of property damage and higher insurance rates (Bin and Polasky, 2004).
Neighborhood variables were estimated for each property using several data sources. First, as access to certain amenities and disamenities may impact home sale prices, Euclidean distances from each parcel centroid to the closest shopping center, central business district, road with a high traffic volume, and college or university were calculated using a GIS. We located shopping centers using GIS polygon files available from the Twin Cities Metropolitan Council depicting major shopping centers and central business districts using Twin Cities Metro Transit downtown fare zones for Minneapolis and St. Paul. College and university locations were identified using the Metro GIS Regional Parcel Dataset described previously and high traffic volume roads were identified from the Met Council and The Lawrence Group Functional Class Roads dataset. As home sale prices may be also influenced by the quality of neighborhood schools, we calculated the average Minnesota Comprehensive Assessment test scores for each neighborhood school at the third, fifth, and seventh grade levels to indicate school quality. We obtained test scores for the year 2005 from the Minnesota Department of Education (http://education.state.mn.us/MDE/ Data/index.html) and averaged scores for each school and grade level and linked them to each residential property by their 2005 elementary and middle school district in a GIS. Because the level of taxation in a property’s community has been found to impact sales prices in past studies (Mahan et al., 2000), we calculated an
additional variable, tax rate, using estimated market values and tax rates from the parcel dataset.
To include the impacts of a property being located in different Ramsey County submarkets, we divided the Ramsey County housing market into a series of market segments. Initially, we delineated market segments using major school districts. This division proved reasonable for the suburbs, where there were relatively few sales and there was relatively little diversity within each district, but not for St. Paul, which has only one major school district, contained more than half of all properties sold in 2005, and has large diversity across neighborhoods. As such, we further divided St. Paul based on middle school districts, then merged adjacent school districts with similar attributes. This resulted in the creation of eight hous- ing submarkets, listed here in order of mean residence sale price from lowest to highest: east St. Paul (reference location), central St. Paul, North St. Paul-Maplewood School District, northwest St. Paul, White Bear Lake School District, St. Anthony-New Brighton and Roseville School Districts, Mounds View School District, and southwest St. Paul. Dummy variables were used to identify each parcel’s market segment.
Open space access may be assessed in several ways in hedonic pricing studies. Some studies use dummy variables to indicate the presence or absence of open space areas within a specified distance of a property (Lutzenhiser and Netusil, 2001; Netusil, 2005). More commonly, however, studies utilize continuous measurements that identify the land area or percent of open space within a specified