GAO warns that feds aren't ready for 2010 census

A trio of reports released by the US Government Accountability Office in the wake of Congressional testimony last week warn that despite numerous warnings, crucial technology still isn't in place for the upcoming decennial (ten-year) census.

The 2010 Census Day is scheduled for April 1 (sound off! 1,2,3,4,5...) and expected to cost over $14 billion. However, before the government counts us, it has to figure out where we are -- not to mention what constitutes a house these days -- and get its counting tools in order.

By "counting tools," I'm referring to an alphabet soup of systems, as detailed in the "Census Bureau Testing of 2010 Decennial Systems Can Be Strengthened" report: Master Address File/Topologically Integrated Geographic Encoding and Referencing (MAF/TIGER), the Decennial Response Integration System (DRIS), the Field Data Collection Automation (FDCA) program, the Data Access and Dissemination System II (DADS II), and even the dead-tree-wrangling Paper-Based Operations (PBO) crew.

These systems, in turn, include tech such as thousands of handheld computers for in-the-field data collection, a massive database of not just addresses but their nature (single-family home? dormitory? shelter? the census needs to know), and all the IT you'd expect to require for an organization charged with gathering some of the most skittish, diffuse data imaginable.

Back in 2008, the General Accounting Office declared the 2010 census a "high-risk area," meaning that the Census Bureau didn't have its act together on various vital IT systems and operations. The Bureau has tried, certainly; back in 2006, it thought it had enough systems in place to do an end-to-end "dress rehearsal" test.

But a few systems weren't ready. Those were removed from the dress-rehearsal process, and two years later they have not been testing and no one knows when they might be. As of now, only the MAF/TIGER and DRIS components of the system have completed dress rehearsal; meanwhile, the Bureau can't yet produce even a master list of systems interfaces. The dress rehearsal period continues until June 2009.

The "Census Bureau Needs to Strengthen Testing of 2010 Decennial Systems" report in particular notes that the new handheld computers -- which will be responsible for as many as 144.7 million addresses for the census -- have had trouble with slow and inconsistent data connections, geolocation, and general hardware freezes. Anecdotal data from subsequent tests in Fayetteville, N.C. run last December indicate progress, according to the "Little Time Remains to Address Operational Challenges" report, but no one's seen the actual results yet.

What to do, what to do? GAO has made recommendations over the years, and the Bureau has generally accepted them (and the reports note, made "commendable progress" on implementation), but time's running out now. As we slide into the home stretch, GAO basically just advises the head of the Census Bureau and the Secretary of Commerce to lean on its people to complete system testing, and to figure out a plan and a schedule for integration testing. (The Bureau, presented with a draft of the GAO reports as is the custom, agreed.) Looking over the three reports, we'd add a suggestion of our own: Lots of pencils, sharpened and ready, just in case.