The Case for a Mandatory ACS: Count on Americans to Get It Right

By Terri Ann LowenthalCensus Project Co-Director Terri Ann Lowenthal

I have to start today’s census musing with an apology of sorts. It occurred to me, as I relay Appropriations Committee proceedings with tongue firmly in cheek, that some readers might think I do not like fish. Or, to be more precise, that I do not appreciate the importance of the fishing industry and coastal zone preservation. Au contraire, mon amis: I love the fishermen (oops, sorry; channeling you-know-who for a moment there), regularly consume fish, and live in a coastal state myself. It’s just that, when one is waiting patiently for a sign — any sign — that lawmakers appreciate the fundamental importance of a successful census and reliable data to the very foundation of our democracy, and the conversation keeps turning to counts of, um, fish, one can become despondent over the possibility that the nation’s largest, most inclusive, and most complex civic activity will get the short end of the budget stick. Again.

I had to get that off my chest before the Senate Appropriations Committee takes up the Census Bureau’s Fiscal Year 2017 funding bill (Commerce, Justice, and Science, or CJS) this week. As CJS Subcommittee Chairman Richard Shelby (R-AL) helpfully noted when Commerce Secretary Penny Pritzker testified in March, the largest chunk of new money for department programs is for 2020 Census planning.

The total request for the 2020 Census is $778 million; related activities, such as the new enterprise data processing system and geographic system updates, will support that effort. The Census Bureau also needs $251 million for the decennial census’ conjoined twin, the American Community Survey (ACS), to maintain a reliable sample size and continue researching new methods that will make it easier for the public to respond — a primary congressional goal.

Chairman Shelby told the Secretary that he is worried the Census Bureau is behind schedule in meeting its milestones to prepare for the census. Vice Chairwoman Barbara Mikulski (D-MD) reminded everyone that past enumerations have been plagued by “techno-boondoggles” (I did not make up that word), but that an accurate census remained the top goal. We will wait with baited (a fishing term!) breath to see how much money the committee is willing to spend on the bedrock of the nation’s democracy, but I find it hard to grasp how cutting the budget at this point will move the needle in the right direction.

At the March hearing, Sen. James Lankford (R-OK) engaged the Commerce chief in the most substantive discussion about Census Bureau programs. In 2012, then-Rep. Lankford helped steer his House colleagues off the cliff into a data black hole, by championing an amendment to eliminate the ACS entirely. His distaste for the survey — whose results help guide the allocation of billions of dollars through other parts of the same appropriations bill, she said without a hint of irony — seems not to have waned; he offered, and then withdrew, an amendment to make ACS response voluntary during consideration of last year’s CJS bill.

This year, the Senator pressed Secretary Pritzker for information on Census Bureau research into making the ACS a more pleasant experience for households fortunate enough to be selected for the nation’s premier survey on the well-being of our communities. Okay, he didn’t exactly say it that way. Nevertheless, we haven’t heard much about the “2015 Summer Mandatory Messaging Test,” so perhaps it would be helpful to consider the findings before lawmakers think again about relieving put-upon Americans of their duty to answer a few queries confidentially for the common good.

Using the September 2015 ACS sample as a test-bed, the bureau evaluated the consequences of several design and messaging changes to ACS materials (e.g. envelopes, questionnaire, reminder postcard, etc.). The modifications included “softening” (by wording choice, visually, or both) or even removing the mandatory response tagline (which currently reads, “Your response is required by law.”) from some or all of the mailings, and highlighting the survey’s importance and benefits.

But before we look at the results, let’s take a quick trip down memory lane, to when lawmakers first got the itch to let Americans opt out of a civic duty that helps ensure a fair, equitable, and informed society. In 2003, at the direction of Congress, the Census Bureau tested the implications of making the ACS a voluntary survey. The outcome was not pretty: mail response rates would plunge; costs — and, ironically, the burden on the public — would rise significantly ($90+ million/year, according to the latest estimate), due to greater reliance on more costly telephone and door-to-door follow-up and the larger sample size necessary to compensate for lower response; and data reliability for smaller communities (towns! neighborhoods! rural counties!) and smaller populations (people with disabilities! veterans! ethnic subgroups!) would drop, perhaps beyond the point of acceptability.

The findings were so stark that Congress barely said another word about mandatory response for almost 10 years. (Our Canadian friends, however, apparently didn’t get the memo. Former Prime Minister Stephen Harper’s Conservative Party pushed through a bill to make their census long form voluntary. Predictably, the renamed 2011 National Household Survey was a data disaster for less populous and geographically smaller areas. The outcry from the business community, municipal officials, and policy researchers was loud enough to prompt a swift restoration of the mandatory survey when Prime Minister Justin Trudeau took office last fall.)

The Census Bureau’s 2003 test and Canada’s experience provided consistent evidence of what Congress and the nation would lose if ACS response were voluntary. Never ones to be deterred by a universally panned idea, however, House Members voted three times to make ACS response optional; Senators, fortunately, refused to go along.

The Census Bureau is taking congressional concerns about the ACS’s perceived intrusiveness seriously. It has consulted with the best survey experts in the business and is researching ways to replace some survey questions with data from other government datasets (administrative records) and ask others less frequently. Then there is last year’s test of design modifications and different ways to convey that response is mandatory.

And whadda’ya know? The experimental design that strengthened the mandatory message and added a plug about the importance of the survey yielded the best outcome in terms of response rates, data reliability, and cost, even compared with the current materials. Self-response and final response rates were “significantly” higher, leading the bureau to project cost savings of $7.3 million per year and better data reliability (i.e. reduced margins of error in the data, which is, after all, derived from a sample). In fact, if the bureau continued to spend the same amount of money on the survey, it could increase the sample size by 4.7 percent, thereby reducing margins of error even more. (If the bureau reduced the sample size but maintained current data reliability, costs would drop even further.)

On the flip side, experimental designs that diminished the emphasis on mandatory response yielded, almost across the board, “significantly” lower response rates. If that were to happen, the Census Bureau would have to decrease the sample size by up to 12 percent, resulting in worse data, or increase the sample size by up to 11 percent to maintain data reliability, at an additional cost of up to $42 million. In other words, pick your poison: pay more, or risk more unusable data, all in the name of keeping the survey mandatory but not letting Americans in on the secret.

Here’s what I think. Most Americans are willing to do their part in support of a strong democracy and good governance. Tell them clearly and honestly why what you’re asking them to do is for the public good, and demonstrate that you don’t take their cooperation lightly and will not abuse the privilege, by continuously innovating and having some faith in your most important customer — the American people.

Congress, are you listening?

A Salute to Our Veterans (Courtesy of the ACS)

by Terri Ann LowenthalTerri Ann Lowenthal

Let me start with a timely salute to our nation’s veterans. All 21 million of them, including 2.4 million African American and 1.2 million Hispanic former service members. Shout-outs to Killeen, Texas, and Clarksville, Tennessee, where veterans comprise a quarter or more of local residents. Hats off to the more than nine in ten veterans with a high school diploma — a greater proportion than the general population. And is it any wonder that these patriotic fellow citizens are twice as likely as non-veterans to hold a job in public administration?

Oh, sorry, I digress from the focus of this blog. But really, people, it’s important that we know this stuff — and more — about those who defend our freedoms. About three-quarters of our living military veterans served worldwide while the country was at war. More than a quarter of both Gulf War and post-9/11 era vets live with a service-connected disability. Nearly 30 percent of veterans reside in rural areas, but rural vets represent 41 percent of those enrolled in the VA health care system. Veterans in rural communities are more likely to have at least one disability compared to non-veteran rural dwellers.

Raise your hand if you know where I’m going with this. That’s right: a lot of what we know about our veterans comes from the Census Bureau’s American Community Survey (ACS). Businesses, nonprofits, and federal, state and local leaders use ACS data to understand and address the needs of veterans — from job training and employment assistance, to health care, to housing, and more. Who among us wouldn’t want that for our former soldiers?

So why, oh why, in the words of Rep. Jeff Duncan (R-SC), sponsor of a bill to cancel the ACS (and just about every other Census Bureau program), are Americans “fed up with these mandatory census surveys and [they’re] asking us to stop the harassment”?

Ummm, no, they’re not. Okay, maybe a few are grumbling. According to the Census Bureau’s new cheerleader for harassed Americans (officially called the Respondent Advocate), roughly 60 percent of households answer the ACS without any prodding at all. With a little encouragement and explanation, by phone or in person, the response rate jumps to 97+ percent (weighted). Of the 3.54 million households in the 2012 ACS sample, less than 8,000 refused to participate (and no one, I can assure you, was hauled off to jail). Let’s see: that’s a refusal rate of (drumroll) two-tenths of a percent. The 535 members of Congress were so deluged with anti-ACS complaints that they sent the bureau (another drumroll, please) 187 letters on behalf of distraught constituents over the past 18 months.

Sure, the ACS questions could use a systematic review and some fine-tuning; thorough training will help ensure positive interaction between survey takers and responding householders. I suspect the Census Bureau has been a little behind the eight ball in acknowledging thoughtful concerns about parts of the survey; it’s finally on the right track, I think. More on these efforts in my next blog.

But let’s stop pretending: ACS critics aren’t falling on their data swords for countless (no pun intended, census fans!) Americans abiding stoically in the shadow of government overreach. Ideology — namely, a belief that government can require little of the governed, coupled with an aversion to the sort of federal assistance dispensed on the basis of ACS data — is driving the campaign to weaken (with voluntary response) or eliminate the survey.

And that’s okay. (Yes, you read that correctly.) If you don’t believe that government has a fundamental interest in producing objective, comprehensive data to inform and guide decision-making, go ahead and make your case. Explain and defend the consequences or propose a practical alternative. Just please drop the cover of phantom citizens cowering behind mailboxes, dreading a nosy questionnaire and the prospect of devoting an hour of time to help the world’s greatest democracy function smartly. Most Americans, it seems, are wiser than you think. And they all love our veterans.

O, Canada! More Lessons From North of the Border

by Terri Ann LowenthalTerri Ann Lowenthal

In my last post, I looked to our northern neighbor to see what lessons we might learn from Canada’s experience with a first-ever voluntary household survey to gather socio-economic data on all communities — data that are used, directly or indirectly, to guide much of public and private sector decision-making. The National Household Survey replaced the mandatory census “long form” after conservative leaders balked at the perceived invasion of privacy and governmental overreach. The result: Increased burden on the public (due to a larger sample size to compensate for falling response rates), increased costs, and no reliable data for a quarter of the country’s localities. This is not an outcome I’d wish upon our venerable democracy.

Canada also takes a census of population, with mandatory response, every five years. In 2011, Canadians answered 10 census questions, compared to the six Americans answered in the 2010 enumeration. Canada first offered the option of answering the census online in 2006; almost a fifth of Canadians did so, leading Statistics Canada (StatCan) to nudge 60 percent of households, via advance letter, to respond on the Internet in 2011. The remaining households received paper questionnaires in the mail or by hand (enumerator drop-off), similar to modified methods used here in rural and remote areas.

Internet response is a money-saver; there’s no scanning and data capture required, as StatCan points out on its website, and there are fewer missing (item non-response) or erroneous answers (yes, some people put down an age and birth date that don’t match, for example!). The U.S. Census Bureau is following suit, using the ongoing American Community Survey (ACS) as a rolling test-bed for Internet response in the 2020 Census. Half of households in the monthly ACS sample (the option became available in January) are ditching the paper form and submitting information online. A promising start, for sure, but the jury is still out on savings ($4 – $5 million a year, the bureau estimates), as more people call the telephone assistance lines for help. And electronic filing of census forms is not a silver bullet. People in low-income households and rural areas are less likely to have broadband access, and there are phishing scams and other data security issues to address. (I received two emails last week purporting to come from a census.gov address. Can’t fool me, but how many others might easily be scammed?)

In 2006, StatCan introduced another operational upgrade to decrease public burden: It asked people for permission to access their tax files for relevant information. Using administrative records is a key part of potential reforms for the U.S. census in 2020, but extensive research is required to overcome considerable hurdles, such as laws that prohibit sharing of personal information between agencies, the need to put people at a physical address (you can’t live in a post office box or with parents who claim you on their tax form if you’re in college), and missing demographic information such as race, age and gender in many databases. And we need to understand how Americans will view this sort of data-sharing — as a smart use of existing information or an example of big brother run amok?

My point here is that there are promising reforms for the 2020 Census, but the Census Bureau needs time and money to vet new methods thoroughly. Congress wants the next count to cost less — a lot less! — yet it is reluctant to invest adequate funds in research, testing and operational development now, so the Census Bureau can realize the significant (billions of dollars!) savings these new techniques will yield down the road, when the bureau starts the enumeration.

President Obama proposed a $983 million budget for the Census Bureau in fiscal year 2014, a small increase over his request for 2013. $245 million of that amount is for 2020 Census planning; another $242 million pays for the ACS, a vital factor in cost-effective 2020 testing. But sequestration and budget cuts set the agency back 11 percent this year, making the funding leap from one year to the next much larger than it should be. Congress needs to get its mind around the concept of ramping up for this uniquely cyclical federal undertaking soon, if it truly wants to see fundamental changes in the way we conduct this nation’s largest peacetime activity and still ensure an accurate count.