Tag: Administrative records

What Price Democracy?

Census Project Co-Director Terri Ann LowenthalBy Terri Ann Lowenthal

The Census Bureau was off to a relatively good start this year in the mysterious and powerful world of those who hold the purse strings, known fondly to many of us as the House and Senate appropriations committees. Or so I thought.

Last week, Commerce Secretary Penny Pritzker took the hot seat before the Senate panel that funds the federal government’s commerce, justice and science programs. This would be the subcommittee (albeit, with several new members) that barely acknowledged the existence of a census at last year’s budget hearing. The panel is heavily populated by lawmakers from coastal states, who apparently have nightmares about uncharted weather catastrophes and depleted fishing stocks.

But the 2020 Census got their attention this year, maybe because the Obama Administration requested a 91 percent funding increase to ramp up planning in Fiscal Year 2016 for the next decennial count. Which, if I haven’t mentioned recently, will be in full swing five years from now.

Panel Chairman Richard Shelby (R-AL) actually led off his opening statement and questioning with census-related concerns. He noted matter-of-factly the need for a significant funding increase to double-down on 2020 Census planning, and he cautioned the secretary to closely watch preparations to avoid future cost-overruns that could leave less money available for other Commerce Dept. programs. The subcommittee’s senior Democrat (and former chairwoman), Sen. Barbara Mikulski (D-MD), briefly mentioned the technology “boondoggle” before the 2010 Census, and that was it. On to New England fisheries, support for U.S. manufacturers and that pesky “polar gap” in weather satellite coverage.

But things went downhill for the Census Bureau from there. The new chairman of the House Commerce/Justice/Science spending panel doesn’t much care for the American Community Survey (ACS), the modern version of the census long form. It’s “intrusive,” he told Secretary Pritzker when she appeared before his subcommittee this week, and the government doesn’t have a right to ask about anything other than the number of people in a household… or ancestry. Ancestry? Where did that come from?

But let’s move on. Ever since the Census Bureau wrapped up the last decennial census, appropriators have indicated that they aren’t willing to spend more on the 2020 Census than they did on the 2010 count. The lifecycle cost of the last population canvass was roughly $13 billion. The Census Bureau thinks it can meet that goal if all of the sweeping reforms it is considering work as envisioned. That’s a big “if,” what with budget shortfalls delaying, cancelling or streamlining critical research and testing of these new initiatives over the past few years. We simply don’t know yet if a markedly redesigned census can ensure an accurate count, especially in historically undercounted communities, and produce the detailed race and ethnicity data needed to implement the Voting Rights Act, as a threshold matter.

But Rep. Culberson apparently isn’t satisfied with those cost-saving efforts. “We don’t have $13 billion to spend on a census,” the chairman told Secretary Pritzker. The congressman wanted to know if the Census Bureau is ready to use Internal Revenue Service records and other government databases to help bring down census costs. The secretary gamely tried to emphasize the importance of testing, testing, testing, to see if that idea, which of course is under consideration, is a viable option. But I’m not sure the chairman has thought this through. If the Census Bureau doesn’t have enough money to thoroughly vet the use of administrative records to supplement or replace direct address canvassing and door-to-door visits, the 2020 Census could cost $1 or $2 billion more than the congressman says we can’t afford to spend. Nevertheless, Rep. Culberson again made it clear that “we won’t have the money next year” to meet the Census Bureau’s budget request.

And that tells me just about everything I need to know. Because if Congress can’t spend $13 billion over the course of a decade to carry out its very first obligation under the U.S. Constitution and to ensure fair political representation for all communities, no matter how difficult to count, then we might have to kiss our storied democracy good-bye and book a seat on that one-way mission to Mars. After all, the Johnson Space Center is pretty darn close to Chairman Culberson’s Houston district. I’m thinking some of that census money will end up fueling a mission to outer space.

Silver Bullets and Red Flags

Census Project Co-Director Terri Ann LowenthalBy Terri Ann Lowenthal

Today, I am going to talk turkey.

No, not the Thanksgiving kind. I had my fill and besides, I am still focused like a laser on the 2020 census. Which, if I haven’t mentioned recently, is only five years away.

That means it’s time to get down to the nuts and bolts. Today’s fascinating topic: multiple response options.

Ladies and gentlemen, the cyber-census is here. Internet response! Email and text message reminders to answer the census! Smartphone apps to fill out your questionnaire! Twitter will be abuzz with daily response rates. It all seems so… so… 21st century! Well, at least more up-to-date than relying solely on paper forms sent and received via U.S. mail. And really, it’s a tad embarrassing that Girl Scouts will get to sell their cookies on the web before our nation’s largest peacetime activity goes high-tech, don’t you think?

Congress is on board with the new approach. Visions of saved dollars are dancing in lawmakers’ heads. So much so that Congress thinks the Census Bureau has it all figured out. Flip the switches and watch the Internet light up with a population count. Why bother with research and field tests and focus groups, when it seems like everyone is plugged in these days. Those activities cost money, and Congress doesn’t seem inclined to pony up a lot of dough to make sure we can do this right.

Truly, the thought of the 2020 Census running as smoothly as the click of a mouse (or tap of a finger) is bliss. (We will not dwell here on the initial failures of healthcare.gov, which crashed under the weight of a few million inquiries, but had a few months breathing room for the first enrollment period while experts fixed the bugs. Because, really, the Census Bureau anticipates up to 8 million hits a day on the 2020 Census website, and the window of opportunity for self-response is a mere several weeks. What could possibly go wrong?)

Congress is so convinced that a cyber-census is a silver bullet to check rising costs, it doesn’t see the wisdom of fully investigating this radical departure from previous counting methods. In their first crack at the FY 2015 Commerce Department funding bill last spring, House members—anointed by the Constitution as the primary beneficiaries of an accurate census—knocked out the entire requested budget increase for 2020 Census research and testing.

I hate to be a glass-half-empty person, but I’m thinking that Congress doesn’t do long-term planning well. Maybe it could start with a report the Census Bureau itself issued last month: Computer and Internet Use in the United States: 2013. About three-quarters of American households have an Internet connection. But that is for the population as a whole. Only 60 percent of black households and 66 percent of Hispanic households have Internet access. The figure drops to under 60 percent for the over-65 crowd. Less than half of households with incomes under $25,000 have home Internet access. The digital divide also affects households led by individuals without a college education and with limited English language proficiency, and those in nonmetropolitan areas.

And then there’s the small matter of cyber-security. People are a little freaked out by the drip, drip, drip of news about data breaches at major U.S. companies— Target, JPMorgan Chase and Home Depot, to name a few—and the hacking of government agency systems (the White House and State Department are the latest apparent victims) and Hollywood conglomerates (Sony). Call me paranoid, but experience tells me that it could take only a whiff of a problem to throw the best census operational plans off track. (As Exhibit A, I give you the 1990 Census, when the U.S. Postal Service returned several million questionnaires to the Census Bureau as “undeliverable” because housing units, primarily in rural areas, received their mail at a P.O. Box, not the street address on the census form. The extensive media coverage—this, when we still received our news slowly, from TV, radio and newspapers—shook public confidence and sent the bureau into full damage control mode.) Picture the consequences in 2020 of even a handful of census phishing scams or, heaven forbid, a cyber-attack on the Census Bureau’s massive digital database, with news pinging around the Internet at lightening speed.

So where Congress sees a silver bullet, I see red flags. Yes, of course there should be an Internet response option for the 2020 Census. Otherwise, we might as well send the marshals out on horseback again. But can the Census Bureau save enough money to keep 2020 costs at or below the 2010 Census budget, as lawmakers have directed, and still produce an accurate count, especially in communities with historically higher undercount rates? I think Congress has its eye on the goalpost without thinking through the plays it will take to get there and score.

This is how I see it. First, at the risk of sending you to bed with nightmares, I will gently remind everyone of the tech failure that added $2 billion to the 2010 Census cost and dashed hopes of sending census takers door-to-door with nifty handheld electronic devices to count reluctant households. If there is a better reason to invest in careful planning, I can’t think of one right now.

And I’m worried about the quarter or more of households that won’t respond in the initial phase of the count. Let’s not pull any punches: most of the people who are more likely to be missed in the census are less likely to have the means to respond electronically. Furthermore, the characteristics of households with lower rates of computer usage (including handheld devices) and Internet access parallel those of households with “low self-response scores” in the Census Bureau’s newly updated planning database. That means many households that don’t respond via the Internet won’t mail back a paper questionnaire either, especially if the strategies for boosting self-response aren’t thoroughly vetted. (In the 2014 Census Site Test, only 3 percent of households that were asked about their preferred method of advance notification chose the email or text option over mailed materials. I’m guessing Americans are wary of electronic messages from unknown sources, as they should be.)

And here’s where the budget comes into play again. Congress wants the Census Bureau to wave a magic wand and plan a census that costs a lot less, without giving the bureau enough resources to make it all work or conducting the informed oversight needed to make sure that it will. What happens to the households that don’t self-respond? Tracking them down is the costliest part of the census, and the bureau is exploring ways to streamline that operation, with fewer boots on the ground and fewer knocks on each recalcitrant door. Congress is pressing the agency to rely more on data the government has already collected through programs such as Social Security, food stamps and Medicaid. The FY 2015 census tests will start to shine a light on whether administrative records can replace much of the pre-census neighborhood address canvassing and some of the door-to-door visits. But with Congress capping the 2020 Census budget in advance—something it has never done in modern census history—the Census Bureau might have no choice but to fill in the blanks with data that are neither acceptably accurate nor sufficiently comprehensive.

That’s a topic for another day. But I see lawmakers chasing a lot of silver bullets when they should be biting the bullet, to make sure the ammunition hits its target. In the meantime, I’ll keep waving the red flags. Maybe Congress will notice before it’s too late.

And with that, we wish our readers and census groupies everywhere a happy, peaceful holiday season. Thanks for being a part of our coalition. See you next year!

Trying to Read the ACS Content Tea Leaves (Good Luck With That)

By Terri Ann LowenthalCensus Project Co-Director Terri Ann Lowenthal

There’s a certain je ne sais quoi to the way Congress deals with the census. One minute, lawmakers are trying to deep-six the Census Bureau’s signature American Community Survey (the modern day census “long form”). The next, they’re ignoring the Census Bureau altogether. Or they’re using it as a piggybank for their favorite programs. Those would be the programs that largely rely on census data to allocate the money legislators from both parties pilfered from the Census Bureau. It’s all very confusing.

But the Census Bureau has tried to rise above the hopelessly mixed signals from Congress (We don’t like what you do. We don’t care what you do. We don’t want to pay for what you do, even though everything else we do depends on it. What exactly is it that you do?), forging ahead with the most rigorous review to date of questions on the ACS.

The Census Bureau is completing the first phase of its multi-year ACS Content Review effort. On October 31, it published a notice in the Federal Register proposing to eliminate several questions that the agency concluded pose a greater burden on the public, relative to the benefits of the data to policymakers and program administrators. The bureau has cool scatter-plots and matrices and charts that show how ACS questions stack up on a cost-benefit analysis, but the bottom line is that Congress itself has asked for most of the data, directly or indirectly, to set policy, allocate resources, and implement programs. A handful of questions tip the scale too far on the cost side and are on the chopping block for the 2016 ACS.

Let’s stipulate that the survey can appear daunting to those who receive it each year. That would be less than 3 percent of American households, although if you believe ACS opponents, you’d think the government had all of us chained to our desks, depriving us of life, liberty and the pursuit of happiness until we answer 72 questions about ourselves, our families and our homes. The range of topics can make it seem like the Census Bureau is being a bit nosy. Naysayers like to point to questions about what time people leave for work or whether people have difficulty dressing or bathing. I am confident these critics do not include legislators who issue triumphant press releases about traffic congestion mitigation projects and services for people with disabilities they secure for folks in the home district.

But, where was I? Oh yes, scrubbing ACS content for errant questions. Turns out that questions on your marital history, what you studied in college and whether there’s a business or medical office on your property don’t produce information that legislators and government agencies use widely.

It’s a good thing, by the way, that the Census Bureau still plans to ask whether you are married or not. For the 2000 Census, the bureau decided to move the “marital status” question from the short form, which everyone gets, to the long form sent to a sample of households. Ultra-conservative Sen. Jesse Helms (R-NC) balked at this slap to a sacred family institution, and his colleagues adopted a resolution to keep the query on the 100-percent form. (Unfortunately for them, timing is everything in a census: the questionnaires had already been printed when the resolution passed.) House members had already jumped on the “more data is better” bandwagon, with more timely bills to add questions on family caregivers, home computer use and Internet access, and to preserve the ancestry question. But once the enumeration started, lawmakers raced to distance themselves from the forms flooding mailboxes; there were seven proposed House bills from March to May 2000 to limit the number of census questions Americans must answer (in most cases, just name and number of people in household). I do not think the law requires consistency in census gripes.

For the current round of questionnaire trimming, the Federal Register comment period closes on Dec. 30, 2014. The agency plans further research on alternative sources for data gathered in the ACS (such as administrative records) and the wording of questions, some of which is problematic. (Millennials, for example, can’t relate to “dial-up service” on the Internet access question. Go figure. Boomers probably have nightmares just seeing the term. Screeeeech ….)

I’m betting that demographers, researchers and policymakers interested in STEM education will fight to save some of the questions the Census Bureau wants to drop. The bureau must finalize all ACS content decisions (adding and dropping questions) before the April 1, 2017, legal deadline for submitting census topics to Congress; the actual questions go to the legislature one year later.

Congress will have the final word on content, which might be difficult to parse when the right hand doesn’t know what the left hand is doing. The House has voted twice to make ACS response voluntary (a stake in the heart of small-area data) and once to eliminate the survey altogether. Yet, lawmakers want the data to divvy up $400+ billion annually for highways and transit, education, emergency preparedness, rural development, food and housing assistance, job training, and much more. Good luck with that when the data disappear.

Putting 2020 Census Innovations to the Test

Census Project Co-Director Terri Ann LowenthalBy Terri Ann Lowenthal

The rubber is about to hit the road.

Or, in the parlance of a 21st century census, fingers are about to hit the keyboard. At least that’s what the Census Bureau is hoping when the 2014 Census Site Test kicks off this spring.

Yes, census fans, the first major 2020 Census field test is on the horizon. According to a Dec. 24, 2013, Federal Register notice, 192,500 lucky households in Rockville, Md., and Washington, D.C., will be a laboratory for enumeration strategies and operational innovations that finally could push the decennial count off its 230 year paper-and-pencil foundation.

The Census Bureau is under orders from Congress to keep the cost of the next census down — way down, as in no more than the cost of the 2010 count. So the bureau really wants people to fire up those desktops, or pull out their laptops, tablets or smartphones, and help reduce spending on paper forms, postage, processing and door-knocking census takers. The 2014 test will ask some households to pre-register for the count and indicate their preferred method of contact with the Census Bureau, such as email or cell phone. People can take it upon themselves to answer the census via the Internet or by phone. The bureau will nudge non-responders by email or snail mail, sending paper forms as needed. (We should pause here to contemplate that today’s high tech gadgets and preferred methods of cyber-communication might be considered ancient when 2020 rolls around.)

Holdouts will fall into the nonresponse follow-up universe, historically the most costly census operation. After an unsuccessful attempt to automate door-to-door operations in 2010, the Census Bureau will have field workers test a range of modern devices, including iPhones and iPads, to gather information at the door. Some census-takers will be invited to bring their own device (BYOD) as part of the field test, and the bureau will evaluate using ubiquitous Google Maps, instead of paper maps, to guide enumerators through neighborhoods.

Census managers also will test their adaptive design strategy, a fancy moniker for deciding, in real time, which homes enumerators should contact, in what order and when, how many times, and using a contact method likely to elicit responses. Previously, census takers set out with a list of addresses, made their rounds without guidance, and kept calling and visiting recalcitrant households up to six times before resorting to proxy sources, such as a neighbor, for information.

The Census Bureau will put its toe in the water of a potentially controversial new approach to reducing the follow-up workload: using administrative records — data from government databases and third-party (commercial) sources — to identify vacant housing units and to enumerate households that don’t respond willingly. Given public angst over the NSA and “big data,” I’m waiting to see how Congress and ordinary Americans react to the idea of a massive sharing of personal information, albeit on a one-way street (into the Census Bureau, but not out). I’m worried that substituting administrative data for the real thing will not yield the detailed race and ethnicity data the census requires. And I wonder how grassroots organizers will structure their “be counted” campaigns in the face of “don’t worry, we counted you another way.” But, hey, someone’s got to lose sleep over this stuff!

If you want to weigh in on the 2014 Site Test design, you have until February 24 to submit comments. Oh, and before I forget, a note to the Census Bureau: Could you please send my dad a paper 2020 Census form from the get-go? He doesn’t use a computer or cell phone (he does still have a slide rule, though), but he’ll be 89 and would sure like to make his daughter proud.

Losing Sleep (While Counting Sheep)

by Terri Ann LowenthalTerri Ann Lowenthal

Census stakeholders, my head is spinning with thoughts of 2020 census planning.

Sure, you might think the next census is too far in the future to keep you up at night. (Plus, I bet most of you would rather let me do the worrying.) But let me quote the Census Bureau’s new counter-in-chief, John Thompson, who told a House subcommittee at a September 11 hearing, “budget uncertainty is causing significant concerns for the 2020 census program as we enter that period during which it is crucial to conduct tests so that we can begin applying new technologies and methods … We have already delayed planned research and testing activities to later years … We cannot further delay critical research that will help us make critical design decisions for those systems.”

Let’s stipulate to one shared goal: The 2020 census can’t look like the 2010 census. For one thing, the nation can’t afford the $30 billion price tag of repeating an outdated census design. Equally important, the way we communicate with each other has changed rapidly.

Automation is the buzzword for 2020, but despite the fact that many of us live on our gadgets, a cyber-census (you heard it here first!) isn’t as simple as it might seem. Will data be secure if census-takers bring their own devices (BYOD)? Can we design a questionnaire that people can navigate as easily on a smartphone as on a computer and that works across all operating systems (the ones we use now and the ones that Google, Microsoft and Apple have yet to dream up)? Will people welcome emails, text messages and cell phone calls from the Census Bureau (where did they get my information — from the NSA?)? And what about people who want to respond online without a unique code tying them to a specific address? The challenges are broad and deep.

There could be significant savings (up to $2 billion, the agency says) if the bureau tapped into demographic, housing and geographic information already in the hands of other government agencies. These administrative records could, potentially, eliminate the need for a universal sweep of the nation’s addresses before the census starts; identify vacant homes, to avoid costly follow-up; determine the best days and times to call or visit unresponsive homes; identify households that might have neglected to report every resident; and yes, even to add people to the count without knocking on their door.

But the Census Bureau must work out separate deals with each federal agency and state holding useful records; much of this data-sharing could require changes in federal or state laws. And each dataset has its strengths and weaknesses — Medicaid records, for example, do not have names or street addresses, only social security numbers and birth dates — which could require linking one set of records to another.

I know I sound like a broken record, but the Census Bureau needs money to figure all of this out in time. The bureau can execute a fundamentally redesigned 2020 census for the 2010 census price tag (plus inflation), Director Thompson says. Invest now, save later — that’s the bottom line.

So let’s review where things stand for the fiscal year (FY2014) that begins today. For starters, we can flip the calendar back to FY2013 for a while. The Census Bureau must make do with last year’s funding level, which was 13 percent below its budget request, while Congress figures out how to … ummm … get its act together. Then we hold our breath while lawmakers decide whether to slide backwards another $45 million (courtesy of House appropriators; H.R. 2787), give the agency most of what it needs (thanks to far-sighted Senate appropriators; S. 1329), or settle on some amount in between.

The census is a 10-year process, a cyclical activity that starts small and builds to the grand finale of enumerating every household in a vast, diverse nation. There is a ramp-up to that denouement, one that starts modestly and escalates as we hurtle towards the “zero” year. But the direction, once planning starts, has to be up.

Wikipedia describes a “ramp up” as the period between product development, and maximum capacity utilization, characterized by product and process experimentation and improvements. Sounds like a logical business practice, right? (Think of the time and investment in research it takes to bring new prescription drugs to the market.) You would think lawmakers would want the Census Bureau to operate like an efficient corporation. But long-term fiscal planning isn’t Congress’ strong suit, now, is it?

The Census Bureau needs $245 million in FY2014 to keep 2020 census planning on track; the House bill cuts that amount by more than a third ($91 million). Already delayed by a year are all of the tests scheduled for this year and next; some tests have been cancelled. The bureau has pushed back the field test of the 2020 Census form to FY2016, which is getting uncomfortably close to the April 1, 2017, and April 1, 2018, deadlines for submitting topics and questions, respectively, to Congress. Perhaps most troubling, the Census Bureau won’t nail down a design framework until late FY2015, a year behind schedule, leaving less time to develop systems and operations.

Did I mention that the next census starts in less than six years? The Census Bureau can do a lot of things, but it cannot stop the clock. I bet Director Thompson is having a few sleepless nights, too.

Slip Sliding Away: The Risk Factor Goes Up As 2020 Census Funding Goes Down

Editor’s note: This blog post was revised on July 11 to reflect more detailed information released by the Appropriations Committee.

by Terri Ann LowenthalTerri Ann Lowenthal

I’m starting this blog post with one of my famous census headaches.

That’s because House appropriators are driving the wrong way — downhill and backwards — on the up ramp to Census 2020. I see a collision in the not too distant future: Congress doesn’t want to pay a lot for the next census, but it won’t put gas in the tank to keep the Census Bureau from stalling on the road to achieving that goal.

The Census Bureau is trying to wrap up its research and testing phase for 2020. Fundamental reforms in methodology and technology are on the table as the agency strives to curtail the big-ticket enumeration items, especially universal canvassing to confirm the address list and reliance on paper forms for every household in the nation. The bureau had hoped to select a design framework in 2014 so it could move forward with operational and systems development; now that’s not going to happen until 2015.

This year (FY2013), lawmakers cut the bureau’s budget request by 11 percent (more if you count another sneaky transfer of money from the Working Capital Fund). And now the House Appropriations Committee is taking another whack at the budget for Fiscal Year 2014; the Commerce, Justice, and Science subcommittee allocated just $845 million in its “committee mark” today, $44 million less than current year funding. The president requested $982 million, only a hair over his request for FY2013. That’s a cut of $137 million, or 14 percent, in discretionary spending.

Talk about rolling backwards down the hill… can I have another Excedrin, please?

Call me crazy (maybe it’s just the pulsing in my temples), but it almost looks like Congress has decided we really don’t need to plan for the next constitutionally prescribed decennial census — coming to a neighborhood near you in less than seven years. The Census Bureau can just envision ways to cut the cost of counting, cross its fingers, and hope for the best.

Field tests scheduled for 2013? A 28 percent funding cut forced the Census Bureau to cancel some and delay others, including key tests of strategies to encourage and facilitate electronic response in ways that preserve public confidence and ensure data security. An Internet response option could save big bucks during the census, but the methodology is far from simple. Do you send a letter to each household first, with a unique code for answering online, or could people somehow “pre-register” to receive a code for online participation? Should you mail a questionnaire if a household doesn’t answer online by a certain date, or just send an enumerator to the door? Would Americans welcome or reject text messages and emails from the Census Bureau, urging them to be counted, and how can everyone avoid the inevitable phishing schemes from false census.gov addresses (I’ve received two myself already)? I can’t answer these questions, and neither can the Census Bureau without thorough evaluation.

Possible changes to census questions? The American Community Survey is a cost-effective test-bed for proposed revisions, but the 2013 funding shortfall will push back a key content test from 2015 to 2016, which — in my opinion — is getting too close for comfort to the statutory April 1, 2017, deadline for submitting census topics to Congress.

The Census Bureau has made the hard choices for 2013: moving some research from the field to a desk; pushing back deadlines; reevaluating projects planned for next year. It is moving forward, for now, with research into a potentially big money-saver: using administrative records to identify possible undercounts and add people who might be missed during the census. But with the National Security Agency data mining program looming large in the American consciousness, that research must be thorough and robust before the Census Bureau can even hope for buy-in from lawmakers and the public. Important research into public trust and confidentiality concerns has been pushed back already. Without a funding increase in 2014, the agency won’t even be able to cover its staffing costs.

The risk of an unacceptably costly or unacceptably inaccurate census, or both, continues to go up as Congress continues to squeeze the Census Bureau’s budget during the critical research, testing and development phases of our decennial enumeration. The census is a 10-year program. Congress can invest in planning now, to keep the life-cycle cost of the enumeration in check ($13 – $18 billion, depending on the design, according to independent auditors), or keep letting the car roll backward — and dig itself out of a census ditch for $30 billion in a few short years.