A Break from the Past
By Lynn Wombold and Edmond Ting, Esri Data Development

Understanding Esri's 2006 demographic updates

Population change is constant—from the slow attrition in declining neighborhoods to the rapid expansion of burgeoning suburbs. The methods of measuring local population change also tend to be constant. New developments in housing are common; new developments in methodology are rare. However, in 2006, Esri implemented address-based allocation (ABA), a new technique that measures population change by block group, which was introduced by Hanley Wood Market Intelligence as a new data source to track residential construction.

A New Method: Address-Based Allocation

Annual population change 2000�2006 for block groups in and around Stapleton, Colorado

What's new about ABA is not the input data but how the data is deployed to measure demographic change by block group. Esri has been modeling the change in households from two primary sources—the InfoBase database from Acxiom Corporation and residential delivery statistics from the U.S. Postal Service (USPS)—in addition to several ancillary sources. The USPS publishes monthly counts of residential deliveries for every postal carrier route. This is the most comprehensive and current information available for small, subcounty geographic areas. The InfoBase database of more than 111 million records is the largest collection of United States consumer data available from one source.

Neither source provides data directly for block groups. The key to using either delivery statistics reported by postal carrier routes or address list counts is transforming the data to block groups. Addresses must be geocoded correctly to the block group. This requires complete street address information and an accurate geocoding database.

Converting delivery statistics from carrier routes to block groups is more complex. Carrier routes are defined to deliver the mail while block groups are constructed to collect and report census data. Comparing two areas that are defined for wholly different purposes reveals a number of correspondence issues. Carrier routes change frequently and commonly overlap multiple block groups. Some carrier routes encompass disjointed areas that can be distant from each other, but block groups are rarely divided into multiple polygons. Carrier route overlaps require an effective method for allocating the postal delivery counts across multiple block groups.

A few techniques are used to distribute delivery statistics among component block groups. One way is to create a correspondence between carrier routes and block groups using boundary files. Changes in postal carrier routes can be tracked through quarterly updates of carrier route boundaries, then delivery statistics can be assigned to block groups with Census 2000 block data. A variation of this approach also employs boundary files but assumes a uniform distribution of households within the area. Another option uses commercial correspondence files such as the block/ZIP+4 file sold by Tele Atlas/TomTom and the ZIP+4/carrier route correspondence available from the USPS.

The best options rely upon Census 2000 blocks to apportion delivery statistics to block groups. However, these techniques require a housing base that existed in 2000 and minimal changes in the carrier routes to apportion delivery statistics effectively. If an area was developed after 2000, these techniques cannot pick up the change. Using standard geodemographic tools, it is possible to estimate the change in households from carrier route delivery statistics and apply that change to any block groups in the area with an existing housing base. But the estimated change is simply being redistributed from one summary area to another. These limitations required a more robust technique that could identify change since 2000 and the location of current households.

Esri has developed another way to link a carrier route to the correct block group(s) using the actual locations of mail deliveries. ABA uses the addresses from Acxiom's household database to apportion delivery statistics. These addresses are geocoded with carrier route and block group, using an enhanced geocoding technique and database, and serve as the foundation for the conversion. This approach is unbounded by geographic borders, arbitrary assumptions about the distribution of households or postal deliveries, and preexisting housing.

Annual population change 2000�2006 for block groups in and around Buckeye, Arizona

The effectiveness of ABA relies on the precision of block group assignments to addresses in Acxiom's InfoBase. Esri improved the geocoding accuracy of the Acxiom file by applying ArcGIS 9.2 with the Tele Atlas/TomTom Dynamap/Address Points database that provides coordinates accurate to individual buildings. It offers a new development in large-scale geographic databases where addresses are represented as points rather than approximations estimated from address ranges or street segments. The database currently covers the most densely populated areas in the United States, with continually increasing geographic coverage. Addresses that fall outside the coverage were geocoded with the conventional approach, based on address ranges.

ABA results have been tested extensively. The tests include benchmarking against Census 2000 data. Manual reviews confirm the capability of the method to identify areas with high growth. The allocation method reveals sprawls and new developments across the country since Census 2000. Assessments based on other data sources verify the efficacy and precision of ABA. For the small portion of block groups where InfoBase addresses are not available, delivery statistics are allocated from a correspondence file. The correspondence between census block groups and postal carrier routes is developed using quarterly updated data from Tele Atlas/TomTom.

Continued on page 2