January 29, 2010

Constance F. Citro on Census Planning and Evaluation

Constance F. Citro, director of the Committee on National Statistics at the National Academy of Sciences, spoke about the challenges of conducting the 2010 Census and the need to plan now for the 2020 count at the Pew Research Center last week. The committee has evaluated major aspects of every census since the 1980 count.

She made her remarks at a Jan. 21 forum on the 2010 Census that was co-hosted by the center, the Washington Statistical Society and the DC chapter of the American Association of Public Opinion Research. In the following edited excerpts, ellipses have been omitted to facilitate reading.

The Committee on National Statistics is a standing committee of the National Academy of Sciences.  Its members serve pro bono. [Census Director] Bob Groves was a distinguished member of the committee for a number of years.  We’ve been around since 1972 and we have done work on the census almost without interruption since that time.

Our first report was a quick review of the plans for the 1980 Census.  And I want to point out that it actually recommended paid advertising as opposed to public service announcements.  Well, that was not done in 1980; it was not done in 1990.  But it was done in 2000 and, from what we have heard, it looks like the 2010 program is a step up from 2000.

Our newest study will look at the 2010 Census and, in particular, look forward to 2020.  The study panel is chaired by Janet Norwood, the former commissioner of the Bureau of Labor Statistics and Tom Cook, who is a systems engineer member of the National Academy of Engineering.

These are some of the recent reports that we’ve done on the census.  We did do a big review of 2000.  We looked forward to 2010.  We dealt with residence rules; that’s things like the snowbirds and the college students and so on.  And we have looked at coverage measurement and, in fact, Jeff Passel is a member of that panel.

We, as I speak, have the final report in the last stages of review. It looks at the experimentation and evaluation program for 2010 [interim report here], which is really important looking to 2020.

The 2010 Census

First of all, we all know this, but we should say it anyway because people like to hear that they’re doing a great job.  The nation owes the dedicated census staff our understanding and appreciation.  They are doing a heroic job under difficult circumstances.

They had a harrowing 2008 when the handheld technology problems blew up out of control.  But now the census is back on track and going forward.  Because of that handheld problem, they had to redesign a lot of their tech and people systems on the fly.  And this makes it all the more important that they have a good evaluation [of the process] so that in looking forward to 2020 you can learn from 2010.

Now, evaluation requires data.  In practically every report we have done, we have emphasized that you’ve got to get the data from all of the operations.  And at first they didn’t think they could do that from 2010.  But, actually, when Bob [Groves] got in there, he said, oh, yes, we can.  We can do that.  And [now] they are planning to capture the inputs and outputs of operations so that you can actually figure out what went on.

Of course, there are lots of expert folks outside as well as inside the bureau.  And, in working on evaluations, the best job will be collaborative.

We have heard from Bob that unfortunately the census coverage measurement program this year probably will not do quite as good a job as it might have because it is starting so late – although Jeffrey [Passel] reminds me that mobility of the population is actually not as pronounced as it was so that may help.

Now, you may ask, why is it starting so late?  Well, that was a decision that got made for what seemed like very good reasons at the time.  And when Bob [Groves] came in I understand he actually looked into this and said, can’t we do something here?

Well, the census is so big; it has to work within the space of just a few months.  Sometimes, when decisions are made, you just have to go with it.

Planning the 2020 Census

My biggest message for 2010 is that it would be easy to forget about 2020.  The current census is already all-consuming for Bob and his staff.  There are appearances like this, congressional hearings that will be more and more frequent, media presentations, operational problems.  There have been unexpected operational problems in every census if you read the history.  There will be some that will need attention.  And, yet, planning for 2020 right now is crucial.

To the bureau’s credit, they have started planning for 2020, and with a charge to rethink what is now a 50-year-old paradigm of mail-out/mail-back that started in the 1970 Census. [The Committee on National Statistics’] new panel has appreciated the opportunity to attend 2020 brainstorming sessions that the Census Bureau has organized.  Out of some of those conversations are coming some messages, one of which is that the planning for 2020 could usefully have a stretch goal of really trying to get a handle on census costs without compromising the completeness of the count.

Congress has always provided generous funding when push comes to shove, but there are surely ways to contain costs that could be the focus of a really directed 2020 planning effort.  And why do I harp on cost?  Well, if you look at inflation, at growth in housing units and you take those factors out, the cost under essentially the same paradigm of mail-out/mail-back from 1970 has gone from about $17, $18 a household, housing unit, an address, to well over $100–both figures in today’s money.

So that’s more than a 600% increase in cost, some of which is undoubtedly needed with our more diverse population, but some of which needs to be looked at.

Well, what do you do to look at ways to have a more cost-effective census?  I have listed some [See slide: 2020 Census Innovations].  None of them are new.  But I think we are at the point where we can push ahead and make some of them come to fruition.

The late Joe Waksberg, whom I imagine many of you had the pleasure of knowing, said that the administrative records were always the solution for the next census.  But they would be worked on and then dropped.  Well, the Census Bureau has made some very wise investment in their administrative records infrastructure and, with further work on it, I think could make greater use of administrative records in 2020 for many census operations including address updating, which is what Joe was talking about, and non-response follow-up.

Now, for the latter, you might need to change Title 13 [law governing privacy of census records] so that an administrative records enumeration, after you’ve made an initial contact, could count as an enumeration [meaning that the person would be included in the census total].  And there is precedent for this.  To do the mail-out/mail-back paradigm, Title 13 was changed in the 1960s to drop the requirement that an enumerator had to visit every house in person instead of just the non-response follow-up that we do now.

Internet Filing

Okay, the Internet:  There was actually an internet option in 2000.  It wasn’t pushed, but it was there.  But we decided to forgo the internet for 2010 and, in so doing, we are now behind the international standard, the world standard.

I think an intensive R&D program in tandem with the American Community Survey could and should really make it feasible for the internet to substitute for mail-back and even mail-out in some areas.  And here we could do worse than to look to our neighbors in the North.  The 2006 Canadian census, without even really trying–they didn’t push it–had an 18% internet response.

They are now planning in 2011 for a 35% to 40% internet response.  About 60% of households in areas that have a lot of broadband will receive only a letter with the URL [of the questionnaire] and they will be asked to respond [online].  If they don’t, they will receive a second letter.  Only then if they do not respond will they receive the full expensive questionnaire.  And then, only then, if they still don’t [comply] will there be in-person follow-up.

Security measures will be appropriately tight, but not so much that you actually can’t get on the site and easily respond.  They also know that they have to build the capacity for peak response because everything is targeted:  Respond on Census Day and if you don’t have the capacity then the whole thing is going to flop because people are going to be turned off.  But they are really working towards that goal.

And while people who respond by the internet might have responded by mail, you save the data capture costs; you have editing advantages, language advantages, et cetera.

Costs and Research

Part of making a more cost-effective census could be to look at every census process–put it under the microscope–because since 1970, we have added stuff to the census. Special group here needs this coverage improvement program; special area there needs a tweak in the basic methodology.  Many of these were well-justified, but it could be time to take a hard look and see, do we need all of the bells and whistles that we’ve added?  Which ones really work and which ones could be scaled back or even dropped?

Handhelds:  Just because we had a problem in the 2010 planning process, of course we should not drop handhelds.  In fact, my understanding was they were used very successfully in the 2009 address canvassing.  We need to look very deeply at what went well there, what didn’t go well.

And based on the past decade’s experience, I think the Census Bureau would be wise to bring in a contractor right now to design the handhelds for 2020 for non-response follow-up, because even if you do use administrative records, even if you do use the internet, there undoubtedly is going to need to be some in-person follow-up.

Okay, I am just about wrapping up here.  I will just make the pitch that is in every single Committee on National Statistics report that is ever done, which is that we are the evaluation people, the statistical methods people.  State-of-the-art R&D is a necessity to move from 2010, learn from it and deal well with 2020.  [See slide 2020 Census—R&D] I’ve listed some of the things there that I think are obvious but can bear repeating.

Where we all come in is the last bullet: Seek adequate early funding for state-of-the-art R&D.  It’s too easy for Congress to decide, okay, we got through that census; let’s forget about it.  Everyone in this room can be part of saying, look, the next census planning needs to start now and you need to fund that if you expect to have a cost-effective 2020 Census.

Earlier postings from this event include the transcript and audio of a presentation by Census Bureau Director Robert Groves. More will be posted soon.