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

6) More than four years after construction.
In contrast to the Base Model, the two DISTANCE categories inside <strong>of</strong> one mile are collapsed
into a single “less than one mile” group. This approach increases the number <strong>of</strong> transactions in
each crossed subcategory <strong>of</strong> data, and therefore enhances the stability <strong>of</strong> the parameter estimates
and decreases the size <strong>of</strong> the standard errors, thus providing an increased opportunity to discover
statistically significant effects. <strong>The</strong>refore, in this model the DISTANCE variable contains four
different options:
1) Less than one mile;
2) Between one and three miles;
3) Between three and five miles; and
4) Outside <strong>of</strong> five miles. 79
<strong>The</strong> number <strong>of</strong> transactions in each <strong>of</strong> the DISTANCE and PERIOD categories is presented in
Table 18.
<strong>The</strong> coefficients <strong>of</strong> interest are 5 , which represent the vector <strong>of</strong> marginal differences between
homes sold at various distances from the wind facility (DISTANCE) during various periods <strong>of</strong>
the development process (PERIOD) as compared to the reference group. <strong>The</strong> reference group in
this model consists <strong>of</strong> transactions that occurred more than two years before the facility was
announced for homes that were situated more than five miles from where the turbines were
ultimately constructed. It is assumed that the value <strong>of</strong> these homes would not be affected by the
future presence <strong>of</strong> the wind facility. <strong>The</strong> VIEW parameters, although included in the model, are
not interacted with PERIOD and therefore are treated as controlling variables. 80
Although the comparisons <strong>of</strong> these categorical variables between different DISTANCE and
PERIOD categories is be interesting, it is the comparison <strong>of</strong> coefficients within each PERIOD
and DISTANCE category that is the focus <strong>of</strong> this section. Such comparisons, for example, allow
one to compare how the average value <strong>of</strong> homes inside <strong>of</strong> one mile that sold two years before
announcement compare to the average value <strong>of</strong> homes inside <strong>of</strong> one mile that sold in the postannouncement-pre-construction
period. For this comparison, a t-Test similar to that in the All
Sales Model is used.
79 For homes that sold in the pre-construction time frame, no turbines yet existed, and therefore DISTANCE is
created using a proxy: the Euclidian distance to where the turbines were eventually constructed. This approach
introduces some bias when there is more than one facility in the study area. Conceivably, a home that sold in the
post-announcement-pre-construction period <strong>of</strong> one wind facility could also be assigned to the pre-announcement
period <strong>of</strong> another facility in the same area. For this type <strong>of</strong> sale, it is not entirely clear which PERIOD and
DISTANCE is most appropriate, but every effort was made to apply the sale to the wind facility that was most likely
to have an impact. In most cases this meant choosing the closest facility, but in some cases, when development
periods were separated by many years, simply the earliest facility was chosen. In general, any bias created by these
judgments is expected to be minimal because, in the large majority <strong>of</strong> cases, the development process in each study
area was more-or-less continuous and focused in a specific area rather then being spread widely apart.
80 As discussed earlier, the VIEW variable was considered most relevant for the post-construction period, so
delineations based on development periods that extended into the pre-construction phase were unnecessary. It is
conceivable, however, that VIEW effects vary in periods following construction, such as in the first two years or
after that. Although this is an interesting question, the numbers <strong>of</strong> cases for the SUBSTANTIAL and EXTREME
ratings – even if combined – when divided into the temporal periods were too small to be fruitful for analysis.
43