekly. someone needs to dog that weekly in terms of what we got done, what remains next week, and are we going to hit those dates. there's key dates in early 2018 that are going to be key to getting this stuff delivered to be included in the end-to-end test. you have to dog the 43 systems from a security perspective, chairman, representative hurd's comments about the process, it's fine to bring in people, but you know what. you need to actually do the assessments. and you need to fix the vulnerabilities and sign off on the atos. that needs to be managed weekly

to insure that gets done and we don't have a major breach. >> last question, the dim uniition of the public trust didn't happen over night. it's not going to be remedied overnight. what is perhaps something members of congress haven't thought about or something you would try to do to restore some modicum of public trust so we can raise the voluntary compliance rate in 2020? >> sure. i don't -- you know, off the top of my head, i don't think there's anything new. i would certainly be leery of trying anything that has not been used before at this point. just because it would be one more risk for the bureau to have to manage. what i think needs to be done is focus on the tried and true methods, things that are known to work, and i would point to the partnership specialists and the whole communication and outreach campaign. the partnership specialists are so critical for engaging the different people that they represent, their different stakeholder groups.

and they're the ones who create that trust. they're the trusted voices. if secretary ross or any of us goes up on tv or, you know, makes -- encourages people to respond to the census, it's not going to have the same level of response as people who know the community, your local minister, your school principal, it was mentioned before that the census bureau hopes to hire as many as 800 partnership specialists. that sounds like a large number, but if you just look at the work load, if you think of over 3,000 counties in this country, that's a huge work load for them. it's a three-dimensional issue. they have to cover both the physical geography, the different channels of communication, plus the deep dive into all of the different communities. and so i think a focus then on the partnership campaign and making sure that that's adequately funded and gets the resources they need will be critical for increasing the hard

to count communities. >> mr. palmer. >> thank you, mr. chairman. ms. rice, can you provide your initial thoughts and observations of the 2018 end-to-end testing? >> sure. so we did visit all three locations. and found, just to remind everybody about 2010, they did use a device in 2010. the field data collection automation worked for canvassing. so using a device in the field for address canvassing, they did last decennial. so it actually went all things considered, okay. we didn't see any major -- >> was there any aspect of the testing, raising any potential issue for future address canvassing? >> we did think that in the rural areas, that there's

something called an optimizer that sends the lister to a location, that it wasn't as efficient as maybe it would have been. we're trying to look into that to find out why, but there were people, some folks traveling a couple hours to get to locations, which seemed fairly inefficient. >> your office raised some issues. you actually published a report questioning the bureau's in-office canvassing, cost estimates, as well as warning that an increase in the in-field canvassing could increase costs. will the percentage of addresses sent for in-field work during the testing be a factor in your review as you consider that issue? >> yes. so the in-office canvassing basically determines whether something grew or ch