Census' hunt for accuracy

The Census Bureau kicked off a procurement Aug. 16 to modernize its database of streets, houses and addresses in a move that agency officials say will allow them to conduct a more accurate and less expensive census in 2010.

The bureau released a draft request for proposals to improve its Master Address File/Topologically Integrated Geographic Encoding and Referencing system, which it relies on to conduct the national head count and other ongoing surveys.

The procurement will improve the accuracy of the MAF/TIGER database by correctly locating every street, specific street addresses and other map features such as roads and rivers. It will also create an "automated change detection process" so the system recognizes where changes, such as a new housing development, have occurred.

"Improving locations is a major objective," said Robert Marx, chief of the Geography Division at Census. "We know from experience, from the field staff and customers that there are significant errors" in existing TIGER location data. "It drives them nuts when they try to [match up] local files with us."

The census counts people and tracks their addresses. "The [action of] placing them where they live is equally important because it is used to divide up pies geographically," said Martha Farnsworth Riche, former Census Bureau director, who now heads Farnsworth Riche Associates, a demographic consulting firm. Census data is used to set policy, distribute federal funds and apportion seats in the House of Representatives.

More accurate data means that census enumerators can easily locate people for follow-up interviews and that the bureau can more easily exchange digital data with other organizations.

For the 2010 census, bureau officials expect to equip 500,000 field employees with mobile computing devices and Global Positioning System receivers to replace traditional pen and paper. At the push of a button, the enumerators would have access to a census questionnaire, local maps, addresses and GPS coordinates and could upload data.

It "will improve the accuracy of the census itself," Marx said. Correct locations in MAF/TIGER and devices equipped with GPS receivers will provide the tools needed by field staff to find the correct housing unit and validate the accuracy of each address. "It's certainly more accurate for government getting credit for what's inside their boundaries."

Census created TIGER in-house in the early 1980s and created MAF following the 1990 census. For the first time, the bureau plans to use outside vendors to help it update the system. Census officials plan to make an award in fiscal year 2002 but would not release the value of the contract.

"This is very major departure for us," Marx said. "We've come to the conclusion we need to take advantage of [vendors'] expertise and understanding."

Rick Ayers, federal account manager for Environmental Systems Research Institute Inc., said this initial phase of the modernization will be the most difficult because there are thousands of local and commercial data sources. ESRI plans to vie for the contract.

"It's a difficult task to make a more accurate integration data set with the timeframe and budget allocated by Congress...but it can be done," Ayers said.

Leveraging the state and local governments' digital geographic information system files and using commercial off-the-shelf products to tie them together could save time and money, Ayers said.

Cliff Kottman, vice president and chief scientist at the Open GIS Consortium Inc., said that Geography Markup Language (GML) — a text-based language that provides a structure for storing and sharing geographic data — should help. Based on Extensible Markup Language, GML is making geographic data sharing and presentation more seamless via the Web and is the key to making projects like the TIGER modernization work.

He said GML will officially arrive when it makes cutting and pasting graphics as easy as cutting and pasting text among different applications.

But doing that with a road or coastline won't work because there are "no protocols for one application to tell the other," Kottman said. "What GML is about is expressing all of those circumstances so the second application understands what the first application is passing to it."MAF/TIGER goals

The Census Bureau has five main objectives for the Master Address File/Topologically Integrated Geographic Encoding and Referencing database modernization program. They are:

* Correct locations of streets, other map features and housing units. Use automated change detection methods to identify new streets and housing units.

* Develop a new MAF/TIGER processing environment using commercial products and geographic information system tools.

* Expand and encourage geographic partnership programs.

* Launch the Community Address Updating System to keep the address list current.

* Institute periodic evaluations.