This application lets the user specify a site (point location, using latitude/longitude coordinates or a ZIP code) anywhere in the U.S., along with one or more radius values in miles. CAPS retrieves small-area data (census block group or MCD/county subdivision level) that is located within the specified circular areas and aggregates them to create (approximate) circular area summaries.
An area (block group or MCD) is selected as being "in" the circle if the coordinates of an internal point for the area (as reported on the Census Bureau's SF3 files) is within the circle. The default output is a summary report with the demographic characteristics of the circular area(s) based on data collected in the 2000 decennial census.
The hardest part of using this application is often determining the coordinates of the site location in degrees of latitude and longitude. An easy alternative, if you do not need great precision as to the exact location, is to just enter a five-digit ZIP code in the box otherwise used for the latitude coordinate.
When entering coordinate values you can use either decimal degrees (e.g., 37.01234) or you can specify the value in degrees, minutes, and seconds using two decimal points (e.g. 37.10.15 for 37 degrees, 10 minutes, 15 seconds). Longitudes are assumed to be west and can be entered with or without a leading minus sign.
There are a number of web sites that can assist in determining coordinates for a location:
There are a number of options that allow you to have more control over how your report is generated and what kind of detail you need in your output. Most are somewhat obvious, but these notes help you know what to expect.
If you are using CAPS for small radii (say, less than three miles), you should be aware that there can be significant geographic rounding errors. CAPS aggregates block groups or MCDs using a method that totally includes or excludes an area from the circle based on a single internal point — the centroid. If a block group containing 1,500 people is near the edge of the circle, it will be entirely ignored if its centroid falls outside the circle, even though in reality many of those 1,500 persons do live within the circle. This is a limitation of the available census data. If all you really need is the total population of a circle you can get a better estimate by using census blocks (for which there is no sample data — nothing about income, poverty, education, etc.). The MCDC Geocorr application has the capability of selecting block-level data for a radius and can thus be used to obtain a more accurate population figure.
When aggregating the data, the program handles averages and medians (as opposed to simple counts) by taking weighted averages. This yields accurate results for averages (such as average household income or per-capita income) but not for medians. The nature of medians prohibits their true aggregation, unless you can aggregate a distribution from which the median can be derived (or at least estimated) in a post-aggregation step. But no such distributions are used here. To get the aggregated median household income (for example), the CAPS program just takes the weighted average of the medians, weighted by the number of households. This generally provides a pretty good estimate, but it is not precise. Note that percentage items (such as percent of persons aged 0-4) can actually be aggregated using weighted averages where the weight is the denominator of the percent calculation.