February 5, 2010

Covering Census 2010: A Workshop for Journalists

Journalists Ron Nixon of the New York Times and Paul Overberg of USA Today presented a workshop for journalists on how to cover the 2010 Census at the Pew Research Center Jan. 21. The session was moderated by D’Vera Cohn, a senior writer at the center and the former demographics reporter for The Washington Post. The workshop  was co-sponsored by Investigative Reporters and Editors.

In the following edited excerpts, ellipses have been omitted to facilitate reading.

D’VERA COHN:  I’d like to welcome you to the second part of our Census 2010 event, a workshop for journalists.  First I want to thank our cosponsor, IRE, and especially its director Mark Horvit for agreeing to put this on with us.  IRE, as many of you know, offers training, data and expertise. I know they’re planning to ramp up their ability to reach out and help journalists with the 2010 Census.

We at the Pew Research Center also have plans and some things that have already been implemented to report on the findings and methods of the census itself.  We launched an All Things Census” page yesterday with postings about census findings and methods, which will include some of our work as well as links to work from around the country, including some news stories.  So please send us links to your stories if you think we should comment on them.

We released a poll yesterday that was mentioned in the first session about census attitudes and awareness.  We’re doing our own line of research on this, I might add, independent of the Census Bureau.  They didn’t have a voice in designing the questionnaire, which is about what people know of the census and what their plans are to answer forms or not answer their forms.

In keeping with our informal spirit, I won’t give you a long introduction for our speakers except to say our first speaker, Ron Nixon, has had years of experience with data, working for IRE, working in Minneapolis and now as a projects journalist for the New York Times in Washington.  Paul Overberg has been the database guru at USA Today for many years.  I think he may know more about the census than some people at the bureau itself.  I think between the two of them we’ve got it covered.

RON NIXON:  I’m going to go over some background, some of the things that the census is doing right now to streamline their process to make the gathering easier than in the 2000 Census.  But also, I will be looking at what their plans are and what they’re doing now in communities along with some of their vendors. I have some documents including their communications plan, contacts with their vendors and things like that.

In ramping up for the census, one of the first things that I did was grab the Census budget justifications and start going through those to see what they’re doing.  You can get a copy of their budget justification at OMB, Office of Management and Budget, or you can get it from the Census themselves.  What it does basically is go through their request for money.  They have to justify why they’re asking for all this money.

So in 2009, they asked for about $2.1 billion for ramping up, which included about 16,739 FTEs (Full-Time Equivalent), not including all those numerous people that are going to be out there knocking on doors and giving out the forms.  In the 2010 budget cycle, they asked for $6.9 billion.  As you can see, the FTEs went up to 105,391.   And in their budget projections, they were saying that for the life cycle of this thing, it’s going to cost about $14.7 billion for them to do a complete and accurate count of every man, woman and child.  And, of course, we know it’s the largest peacetime mobilization.

They sort of reengineered the census this time around.  Some of this stuff was talked about earlier today.  The ACS, the American Community Survey, is one of the big changes that impacts the 2010 Census because they don’t have to do the long form any more.  So with the ACS, there’s a cost savings.

They’re going to change some of the maps and the boundaries that we saw the last time in the 2000 Census to make political, legal and statistical boundaries more accurate.  And proper allocation of more than $200 billion is based on these.  It’s going to improve the accuracy of redistricting.  Of course, a lot of that’s going to hit the states–and improve the accuracy of reapportionment.

They’re going to mail this second questionnaire to nonresponsive households.  They’ve tested that in 2003 and 2005.

Census Outreach Campaign

The biggest part of this is their outreach.  There are various minority community organizations – churches, groups like the NAACP, the Hispanic groups, Native American groups, Asian-American groups – that they are reaching out to.  The [communications] vendors are supposed to have strategic plans to engage these communities in helping to get out the word about the importance of the census and allaying some of the concerns that people have about the census because there is this huge worry within the Hispanic community about immigration status.

Minority media. I know I was talking to the National Newspaper Publishers Association, which is a group of about 200 different black newspapers throughout the country.  And they haven’t heard much from the census or their vendors about outreach on this. And so that’s a line of reporting that I actually hope to get into myself.

Religious organizations are, of course, playing a big role–churches, mosques, synagogues–and that’s part of their communications and outreach plans.  State and local governments have their own Web sites.  And I know that Mount Pleasant in New York actually put out a request for proposal for vendors to do their community outreach.  So you have different layers of this, both from the federal, state and local level where everybody is trying, because again there is a lot at stake, both in terms of funding, political representation.

There’s these multi-language instructions and forms that are going out now.  So you’ll have in some communities English, Spanish, I think it’s like eight different languages.

D’VERA COHN:  Let me read some numbers that Census Director Groves presented during his talk before this workshop began.  He was talking about how 13 million households will get a bilingual English standard form offering the option to reply in either language.  Forms will go out in six languages as Ron mentioned.  But there will be 59 language assistance guides, which will essentially show how a question translates from English into another language.  And those guides will enable the census, the bureau believes, to reach more than 99% of the population.

They’ll also have promotional materials, ads in 28 languages.  They’ll have capabilities with their non-census partners for reaching out to people in 101 languages.

RON NIXON:  The replacement questionnaires for those folks who didn’t respond to the initial ones, they’re going to mail out a second one and then a reminder postcard.  And if they still don’t get responses, by May 1st through July 10th, they’re going to do in-home follow-ups where people actually go to these homes [and]  knock on the door to make sure someone is there.

The communications vendor is Draftfcb.  It’s based out of Chicago.  The contract is a $250 million to $300 million contract to do this outreach and they subcontracted it out to these various firms to deal with the various ethnic groups.

For African-Americans, Black Africans, Caribbean folks and Haitians, GlobalHue is this firm that they’re using.  For Hispanics, there’s two–GlobalHue Latino and D’Exposito & Partners.  For the Asian groups, there’s IW Group.  Native Americans, there’s a G&G Advertising.

And then for these emerging audiences, groups that they think [are going to be more prevalent] this time in the census count, they’ve contracted with Allied Media Group. And then Draftfcb is doing the overall stuff for Puerto Rico as well.

There is a rollout plan that’s about this thick for the entire thing, including measures of success:  How will we know that we’ve been successful in getting the information out to people?  And then following through on that, the kinds of response rates they want, making sure that they got the proper media.  You know, the Super Bowl is going to be a big one.  They’re supposed to have a big blowout there.  Various ethnic events, they are having huge advertising there.

There was a little blowup with Draftfcb about a year ago where people with the Census advisory committees said that the company was not doing an adequate job in providing outreach to the various communities.  And there was some concern about it.  I’ve asked the Census if they would give me the updates–when you’re a contractor you have to give these updates, progress reports.

I’ve asked for all the contracts and the subcontracts for each one of these particular vendors to see how they’re actually going to go about getting this information.  And if I do get it, I’ll be more than happy to share it rather than having people file additional FOIAs for it.  So just shoot me an e-mail at nixon@nytimes.com.

Problems for the Census

One of the things that’s really impacting this census is foreclosures.  As you know, because of the global recession and the national recession a lot of people have been forced to move, so that the mailing list that they have generated over time may no longer be accurate.

Fear over immigration status–particularly for Hispanics.  That’s something that they’ve tried to deal with for years, telling people that this is not looking at your immigration status, but there’s still fear of that.  It keeps people from filling out these forms and keeps the response rates in that particular community low.

Talking about the undercount and the lower response rates, I believe the overall response rate in 2000 was about 67%.  Of course, that varies across the country.  But all the stuff that we just talked about they’re hoping will  help them increase that response rate [especially] by not having that long form.

Continuing problems with rural areas–even though urban areas traditionally have a lot of undercounting too, rural areas have been a problem because of all these little places.  I used to be a reporter in Roanoke, Virginia.  And there are all these little places that have names but they’re not really like places.  There is such and such hollow and people know what they are.  And they’re not really on a map.  In urban areas, even though there are large undercounts, you still have places that you can go to.  You have streets that you can go to.  In all these [rural]  places, there aren’t those streets.

Scams–of course, any time the government collects information on something, someone is involved in it.  I saw a couple of things online where people were offering to fill out your census form for you for a fee.  It’s kind of the H&R Block of census.  Come in, we’ll fill it out for you.  Those kinds of things are going to be going on in the communities as well.  And this is a line of reporting that I certainly hope to do.

The ACORN effecteverybody is familiar with ACORN by now, right?  But it’s having an effect–and someone told me this on background– in that people feel like they’re being vetted by the Census because of what happened with ACORN.  The partner organizations that would do what ACORN and other groups have done for the longest time–that is help with outreach to communities, help people understand what the census is about, why it’s important–feel like there’s more pressure.  And so some groups have chosen not to partner with the Census this time around because of this ACORN effect.

And just from talking to people in some of the groups here in D.C., that’s been the case.  In Minnesota, my home state, I talked to a few people.  But a lot of the community groups that sort of take positions that are deemed to be somewhat more radical that would have partnered with the Census previously are not doing so this time around.  And I don’t know what effect that will have on the collection of this information.

The bottom line is all of these things are not addressed in the census communications plans that Draftfcb has come up with.  [A good question for journalists to pursue is] what impact will that have on the overall count?  So that’s sort of what I have done in terms of prepping myself for the census and all of these numbers that are coming.

I want to see what communities are doing, what state and local governments are doing, what the Census itself is doing.  And then how effective their vendors have been and seeing their advertising plan.  The other thing is to do a sample of the advertising in the various communities by going to the TV station and just looking at the paid ads to see how much they’ve actually paid on this and when those ads are. You can do that; it’s public information. See when they’re targeting the ads, how they’re targeting, times, events, and that kind of thing.  So you’ll get a sense of how effective this will be.

PAUL OVERBERG:  This is Christmas Eve for me.  You know, this is the last day of school and summer vacation is stretching out in front of me.  It doesn’t get any better than this.

What I’m going to do is focus on what’s different this time than last time, since if any of you have a framework for covering this [census], it’s from the last time. So what’s different?  Let’s talk about sampling, which was a huge issue last time and is not at all this time.  And the reasons why are political and statistical and economic.

Last time there was a huge issue which made Congress’ role in the 2000 Census much more meddlesome and troublesome in some ways in terms of getting the census done on the schedule that it has to get done on.

In the 1990s, the Census Bureau said we can actually take the census and where the places are that we don’t get good response we can adjust for that with a statistical sample.  And then we’ll take a big post-census survey and, in the places where we know we didn’t get good response, we’ll adjust the totals up and down based on the survey.  This is sort of pretty standard stuff in statistical science.

And as this plan went forward in the ’90s, it got to a Supreme Court case in 1999.  The Supreme Court looked at the law that governs how the Census Bureau is supposed to do the census. The Supreme Court said, well, you can’t use statistical adjustments for reapportionment, which is just which state gets how many seats in Congress.  But for everything else the census law seems to permit that.

So the Census Bureau said, well, okay, that scraps half of our plan.   They scrapped part of the plan and they changed things around and they retained a way to go ahead and still do a post-census adjustment for things like redistricting data and all of the other uses that the Census Bureau data get released for.

The unadjusted data came out.  The decision had to be made by March of 2001 whether or not the adjusted data was going to be released.  Was it better?  They found huge problems with it.  And they had to bite the bullet and say, no, sorry, we’re going to go with the unadjusted data.

They had to make another decision in October of 2001, not for redistricting stuff, but for everything else for the rest of the decade:  Should we use this adjusted data?  At that point they had done more feverish research and found some of the big problems that they had seen in the 2000 Census that this post-census survey had not been able to catch.  And they said, no, we can’t adjust.  We cannot adjust and still be confident that, at the levels at which people are going to use the data, down at the tract level, it’s going to be more accurate than the unadjusted data.

Operational Improvements

So that’s off the table completely this time.  The salutary effect of that may be that a lot of the effort all through the planning for the 2010 Census was to get it better by operations improvements, not statistical adjustment and things like that.

So there has been a lot of focus on things like fixing the address list and cleaning it up, and there has been a lot of improvements in terms of things like, let’s go ahead and innovate, and let’s send a replacement questionnaire out, if people don’t send back a questionnaire.  Survey people have been doing this for 30, 40 years, and they know that it boosts your response rate.

The Census Bureau was way behind, very conservative on this sort of thing.  That’s one of the big innovations, this time.  There will be a replacement questionnaire–not everywhere, but in the places that they know are hard to count.  One of the things that they’ve done, that they spent a lot of time and effort on last year:  They literally had somebody walk every street in the country, to make sure that where they had an address, the building was still there.

In the previous session, Joe Salvo from New York City’s Department of Planning put up a picture of a typical house you might see in Queens, where there are five mailboxes at the door.

So last summer they had 140,000 people walk every street in the country, and map every address, and find new ones, and take out old ones, and check things like subdivided ones.  This is the only thing that they did with these handheld computers that they spent a fortune on and ended up getting not much out of. They were able to stand there, they literally put a person on each doorstep, and they stood there, and they grabbed a GPS latitude-longitude off the global positioning system satellite.

So now they not only have a clean address list, they can send somebody back there when they are building all of the follow-up lists for the census this time, for the people who don’t send back the forms. That can not only speed up the follow-up, but it can make it more accurate, because they’re not rushing, and they’re sending the right people to the right places.

Hard-to-Count Database

One of the things that they focused on is this hard-to-count database.  They built a tremendous amount into the planning of the census this time–how do we target better? Not just in the advertising, but in the planning of the advertising.  So they took the response rates from 2000– at the census tract and census block level–and they matched that and did a lot of analysis against demographic data.

They found a good way to predict–by basic demographics, things like language spoken, and income, and is it a rental unit–how likely people are to respond to the census.  There’s a hard-to-count score for every census tract in the country, of which they are about 65,000.

It’s on their Web site. You can download it – it’s a spreadsheet, and you can put it into software and you can analyze it yourself.  We did a storyUSA Today offered it a couple of months ago– and we put a map on our Web site and in the paper.  So every census tract in the country has a score, and then, obviously, they have looked at things like, okay, why is it hard in this census tract?

This summer, what they did with the stimulus money–the extra billion dollars they got from the stimulus–was that they hired a lot more people to staff.  There’s a staff person responsible for each of those census tracts, and the volunteers who are there.

This summer they did a walk-around in every tract.  Where are the churches?  What are the important employers?  Where do people hang out? How are we going to target the people here?

So there’s been a tremendous amount of that, that didn’t happen last time.  The same thing happened in the buying of the media–the paid media–that you’re going to see in the next two months.  Where do we need to put ads out, and in what language, and in what medium?  Should we be buying billboards?  Should we be buying ads in the local black newspaper?  Is it fliers that we hand out at bus stops? How do people move around?  What are the trust points and institutions in each neighborhood?

That is all built into their communications plan.

One of their big slogans for the ad campaign is “10 questions, 10 minutes.”  If you take out the screening questions, you know, can you please put down your phone number in case we have to follow up–six questions about each person in the household get answered.  That’s all.  Just six.  It’s the name, the birth date and age – they want the birth date, because people tend to round, if they just ask for the year.  It’s the birth date and age; it’s the gender; it’s Hispanic or not, and then it’s race.  And then the final question is, what is your relationship to the house-holder or the person who is filling out the form?

Shorter Short Form

This is the shortest mail census ever.  There is no long form, which used to go, as Ron said, to one household in seven, the thing that would ask all about commuting and income and education and occupation and industry and that sort of thing.  That’s not there.  That’s all in the ACS; that’s a whole separate thing now.  So the whole thing fits on one page per person.  And you literally can do it in two minutes if you move along.  You can do it in one standard commercial break, I think.

This has all sorts of beneficial effects.  They found last time a tremendously lower response rate for the long-form households than short-form households since the thing went on for pages and pages and pages.  They think that will help improve response this time.

Precision. Ron talked about improving the mapping stuff.  And I told you about the GPS coordinates they now have for every address in the country.  So they have a clean address list, which is good if you’re mailing stuff.

Last time they got about two-thirds of the forms back.   If they do that well this time, a third of the [non-responding] households in this country will be about 40 million households; they’re going to have to knock on the door.  So if you’re trying to optimize 40 million households, it’s great if you’ve got a really precise location and you can run that through some software, to make it easier for the census-takers who are going to have to do that.

QUESTION about whether the address locations are public information.

PAUL OVERBERG:  That’s a good question.  The association of state chief geographical officers [National States Geographic Information Council] is actually in discussions, as they call it in polite circles, with the Census Bureau.  They want the latitude-longitude file for all those addresses.  And the Census Bureau, because this all gets into the confidentiality of the census and the census process, they’re ready to fall on their sword before they give up anything that might impede the idea that the census is absolutely confidential.  So even the geographical data around it, like those lat-longs, they don’t want to give up.

You can imagine about how important that data could be, and helpful for local government operations, especially ones that don’t have sophisticated planning and GIS systems. At some point, I think, they’re going to come up with some compromise, and the state geographical officers are even willing to lobby Congress now to get the law changed–Title 13– which spells out the parameters of how the Census Bureau does its stuff, and what’s confidential.

Foreclosure Crisis Impacts

Recession. Obviously the recession is going to have a big effect on the Census Bureau.  One of the things that they’ve built into their hard-to-count database now, in addition to all the scores,  is a tract-level foreclosure database and a tract-level vacancy database.  They are going to watch, as the forms start coming back, at the tract level, how much that sort of things affects it.  They’ve got unemployment rates at the county level, or as local as they can get them, and they’re going to be watching that sort of thing.

The biggest worry they have is that foreclosures have doubled up households.  It’s epidemic in some parts of the country.  The problem is that they go and they mail the form–the postal carrier takes it and says, well, okay, this house has been boarded up for two months, it’s foreclosed.  They return it as undeliverable. That’s not the problem.

The problem is where did those people go?  Where are they now?  And I know that if I have to move out of my house–it gets foreclosed–and I have to live in my brother’s basement for the next six months, I’m going to feel a little ashamed.  I’m not going to be bragging about that with my friends.  And when the census form comes to my brother, is he going to make sure that he counts me and my wife and my two kids living in the basement?

Not only is there a stigma issue, there’s a logistical issue:  Oh, he’s only here for a couple months, so it doesn’t really matter–we don’t count him here.  Well, they’re supposed to be counted, wherever they’re living on April 1.  It’s a huge issue that’s now been moved into the communications plan–making sure, especially in those places where they know foreclosures are rampant, to make sure everybody knows, even though he’s only here for a couple months, you count this family, or this guy, or whoever it is.

The targeting system. In addition to producing this hard-to-count score for every census tract, the Census Bureau had their contractor do a huge, long survey about a year-and-a-half ago – 4,000 people. In addition to basic demographics of these people, they asked dozens and dozens of questions about their attitudes toward the government and their attitudes toward the census and their knowledge about the census.

They produced their own segmentation system.  Claritas and a couple other businesses do a job of categorizing Zip codes and even smaller areas with cute names like “station wagons and furs,” or “upscale starters,” and things like that.

The Census Bureau has done that as well with the results of this survey.  So they have about eight groups on which they have done a cluster analysis and found how people categorize in terms of their attitudes towards the census, and the kinds of media they consume.  That’s also a piece of this communications plan.

Census 2000 was the first paid advertising campaign for the census.  It was pretty much, sort of a mass-market, plain-vanilla campaign.  There was a lot of volunteer effort as well.  This one is leagues beyond that in its sophistication and in its depth and in its diversity.

They literally can target people–even, say, middle-aged, white, middle-income householders, who may be at various stages of their life and have very different political opinions. They know where those people are on the ground, and they can target them in terms of the media that they get and the message they get and why they would want to fill out the census.

This shows a map of that hard-to-count database [see slide] at the census tract level.  This is Washington, D.C.; Baltimore’s up about here.  You can see that – and I colored it so the hardest-to-count tracts are the darkest colors, and the easiest to count are like gray, here.  They took 12 variables and gave them each a score of one to 11, so the total was 132 was the highest score you could get in this database.  And the hardest ones in this area, I think, are about 80, 85.

Hard-to-count map of Washington, D.C. area

So you can see that places with a lot of poverty are hard to count.  But look at how close they are to places that are relatively easy to count – right in the middle of, you know, certain wards in Washington, D.C.  You can see all kinds of variety just in Southeast D.C.  You can see the same sort of thing in Arlington County, all over southern Montgomery County – a lot of variation.  Some of this is based not just on poverty but on stuff, obviously, like language and the percentage of rental units versus owned units, and things like that.

QUESTION about where people should be counted if they live part of the year in one place and part of the year in another, such as college students and snowbirds.

PAUL OVERBERG:  There’s a whole set of rules basically, a primer that they give the telephone operators who are going to be at their call centers, detailed sets of rules about what they should say when they get a call like:

Well, should I fill it out?  My husband’s a consultant and he’s away on a two-month assignment.  Should I count him here?  Or, my son’s at college.  Or, I live exactly six months a year in Michigan and six months a year in Fort Lauderdale; where should I fill out the census form?  Or, we have an au pair here for five months; should we count her or not?

You can download the basic script that the phone operators are going to use and the rules; it’s on the 2010 Census Web site.

QUESTION about counting college students living in dormitories.

D’VERA COHN:  The question was, who’s in charge of counting kids in dorms, and Paul can go into that.  A dorm is a group quarter just as is a nursing home or a prison or a military barracks.  Generally, they work with the Bureau to count folks there.  There were some big problems with group quarters counts from 2000, which, by the way, keep in mind when you’re using that data currently, and they’re hoping to make it better this time.

PAUL OVERBERG:  Group quarters is any time there are more than 10 people living together who aren’t related.  About 3% of the people in this country are living in group quarters.  The biggest ones obviously are prisons, dorms and barracks.  But it’s the hard-to-define, more blend-in kind that create problems in places like college towns, but also, urban areas – halfway houses, shelters for battered women, and group homes for the mentally retarded, and all these sorts of places.

The Census Bureau has a list just like their master address list.  And actually, this is the first census where they’ve actually merged the two.  They don’t have separate divisions trying to track this stuff.  They think this will make it a lot easier to process.  And they went to all the local governments and said, give us your list.  And they’ve added a bunch of them.

It’s easy with a dorm at Georgetown.  It’s been there every census for the last 50 years.  It’s harder with the crisis pregnancy home that comes and goes, those sorts of places the city knows about but the Census Bureau doesn’t.

So they’ve compiled a list and they visited each of those – is this still there? who’s in charge of this place? who can we come to when we get back to the site and hand out the census forms there? et cetera, et cetera.

It’s easy when you’ve got a prison.  Pretty much everybody is under control there.  Not so easy with a dorm.  It’s really hard with places where the people come and go; even the coordinator may come and go with a shelter or something like that.

But they’ve got a much better list of those now this time, they think, and they will go and hand out the forms and they will get them back.  And then depending on the degree of, as the Census Bureau calls it, “institutionalization”– prisons being one extreme and shelters being quite the other extreme – they’ll have more or less ability to make sure that they get all the forms back.

You’ve got 100 dorm rooms; we need 100 forms back.  We don’t get all the forms back, we need to go and find the persons who live in room 58 of that dorm and we at least need to know that they’re there, and two or three basic things about them.

If we know their gender, we know their age and we know their race, we can do a lot of fleshing out of the details with statistics.

They almost didn’t release the 2000 group quarters data it was so bad.  It had so many problems, partly because of duplication issues – college kids counted in the dorm and back home.  Their de-duping process didn’t work well at all in 2000.

QUESTION about whether Americans overseas are included in the census count.

PAUL OVERBERG:  Yes, but only for the purposes of apportionment. This is the census of the people who live in the United States, except starting in 1990, federal employees who are overseas on Census Day get counted through their agency and assigned back to a state–not to a locality or any address.  They get assigned to a state.

So the Pentagon has a list of the hometown of record of every serviceman.  And everybody who is deployed, National Guard or regular, on Census Day, the Pentagon will give the Census Bureau a list that basically says, North Carolina, 12,408 people; Ohio, 6,107 people.  Those will only get added to the totals for the formula that cranks out the House apportionment–nothing else.  No funding, no redistricting inside the state counts, anything else.

Let’s say they live in North Carolina just off the base, say, just off Camp Lejeune, whoever is in that household would get the form and they should count that person if that’s where they normally live do.

QUESTION about counting troops deployed overseas.

PAUL OVERBERG:  If your husband’s overseas, don’t count him.  If he’s just, say, over at Fort Hood for training for a month, you can go ahead and count him, but if he’s deployed in Afghanistan, don’t count him; we’ll catch him on that count.

RON NIXON: Just to add, I was in the military in 1990 for the Census and I was not counted because I was off in the first Gulf War. So my wife counted people in the household then but not me, so I was absent.

D’VERA COHN:  Let me tell you who else overseas is not counted because this is a perennial controversy.  If you’re working for American Express in London, you’re not counted.  If you’re a Mormon missionary in Africa, you’re not counted.  And this was actually one of the subjects Utah brought up in its litigation over the fact that it didn’t win an extra congressional seat [after the 2000 Census].  North Carolina did.  Utah has been pushing for LDS missionaries to be counted as part of the census.  So far, they’ve lost that argument.

PAUL OVERBERG:  The Census Bureau actually, after the 2000 Census, did a test and included overseas, trying to count Americans.  And as they predicted, it was horribly expensive and just horribly incomplete.  They did a couple cities; I think Paris was one.

Nobody is counted overseas except federal employees because obviously, there, you’ve got the administrative records to know exactly how many there are, and you’ve got to assign them back to some place.

This reminds me, if you do nothing else in covering the taking of the census, and also then, reporting out on the results, remember the three rules of taking a census:  You have to count everybody, count them once and count them in the right place. That’s a really hard job, as we found out in 2000; is a really hard job.

The duplication is tremendous and you have to be able to catch them.  And they didn’t do a good job of de-duping in 2000.

And, in the right place. Now that we’ve got the precision of GIS systems drawing the district lines and satellites placing each address, it matters down to within 100 yards or so.

QUESTION about whether prisoners are included in the census count in their home state or in the state in which they are incarcerated.

PAUL OVERBERG:  The prisoners are counted where they’re incarcerated as opposed to where they lived, wherever that was, before they were incarcerated.

The Prison Policy Initiative has done a lot of the work.  But a lot of the cities that have a substantial portion of convicts who are somewhere else in the state at a prison in some remote rural area say, this is just depriving us of their presence, their representation in terms of funding, but also it de-citizenizes us of them.  It’s not fair to count them when they’re in rural Upstate New York when they really live in New York City.

There was some back-and-forth study that got done on this, this decade.  The counting procedure this time is the same as last time.  They get counted where they live, which is in the prison.

There are a raft of difficulties in trying to figure out where somebody who’s been living in a prison for six months or a year or five or 10 years–where you would assign them to back in whatever place, and who would you believe, where they actually lived?

And it’s just not practical, just like trying to track down anybody who’s not a federal employee overseas to actually get them in and get them assigned to some place because it’s got to be precise.  You can’t just assign them to the borough of Brooklyn or to the city of Yonkers.  You’ve got to assign them to an address or you’re basically creating a secondary system that different from everybody else in terms of the precision of where they get counted.  So it isn’t going to change any time soon, I don’t think.

Let me just keep moving along with some of the differences.  This census is much more polyglot than last time.  Last time, the form came in one language–English.  This time, the form will come in six languages. Spanish is the most common one.  They’re going to send 13 bilingual forms.  The other languages include Chinese and Russian and, I think, Vietnamese and Arabic and Korean.

If you go to the Census web site, you’ll see that there is a scrolling set of other-language prompts that may prompt you to grab a dropdown, which are in 59 different languages and more alphabets than I’ve ever seen.  And what that pulls up is what’s called “the assistance guide.”  It’s an actual replica of the census form, and all the wording is exactly the same as the census form but it’s done in that language and that alphabet.

It basically is designed so if you put it side-by-side with the English form, even though all you speak is Bengali, you can actually fill out the English census form because you’ve got the replica in Bengali right next to it, and Dari and Dinka and Navajo, and all kinds of languages, Native American languages, and languages I didn’t know existed.

QUESTION about how the Census Bureau chose the languages in which translated forms would appear.

PAUL OVERBERG:  One of the things that they did–and this is one of the benefits of the American Community Survey–is they ask you every year on the ACS, what language is spoken in this household?  And they tabulate more than a hundred of them  nationwide.

So they look at the list and they start down from the most common and they go down as far as they can.  The media campaign is in 28 different languages, if you look at the bus shelter ads and the handbills and the newspaper ads and the TV ads and all that stuff.

So it’s 28 languages for the media, 59 for the assistance guides, and they have partner organizations that speak something like 101 languages.  That count includes eight forms of Chinese, various dialects and writing systems.  So maybe it’s only 90 instead of 100, depending on how you count.  But it’s much more polyglot than it used to be because the country is.

And in some ways, the Census Bureau was behind the times in 2000.  They’re doing a lot of catch-up now, but this country is obviously getting more diverse, and I think you’ll see even more of that in the years to come.  This means, obviously, that they have a better chance at getting better response rates, especially from these hard-to-count people.

There are some things that are the same this census as the last census.  We hand out something like $400 billion a year that’s tied to census counts, and the political power in the House of Representatives, state by state, is apportioned by the counts, but also the lines of state assembly districts and city wards are drawn based on them.

Power shifts are a constant and there are  going to be more power shifts in this census, just like there were last time.  And so much of what we do in 2011 will be chronicling that.  Right now, it’s a lot more fun to focus on this massive civic exercise and the operational details of it, because, if you think about it, there’s nothing else in this country that we all do, or have to do.

We don’t all pay taxes; we don’t all watch the Super Bowl; we don’t all vote; we don’t all go to public schools.  You know, think of something else that’s pretty embracing in American life.  We all don’t watch “American Idol,” that’s for sure.  The census, though, is everybody, by design and by dint of $14 billion this time.  You can still get a lot for $14 billion in terms of covering a country that’s this big and this diverse and this stubborn about not wanting to be counted, sometimes.

But the power shifts are going to be there.  We can pretty well tell how Congress is going to line up after the data come out.  You can see pretty much which states are going to lose a seat–Iowa, Pennsylvania, New York are pretty guaranteed.  Texas will gain three or maybe four.  California will not gain one for the first time ever.  Arizona and Florida will gain a seat, probably.

There are a few states that are on the bubble.  And actually, that’s where the precision of the count matters.

QUESTION about a recently defeated congressional proposal to ask citizenship status in the 2010 Census, and about the role that members of Congress play in the design and operations of the census.

PAUL OVERBERG:  One of the successes of the Census Bureau over the last half-century has been creating this culture and this mystique that it’s a sort of an apolitical priesthood of data.  And if you’re trying to design a massive organization and keep them all focused, in a variety of complicated missions, that’s not a bad way to go if that is the primary purpose.

Inevitably, there is some of that.  There was much less of that this time, in terms of Congress.  This Congress didn’t have to worry about sampling and the whole method of how that might affect apportionment, compared to 2000.  Last time when the 2000 Census numbers were released it was a Republican administration, but most of the census was built from 1995 to ’99 under a Democratic administration.  And actually, that’s one of the reasons why sampling went forward as far as it did.

President Clinton believed the people at the Census Bureau who said we can do this, we can make this work.  And he trusted them.  As it turned out, they were the ones who ended up calling themselves off and saying no, we couldn’t pull it off; we can’t do it.  Then, I’m sure, President Bush and the Republicans had some private thoughts about that.  But as it turned out, it was the priests of the data who said no, we can’t do this.

This time, interestingly, the idea of putting a question on the form [about citizenship status] at the last minute was ludicrous in its lateness and its misguidedness because I don’t think that people who are sponsoring the thing even understood that there were literally warehouses full of forms that had already been printed.  They certainly didn’t understand that when you talk about somebody’s citizenship status, that does nothing to go into whether or not they’re legally here or not, because you can be an immigrant here and here legally and not be a citizen.  I think that whole question was fairly garbled and it was sort of indicative of how misguided that effort was.

But there are political influences.  They’re very subtle and they tend to play out in the middle years of the decade.  But, you know, the census, because it’s so expensive and it’s so complicated, is an intensely empirical exercise.  That means it has to work and you have to be able to prove that it’s going to work.  There’s no do-overs; there’s no mulligans; there’s no “we’ll push it down the road a year and then we’ll get to it,” or “we’ll fix it in version 1.5.”  It has to happen right and those 50 state numbers that go to the president on December 31st this year have to be right.

And all the detailed data that comes flooding out a year from March has to be right.  It has to be so right that everybody buys it and commits billions of dollars of commerce to it, and all the politicians have to buy into it because it determines who gets what district and where.  So even if you’re trying to influence the process, it has to be able to work.  And that was one of the laughable things about that citizenship status question.  Even if it had come up five years earlier, the first thing they would have said was, all right, let’s see, should we add a question to the census form that says, “Are you here legally or not?”  Do you know what that would do to response rates?

Even if it got to the point where the Census Bureau was thinking about putting that question on the form somebody would have gently told those folks, even before they got to the point of having to test it, that this is not going to be a good idea because you won’t catch the people you want to catch – you won’t find them.  You’ll get bad data and you’ll pollute the good data that you’ve got because it’s more incomplete and more sketchy.

The Census Bureau made a good point when that came up this year, dusting off the history files.  There was not, even in 1790, but certainly in every census since then–there’s never been an idea that you wouldn’t count everybody.  When it got down to it, you would count everybody.  The big debate in the Constitution was: How do you count slaves?  We’re not going to not count them; it’s just a question of, how do you count them.  And ever since the 14th Amendment, if you read the language, everybody gets counted.

Sources for Census Reporters

D’VERA COHN:  I wanted to cover one thing that we didn’t address explicitly, which is, what are good sources to know about as you try to address how well the census is being conducted.

One is the GAO–the Government Accountability Office, the investigative arm of Congress.  It does regular work on the census–has done for decades–looking at the operations in place.  At one point, they deemed the census an at-risk operation. It’s inherently a high-risk operation.  But they went into some detail about what concerns they have.  So if you familiarize yourself with the GAO’s work, you can get an idea of what problems they think might happen and track whether those problems actually do happen.  There’s also an inspector general in the Census Bureau who issues reports, and there may be some meat for you in that. The census, as Ron mentioned, has a series of advisory committees that are supposed to give the census advice on what to do.

And many of them have opinions of when the census is being done right or wrong.  There are, for example, advisory groups for professional associations, for people from different racial groups, for local government officials.  So that will give you some sense – give you some people to talk to who can help truth-check whether the census is being conducted well.

The names of those advisory committee members are on the census Web site, yes.  Their minutes also are up there, although sometimes it takes a while to get there.  There also are the partner organizations, which can be a source for you, locally especially, of whether they feel the census is reaching out as broadly and appropriately, and what kinds of things is the census doing in those hard-to-count tracts.  And you know, politicians are always happy to talk.

RON NIXON:  Two others–the Congressional Research Service has done and will do, I’m sure, more reports on that.  The problem with the Congressional Research Service is that not all of their reports are public, but sometimes, the congressional offices can get the ones that aren’t public.

The other thing is the Office of Management and Budget.  Of course, they deal with the budget and spending, but they also have this program that’s  evaluating the performance of the various agencies in carrying out their missions.  And they rate them based on a score.  Like the list that the GAO had, they also have this list of problematic programs and agencies.

PAUL OVERBERG:  I would add a couple others.  There’s a group of stakeholder organizations  – groups that, you know, count a lot on census data, and good census data.  And they’re sponsoring a Web site, which has a blog, called thecensusproject.org.  Terri-Ann Lowenthal, who spearheads that is a former Hill staffer and very knowledgeable about all the ins and outs of the appropriations process, the operations of the census, etc.  And she does a good job of turning a lot of this arcane stuff into English.

[Census Director] Bob Groves actually has a blog on the Census Bureau site, and it’s fairly well-written and he’s keeping up with it, so far at least.  It hits a lot of the, sort of, high-point question issues as they come up, like the residence rules and things like that.

D’VERA COHN:  And the other thing, actually, is that every community and state is supposed to have a locally based complete count committee to help promote the census.  Now, our parent organization–Pew Charitable Trusts–did a report about how there have been a lot of budget cutbacks at the state and local level that have hurt the ability of state and local governments to do this kind of work.

Some of that’s being made up in the nonprofit sector, but that could become an issue.  But meanwhile, there should be folks in state and local governments whose job is to promote a good count, and they may also be good sources about  what’s going on.

RON NIXON:  The stimulus has been used to deal with some of those shortfalls in funding.

QUESTION about whether the availability of regularly updated American Community Survey data availability has made up for the loss in dropping the census long form.

PAUL OVERBERG:  We could do a whole ’nother panel on that.  I think that ACS gives us a lot that we never even had with the long form.  It’s not as good as the long form on some of the things that the long form gave us.  The sample’s not as rich and, certainly for smaller places, whether it’s neighborhoods of a city or smaller towns, you don’t get the data in as timely a fashion.

However, there’s a lot more to the ACS that we never used to have.  And the ACS is actually, now, improving the census and the annual estimates programs, because with a national sample, every year, of 3 million people, it’s giving us a timely source of lots of detail, like what languages are we speaking and how many people in this country are foreign-born, never mind their legal status.  Just that alone helps improve the estimation process – the annual population estimates, for instance – in terms of immigration rates and things like that.

So there are some tradeoffs.  There’s no going back, obviously.  And one of the big challenges the Census Bureau has for 2010 is, among other things, to make sure everybody knows that there are two separate things.  There’s the 2010 Census – and when that data comes out in 2011, it will be coming out at the same time as the 2010 ACS data.

They want to make sure everybody – especially those who are helping people disseminate data, like us — makes it clear that there’s two different programs, two very different kinds of data for very different kinds of uses.  So I think that the ACS will get better as it goes on, but we tend to think in, sort of, mayfly kind of timescales compared to the Census Bureau.  They tend to think in geological timescales.

Data-Release Schedule

QUESTION about the data-release schedule.

PAUL OVERBERG:  Some day right around Christmas this year, President Obama will get a list of 50 numbers–one for each state.  That’s the apportionment count.  That will tell us how many seats in the House each state gets.  And that does not include D.C., by the way, because, guess what?  D.C. isn’t in the House, yet, as a voting member.

By law, that has to get out by New Year’s Day or New Year’s Eve.  So typically, they take as much time as they can and they zip it out right at the end of December, right around Christmas–maybe the week before.

By law, within a year of Census Day, they have to get the raw material out to the state legislatures to do the redistricting, especially the two states [New Jersey and Virginia] that have 2011 elections.  So by April 1, 2011, all that data has to be out.  And that’s really simple data, in terms of the types of characteristics, but it’s, geographically, everything from all the cities and counties down to every census block.  There are 9 million census blocks in this country.  And all the intermediate levels of geography that anybody cares about – state legislative districts are in that file, voting districts, cities, counties, American Indian reservations, census tracts – everything else.  That file includes total population and the over-18 population for every race category.

D’VERA COHN:  The race categories are necessary, in part, because the Voting Rights Act still is in effect in some states and requires that changes cannot be made to voting districts if they would violate equal representation.

PAUL OVERBERG:  So those districts have to be fine tuned to within–I think case law, now, pretty much says zero discrepancy unless you would have to split key entities to do it.  And with powerful GIS software, you can do that.

So every state, even if the Voting Rights Act doesn’t require them to sort of pre-clear their districts, they have to get their districts down and they have to be precise, to within, like, a few people each.  So that data is geographically very rich.  All there is, though, is the race totals for each of the 64 race combinations, and Hispanic and not.  So you get 120 columns of data times every single row of geographical possibilities.

That all has to come out – it flows out state by state.  So they’ll start sometime around March 1st next year and launch a state or two, or three, every day through March.  That way, they take all the time that they can get – and it’s a rush to get it done that fast – and still get it out by the statutory deadline of April 1st.

After that, they take a breather, and then, next summer, they will start releasing more detailed stuff off the short form. The data released by April 1 doesn’t include any detailed racial breakdowns by age [except 18 and over]– all the other stuff that you can do with just six questions – or the relationship question tabulations.  All that stuff will start coming out next summer in a planned series of products. If you covered Census 2000, it would be SF1 and SF2, and the profiles that preceded those.

D’VERA COHN:  There’s also an own/rent question on the short form, which can give you a pretty detailed homeownership rate.

PAUL OVERBERG:  Right, there’s that one question on that.  And once you cross-tabulate everything by that, you get a lot of insight into the community, as well.  So that’s the rough schedule.  But right now, I’m just hoping to survive the next few months.

D’VERA COHN:  Let me offer you one more story idea that many of you, if you don’t know about it already, will get thrown at you.  The Census Bureau is promising that by sometime in March – by the end of March, in fact – they will be giving you data that will allow you to evaluate real-time participation rates in the census down to the Zip Code or census tract level, and to compare those participation rates with the national participation rate or with any other geographic entity.

So if you’re living in Chicago, you can compare Chicago with Los Angeles or you can compare one Zip Code in Chicago with another Zip Code in Chicago.  And they’re promising to update those frequently.

PAUL OVERBERG:  I think on the response rate tool that they’re building, it’s only going to run for a month.  So I’m not sure how much they’re going to program out beyond what they’re trying to put on their Web site.  But they are going to make that data available.  So if you build something, you can update it if you bring the data into whatever you’ve built, for whatever works for you.

One of the things that they’re doing, as well, with these partner organizations is, they’re using those daily response rate things to say, all right, here we are in Yonkers and we have 12 census tracts and clearly, we need to send the people out into the streets in these two census tracts and not over here, because they’ll have frequent updates during that crucial late March-to-end-of-April kind of period when people haven’t thrown the forms out, they can still get them a replacement form and they can actually have an effect on the ground.

D’VERA COHN:  I want to thank everyone for coming to today’s journalism workshop.  We’ll be putting up some of the material from the workshop on our Web site, and I’m sure IRE will, as well.  And you know, keep watching.  Read our “All Things Census” page, on which we’ll be posting some commentary and descriptions of what’s going on with Census methods and findings and some resources that might be helpful in your coverage.  And good luck with your stories.