We used a geographic information system (GIS) (ArcMap 9.0, ESRI, Redlands, California) to calculate estimates of the areas and populations that could be affected by SRTS funding. The primary spatial data sources were geographic boundaries and area characteristics from the 2000 U.S. Census (5
) and a 2003–2004 database of elementary, middle, and high schools in the United States coded by geographic location (6
We assigned land area in the United States to 1 of 4 categories: large urban areas, defined as having a population of at least 1 million; small urban areas, defined as having a population of 50,000–999,999; metropolitan counties (excluding urban areas); and nonmetropolitan counties. Urban areas are defined by the U.S. Census Bureau as being densely settled territories containing more than 50,000 people, with core census block groups or blocks that have a population density of at least 1000 people per square mile, and with surrounding census blocks that have an overall density of at least 500 people per square mile (7
). We imported into the GIS a data file containing boundaries and other information about U.S. Census-defined urban areas, last updated in 2000.
We then classified land areas outside of urban areas as being in metropolitan statistical areas (MSAs) or outside of MSAs, which are defined at the county level. Using information from the U.S. Census Bureau, the federal Office of Management and Budget (OMB) defines MSAs as places associated with an urban area of more than 50,000 people and with a high degree of integration within the urban core (8
). Note that we did not count urban clusters, defined as urban areas having fewer than 50,000 people, as urban areas. Urban clusters such as Curtis, Nebraska, are small towns in rural areas and are not associated with an MSA. We included small rural town areas in the nonmetropolitan areas category.
To categorize places outside of urban areas as metropolitan or nonmetropolitan, we imported a file containing information on U.S. counties into the GIS and joined it to a file containing 2003 rural-urban continuum codes (RUCCs). These codes, based on the OMB definition for MSAs, are available for every county in the United States on the U.S. Department of Agriculture Web site (9
). We consolidated the original 9 RUCCs into 2 codes: counties in metropolitan areas and counties not in metropolitan areas. This consolidation resulted in 1088 metropolitan counties and 2048 nonmetropolitan counties. We counted only the portions of metropolitan counties outside of urban areas as metropolitan areas; portions of counties within urban areas were counted as urban areas. Therefore, a metropolitan county may be divided in these analyses as a part-urban area and a part-metropolitan area outside an urban area.
Finally, we imported into the GIS a data file, or layer, for 2003–2004 that contained the location and other information about U.S. public schools with grades pre-kindergarten through 12th (6
). Approximately 90% of children in the United States attend public schools (10
). We did not include private and parochial schools in the study because data for them are less available and less reliable. In addition, we excluded 14,675 special education schools, vocational schools, other/alternative schools, and schools having no students (such as new schools that are not yet operational or schools in the process of closing). The remaining 85,919 schools were then classified into 4 mutually exclusive categories: 25,938 were in large urban areas, 19,740 were in small urban areas, 16,142 were in metropolitan counties (outside of an urban area), and 24,099 were in nonmetropolitan counties.
To analyze the area and population potentially affected by SRTS programs, we created buffers with a radius of 0.5 mile around a point at the center of each school. These buffers counted only the area within 0.5 mile of a school and did not take into account the actual distance to a school by the road network or the connectivity of roads or paths. One mile is considered a reasonable distance to walk; 2 miles is considered a reasonable distance to bike (11
). Students living within 1 mile of school are more likely to walk to school than are students living farther away (12
). However, we were not measuring distance along road networks, so we chose 0.5-mile buffers for this analysis because distance along the road network is generally longer than a direct route. Those living within the 0.5-mile buffer are more likely to be close enough to walk than are those living farther away. Given that only 31% of trips made to school by students living within 1 mile of school in 1995 were made by walking (11
), it is unlikely that many of those living farther than 1 mile would choose to walk. The Healthy People 2010
goal that relates to walking to school aims to increase the number of walkers living within 1 mile of a school, but this distance is measured by self-reports (11
). We assumed that most people reported the distance by measuring along streets and roads. Using 1-mile buffers could include many people who, in fact, live more than 1 mile away from the school when actual routes along streets and roads are measured. If 2 or more schools within a county or urban area had overlapping half-mile buffers, we combined these buffers to prevent double counting of any area or population. The total land area covered by the half-mile buffer around all public schools within an urban area or county was calculated. This figure was then divided by the total land area within each urban area or county to calculate the percentage of the total land area within 0.5 mile of the local schools. To estimate the total population potentially affected, the percentage of land area was multiplied by the total population of the county or urban area.