Census director focuses on putting IT to the test before the big count

Census Director Robert Groves is busy checking the nature of the infrastructure for the decennial count. ROGER L. WOLLENBERG/UPI/Landov

Since his confirmation in July, Census Bureau Director Robert Groves has found himself in charge of the costliest and most controversial census to date.

Well-publicized technology issues and budget overruns have hampered the bureau's preparations for the 2010 count. Last month, Groves told lawmakers that the budget overruns leading to the decennial count's $15 billion price tag were "intolerable."

But he told Nextgov on Monday that the bureau plans to push the limits of new technology in tests scheduled for after the Thanksgiving break in hope of making sure the census goes as hitch-free as possible in April 2010.

Groves was the director of the Survey Research Center at the University of Michigan's Institute for Social Research and served as associate director of the 1990 census in 1990. Nextgov reporter Gautham Nagesh spoke with him on Monday about the preparations for the 2010 census and the bureau's progress on solving some of the technology problems that the Government Accountability Office and the inspector general found.

Nextgov: What is the status of the bureau's preparations for the 2010 decennial census, especially concerning the information technology systems needed to support it?

Groves: I came in July and had not been there since 1990. There are a couple things to note on the IT side: First, I've been focused clearly on decennial IT issues, not on looking backwards. We have a new chief information officer, Brian McGrath, who came in weeks before me, and he was engaged in sort of the same thing I'm doing -- checking the nature of the infrastructure for the decennial.

We had on our table GAO and IG reports concerning the lack of testing in an integrated way of the various subsystems used for the decennial census. We had some outside folks take a look at whether core subsystems were being tested in an integrated way.

We also have a new set of software we're building as a result of the abandonment of the handhelds that will support paper-based nonresponse follow-up. That is the critical task on the software side I spend the most time on. After Thanksgiving we will perform a load test on systems that will be in action during nonresponse follow-up. We're going to make sure we break the system to measure the capacity.

The other thing that's notable from your readers' perspective: We're 80 percent through opening 500 different local Census offices, each of which has its own computer network issues. That was done through Harris Corp., part of the Field Data Collection Automation contract. We've got 400 local offices up and running, each site is its own little story. After some initial bumps that seems to be going well.

Nextgov: What was the situation like when you arrived regarding IT systems development? What in your view caused or contributed to the IT challenges at the Census Bureau?

Groves: I haven't spent much time going back and diagnosing those problems. I have to focus on the future.

But I am a believer in certain philosophies when you develop software and hardware products for large, diverse sets of users, including that a user has to be at the table from Day One. The user has to be part of the inspection process for all the intermediate products as they are developed. The notion of writing down all the specs for complex systems and getting them right the first time, having programmers go away for a while and code those specs, that's an approach that brings with it big risks.

In my past life in software development I have learned from a management perspective that you've got to get the user there all the time. They have to be part of the development. Humans can't anticipate all the features of a software system before they see the first version of it.

But I need to emphasize that my job hasn't been postmortem on handhelds, I have just not done that.

Nextgov: There were reports that the handhelds had some problems during address canvassing, particularly regarding their mapping function. How are you dealing with those?

Groves: There are two parts of the master address file: the geographical information that provides boundaries for aerial units and the address records. The big good news is that after this gigantic address canvassing operation, the number of records we have is very close to independent estimates of what it should be: 134 million households. That's a good thing, based on the independent benchmark we get from sample surveys.

Now we're going out and checking for clusters of records deleted [during address canvassing]. If you were listed in address canvassing and you noted that an address was improperly placed in a block, your job [as a canvasser] was to delete that one address and add it in the correct place [using the handheld]. We're scrutinizing any clusters of deletes. In some regions we've reinspected areas that look suspicious.

Nextgov: What do they find upon reinspection?

Groves: We're getting spotty results. It's not a slam-dunk one way or the other. When we go out and have a whole group of addresses deleted, sometimes everything looks fine, sometimes the ones that were deleted were duplicates, and sometimes they were deleted in error. There's no typical result.

Nextgov: The idea of using the Internet to collect responses was proposed and rejected last year , despite conducting a pilot in 2000. What's your opinion on allowing responses online? Is that something you think should be explored for 2020?

Groves: My son filled out a questionnaire for the 2000 census on the Internet. The decision to eliminate the Internet option for 2010 was made before I got here. I haven't diagnosed that decision. I know the most commonly cited reason is concerns about security, which are indeed real and completely legitimate.

Looking forward, I can say I can't imagine a 2020 census without some Internet use. At the same time, in the same breath we have to know that neither you nor I have any idea what the 2020 Internet is actually going to be capable of. When I say we must have an Internet option, I must admit I'm not quite sure myself. We have to take advantage of the technology; other countries already are. In 2006, 18 percent of Canadian households responded to their census on the Internet.

Nextgov: Do you plan to serve beyond next year? Would you like to be involved in the planning for the 2020 count?

Groves: I serve at the pleasure of the president and will serve as long as he is pleased with my service. I'm terribly interested in 2020 and also interested in innovation in all of the other surveys the Census Bureau does, thousands depending on how you count. The challenge of doing economic and social measurement in this country is never-ending. The rate of innovation lets us use technology in new and important ways; it can change the way we measure the country. That pace has to pick up in any organization like the Census Bureau. I'm terribly interested in being part of that.