DGR and the NM HPC are currently evaluating demographic data obtained from several sources. This data will be used to enhance the gravity model analyses that are currently in preliminary stages of development. Work to date has used only driver counts as a surrogate measure for population in ZIP Codes. Hopefully, these commercially available demographic data sets will provide more realistic counts of persons per ZIP Code along with a host of other socio-economic information about the population that will be useful in developing "risk factor" enhancements to the basic gravity model.

DGR has currently obtained demographic data from the following sources:

ESRI's Business Analyst

DGR and the NM HPC have recently acquired a copy of ESRI's ArcView Business Analyst extension for evaluation. The ArcView Business Analyst extension is a suite of integrated software tools and data that works in conjunction with the ArcView GIS software package. At present it is designed primarily for market area analysis applications mostly for retailing and store location but can also be applied to health care applications (facility location analysis) that are based on similar principles and assumptions.

The gravity model application being developed by DGR and the NM HPC has a slightly different focus than the traditional business oriented applications supported by ESRI's ArcView Business Analyst. These traditional business applications are focused on the concepts of "market areas" and "consumer behavior". The ability to measure and predict consumer behavior and to delineate existing and proposed market areas is of paramount importance. Also, the ability to evaluate potential scenarios based on the competition between facilities (usually the introduction of new stores) is given special attention.

DGR's gravity model is not primarily concerned with "consumer/patient behavior" or the delineation of "market areas" for health care providers and facilities. Instead, it is designed to be a simple and basic measurement of the distribution of facilities and providers. As such, it should provide a relatively objective picture of the geographic access to health care providers and facilities. It does not try to measure or predict where consumers/patients will go to obtain health care or where the best place is to locate new facilities or providers. It does, however, provide a useful picture of how facilities and providers are distributed throughout the landscape (state) and what areas are relatively under or overserviced with respect to all others.

Urban Decision Systems (UDS) Data

The primary utility of ESRI's Business Analyst extension to the work being conducted by the NM HPC and DGR is the demographic data that is packaged with it. One of the databases included are nationwide (1996) market base demographic estimates prepared by Urban Decision Systems Inc. (UDS) . It contains 240 demographic variables for the current year (1996) with five- and 10-year projections by state, county, ZIP Code , and block group.

The first problems that DGR addressed in the use of this demographic data are those of completeness and reliability. The ESRI Business Analyst/UDS data needs to be compatible with DGR's Zip Code coverage (map database), the geographic underpinning for the gravity model analysis. In other words, as DGR's zipcode coverage has 400 unique zipcodes, it would be wonderful if the ESRI Business Analyst/UDS coverage (map database) provided data for all of these 400 unique zipcodes.

Unfortunately, this was not exactly the case when we examined and compared the ESRI Business Analyst/UDS coverage (a shapefile) with DGR's zipcode coverage(a polygon based ARC/INFO coverage). The ESRI Business Analyst/UDS coverage contained only 373 unique zipcodes (the number of polygons was not able to be determined because it is in shapefile format). Of the 373 unique zipcodes one was a misassigned (79045) Texas zipcode and two were for the unpopulated areas of White Sands Missle Range in Otero County and the Bandelier National Monument in Sandoval County (250 and 252). Hence, there were 370   zipcodes in the ESRI Business Analyst/UDS coverage for New Mexico. But, 103 of these had no data.  This resulted in 267 useable zipcodes from the ESRI Business Analyst/UDS coverage that had demographic data. The following maps of The ESRI Business Analyst/UDS coverage illustrate this breakdown.

ESRI Business Analyst/UDS Coverage Maps:

Having 400 zipcodes as the basis for the gravity model analysis and only 267 zipcodes availabel with demographic data does not present a completely unworkable problem. In fact, many of the zipcodes with no demographic data are also ones that have no postal delivery and were estimated (polygons added by DGR) in preparing DGR's Zipcode coverage.

In essence, The zipcodes that DGR added were basically "carved out" from another zipcode (the mother zipcode). When no demographic data is available for these (child zipcodes) it is possible to estimate demographic data for them using a proportional allocation method. For instance, if a child zipcode accounts for half the known population (proportional estimation factor of .50) of a mother zipcode a reasonable estimate would be to assign demographic data for the child zipcode based on this known proportion.

Fortunately, DGR has good (recent) counts of licensed drivers and other ID cards issued by zipcode provided by the New Mexico Motor Vehicle Division. This information is a reasonable surrogate measure for population and allowed for a proportional estimation factor to be derived for all zipcodes where demographic data needed to be estimated. In some cases there were multiple child zipcodes "carved out" of a single mother zipcode and this estimation method was able to efficiently produce a solution. Also, there were a few zipcodes that had missing demographic data and were not added by DGR. These zipcodes were assigned as children of an adjacent mother zipcode and estimates were readily derived.

The results of this proportional allocation method are, however, only as reliable as the input data used (both driver counts and demographic). In fact, a simple ratio of the count of persons (estimated or actual) divided by the driver counts (from MVD) helps in the evaluation of both the actual and estimated demographic data. A ratio value close to one indicates a reasonable correspondence between the demographic data (numerator) and the total number of licences and other identification permits issued (denominator). Excessive variation would indicate particular zipcodes where the demographic and perhaps driver counts (at this point more reliable) are questionable. The following map depicts this ratio of total persons divided by total driver counts:

ESRI Demographic Data Ratio Test Map:

Ratio Test Map (24kb)

Note that the majority of the state's zipcodes (in blue) appear to be reasonably in line with what one would expect if the data were to be trusted. That is the counts of persons in the demographic database seems to reflect reasonably well what we know about the number of licensed drivers and other ID cards issued. However, there are some zipcodes (green) where there are less persons than the number of licensed drivers and other IDs issued - and others (pink and red) where there are considerably more persons than licensed drivers and other IDs issued.  These are the areas where the results of the gravity model that employ this demographic data may be questionable. For now, this ESRI-UDS based data enhanced with DGR's estimates are the best demographic data we have and we intend to use (experiment further) it until we prove that the results are very questionable or we come up with a better source for demographic data.

Only selected elements from the original ESRI-UDS demographic data were included in the proportional allocation estimation method. These elements were selected because they are potentially useful in deriving "risk factors" for subsequent gravity model applications. These data elements have been maintained in both ArcView and SAS formats for subsequent use.

ESRI-UDS Demographic Data with DGR Estimates:

Metromail Data

Another database included with ESRI's ArcView Business Analyst is nationwide household level consumer information provided by Metromail-Experian.

Microsoft MapPoint 2000

One possibility for better data is a new product by Microsoft called MapPoint 2000. It is scheduled for release June 10, 1999. As soon as we get a copy we will evaluate it and compare it with ESRI's Business Analyst (UDS) data and Metromail data  we have.

Claritas Data

Demographic Data Results

As we continue the evaluation of each of these sources of demographic data, DGR will continue to develop some "risk based" gravity model applications. Please see the section entitled "Recent Data and Discussion of Results" for more information and example applications.