The Impact of Wind Power Projects on Residential Property Values ...

5 is a vector <strong>of</strong> d parameter estimates for the DISTANCE variables as compared to homes
sold situated outside <strong>of</strong> five miles for S study areas, and
is a random disturbance term.
To refresh, the fully restricted equation (1) takes the following form:
ln P N S X VIEW DISTANCE
(1)
0 1 2 3 4 5
s k v d
where
P represents the inflation-adjusted sale price,
N is the spatially weighted neighbors’ predicted sale price,
S is the vector <strong>of</strong> s Study Area fixed effects variables (e.g., WAOR, OKCC, etc.),
X is a vector <strong>of</strong> k home and site characteristics (e.g., acres, square feet, number <strong>of</strong> bathrooms,
condition <strong>of</strong> the home, age <strong>of</strong> home, VISTA, etc.),
VIEW is a vector <strong>of</strong> v categorical view <strong>of</strong> turbine variables (e.g., MINOR, MODERATE, etc.),
DISTANCE is a vector <strong>of</strong> d categorical distance to turbine variables (e.g., less than 3000 feet,
between one and three miles, etc.),
0 is the constant or intercept across the full sample,
1 is a parameter estimate for the spatially weighted neighbor’s predicted sale price,
2 is a vector <strong>of</strong> s parameter estimates for the study area fixed effects as compared to homes sold
in the Washington/Oregon (WAOR) study area,
3 is a vector <strong>of</strong> k parameter estimates for the home and site characteristics,
4 is a vector <strong>of</strong> v parameter estimates for the VIEW variables as compared to homes sold with
no view <strong>of</strong> the turbines,
5 is a vector <strong>of</strong> d parameter estimates for the DISTANCE variables as compared to homes sold
situated outside <strong>of</strong> five miles, and
is a random disturbance term.
<strong>The</strong> significant change between equations (1) and (F13) is that each <strong>of</strong> the primary groups <strong>of</strong>
variables in equation (F13) is interacted with the study areas (S) so that parameters can be
estimated at the study area level. For example, whereas ACRES is estimated in equation (1)
across all study areas, in equation (F13) it is estimated for each study area (i.e., Acres_WAOR,
Acres_TXHC, etc). 117 Similarly, when considering the possible impact <strong>of</strong> wind facilities on
residential sales prices, equation (1) seeks average effects that exist over the entire sample, while
equation (F13) instead looks for differential effects in each individual study area. Additionally,
in equation (F13), instead <strong>of</strong> estimating fixed effects using inter-study area parameters alone (e.g.,
WAOR, TXHC), a set <strong>of</strong> intra-study area effects (Y) - school district and census tract
delineations - are added. 118 <strong>The</strong>se latter coefficients represent not only effects that are presumed
117 This change is made because, theoretically, the contribution to sales prices <strong>of</strong> home or site characteristics may
differ between study areas – for instance Central_AC in Texas vs. New York – and therefore estimating them at the
study area level may increase the explanatory power <strong>of</strong> the model.
118 In the evaluation and selection <strong>of</strong> the best model to use as the “Base Model” a set <strong>of</strong> census tract and school
district delineations were used instead <strong>of</strong> the study area fixed effects. <strong>The</strong>se more-granular fixed effects were
extracted from GIS using house locations and census tract and school district polygons. Often, the school district
and census tract delineations were not mutually exclusive. For example, in Wisconsin the WIKCDC study area
contains four school districts and six census tracts, none <strong>of</strong> which completely overlap. Alternatively, in some study
125