Over the years, we GIS users have acquired data from a number of sources. No matter what was obtained or from where, we had to be certain the data were suited for the task at hand. Positional accuracy, vintage, coverage area, price and other elements had to be evaluated.

But sometimes merely looking over the marketing material, or even the sample metadata or documentation may not be enough, or may even be misleading.

Many vendors sell data products that are derived from U.S. Census Bureau's TIGER data. Along with much else, the TIGER files contain a set of complex polygons resembling municipalities. TIGER refers to these statistical entities as Census Designated Places (CDPs). They are defined by the Census Bureau as statistical units for use with the decennial census.

Just as the actual decennial census numbers are fixed in time, the geographic portion of those files are frozen as well to keep the relationship meaningful. Even when TIGER issues newer releases (note that they are officially referred to as TIGER/Line files), the polygon files are still where they were as of the last full census.

The opportunity for user error may come about when purchasing "city'' data from a vendor. It has been discovered that some vendors are referring to their translated TIGER CDP coverages as "City Boundary'' files. If you have mission critical applications that depend on knowing if a given location is within a municipality's borders, you could be analyzing against data that is now nine years old. What may further reinforce a misplaced belief that you are working with current data is the files own metadata. In my case, it was discovered that simple cartographic error corrections led to the vendor's re-dating of the vintage shown in the metadata.

There is another misconception that providing CDPs as City Boundaries leads to -- that is that all the records are actually incorporated municipalities. Many of the TIGER CDP records are simply unincorporated communities that were demarcated by the Census Bureau. When doing analysis of what areas are incorporated versus unincorporated, this can cause major problems. For instance, Hillsborough County, Florida (home of Tampa) only has three incorporated cities; however, there are a total of 26 CDPs in that county.

I'm no Dickens, but my "Tale of Two Cities'' has these messages:

• Users must ensure that their data is adequate for the intended purpose:
  • If accurate city boundaries are even somewhat critical to your GIS project's success, make sure your City Boundaries are truly up to date and to an acceptable precision.
  • If cities are merely for casual orientation reference, vintage and accuracy should not matter.
• Vendors need to call coverages what they really are:
  • Cities are incorporated municipalities that have life and often change their geographical footprint over time.
  • CDPs are statistical units that only have relevance when viewed with the decennial Census Data for which they were designed and intended.