Those of us who make maps, and companies that use those maps, often forget ZIP Codes are not really boundaries at all. ZIP Codes are the 5-digit codes the US Postal Service assigns to help make mail delivery more efficient (and ZIP+4 codes are a further refinement of that numbering plan). In reality, “ZIP Codes” are simply collections of addresses that have the same 5-digit code. They usually don’t follow legal town/city boundaries and usually don’t follow the set of geography defined by the census bureau. And the USPS doesn’t produce maps for the public. So how can companies make maps that show ZIP Code boundaries?

By grouping together those addresses, you can probably start seeing how a boundary could be formed. But that’s easier to envision than it is to generate accurately. The first step is to assign those address to the earth in a mapping system. That’s typically done through a street network, with address ranges assigned to each street segment. Those street segments that have the same ZIP Code are then “collapsed” to form contiguous boundaries.

Is There a Way to Equate Postal and Census Geographies?

Postal geography – ZIP Codes, carrier routes – and Census Geography – tracts, blockgroups – were designed for very different purposes and thus postal and census geographies do no line up with each other. Depending on your business needs, one may make more sense than the other. But what if you have some customer data by ZIP Code but need to produce more stable territories using census tracts? Or what if you have demographic data by blockgroup but want to coordinate a direct mail campaign by postal carrier route?

There are several approaches that can be used to equate postal geography with census geography. In fact, companies employ a unique methodology to generate accurate demographics at a carrier route level. (Some actually combine several methods in an intelligent system to integrate the two and for proprietary reasons that won’t be fully explained here.)

One way that some other companies use is to just assign values based on the center points of each polygon. For example, if the center point of a postal carrier route falls inside a particular census tract, then that carrier route would inherit the values from the tract. Clearly, there are limitations to the accuracy of this approach most notably the fact that using center points alone may ignore large portions that are outside.

Conversely, another approach is to assign values based on cases where the areas of each polygon are used to produce a proportional overlay of the values. For example, if 30% of a particular postal ZIP Code overlaps with a particular census blockgroup, then assign that ZIP Code 30% of the values from the blockgroup. But this approach, in isolation, would ignore the fact that the blockgroup itself may overlap multiple ZIP Codes. Mathematically, this problem becomes exponentially difficult to calculate.

However, there are limitations to the types of values that can be cross-correlated between two fundamentally different sets of geography. Most notably, only statistical values can be assigned. Attributes such as average household income can be assigned accurately this way but exact population could not. Median age could be assigned this way, but the number of males over the age of 50 could not. That’s because in the real world, people don’t live in polygons. They live in houses and apartments and cartographers/demographers group them into polygons for ease of analysis. If 90% of the people live in the northeast corner of a polygon, raw quantities could not be assigned to other geographies that have a different origin. But since averages and medians are already statistical beasts, they lend themselves to an accurate representation by skilled mapping experts.

About the Author:

The author, Mark Friend, works with Maponics, which focuses on map data licenses for neighborhood boundaries and ZIP Code maps. You can find out more at www.maponics.com.