Some years ago, a new player in the demographic forecasting world came onto the scene with two key points of differentiation. First, their estimates were released quarterly which somehow suggests improved accuracy. Second, their estimates are based primarily on what they called “a new source of population data – the U.S. Postal Service’s ZIP+4”.
It was, of course, not a “new” source. AGS, like most, had been using the ZIP+4 databases from the post office for many years, so it wasn’t like they discovered a new data source, or for that matter, even a new use for it. So, the question is, if AGS knew about this remarkable data source, why did we not use it as our primary source? Or, for that matter, make a big deal of it from a marketing perspective?
Over the long term, there is no doubt that the total count of ZIP+4 units nationwide is roughly parallel to the national population. It is not unreasonable at first glance to assume that since the ratio of population to ZIP+4 units is fairly constant, that the addition of a new ZIP+4 would indicate growth.
There are several reasons why AGS doesn’t use the ZIP+4 roster as one of its primary sources –
- The creation of a ZIP+4 usually occurs very early within the development process. A large development in Thousand Oaks, California (our HQ) of about 1,100 homes occurred over a twenty-five-year time period, but the ZIP+4 records were created early on when the streets (and underground services) were created. For many years, the use of a count of ZIP+4 records would have grossly overstated the population in the new area while simultaneously underestimating the population in existing nearby neighborhoods.
The larger the development, the longer it takes between the time that a ZIP+4 is created and when it is populated. At best, we view the new ZIP+4 as an indicator of future not current growth. Used in conjunction with building permits data, they are useful in building projections.
- The post office can arbitrarily reorganize the ZIP+4’s in an area which can increase or decrease the number of records, in which case we are tracking false signs of change.
- The number of mailboxes represented by a ZIP+4 has a large range which cannot be immediately ascertained from the data provided by the post office. A new ZIP+4 which is a small apartment building is not the same as a new ZIP+4 that is single family housing.
- Mailboxes are not households, as a significant percentage of households own one or more properties, usually seasonally vacant or used as vacation rentals.
In order to be usable as the primary indicator of change, the following would all need to be true –
- The lag between the creation of the ZIP+4 and actual new households must be short, predictable and consistent so that it can be modeled into the estimates appropriately.
- A ZIP+4 should represent a constant number of households (or at least mailboxes)
- Removal of housing should result in a removal of ZIP+4 records, which it rarely does. Pull up a satellite image of Flint, Michigan and look at the number of housing units which have simply been removed over the last few years. The ZIP+4 coding has not changed, but a majority of the housing has actually been demolished, leaving a patchwork of houses and open fields which were formerly fully built out.
None of these are remotely close to being true. Mailboxes do not translate well into on the ground household counts at the local level.
At the outset, we stated that we have been using the file for years as a source. A more robust methodology will incorporate the ZIP+4 record counts into a broad range of indicators of change that include changes in street files, the use of multiple residential mailing lists, and building permits. AGS believes that growth (or decline) should be based not on a single source, but on the concurrence of several sources, and as a final check, should be identifiable by analyzing changes to satellite imagery.
We remain convinced that the method of counting ZIP+4’s as the core building block of a methodology for current estimates is fundamentally flawed.
Trackbacks/Pingbacks