Microsoft Word Hoen et al (2013) A Spatial Hedonic Analysis of the Effects of Wind Energy Facilities on Surrounding Property

Transcript

1 LBNL-6362E AWRENCE L RLANDO O RNEST E L ATIONAL N ERKELEY ABORATORY B A Spatial Hedonic Analysis of the Effects of Wind Energy Facilities on Surrounding Property Values in the United States n, Thomas Jackson, Ben Hoen, Jason P. Brow yer and Peter Cappers Ryan Wiser, Mark Tha Environmental Energy Technologies Division August 2013 Download from http://emp.lbl.gov /sites/all/files/lbnl-6362e.pdf This work was supported by the Office of Energy Efficiency and Renewable Office) of the U.S. Department Energy (Wind and Water Power Technologies of Energy under Contract No. DE-AC02-05CH1123.

2 Disclaimer This document was prepared as an account of work sponsored by the United States Government. While this document is believed to contain correct information, neither the United States Government nor any agency thereof, nor The Regent California, nor any of s of the University of their employees, makes any warranty, express or implied, or assumes any legal responsibility for the accuracy, completeness, or usefulness of a ny information, apparatus, product, or process ghts. Reference herein disclosed, or represents that its use would not in fringe privately owned ri to any specific commercial product, process, or service by its trade name, trademark, or imply its endorsement, does not necessarily constitute manufacturer, or otherwise, recommendation, or favoring by the United States Government or any agency thereof, or The Regents of the University of California. The vi ews and opinions of authors expressed herein do not necessarily state or reflect t hose of the United States Governme nt or any agency thereof, The Regents of the University of California, the Fede ral Reserve Bank of Kansas City, or the Federal Reserve System. equal opportunity employer. Ernest Orlando Lawrence Berk eley National Laboratory is an

3 LBNL-6362E A Spatial Hedonic Analysis of the Effects of Wind Energy Facilities on Surrounding Property Values in the United States Prepared for the Office of Energy Efficiency and Renewable Energy Wind and Water Power Technologies Office U.S. Department of Energy Principal Authors: † Ben Hoen , Ryan Wiser, Peter Cappers Lawrence Berkeley National Laboratory, 1 Cyclotron Road, MS 90R4000, Berkeley, CA 94720-8136 Jason P. Brown Federal Reserve Bank of Kansas City 1 Memorial Drive, Kansas City, MO 64198-0001 Thomas Jackson, AICP, MAI, CRE, FRICS Real Analytics Inc. and Texas A&M University 4805 Spearman Drive, College Station, TX 77845 ‐ 4412 Mark A. Thayer San Diego State University 5500 Campanile Dr., San Diego, CA 92182-4485 August 2013 This work was supported by the Office of Energy Efficiency and Renewable Energy (Wind and Water Power Technologies Office) of the U.S. Department of Energy under Contract No. DE-AC02-05CH1123. † ; Mailing address: 20 Sawmill Road, Milan Corresponding author: Phone: 845-758-1896; Email: [email protected] NY 12571. i

4 Acknowledgements This work was supported by the Office of Ener gy Efficiency and Renewable Energy (Wind and Water Power Technologies Office) of the U.S. Department of Energy under Contract No. DE- AC02-05CH11231. For funding and suppor ting this work, we especia lly thank Patrick Gilman, Cash Fitzpatrick, and Mark Higgins (U.S. DOE). Fo r providing the data that were central to the analysis contained herein, we thank Cameron R ogers (Fiserv) and Joshua Tretter (CoreLogic Inc.), both of whom were highly supportive a nd extremely patient thr oughout the complicated many external reviewers for data-aquistion process. Finall y, we would like to thank the providing valuable comments on an earlier draft version of the re port. Of course, any remaining errors or omissions are our own. ii

5 Abstract Previous research on the effects of wind energy facilities on surrounding home values has been limited by small samples of relevant home-sale data and the inability to account adequately for confounding home-value factors a nd spatial dependence in the data. This study helps fill those gaps. We collected data from more than 50,000 home sales among 27 counties in nine states. These homes were within 10 miles of 67 different wind facilities, and 1,198 sales were within 1 mile of a turbine—many more than previous st udies have collected. The data span the periods well before announcement of the wind facilities to well after their construction. We use OLS and fference hedonic models to estim ate the home-value impacts of spatial-process difference-in-di lue factors existing before the wind facilities’ the wind facilities; these models control for va announcements, the spatial dependence of unobserve d factors effecting home values, and value changes over time. A set of robustness models a dds confidence to our results. Regardless of tical evidence that home values model specification, we find no statis near turbines were affected in the post-construction or pos t-announcement/pre-construction pe riods. Previous research on potentially analogous disamenities (e.g., high-voltage transmission lines, roads) suggests that the property-value effect of wind turbines is likely to be small, on aver age, if it is present at all, an effect was found in the present research. potentially helping to explain why no evidence of iii

6 Table of Contents ... 1 1. Introduction ... 2. Previous Literature ... ... 2 Methodology ... ... 7 3. 3.1. Basic Approach and Models ... 8 3.2. Spatial Dependence ... 12 3.3. Robustness Tests ... 14 3.3.1. Outliers and Influential Cases ... 15 3.3.2. Interacting Sale Year at the County Level ... 15 Using Only the Most Recent Sales ... 15 3.3.3. 3.3.4. Using Homes between 5 and 10 Mi les as Reference Category ... 16 3.3.5. Using Transactions Occurring More than 2 Years before Announcement as Reference Category ... ... 16 4. Data ... ... 17 4.1. Wind Turbine Locations ... 17 Real Estate Transactions ... 17 4.2. 4.3. Home and Site Characteristics ... 18 Census Information ... 19 4.4. Distances to Turbine ... 19 4.5. Wind Facility Development Periods ... 19 4.6. Data Summary ... 20 4.7. Comparison of Means ... 23 4.8. 5. Results ... ... 25 5.1. Estimation Results for Base Models ... 25 5.1.1. Control Variables ... 26 Variables of Interest ... 28 5.1.2. 5.1.3. Impact of Wind Turbines ... 32 5.2. Robustness Tests ... 34 6. Conclusion ... ... 37 7. References ... ... 39 . 44 8. Appendix – Full Results ... iv

7 Tables Table 1: Interactions between Wind Facility Development Periods and Distances – ½ Mile ... 12 Table 2: Interactions between Wind Facility Development Periods and Distances - 1 Mile ... 12 Table 3: Summary Statistics ... ... 21 Table 4: Summary of Transactions by County ... 22 Table 6: Wind Facility Summary ... .. 23 Table 7: Dependent and Inde pendent Variable Means ... 25 Table 8: Levels and Significance for County- an d State-Interacted Controlling Variables ... 28 fdp and tdis ... 31 Table 9: Results of Interacted Variables of Interest: Table 10: "Net" Differe nce-in-Difference Impacts of Turbines ... 34 Table 11: Robustness Half-Mile Model Results ... 36 Figures Figure 1: Map of Transactions, States, and Counties ... 21 v

8 1. Introduction turbines were installed in the United States, In 2012, approximately 13 gigawatts (GW) of wind bringing total U.S. installed wind capacity to from more than 45,000 approximately 60 GW turbines (AWEA, 2013). Despite uncertainty about future extensions of the federal production me to continue growing by approximately 5–6 tax credit, U.S. wind capacity is expected by so andards and areas where wind can compete with GW annually owing to state renewable energy st natural gas on economics alone (Bloomberg, 2013) ; this translates into approximately 2,750 1 Much of that development is expected to occur in relatively populated areas turbines per year. (e.g., New York, New England, Midwest) (Bloomberg, 2013). the Mid-Atlantic and upper In part because of the exp ected wind development in more-populous areas, empirical investigations into related comm unity concerns are required. One concern is that the values of properties near wind developments may be reduced; after all, it has been demonstrated that in 2 are capitalized some situations market perceptions a bout an area’s disamenities (and amenities) into home prices (e.g., Boyle and Kiel, 2001; Jackson, 2001; Simons and Saginor, 2006). The published research about wind ener gy and property values has la rgely coalesced around a finding been constructed do not that homes sold after nearby wind turbines have experience statistically significant property value impacts. Additional research is required, however, especially for es, where impacts would be expected to be the homes located within about a half mile of turbin ese proximate homes in part because setback largest. Data and studies are limited for th requirements generally result in wind facilities be ing sited in areas with relatively few houses, limiting available sales transacti ons that might be analyzed. This study helps fill the research gap by collec ting and analyzing data from 27 counties across nine U.S. states, related to 67 different wind faci lities. Specifically, using the collected data, the study constructs a pooled model th at investigates aver age effects near the turbines across the sample while controlling for the local effect s of many potentially correlated independent variables. Property-value effect estimates are deri ved from two types of models: (1) an ordinary 1 Assuming 2-MW turbines, the 2012 U.S. average (AWEA, 2013), and 5.5 GW of annual capacity growth. 2 Disamenities and amenities are defined respectively as disadvantages (e.g., a nearby noxious industrial site) and advantages (e.g., a nearby park) of a location. 1

9 least squares (OLS) model, which is standard fo r this type of disamenity research (see, e.g., ), and (2) a spatial-process model, which discussion in Jackson, 2003; Sirmans et al., 2005 model is used to construct a difference-in- accounts for spatial dependence. Each type of difference (DD) specification—which simultaneous ly controls for preexisting amenities or disamenities in areas where turbines were sited and changes in the community after the wind facilities’ construction was announced—to estimate ties after the turbines effects near wind facili 3 were announced and, later, after the turbines were constructed. The remainder of the report is structured as follows. Section 2 reviews the current literature. Section 3 details our methodology. Section 4 desc ribes the study data. Section 5 presents the results, and Section 6 provides a di scussion and concluding remarks. 2. Previous Literature Although the topic is relatively ne w, the peer-reviewed literature investigating impacts to home values near wind facilities is growing. To date, results larg ely have coalesced around a common set of non-significant findings generated from ho me sales after the turbines became operational. this area (Hoen et al., 2009, onal Laboratory (LBNL) work in Previous Lawrence Berkeley Nati 2011) found no statistical evidence of adverse pr operty-value effects due to views of and proximity to wind turbines after the turbines we re constructed (i.e., post-construction or PC). Other peer-reviewed and/or academic studies al so found no evidence of PC effects despite using a variety of techniques and residential transa ction datasets. These include homes surrounding wind facilities in Cornwall, United Kingdom (S ims and Dent, 2007; Sims et al., 2008); multiple wind facilities in McLean County, Illinois (Hin man, 2010); near the Maple Ridge Wind Facility in New York (Heintzelman and Tuttle, 2011); and, near multiple facilities in Lee County, Illinois 2012 Canadian case found a lack of (Carter, 2011). Analogously, a evidence near a wind facility in Ontario to warrant the lowering of surr ounding assessments (Kenney v MPAC, 2012). In contrast, one recent study did find impacts to la nd prices near a facility in North Rhine- Westphalia, Germany (Sunak and Ma dlener, 2012). Taken together, these results imply that the 3 Throughout this report, the terms “announced/announcemen t” and “constructed/constru ction” represent the dates on which the proposed wind facility (or facilities) ente red the public domain and the dates on which facility t (PA), post-announcement/pre- construction began, respectively. Home transactions ca n either be pre-announcemen construction (PAPC), or post-construction (PC). 2

10 PC effects of wind turbines on surrounding home valu es, if they exist, are often too small for all percentage overall), or app detection or sporadic (i.e., a sm earing in some communities for some types of properties but not others. In the post-announcement, pre-c onstruction period (i.e., PAPC), however, recent analysis has operty value effects: by theori found more evidence of potential pr zing the possible existence of, but not finding, an effect (Laposa and Mueller, 2010; Sunak and Madlener, 2012); potentially 4 finding an effect (Heintzelman and Tuttle, 2011) ; and, consistently finding what the author terms an “anticipation stigma” effect (Hinma n, 2010). The studies that found PAPC property- value effects appear to align with earlier studi es that suggested lower community support for proposed wind facilities before c onstruction—potentially indicating a ri sk-averse (i.e., fear of the unknown) stance by community member s—but increased sup port after faciliti es began operation (Gipe, 1995; Palmer, 1997; Devine-Wright, 2005; Wolsink, 2007; Bond, 2008, 2010). Similarly, researchers have found that surv ey respondents who live closer to turbines suppor t the turbines otland, 2003; Baxter et (Braunholtz and MORI Sc more than respondents who live farther away al., 2013), which could also indi of the unknown effects (these cate more risk-adverse / fear among those who live farther away). Analogously , a recent case in Canada, although dismissed, highlighted the fears that nearby residents have for a planned facility (Wiggins v. WPD Canada Corporation, 2013) Some studies have examined property-value co nditions existing before wind facilities were announced (i.e., pre-announcement or PA). This is important for explorin g correlations between wind facility siting and pre-existi ng home values from an enviro nmental justice perspective and also for measuring PAPC and PC effects more accurately. Hoen et al. (2009, 2011) and Sims and r homes that sold before a wind facility’s Dent (2007) found evidence of depressed values fo announcement and were located near the facility’s eventual location, but they did not adjust their e reductions of 12%–20% PC estimates for this finding. Hinman (2010) we nt further, finding valu for homes near turbines in Illinois, which sold prior to the facilities’ announcements; then using these findings to deflate their PC home-value-effect estimates. 4 Heintzelman and Tuttle do not appear convinced that the effect they found is related to the PAPC period, yet the d transaction data produced two counties in which they found an effect (Clinton an d Franklin Counties, NY) ha almost entirely in the PAPC period. 3

11 Some research has linked wind-related property-valu e effects with the effects of better-studied samenity literature (e.g., Boyle and Kiel, 2001; disamenities (Hoen et al., 2009). The broader di ts that, although property-value effects might Jackson, 2001; Simons and Saginor, 2006) sugges r disamenities, those effects (if they do exist) are occur near wind facilities as they have near othe ist some distance from a facility, and might fade likely to be relatively small, are unlikely to pers over time as home buyers who are more accepting of the condition move into the area (Tiebout, 1956). For example, a review of the l iterature investigating effects n ear high-voltage transmission lines (a largely visual disturbance, as turbin es may be for many surrounding homes) found the following: property-value reducti ons of 0%–15%; effects that fa de with distance, often only affecting properties crossed by or immediately ad jacent to a line or tower; effects that can increase property values when the right-of-way is considered an amenity; and effects that fade with time as the condition becomes more accepte d (Kroll and Priestley, 1992). While potentially much more objectionable to residential communitie s than turbines, a review of the literature on affic, and groundwater-contamina tion issues) indicates effects landfills (which present odor, tr that vary by landfill size (Ready, 2010). Large-volume operations (accepting more than 500 tons per day) reduce adjacent property values by 13.7% on average, fading to 5.9% one mile from the landfill. Lower-volume operations reduce adjacent property values by 2.7% on average, fading to 1.3% one mile away, with 20%–26% of lower-vol ume landfills not having any statistically significant impact. A study of 1,600 toxic industr ial plant openings found adverse impacts of 1.5% within a half mile, which disappeared if th e plants closed (Currie et al., 2012). Finally, a review of the literature on road noise (which might be an alogous to turb ine noise) shows property-value reductions of 0% –11% (median 4%) for houses adjacent to a busy road that experience a 10-dBA noise increase, compared with houses on a quiet street (Bateman et al., 2001). It is not clear where wind turbines might fit in to these ranges of impacts, but it seems unlikely that they would be considered as severe a disamenity as a large-volume landfill, which present odor, traffic, and groundwater-contamination issu es. Low-volume landfills, with an effect near 3%, might be a better comparison, because they ha ve an industrial (i.e ., non-natural) quality, similar to turbines, but are less likely to have clear health effects. If sound is the primary 4

12 concern, a 4% effect (correspondi pplied to turbines, which might ng to road noise) could be a a half mile of a turbine (see e.g., Hubbard and correspond to a 10-dBA increase for houses within s, if houses are in sight but not within sound Shepherd, 1991). Finally, as with transmission line lue effects unless those homes are immediately distance of turbines, there may be no property-va adjacent to the turbines. In summary, assumi ng these potentially analogous disamenity effects can be entirely transferred, turbine impacts migh t be 0%–14%, but more likely might coalesce closer to 3%–4%. Of course, wind turbines have certain positive qu alities that landfills, transmission lines, and roads do not always have, such as mitigating greenhouse gas emissions. no air or water pollution, no use of water during the generation of energy, and no generation of solid or hazardous waste that requires permanent storage/ disposal (IPCC, 2011). Moreover, wind facilities can, and often do, provide economic benefits to local commun ities (Lantz and Tegen, 2009; Slattery et al., 2011; Brown et al., 2012; Loomis et al., 2012), which might not be the case for all other disamenities. Similarly, wind facilities can ha ve direct positive effects on local government budgets through property tax or other similar payments (Loomis and Aldeman, 2011), which might, for example, improve school quality and t hus increase nearby home values (e.g., Haurin and Brasington, 1996; Kane et al., 2006). Thes e potential positive qua lities might mitigate entirely. Therefore for the purposes of this potential negative wind effects somewhat or even research we will assume 3-4% is a maximum possible effect. The potentially small average property-value eff ect of wind turbines, pos sibly reduced further by wind’s positive traits, might help explain why eff ects have not been discov ered consistently in previous research. To discover effects with smal l margins of error, la rge amounts of data are needed. However, previous datasets of homes very near turbines have been small. Hoen et al. (2009, 2011) used 125 PC transactions within a mile of the turbines, while others used far fewer n ~ PC transactions within a mile: Heintzelman and Tuttle (2012) ( 35); Hinman (2010) ( n ~ 11), Carter (2011) ( 41), and Sunak and Madlener (2012) ( n ~ 51). Although these numbers of n ~ observations are adequate to examine large im pacts (e.g., over 10%), they are less likely to reveal small effects with any reasonable degree of statistical significan ce. Using results from Hoen et al. (2009) and the confidence intervals for the various fixed-effect variables in that study, to find effects of various sizes were obtained. estimates for the numbers of transactions needed 5

13 Approximately 50 cases are needed to find an ef fect of 10% and larger, 100 cases for 7.5%, 200 cases for 5%, 350 cases for 4%, 700 cases for 3%, and approximately 1,000 cases for a 2.5% 5 Therefore, in order to detect an effect in the range of 3%–4%, a dataset of approximately effect. turbines will be required to de 350–700 cases within a mile of the tect it statistically, a number that to-date has not been amassed by any of the previous studies. As discussed above, in addition to being relatively small on averag e, impacts are likely to decay with distance. As such, an appropriate empirica l approach must be able to reveal spatially diminishing effects. Some researchers have used continuous variables to capture these effects, such as linear distance (Hoen et al., 2009; Sims et al., 2008) and inverse distance (Heintzelman and Tuttle, 2012; Sunak and Madlener, 2012), but doing so forces the model to estimate effects at the mean distance. In some cases, those mean s can be far from the area of expected impact. For example, Heintzelman and Tuttle (2012) estimat ed an inverse distance effect using a mean distance of more than 10 miles from the turbines, while Sunak and Madlener (2012) used a mean distance of approximately 1.9 miles. Using this approach weakens the ability of the model to be stronger. More importantly, rbines, where they are likely to quantify real effects near the tu this method encourages researchers to extrapolate their findings to the ends of the distance curve, e distances to support thes near the turbines, despite having few data at thos e extrapolations. This was the case for Heintzelman and Tuttle (2012), w ho had fewer than 10 cases within a half mile in the two counties where effects were found and onl in those counties after y a handful that sold the turbines were built, yet they extrapolated thei r findings to a quarter mile and even a tenth of a mile, where they had very few (if any) cases. Similarly, Sunak and Madlener (2012) had only six PC sales within a half mile and 51 within 1 mile , yet they extrapolated their findings to these distance bands. estimate effects at all distances is to use a One way to avoid using a single continuous function to spline model, which breaks the distances into continuous groups (Hoen et al., 2011), but this method still imposes structure on the data by forci ng the ends of each spline to tie together. A second and more transparent method is to use fixed- effect variables for disc rete distances, which imposes little structure on the data (Hoen et al., 2009; Hinman, 2010; Cart er, 2011; Hoen et al., 5 This analysis is available upon request from the authors. 6

14 2011). Although this latter method has been used in a number of studies, because of a paucity of n ineffective at detecting what data, the resulting models are ofte might be relatively small effects very close to the turbines. As su ch, when using this method (or any other, in fact) it is important that the underlying datase ticipated magnitude of the effect t is large enough to estimate the an sizes. Finally, one rarely inve stigated aspect of potential wi nd-turbine effects is the possibly idiosyncratic nature of spatially averaged transa ction data used in the hedonic analyses. Sunak and Madlener (2012) used a geographically we ighted regression (GWR), which estimates different regressions for small clusters of da ta and then allows the investigation of the distribution of effects acro ss all of the clusters. Alt hough GWR can be effective for understanding the range of impacts across the study area, it is not as effective for determining an average effect or for testing the statistical sign ificance of the range of estimates. Results from 6 As is discussed in more studies that use GWR methods are also sometimes counter-intuitive. to estimate a spatial-process a potentially better approach is detail in the methodology section, model that is flexible enough to tial heterogeneity and spatial simultaneously control for spa dependence, while also estimating an averag e effect across fixed discrete effects. In summary, building on the exis ting literature, further research is needed on property-value ines. Specifically, research is needed that uses effects in particularly close proximity to wind turb turbines, accounts for home values before the announcement of the a large set of data near the facility (as well as after announ cement but before construction), accounts for potential spatial dependence in unobserved factors effecting home va lues, and uses a fixed-effect distance model that is able to accurately es timate effects near turbines. 3. Methodology The present study seeks to respond to the identified research needs noted above, with this section describing our methodological framework for estima ting the effects of wind turbines on the value of nearby homes in the United States. 6 For example, Sunak and Madlener (2012) find larger effects related to the turbines in a city that is farther from the lly, they find stronger effects in turbines than they find in a town which is closer. Additiona the center of a third town than they do on the outskirts of that town, which do not seem related to the location of the turbines. 7

15 3.1. Basic Approach and Models answer the following questions: Our methods are designed to help Did homes that sold prior to the wind fac ilities’ announcement (PA)—and located within 1. om where the turbines were eventually a short distance (e.g., within a half mile) fr located—sell at lower prices th an homes located farther away? 2. Did homes that sold after the wind facil ities’ announcement but before construction (PAPC)—and located within a short distance (e.g., within a half mile)—sell at lower prices than homes located farther away? 3. Did homes that sold after the wind faciliti es’ construction (PC)—and located within a short distance (e.g., within a half mile)—sell at lower prices than homes located farther away? 4. For question 3 above, if no statistically iden tifiable effects are found, what is the likely maximum effect possible given the margins of error around the estimates? To answer these questions, the hedonic prici ng model (Rosen, 1974; Freeman, 1979) is used in el, 2001; Jackson, 2001; this paper, as it has been in other disamenity research (Boyle and Ki Simons and Saginor, 2006). The value of this appro ach is that is allows one to disentangle and nces of home, site, neighborhood, and market control for the potentially competing influe ly determine how home values near announced characteristics on property values, and to unique 7 To test for these effects, two pairs of “base” models are or operating facilities are affected. estimated, which are then coupled with a set to test and bound the of “robustness” models estimated effects. One pair is estimated using a standard OLS model, and the other is estimated using a spatial-process model. The models in eac h pair are different in that one focuses on all homes within 1 mile of an existing turbine ( one-mile models), which allows the maximum number of data for the fixed effect to be use on homes within a half d, while the other focuses mile ( half-mile models), where effects are more likely to appear but fewer data are available. We assume that, if effects exist near turbines, they are larger for the half-mile models than the one- mile models. 7 See Jackson (2003) for a further discussion of the Hedonic Pricing Model and other analysis methods. 8

16 As is common in the literature (Malpezzi, 2003; Sirmans et al., 2005), a semi-log functional form of the hedonic pricing model is used for all mode ls, where the dependent variable is the natural log of sales price. The OLS half-mile model form is as follows: (1) T S ln( ( ) ( ) ( ) ( )   SP ) W X C D P           iiiiiiiii 12 3 4 ab where represents the sale price for transaction SP i , i is the constant (intercept) across the full sample, α ar and if the sale occurred in winter) variables (e.g., sale ye is a vector of time-period dummy T i in which transaction i occurred, is the state in which transaction S occurred, i i occurred, W is the census tract in which transaction i i (e.g., square feet, X i is a vector of home, site, and neig hborhood characteristic s for transaction i age, acres, bathrooms, condition, percent of bloc k group vacant and owned, median age of block 8 group), C is the county in which transaction i occurred, i D o the nearest turbine) bin (i.e., is a vector of four fixed-effect variables indicating the distance (t i i is located (e.g., within a half mile, between a half and 1 mile, group) in which transaction between 1 and 3 miles, and between 3 and 10 miles), P is a vector of three fixed-effect variables i ndicating the wind projec t development period in i which transaction i occurred (e.g., PA, PAPC, PC), for the controlling variables, is a vector of estimates B 1-3 Β od interacted variables is a vector of 12 parameter estimates of the distance-development peri 4 of interest, ε is a random disturbance term for transaction i. i This pooled construction uses all pr operty transactions in the entire dataset. In so doing, it takes advantage of the large dataset in order to estimat e an average set of turbine-related effects across all study areas, while simultaneously allowing for th e estimation of controlling characteristics at 8 A “block group” is a US Census Bureau geographic delineation that contains a population between 600 to 3000 persons. 9

17 9 the local level, where they are likely to vary substantially across the study areas. Specifically, the interaction of count C ) with the vector of home, site, and neighborhood y-level fixed effects ( i ) allows different slopes for each of thes e independent variables to be estimated characteristics ( X i for each county. Similarly, interacting the state fixed-effect variables ( S ) with the sale year and i sale winter fixed effects variables ( T either Q1 or Q4) allows the ) (i.e., if the sale occurred in i estimation of the respective inflation/deflati on and seasonal adjustments for each state in the 10 Finally, to control for the potentially unique collection of neighborhood characteristics dataset. 11 Because a pooled model is that exist at the micro-level, censu s tract fixed effects are estimated. used that relies upon the full dataset, smaller eff ect sizes for wind turbines will be detectable. At the same time, however, this approach does not allow one to distinguish possible wind turbine effects that may be larger in some communities than in others. As discussed earlier, effects might predate the announcement of the wind facility and thus must be controlled for. Additionally, the area surrounding the wind fac ility might have changed over time simultaneously with the arrival of the tu rbines, which could affect home values. For example, if a nearby factory closed at the same time a wind facility was constructed, the influence of that factor on all homes in the general area would ideally be controlled for when estimating wind turbine effect sizes. To control for both of these issues simulta neously, we use a difference-in-difference ( DD) specification (see e.g., Hinman, 2010; Zabel and Gui gnet, 2012) derived from the interaction of 9 The dataset does not include “participating” landowners, thos ed on their land, but does e that have turbines situat include “neighboring” landowners, those adjacent to or n earby the turbines. One review er notes that the estimated average effects also include any effect s from payments “neighbori ng” landowners might receive that might transfer with the home. Based on previous conversations with developers (see Hoen et al, 2009), we expect that the frequency of these arrangements is low, as is the right to transfer the payments to the new homeowner. Nonetheless, our results should be interpreted as “net” of any influence whatever “neighboring” landowner arrangements might have. 10 Unlike the vector of home, site, and neighborhood char acteristics, sale price inflation/deflation and seasonal changes were not expected tovary substantially across va rious counties in the same states in our sample and therefore the interaction was made at the state level. Th is assumption was tested as part of the robustness tests though, where they are interact ed at the county level and found to not affect the results. 11 In part because of the rura l nature of many of the study areas included in the research sample, these census tracts are large enough to contain sales that ar e located close to the turbines as well as those farther away, thereby ensuring that they do not unduly absorb effects th at might be related to the turbines. Moreover each tract contains sales from wind facilities’ announcement and construction, further throughout the study periods, both before and after the ensuring they are not biasing the variables of interest. 10

18 the spatial ( D ) and temporal ( ) terms. These terms produce a vector of 11 parameter estimates P i i β ( ) as shown in Table 1 for the half-mile models and in Table 2 for the one-mile models. The 4 omitted (or reference) group in both models is th e set of homes that so ld prior to the wind facilities’ announcement and which were locate d more than 3 miles away from where the turbines were eventually located (A3). It is a ssumed that this referen ce category is likely not affected by the imminent arriva l of the turbines, a lthough this assumption is tested in the robustness tests. Using the half-mile models, to test whether the homes located near the turbines that sold in the PA period were uniquely affected ( research question 1 ), we examine A0, from which the null ld in the PAPC period ear the turbines that so hypothesis is A0=0. To test if the homes located n were uniquely affected ( ), we first determine the difference in their values as research question 2 compared to those farther away (B0-B3), while also accounting for any pre-announcement (i.e., in the local market over the development pre-existing) difference (A0-A3) and any change rmined in relation to the omitted category (A3), period (B3-A3). Because all covariates are dete y, in order to determine if homes near the the null hypothesis collapses B0-A0-B3=0. Finall turbines that sold in the PC period were uniquely affected ( research question 3 ), we test if C0- A0-C3=0. Each of these DD tests are estimated using a linear combination of variables that produces the “net effect” and a measure of th e standard error and corresponding confidence intervals of the effect, which enables the es timation of the maximum (and minimum) likely impacts for each research question. We use 90% confidence intervals both to determine significance and to estimate maximum likely effects ( research question 4 ). Following the same logic as above, th e corresponding hypothesis tests for the one-mile models C1-A1-C3=0. PC, are as follows: PA , A1=0; PAPC , B1-A1-B3=0; and, 11

19 Table 1: Interactions between Wind Facility Development Periods and Distances – ½ Mile Distances to Nearest Turbine Between Between Within Outside of 1/2 and 1 1 and 3 1/2 Mile 3 Miles Mile Miles Wind Facility Development P eriods A3 A1 P rior to Announcement A0 A2 (Omitted) After Announcement B3 B2 but Prior to B0 B1 Construction Post Construction C0 C1 C2 C3 Table 2: Interactions between Wind Facility Development Periods and Distances - 1 Mile Distances to Nearest Turbine Between Within 1 Outside of 1 and 3 Mile 3 Miles Miles Wind Facility Development Periods A3 Prior to Announcement A2 A1 (Omitted) After Announcement B2 but P rior to B3 B1 Construction C1 C2 C3 Post Construction 3.2. Spatial Dependence e data used in hedonic models is the spatially As discussed briefly above, a common feature of th dense nature of the real estate unique insights nsity can provide transactions. While this spatial de often raised is the impact of potentially omitted into local real estate markets, one concern that is variables given that this is im possible to measure all of the lo cal characteristics that affect housing prices. As a result, sp odel is likely because houses atial dependence in a hedonic m located closer to each other typically have similar unobserva ble attributes. Any correlation between these unobserved factors and the explanat ory variables used in the model (e.g., distance to turbines) is a source of omitted-variable bias in the OLS models. A common approach used in 12

20 rrect this potential bias is to incl ude local fixed effects (Hoen et al., the hedonic literature to co 2009, 2011; Zabel and Guignet, 2012), which is ou r approach as described in formula (1). In addition to including local fixed effects, sp atial econometric methods can be used to help further mitigate the potential impact of spatially omitted variables by modeling spatial dependence directly. When spatial dependence is present and appropriately modeled, more accurate (i.e., less biased) estimates of the fact ors influencing housing values can be obtained. These methods have been used in a number of previous hedonic price st udies; examples include the price impacts of wildfire risk (Donovan et al., 2007), residential community associations lin and Lozano-Gracia, 2009), and (Rogers, 2006), air quality (Anse spatial fragmentation of land been applied to studies of the however, these methods have not use (Kuethe, 2012). To this point, impact of wind turbines on property values. Moran’s I is the standard statis tic used to test for spatial de pendence in OLS residuals of the hedonic equation. If the Moran’s I is statistically significant (as it is in our models – see Section 5.1.2), the assumption of spatial independence is re jected. To account for this, in spatial-process models, spatial dependence is ro utinely modeled as an additiona l covariate in the form of a is an ε where Wy , or in the error structure spatially lagged dependent variable μ λ W με   , identically and independently distributed di sturbance term (Anselin, 1988). Neighboring criterion determines the structure of the spatial weights matrix W , which is frequently based on contiguity, distance criterion, or k -nearest neighbors (Anselin, 2002) . The weights in the spatial- weights matrix are typically row standardized so that the elements of each row sum to one. 12 , allows e SARAR model (Kelejian and Prucha, 1998) The spatial-process model, known as th autoregressive process in the lag-dependent and for both forms of spatial dependence, both as an in the error structure, as shown by: , yWyX    (2)  W  . 12 odel with spatial autoregressive residuals”. SARAR refers to a “spatial-autoregressive m 13

21 Equation (2) is often estimated by a multi-step procedure using generalized moments and et al., 2009), which is our appr instrumental variables (Arraiz oach. The model allows for the innovation term ε in the disturbance process to be hete roskedastic of an unknown form (Kelejian λ or ρ and Prucha, 2010). If either reduces to the respective spatial are not significant, the model lag or spatial error model (SEM). In our case, as is discussed later, the spatial process model reduces to the SEM, therefore both half-mile and one-mile SEMs are estimated, and, as with the OLS models discussed above, a similar set of DD “net effects” are estim ated for the PA, PAPC, and PC periods. One requirement of the spatial m dinates be unique across odel is that the x/y coor the dataset. However, the full set of data (as de scribed below) contains, in some cases, multiple 13 sales for the same property, which conseque ntly would have non-uni que x/y coordinates. Therefore, for the spatial models, only the most recent sale is used. An OLS model using this limited dataset is also estimat ed as a robustness test. In total, four “base” mode ls are estimated: an OLS one-mile model, a SEM one-mile model, an es of robustness models are OLS model. In addition, a seri half-mile model, and a SEM half-mile estimated as described next. Robustness Tests 3.3. To test the stability of and potentially bound the re sults from the four base models, a series of robustness tests are conducted that explore: the effect that outliers and influential cases have on the results; a micro-inflation/deflation adjustment by interacting the sale-year fixed effects with the county fixed effects rather than state fixed e ffects; the use of only th e most recent sale of homes in the dataset to compare results to the SEM models that use the same dataset; the application of a more conserva tive reference category by using transactions between 5 and 10 miles (as opposed to between 3 and 10 miles) as the reference; and a more conservative 13 e occurring after announcement and construction, that are The most recent sale weights the transactions to thos more recent in time. One reviewer wondered if the frequenc y of sales was affected near the turbines, which is also outside the scope of the study, though th is “sales volume” was investigated in Hoen et al. (2009), where no evidence of such an effect was discovered. An other correctly noted that the most re cent assessment is less accurate for older sales, because it might overestimate some characteristics of the home (e.g., sfla, baths) that might have changed (i.e., increased) over time. This would tend to bias those characteristics’ coefficients downward. Regardless, it is y to turbines and therefore would not bias the variables assumed that this occurrence is not correlated with proximit of interest. 14

22 reference category by using transact ions more than 2 years PA (as opposed to simply PA) as the sts is discussed in detail below. reference category. Each of these te 3.3.1. Outliers and Influential Cases Most datasets contain a subset of observations with particularly high or low values for the dependent variables, which might bias estimates in unpredictable ways. In our robustness test, we assume that observations with sales prices above or below the 99% and 1% percentile are potentially problematic outliers. Similarly, indivi and the values of the dual sales transactions corresponding independent variable s might exhibit undue influence on the regression coefficients. In our analysis, we therefore estimate a set of Cook’s Distance sta tistics (Cook, 1977; Cook and Weisberg, 1982) on the base OLS half-mile model and assume any cases with an absolute value of this statistic greater than one to be potentially problematic influential cases. To examine the ate a model with both th e outlying sales prices influence of these cases on our results, we estim and Cook’s influential cases removed. Interacting Sale Year at the County Level 3.3.2. It is conceivable that housing in flation and deflation varied drama tically in different parts of the same state. In the base models, we interact sale year with the state to account for inflation and deflation of sales prices, but a potentially more-accurate adju stment might be warranted. To explore this, a model with the interaction of sale year and county, instead of state, is estimated. 3.3.3. Using Only the Most Recent Sales The dataset for the base OLS mode ls includes not only the most recent sale of particular homes, but also, if available, the sale prior to that. So me of these earlier sales occurred many years prior to the most recent sale. The home and site characte ristics (square feet, acre s, condition, etc.) used in the models are populated via assessment data for the home. For some of these data, only the most recent assessment information is available (rather than the assessment from the time of ght be more prone to error as their characteristics might have sale), and therefore older sales mi 15

23 14 changed since the sale. ll x/y coordinates Additionally, the SEMs require that a entered into the model are unique; therefore, for those models onl y the most recent sale is used. Excluding older measurement error, and also enables a more-direct comparison sales therefore potentially reduces OLS model and SEM results. of effects between the base 3.3.4. Using Homes between 5 and 10 Miles as Reference Category The base models use the collection of homes betw een 3 and 10 miles from the wind facility (that sold before the announcement of the facility) as in which wind facility the reference category effects are not expected. However, it is conceivable that wind turb ine effects extend farther than 3 miles. If homes outside of 3 miles are affected by the presence of the turbines, then effects estimated for the target group (e.g., those inside of 1 mile) will be biased downward (i.e., smaller) in the base models. To test this possibili ty and ensure that the re sults are not biased, the used as a reference category as a robustness group of homes located between 5 and 10 miles is test. 3.3.5. Using Transactions Occurring More th an 2 Years before Announcement as Reference Category The base models use the collection of homes that sold before the wind facilities were announced s) as the reference ca tegory, but, as discussed (and were between 3 and 10 miles from the facilitie in Hoen et al. (2009, 2011), the announcement date of a facility, when news about a facility enters the public domain, might be after that project was known in private. For example, wind facility developers may begin ta lking to landowners some time before a facility is announced, and these landowners could share that news w on, the developer might ith neighbors. In additi erect an anemometer to collect wind-speed data well before the facility is formally “announced,” which might provide concrete evidence that a facility may soon to be announced. In either case, this news might enter the local real estate mark et and affect home pri ces before the formal facility announcement date. To expl ore this possibility, and to ensu re that the reference category 14 As discussed in more detail in the Section 4, approximately 60% of all the data obtained for this study (that obtained from CoreLogic) used the most recent assessmen t to populate the home and site characteristics for all transactions of a given property. 16

24 is unbiased, a model is estimated that uses transa ctions occurring more than 2 years before the (and between 3 and 10 miles) wind facilities were announced as the reference category. Combined, this diverse set of robustness tests a llows many assumptions used for the base models ter confidence in the final results. to be tested, potentially allowing grea 4. Data The data used for the analysis are comprised of f our types: wind turbine lo cation data, real estate transaction data, home and site characteristic da ta, and census data. From those, two additional sets of data are calculated: di stance to turbine and wind facility development period. Each data type is discussed below. Where appr italics . opriate, variable names are shown in 4.1. Wind Turbine Locations Location data (i.e., x/y coordinates) for instal led wind turbines were obt ained via an iterative process starting with Federal Av iation Administration obstacle data , which were then linked to 15 specific wind facilities by Ventyx and matched with facility-lev el data maintained by LBNL. Ultimately, data were collected on the location of almost all wind turbines installed in the U.S. n acility’s announcement, construction, through 2011 ( ~ 40,000), with information about each f and operation dates as well as turb , rotor diameter, and facility ine nameplate capacity, hub height size. 4.2. Real Estate Transactions Real estate transaction data we re collected through two source s, each of which supplied the home’s sale price ( sp ), sale date ( sd ), x/y coordinates, and addr ess including zip code. From of sale price ( those, the following variables were calc ulated: natural log ), sale year ( lsp sy ), if the sale occurred in winter ( ) (i.e., in Q1 or Q4). swinter The first source of real estate transaction da ta was CoreLogic’s extensive dataset of U.S. 16 residential real estate information. Using the x/y coordinates of wind turbines, CoreLogic 15 See the EV Energy Map, which is part of the Velocity Suite of products at www.ventyx.com . 16 . www.corelogic.com See 17

25 selected all arms-length single- between 1996 and 2011 within 10 family residential transactions ey maintained data (not including New York – miles of a turbine in any U.S. counties where th 17 The full set of counties for which data were see below) on parcels smaller than 15 acres. collected were then winnowed to 26 by requiring at least 250 transacti ons in each county, to ensure a reasonably robust estimation of the c ontrolling characteristic s (which, as discussed above, are interacted with county-level fixed effects), and by requiring at least one PC transaction within a half mile of a turbine in each county (because this study’s focus is on homes that are located in close proximity to turbines). The second source of data was the New Yo rk Office of Real Property Tax Service 18 (NYORPTS), which supplied a set of arms-length si ngle-family residential transactions between 2001 and 2012 within 10 m ines in any New York county in which iles of existing turb wind development had occurred prior to 2012. As before, only parcels smaller than 15 acres tions and at least one PC transaction within a were included, as were a minimum of 250 transac half mile of a turbine for each New York c ounty. Both CoreLogic and NYORPTS provided the available, the prior sale. most recent home sale and, if 4.3. Home and Site Characteristics A set of home and site characteristic data was also collected from both data suppliers: 1000s of square feet of living area ( sfla1000 ), number of acres of the parcel ( acres ), year the home was built (or last renovated, whichever is more recent) ( yrbuilt ), and the number of full and half 19 the other variable s as well: log of Additional variables were calculated from ). bathrooms ( baths 20 21 1,000s of square feet ( ), the number of acres less than 1 ( lt1acre) , age at the time of lsfla1000 22 age ), and age squared ( agesqr ). sale ( 17 The 15 acre screen was used because of a desire to excl ude from the sample any trans action of property that might be hosting a wind turbine, and therefore directly benef itting from the turbine’s presence (which might then increase property values). To help ensure that the screen was e ffective, all parcels within a mile of a turbine were also visually inspected using satellite and ortho imagery via a geographic information system. 18 See www.orps.state.ny.us 19 Baths was calculated in the following manner: full bathrooms + (half bathrooms x 0.5). Some counties did not have baths data available, so for them baths was not used as an independent variable. 20 The distribution of sfla1000 is skewed, which could bias OLS estimates, thus lsfla1000 is used instead, which is was used instead. sfla1000 more normally distributed. Regression results, though, were robust when 18

26 Logic supplied the related home and site Regardless of when the sale occurred, Core characteristics as of the most recent assessmen t, while NYORPTS supplied the assessment data 23 as of the year of sale. 4.4. Census Information Each of the homes in the data was matched (b ased on the x/y coordinates) to the underlying census block group and tract via ArcGIS. Using the year 2000 block group census data, each transaction was appended with neighborhood characterist ics including the me dian age of the ), the total number of housing units ( residents ( medage ) homes, units ), the number vacant ( vacant ) homes. From these, the percen tages of the total number of owned and the number of owned ( and housing units in the block group that were vacant and owned were calculated, i.e., pctvacant pctowned . 4.5. Distances to Turbine Using the x/y coordinates of both the homes and th e turbines, a Euclidian distance (in miles) was ), regardless of when the sale occurred calculated for each home to the nearest wind turbine ( tdis 24 (e.g., even if a transaction occurred prior to the wind These were then facility’s installation). exclusive distance bins (i half-mile models: broken into four mutually .e., groups) for the base inside a half mile, between a ha lf and 1 mile, between 1 and 3 m iles, and between 3 and 10 miles. lly exclusive bins for the base one-mile They were broken into three mutua models: inside 1 mile, between 1 and 3 miles, and between 3 and 10 miles. 4.6. Wind Facility Development Periods After identifying the nearest wind turbine for each home, a match could be made to Ventyx’ dataset of facility-develop ment announcement and construc tion dates. These facility- development dates in combination with the dates of each sale of the homes determined in which 21 st st This variable allows the separate estimations of the 1 acre and any additional acres over the 1 . 22 Age and agesqr together account for the fact that, as homes age, their values usually decrease, but further increases in age might bestow count ervailing positive “antique” effects. 23 See footnote 13. 24 Before the distances were calculated, each home inside of 1 mile was visually insp ected using satellite and ortho imagery, with x/y coordinates corrected , if necessary, so that those coordinates were on the roof of the home. 19

27 of the three facility fdp ) the transaction occurred: pre-announcement (PA), -development periods ( (PAPC), or post-announcement-pre-construction (PC). post-construction 4.7. Data Summary After cleaning to remove missing or erroneou s data, a final dataset of 51,276 transactions 25 As shown in the map of the stud was prepared for analysis. y area (Figure 1), the data are arrayed across nine states and 27 counties (see Table 4), and surround 67 different wind facilities. Table 3 contains a summary of those data. The av erage unadjusted sales price for the sample is $122,475. Other average house characteristics include the following: 1,600 square feet of living 26 ; land parcel size of 0.90 acres; 1.6 bathrooms; in a block group in space; house age of 48 years which 74% of housing units are owned, 9% are vacan t, and the median resident age is 38 years; located 4.96 miles from the near est turbine; and sold at th e tail end of the PA period. ance bins as would be expected, with smaller The data are arrayed across the temporal and dist numbers of sales nearer the turbines, as shown in Table 5. Of the full set of sales, 1,198 occurred within 1 mile of a then-current or future turbine location, and 376 of these occurred post construction; 331 sales occurred within a half mile, 104 of whic h were post construction. Given these totals, the models should be able to dis cern a post construction eff ect larger than ~3.5% within a mile and larger than ~7.5% within a half mile (see discussion in Section 2). These effects are at the top end of the expected range of effects based on other disamenities (high- voltage power lines, roads, landfills, etc.). 25 Cleaning involved the removal of all data that did not have certain core characteristics (sale date, sale price, sfla, yrbuilt, acres, median age , etc.) fully populated as well as the removal of any sales that had seemingly miscoded that was greater than acres, having a yrbuilt more than 1 year after the sale, having less than sfla data (e.g., having a one bath ) or that did not conform to the rest of the data (e.g., had acres or sfla that were either larger or smaller, respectively, than 99% or 1% of the data). OLS models were rerun with those “nonconforming” data included with to the screened data presented in the report. no substantive change in the results in comparison 26 Age could be as low as -1(for a new home) for homes that were sold before construction was completed. 20

28 Figure 1: Map of Transactions, States, and Counties Table 3: Summary Statistics Variable De s criptio n M e an Std. De v . M in M ax s a le pric e in dolla r s 122,475 $ 80,367 s p 9,750 $ 690,000 $ $ lsp natural log of sale price 11.52 0.65 9.19 13.44 sd sale date 1/18/2005 1,403 days 1/1/1996 9/30/2011 sy sale year 3.84 1996 2011 2005 sfla1000 1.60 0.57 0.60 4.50 living area in 1000s of square feet lsfla1000 natural log of sfla1000 0.41 0.34 -0.50 1.50 acres number of acres in parcel 0.90 1.79 0.03 14.95 0.00 -0.97 0.34 acreslt1* -0.58 acres less than 1 48 37 -1 297 age of home at time of sale age 4925 age squared 3689 88209 0 agesq baths** number of bathrooms 1.60 0.64 1.00 5.50 0.63 pctowner fraction of house units in block group that are owned (as of 2000) 0.74 0.17 0.98 0.38 0.00 pctvacant 0.10 fraction of house units in block group that are vacant (as of 2000) 0.09 63 median age of residents in block group (as of 2000) 38 6 20 med_age distance to nearest turbine (as of December 2011) in miles tdis 4.96 2.19 0.09 10.00 3.00 0.87 1.00 fdp*** facility development period of nearest turbine at time of sale 1.94 Note: The number of cases for the full dataset is 51,276 * acreslt1 is calculated as follows: acres (if less than 1) * - 1 ** Some counties did not have bathrooms populated; for those, these variables are entered into the regression as 0. *** fdp periods are: 1, pre-announcement,; 2, post-announcement-pre-construction; and, 3, post-construction. 21

29 Table 4: Summary of Transactions by County State <1/2 mile 1-3 miles 3-10 miles Total County 1/2-1 mile 12 1,065 56 331 666 IA Carroll 119 IA 2 402 526 Floyd 3 IA 8 1 9 322 Franklin 340 Sac 6 77 78 485 646 IA IL 44 8 DeKalb 605 661 4 Livingston 6 237 16 1,883 2,142 IL McLean IL 18 88 380 4,359 4,845 Cottonwood MN 10 126 1,012 1,151 3 MN 16 117 Freeborn 2,521 2,671 17 Jackson 28 36 19 149 232 MN Martin MN 7 25 332 2,480 2,844 Atlantic 96 1,532 34 6,211 7,873 NJ Paulding OH 15 58 115 309 497 563 Wood 5 31 OH 4,844 5,443 2,252 349 1,834 Custer OK 45 24 Gra dy 6 97 874 978 1 OK 297 2 10 284 Fayette PA 1 PA 100 1,037 23 2,144 3,304 Somerset Wayne PA 4 29 378 739 1,150 Kittitas 6 61 2 349 418 WA N Y 4 6 49 1,419 1,478 C linton Franklin NY 16 41 75 149 281 NY 2,248 1,874 Herkimer 3 17 354 Lewis 6 93 5 732 836 NY Madison NY 5 26 239 3,053 3,323 5 Steuben NY 52 140 1,932 2,129 Wyoming 1,646 NY 50 50 250 1,296 Total 331 867 8,919 41,159 51,276 ine Distance and Development Period Bins Table 5: Frequency Crosstab of Wind Turb <1/2 mile 1/2-1 mile 3-10 miles total 1-3 miles PA 383 3,892 16,615 21,033 143 PAPC 84 212 1,845 9,995 12,136 PC 104 272 3,182 14,549 18,107 41,159 51,276 8,919 total 331 867 22

30 As shown in Table 6, the home sales occurred at range from a single- around wind facilities th turbines of 290–476 feet (averaging almost 400 turbine project to projects of 150 turbines, with to tip of blade and with an average nameplate capacity of 1,637 feet) in total height from base kW. The average facility was announced in 200 4 and constructed in 2007, but some were announced as early as 1998 and others were constructed as late as 2011. Table 6: Wind Facility Summary 75th 25th max percentile me dia n pe r c e ntile min mean 262 turbine rotor diameter (feet) 154 253 253 269 328 328 256 197 256 262 262 turbine hub height (feet) 388 290 387 turbine total height (feet) 389 397 476 turbine capacity (kW) 1637 1500 1500 1800 2500 660 f a c ility a 2010 2005 2003 2002 1998 2004 nnouncement year 2006 2010 f a c ility c ons tr uc tion ye a r 2007 2000 2004 2011 numbe r of tur bine s in f a c ility 5 1 48 35 84 150 300 79 1. 5 7. 5 53 137 na me pla te c a pa c ity of f a c ility ( MW) Note: The data correspond to 67 wind facilities located in the study areas. Mean values are rounded to integers 4.8. Comparison of Means To provide additional context for the analysis discussed in the next section, we further summarize the data here using four key variab development period ( fdp ) and les across the sets of 27 distance bins ( one-mile models. tdis The variables are the de pendent variable log of ) used in the sale price (lsp) and three independent variables: lsfla100, acres . These summaries are age , and provided in Table 7; each sub- fdp table gives the mean values of the variables across the three bins and three bins, and the corresponding fi gures plot those values. tdis The top set of results are focused on the log of the sales price, and show that, based purely on price and not controlling for differences in homes , homes located within 1 mile of turbines had lower sale prices than homes farther away; this is true across all of the three development periods. Moreover, the results also show that, over the th ree periods, the closer homes appreciated to a somewhat lesser degree than homes located farthe r from the turbines. As a result, focusing only on the post-construction period, these results might suggest that home prices near turbines are 27 models reveal a similar relationship, so only the half-mile Summaries for the model summaries are one-mile shown here. 23

31 adversely impacted by the turbines. After all, the logarithmic values for the homes within a mile ee miles (11.72) translate of the turbines (11.39) and those outside of a thr into an approximately 40% difference, in comparison to an 21% diffe rence before the wind facilities were announced 28 (11.16 vs. 11.35). values between the pre-announcement Focusing on the change in average and post-construction periods might also suggest an adverse effect due to the turbines, because homes inside of 1 mile appreciated more slowly (11.16 to 11.39, or 25%) than those outside of 3 miles (11.35 to 11.72, or 45%). Both conclusions of adverse turbine effects, however, disregard other important differences between the homes, wh ich vary over the periods and distances. Similarly, comparing the values of the PA inside 1 mile homes (11.16) and the PC outside of 3 miles homes (11.72), which translates into a di fference of 75%, and which is the basis for comparison in the regressions discussed below, bu t also ignores any diffe rences in the underlying characteristics. The remainder of Table 7, for example, indicates th at, although the homes that sold within 1 mile are lower in value, they are also generally (in a r, on larger parcels of ll but the PA period) smalle land, and older. These differences in home size and age across the periods and distances might explain the differences in price, while the differences in the size of the parcel, which add value, further amplifying the differences in price. Wit hout controlling for these possible impacts, one cannot reliably estimate the impact of wind turbines on sales prices. In summary, focusing solely on trends in home price (or price per squa re foot) alone, and for r analysis, might incorre only the PC period, as might be done in a simple ctly suggest that wind turbines are affecting price wh en other aspects of the markets, and other home and sites characteristic differences, could be driving the observed price di fferences. This is precisely why researchers generally prefer the hedonic model approach to cont rol for such effects, and the results from our hedonic OLS and spatial modeling de tailed in the next section account for these and many other possible in fluencing factors. 28 exp(11.72-11.39)-1=0.40 and exp(11.35-11.16)-1=0.21. Percentage differences are calculated as follows: 24

32 Table 7: Dependent and I ndependent Variable Means Results 5. This section contains analysis results and discussion for the four base models, as well as the results from the robustness models. 5.1. Estimation Results for Base Models 25

33 29 Estimation results for the “base” models are shown in Table 8 and Table 9. In general, given 2 the diverse nature of the data, the mode values ranging ls perform adequately, with adjusted R from 0.63 to 0.67 (bottom of Table 9). 5.1.1. Control Variables s, which are interacted at the county level, The controlling home, site, and block group variable one-mile OLS are summarized in Table 8. Table 8 focuses on only one of the base models, the 30 To concisely summarize model, but full results from all models are shown in the Appendix. results for all of the 27 counties, the table contains the percenta ge of all 27 counties for which each controlling variable has statis tically significant (at or below the 10% level) coefficients for the OLS model. For those controlling variab les that are found to be statistically one-mile significant, the table further contains mean values, standard deviations, and minimum and maximum levels. Many of the county-interacted controlling variables (e.g., lsfla1000, lt1acre, age, agesqr, baths, and swinter ) are consistently (in more than two thirds of the counties) statistically significant (with a p -value < 0.10) and have appr opriately sized mean values. The seemingly spurious minimum and maximum values among some of the county-level contro lling variables (e.g., ar counties are highly lt1acre variables in particul minimum of -0.069) likely arise when these correlated with other variable s, such as square feet ( ), and also when sample size is lsfla1000 31 pctvacant, acres and the three block group level census variables: The other variables ( limited. pctowner, med_age ) are statistically significant in 33-59% of the counties. Only one and variable’s mean value—the pe rcent of housing units vacant in the block group as of the 2000 census ( pctvacant )—was counterintuitive. In that instan ce, a positive coefficient was estimated, when in fact, one would expect vacant housing would lower prices; that increasing the percent of 29 The OLS models are estimated using the areg procedure in Stata with robust (White’s corrected) standard errors (White, 1980). The spatial error models are estimated using the gstslshet routine in the sphet package in R, which also allows for robust standard erro rs to be estimated. See: http://cran.r-project.org/web/packages/sphet/sphet.pdf 30 The controlling variables’ coefficients were similar across the base models, so only the one-mile results are summarized here. 31 The possible adverse effects of these collinearities were fully explored both via the removal of the variables and by examining VIF statistics. The VOI results are robust to controlling variable removal and have relatively low (< 5) VIF statistics. 26

34 due to collinearity with one or more of the other variables, or this counter-intuitive effect may be 32 possible measurement errors. The sale year variables, which ar e interacted with the state, are also summarized in Table 8, with the percentages indicating the num ber of states in which the co efficients are statistically significant. The inclusion of these sale year variables in the regr essions control for inflation and deflation across the various states over the study period. The coefficients represent a comparison to the omitted year, which is 2011. All sale year state-level coefficients are statistically years except 2010, and they are significant in two significant in at least 50% of the states in all thirds of the states in all except 3 years. The s are appropriately signed, mean values of all year showing a monotonically ordered peak in values in 2007, with lower values in the prior and following years. The minimum and maximum valu es are similarly signed (negative) through 2003 and from 2007 through 2010 (positive), and are both positive and negative in years 2003 through 2006, indicating the differen ces in inflation/deflation in those years across the various states. This reinforces the appropr iateness of interacting the sale ye ars at the state level. Finally, although not shown, the model also contains 250 fi xed effects for the census tract delineations, of which approximately 50% we re statistically significant. 32 The removal of this, as well as the other block group census variables, however, did not substantively influence the results of the VOI. 27

35 33 Table 8: Levels and Significance for County- and State-Interacted Controlling Variables % of Counties/States Having Significant Statistics for Significant Variables -value <0.10) ( p Coefficients Mean Min Max Variable St Dev 100% 0.604 0.153 0.332 0.979 lsfla1000 acres 48% 0.035 -0.032 0.091 0.025 lt1acre 0.280 0.170 -0.069 0.667 85% 81% 0.010 -0.021 age 0.008 -0.006 agesqr -0.006 0.063 -0.113 0.108 74% 0.366 85% 0.156 0.088 0.083 baths* pctvacant 48% 1.295 3.120 -2.485 9.018 pctowner 33% 0.605 0.811 -0.091 2.676 -0.508 0.132 0.066 med_age 59% -0.016 swinter -0.020 -0.053 0.012 -0.034 78% sy1996 100% -0.481 -0.820 -0.267 0.187 -0.448 sy1997 100% -0.242 -0.791 0.213 sy1998 -0.404 0.172 -0.723 -0.156 100% -0.156 100% -0.359 0.169 -0.679 sy1999 sy2000 -0.088 88% -0.298 0.189 -0.565 -0.080 sy2001 88% -0.286 0.141 -0.438 67% sy2002 -0.330 0.074 -0.261 -0.128 -0.119 67% -0.218 0.069 -0.326 sy2003 75% sy2004 0.087 -0.208 0.133 -0.084 0.278 sy2005 67% 0.082 0.148 -0.111 sy2006 67% 0.340 -0.066 0.158 0.128 sy2007 0.297 0.143 0.057 0.196 67% sy2008 0.084 0.051 0.160 56% 0.218 50% 0.071 sy2009 0.138 0.065 0.219 0.231 0.105 0.063 0.172 33% sy2010 * % of counties significant is reported only for counties that had the baths variable populated (17 out of 27 counties) 5.1.2. Variables of Interest fdp bins, are shown in Table 9 for tdis and The variables of interest, the interactions between the the four base models. The reference (i.e., omitted) case for these variables are homes that sold prior to the wind facilities’ a nnouncement (PA) and are located be tween 3 and 10 miles from the 33 one-mile Controlling variable statistics are provided for only the OLS model but did not differ substantially for other models. All variables are interacted with counties, excep t for sale year (sy), which is interacted with the state. 28

36 ansactions, three of the eight In relation to that group of tr wind turbines’ eventual locations. interactions in the one-mile models and four of th e 11 interactions in the half-mile models cally significant (at the 10% level). produce coefficients that are statisti cients show statistically significant differences Across all four base models none of the PA coeffi between the reference category (out side of 3 miles) and the group of transactions within a mile 34 or within a models (OLS: -1.7%, p -value 0.48; SEM: -0.02%, p -value 0.94) for the one-mile half- or between one-half and one-mile for the models (OLS inside a half mile: 0.01%, half-mile p -value 0.97; between a ha lf and 1 mile: -2.3%, p -value 0.38; SEM inside a half mile: 5.3%, p - value 0.24; between a half and 1 mile: -1.8%, p -value 0.60). Further, none of the coefficients are partially explains thei r non-significance). Given significant, and all are relatively small (which these results, we find an absence of evidence of a PA effect for homes close to the turbines ( ). These results can be contrasted wi th the differences in prices between research question 1 es as summarized in Section 4.8 when no within-1-mile homes and outside-of-3-miles hom , the neighborhood, etc. are accounted for. The differences in the homes, the local market the pre-announcement period 1-mile homes, as approximately 75% difference in price (alone) in Section 4.8, is largely explained by differences compared to the PC 3-mile homes, discussed in why the pre-announcement distance coefficients in the controlling characteristics, which is shown here are not sta tistically significant. Turning to the PAPC and PC periods, the result s also indicate statistically insignificant differences in average home values, all else being equal, between the reference group of transactions (sold in the PA pe more than 3 miles from the riod) and those similarly located turbines but sold in the PAPC or PC periods. Those differences are estimated to be between - 0.8% and -0.5%. The results presented above, and in Table 8, in clude both OLS and spatial models. Prior to estimating the spatial models, the Moran’s I was calculated using the residuals of an OLS model that uses the same explanatory variables as th e spatial models and the same dataset (only the most recent transactions). The Moran’s I statistic (0.133) was highly significant ( p -value 0.00), 34 p-values are not shown in the table can but can be derived from the standard errors, which are shown. 29

37 e residuals are spatially independent. Therefore, which allows us to reject the hypothesis that th there was justificati on in estimating the spatial models . However, after estimation, we determined that only the spatial error process wa s significant. As a result, we estimated spatial error models (SEMs) for the fina l specification. The sp atial autoregressive coefficient, lambda (bottom of Table 9), which is an indication of spa tial autocorrelation in th e residuals, is sizable -value 0.00). The SEM models’ variable-of- p and statistically significant in both SEMs (0.26, interest coefficients are quite similar to those of the OLS models. In most cases, the coefficients are the same sign, approximately the same level, a nd often similarly insignif icant, indicating that although spatial dependence is presen t it does not substan tively bias the variables of interest. The one material difference is the coefficient size and significance for homes outside of 3 miles in the PAPC and PC periods, 3.3% ( p -value 0.000) and 3.1% ( p -value 0.008), indicating there are important changes to home values over the periods that must be accounted for in the later DD models in order to isolate the pot ential impacts that occur due to the presence of wind turbines. 30

38 ed Variables of Interest: and tdis Table 9: Results of Interact fdp one-mile half-mile half-mile one-mile OLS SEM OLS SEM fdp tdis β (se) β (se) β (se) β (se) -0.017 0.002 < 1 mile PA (0.031) (0.024) 0.008 -0.015 1-2 miles PA (0.016) (0.011) Omitted Omitted PA > 3 miles n/a n/a -0.035 -0.038 < 1 mile PAPC (0.033) (0.029) -0.033. -0.001 1-2 miles PAPC (0.014) (0.018) -0.006 -0.033* ** > 3 miles PAPC (0.01) (0.008) -0.022 0.019 < 1 mile PC (0.032) (0.026) -0.001 0.044** * PC 1-2 miles (0.014) (0.019) -0.031** -0.005 PC > 3 miles (0.010) (0.012) 0.053 0.001 < 1/2 mile PA (0.045) (0.039) -0.023 -0.018 1/2 - 1 mile PA (0.035) (0.027) 0.008 -0.015 PA 1-2 miles (0.016) (0.011) Omitted Omitted PA > 3 miles n/a n/a -0.065 -0.028 PAPC < 1/2 mile (0.049) (0.056) -0.027 -0.038 PAPC 1/2 - 1 mile (0.033) (0.036) -0.001 -0.034. PA PC 1-2 miles (0.014) (0.017) -0.033* ** -0.006 PAPC > 3 miles (0.009) (0.008) -0.036 -0.016 PC < 1/2 mile (0.046) (0.041) 0.032 -0.016 1/2 - 1 mile PC (0.031) (0.035) 0.044** * -0.001 1-2 miles PC (0.018) (0.014) -0.031** -0.005 > 3 miles PC (0.010) (0.012) 0.247 ** * 0.247 ** * lamb d a (0.008) (0.008) Note: p-values: < 0.1 *, < 0.05 **, <0.01 ***. 38,407 51,276 n 51,276 38,407 adj R-sqr 0.670.640.670.64 31

39 Impact of Wind Turbines 5.1.3. As discussed above, there are important differen ces in property values between development periods for the reference group of homes (those located outside of 3 miles) that must be accounted for. Further, although they are not si gnificant, differences between the reference side of 1 mile in the PA peri od still must be accounted for if category and those transactions in turbine effects are to be estimated. The DD accurate measurements of PAPC or PC wind specification accounts for both of these critical effects. Table 10 shows the results of the DD tests across the four models, based on the results for the 35 For example, to determine the net difference for variables of interest presented in Table 9. homes that sold inside of a half mile (drawing from the OLS model) in the PAPC half-mile period, we use the following formula: PAPC half -mile coefficient (-0.028) less the PAPC 3-mile coefficient (-0.006) less the PA half-mile co efficient (0.001), which equals -0.024 (without 36 rounding), which equates to 2.3% difference, and is not statistic ally significant. None of the DD effects in either the OLS or SEM specifications are statistically significant in the PAPC or PC periods, indicating th at we do not observe a statistical ly significant impact of wind are apparent in the calculated coefficients, turbines on property values. Some small differences with those for PAPC being gene rally more negative/less positive than their PC counterparts, that declines once a facility is constructed. perhaps suggestive of a small announcement effect more negative/less positive than their between-a- Further, the inside-a-half-mile coefficients are aps suggestive of a small propert y value impact very close to half-and-1-mile counterparts, perh 37 turbines. However, in all cases, the sizes of these differences are smaller than the margins of error in the model (i.e., 90% c onfidence interval) and thus are not statistically significant. Therefore, based on these results, we do not find evidence supporting either of our two core hypotheses ( ). In other words, there is no statistical evidence that research questions 2 and 3 homes in either the PAPC or PC periods that sold near turbines (i.e., within a mile or even a half 35 All DD estimates for the OLS models were calculated usi ng the post-estimation “lincom” test in Stata, which uses the stored results’ variance/covariance matr ix to test if a linear combination of coefficients is different from 0. For the SEM models, a similar test was performed in R. 36 All differences in coefficients are converted to percentages in the table as follows: exp(coef)-1. 37 Although not discussed in the text, this trend continues with homes between 1 and 2 miles being less negative/more positive than homes closer to the turbines (e.g., those within 1 mile). 32

40 mile) did so for less than similar homes that sold between 3 and 10 away miles in the same period. Further, using the standard errors from the DD models we can estimate the maximum size an average effect would have to be in ou research question 4 ). r sample for the model to detect it ( For an average effect in the PC period to be found for homes within 1 mile of the existing turbines (therefore using the one-mile model results), an effect gr eater than 4.9%, either positive 38 or negative, would have to be pr In other words, it is highly esent to be detected by the model. unlikely that the true average effect for homes that sold in our sample area within 1 mile of an existing turbine is larger than +/-4.9%. Similarly, it is highly unlikely that the true average effect for homes that sold in our sample area within a ha lf mile of an existing turbine is larger than +/- 39 Regardless of these maximum effects, however, 9.0%. weak suggestion of a as well as the very possible small announcement effect and a possible small effect on homes that are very close to turbines, the core results of these models show e ffect sizes that are not statistically significant 40 from zero, and are considerably smaller than these maximums. 38 Using the 90% confidence interval (i.e., 10% level of significance) and assuming more than 300 cases, the critical t-value is 1.65. Therefore, using the st andard error of 0.030, the 90% confidence intervals for the test will be +/- 0.049. 39 Using the critical t-value of 1.66 for the 100 PC cases within a half mile in our sample and the standard error of 0.054. 40 It is of note that these maximum effects are slightly larg er than those we expected to find, as discussed earlier. sample, causing relatively higher standard errors for the This likely indicates that there was more variation in this same number of cases, than in the sample used for the 2009 study (Hoen et al., 2009, 2011). 33

41 Table 10: "Net" Difference-in-Difference Impacts of Turbines < 1/2 Mile < 1/2 Mile < 1 Mile < 1 Mile SEM SEM OLS OLS b/se fdp tdis b/se b/se b/se NS NS -1.2% -0.7% PAPC < 1 mile (0.033) (0.037) NS NS 4.2% 0.7% < 1 mile PC (0.030) (0.035) NS NS -8.1% -2.3% PAPC < 1/2 mile (0.060) (0.065) NS NS -0.8% 2.5% PAPC 1/2 - 1 mile (0.039) (0.043) NS NS -5.6% -1.2% < 1/2 mile PC (0.057) (0.054) NS NS 6.3% 3.4% PC 1/2 - 1 mile (0.036) (0.042) NS , < 10% *, < 5% **, <1 % *** Note: p-values: > 10% Robustness Tests 5.2. obustness tests. For simplicity, only the DD Table 11 summarizes the results from the r 41 The first two columns show the coefficients are shown and only for the half-mile OLS models. , in Table 9), and the remaining DD results (also presented earlier base OLS and SEM half-mile columns show the results from the robustness models as follows: exclusion of outliers and influential cases from the dataset ( outlier ); using sale year/county in teractions instead of sale ); using only the most recent sales instead sycounty year/state ( of the most recent and prior sales ( recent the reference category, instead of homes ); using homes between 5 and 10 miles as ); and using transactions occurring more than 2 years before between 3 and 10 miles ( outside5 before stead of using transactions simply announcement as the reference category in announcement ( prior ). 41 OLS models for each of the robu stness tests and are available upon one-mile Results were also estimated for the models. Because of the half-mile ffer from what is presented here for the request: the results do not substantively di similarities in the results between the OLS and SEM “base” models, robustness tests on the SEM models were not prepared as we assumed that differences between the two models for the robustness tests would be minimal as well. 34

42 e base model results: none of the coefficients The robustness results have patterns similar to th are statistically different from zero; all coeffi cients (albeit non-significant) are lower in the PAPC period than the PC period; and, all coeffici ents (albeit non-significa nt) are lower (i.e., less 42 mile than outside a half mile. negative/more positive) within a half In sum, regardless of dataset or specification, there is no change in the basi c conclusions drawn from the base model results: there is no evidence that homes near operating or announced wind turbines are impacted t, either the average impacts are ion. Therefore, if effects do exis in a statistically significant fash relatively small (within the margin of error in the models) and/or sporadic (impacting only a small subset of homes). Moreover, these results seem to corroborate what might be predicted given the other, potentially analogous disamenity literature that was re viewed earlier, which might be read to suggest that any property valu e effect of wind turbines might coalesce at a maximum of 3%–4%, on average. Of course, we cannot offer that corroboration directly because, although the size of the coefficients in the models presented here are reasonably consistent with effects of that magnitude, none of our models offe r results that are statis tically different from zero. 42 This trend also continues outside of 1 mile, with those coefficients being less negative/more positive than those within 1 mile. 35

43 Table 11: Robustness Half-Mile Model Results Robus tne s s OLS M ode ls Base Base OLS SEM outlier sycounty recent outside5 prior β fdp tdis (se) (se) β β (se) β (se) β (se) β (se) β (se) NS NS NS NS NS NS NS -4.7% 0.1% -4.2% -8.1% -5.6% -1.7% -2.3% < 1/2 mile PA PC (0.060) (0.065) (0.056) (0.060) (0.066) (0.060) (0.062) NS NS NS NS NS NS NS -2.5% 2.5% 2.3% -1.7% -0.2% -0.8% 0.4% 1/2 - 1 mile PA PC (0.039) (0.043) (0.036) (0.039) (0.043) (0.039) (0.044) NS NS NS NS NS NS NS -5.6% -1.2% -0.3% -1.8% 1.3% -0.5% -4.3% PC < 1/2 mile (0.047) (0.054) (0.056) (0.054) (0.054) (0.057) (0.056) NS NS NS NS NS NS NS 7.1% 4.1% 7.5% 3.8% 6.2% 3.4% 6.3% 1/2 - 1 mile PC (0.042) (0.036) (0.041) (0.036) (0.041) (0.033) (0.036) NS Note: p-values: > 0.1 , < 0.1 *, <0.5 **, <0.01 *** 51,276 38,407 51,276 51,276 51,276 50,106 n 38,407 0.67 0.67 0.67 0.67 adj R-sqr 0.64 0.66 0.66 36

44 6. Conclusion to continue to be developed Wind energy facilities are expected in the United States. Some of ed regions, raising concer ns about the effects of this growth is expected to occur in more-populat wind development on home values in surrounding communities. topic has tended to indicate that wind facilities, Previous published and academic research on this after they have been constructe d, produce little or no effect on home values. At the same time, some evidence has emerged indicating potential home-value effects occurring after a wind facility has been announced but be studies, however, have been fore construction. These previous limited by their relatively small sample sizes, partic ularly in relation to the important population of homes located very close to wind turbines, and have sometim es treated the variable for distance to wind turbines in a problematic fashion. Analogous st udies of other disamenities— including high-voltage transmissi on lines, landfills, and noisy road s—suggest that if reductions in property values near turbines were to occu r, they would likely be no more than 3%–4%, on rger amounts of data are needed average, but to discover such sm all effects near turbines, much la than have been used in previous studies. es have not accounted Moreover, previous studi adequately for potentially confoundi ng home-value factors, such as those affecting home values before wind facilities were announced, nor have they adequately controlled for spatial dependence in the data, i.e., how the values a nd characteristics of ho mes located near one another influence the value of those homes (independent of th e presence of wind turbines). This study helps fill those gaps by collecting a very large data sample and analyzing it with methods that account for confounding factors and spatial dependence. We collected data from more than 50,000 home sales among 27 counties in nine states. These homes were within 10 miles of 67 different then-current ith 1,198 sales that were within 1 or existing wind facilities, w mile of a turbine (331 of which were within a half mile)—many more than were collected by previous research efforts. The data span the periods well before announcement of the wind facilities to well afte r their construction. We use OLS a nd spatial-process difference-in- difference hedonic models to estimate the home- value impacts of the wind facilities; these models control for value factors existing prior to the wind facilities’ announcements, the spatial r time. We also employ a series of robustness dependence of home values, and value changes ove 37

45 models, which provide greater confidence in our re s of data outliers and sults by testing the effect tion across regions, older sales data for multi-sale influential cases, heterogeneous inflation/defla our reference case, and th e amount of time before homes, the distance from turbines for homes in wind-facility announcement for homes in our reference case. Across all model specifications, we find no stat istical evidence that home prices near wind turbines were affected in either the pos t-construction or post-announcement/pre- construction periods . Therefore, if effects do exist, either the average impacts are relatively small (within the margin of error in the models ) and/or sporadic (impacting only a small subset of homes). Related, our sample size and analy tical methods enabled us to bracket the size of effects that would be detected, if those effects were present at al l. Based on our results, we find that it is highly unlikely that the actual average effect for homes that sold in our sample area within 1 mile of an existing turbine is larger th an +/-4.9%. In other word s, the average value of these homes could be as much as 4.9% higher th an it would have been without the presence of wind turbines, as much as 4.9% lower, the same (i.e., zero effect), or anywhere in between. the average actual effect for homes that sold in our sample Similarly, it is highly unlikely that area within a half mile of an ex than +/-9.0%. In othe r words, the average isting turbine is larger value of these homes could be as much as 9% higher than it would have been without the presence of wind turbines, as much as 9% lower, the same (i.e., zero eff ect), or anywhere in between. Regardless of these potential maximum effects, the co re results of our analysis consistently show no sizable statistically significant impact of wind turbines on nearby property values. The gous disamenities (hi maximum impact suggested by potentially analo gh-voltage transmission lines, landfills, roads etc. the models presented in this study ) of 3%-4% is at the far end of what would have been able to discer n, potentially helping to explain why no statistically significant effect was found. If effects of this size are to be discovered in future research, even larger samples of data may be required. For those intere sted in estimating such effects on a more micro (or local) scale, such as appr aisers, these possible data require ments may be especially daunting, though it is also true that the inclusion of additional market, neighborhood, and individual sessments may sometimes improve model fidelity. property characteristics in these more-local as 38

46 7. References operty Assessment Corporation (MPAC), Edward and Gail Kenney v. The Municipal Pr (2012) Ontario Assessment Review Bo ard (ARB). File No: WR 113994. (2013) Superior Court of Justice - Ontario, CA. May 22, Wiggins v. WPD Canada Corporation, 2013. File No: CV-11-1152. American Wind Energy Association (AWEA) ( 2013) Awea U.S. Wind Industry - Fourth Quarter 2012 Market Report - Executive Summary. American Wind Energy Association, Washington, DC. January 30, 2012. 11 pages. Anselin, L. (1988) Spatial Econometrics: Me thods and Models. Springer. 304 pages. 9024737354. Anselin, L. (2002) Under the Hood Issues in th e Specification and Inte rpretation of Spatial Regression Models. Agricultural Economics. 27(3): 247-267. Anselin, L. and Lozano-Gracia, N. (2009) Errors in Variables and Spatial Effects in Hedonic House Price Models of Ambient Air Quality. Spatial Econometrics : 5-34. Arraiz, I., Drukker, D. M., Keleji an, H. H. and Prucha, I. R. (2009) A Spatial Cliff-Ord-Type Journal of : Small and Large Sample Results. Model with Heteroskedastic Innovations Regional Science. 50(2): 592-614. Bateman, I., Day, B. and Lake, I. (2001) The E ffect of Road Traffic on Residential Property Values: A Literature Review and Hedonic Pricing Study. Prepared for Scottish Executive and The Stationary Office, Ed inburgh, Scotland. January, 2001. 207 pages. Baxter, J., Morzaria, R. and Hirsch, R. ( 2013) A Case-Control Study of Support/Opposition to Wind Turbines: The Roles of Health Risk Perception, Economic Benefits, and Community Conflict. Energy Policy. Forthcoming: 40. Bloomberg New Energy Finance (Bloomberg) (2013) Q1 2013 North America Wind Market Outlook. Bloomberg New Energy Finan 25 pages. ce, New York, NY. March 11, 2013. Bond, S. (2008) Attitudes Towards the Deve lopment of Wind Farms in Australia. Journal of Environmental Health Australia. 8(3): 19-32. Bond, S. (2010) Community Perceptions of Wind Farm Development and the Property Value Impacts of Siting Decisions. Pacific Rim Property Research Journal. 16(1): 52-69. Boyle, M. A. and Kiel, K. A. (2001) A Survey of House Price Hedonic Studies of the Impact of 9(2): 117-144. Journal of Real Estate Research. Environmental Externalities. 39

47 Braunholtz, S. and MORI Scotland (2003) Public Attitudes to Windfarms: A Survey of Local Residents in Scotland. Prepared for Scottish Executive, Edinburgh. August 25, 2003. 21 pages. 0-7559 35713. Brown, J., Pender, J., Wiser, R., Lantz, E. and Hoen, B. (2012) Ex Post Analysis of Economic lopment in U.S. Counties. Impacts from Wind Power Deve 34(6): Energy Economics. 1743-1745. Carter, J. (2011) The Effect of Wind Farms on Residential Property Values in Lee County, Illinois. Thesis Prepared for Masters Degr ee. Illinois State University, Normal. Spring 2011. 35 pages. tial Observations in Linear Regression. Technometrics. Cook, R. D. (1977) Detection of Influen 19(1): 15-18. Cook, R. D. and Weisberg, S. ( 1982) Residuals and Influence in Regression. Chapman & Hall. New York. Currie, J., Davis, L., Greenstone, M. and Wa lker, R. (2012) Do H ousing Prices Reflect Environmental Health Risks? Evidence from More Than 1600 Toxic Plant Openings and Closings. Working Paper Series. Prepared for Massachusetts Institute of Technology, Department of Economics, Cambridge, MA. December 21, 2012. Working Paper 12-30. Devine-Wright, P. (2005) Beyond Nimbyism: Towards an Integrated Framework for Understanding Public Perceptions of Wind Energy. Wind Energy. 8(2): 125-139. Donovan, G. H., Champ, P. A. and Butry, D. T. (2007) Wildfire Risk and Housing Prices: A Case Study from Colorado Springs. Land Economics. 83(2): 217-233. Freeman, A. M. (1979) Hedonic Pr ices, Property Values and Measuring Environmental Benefits: A Survey of the Issues. Scandinavian Journal of Economics. 81(2): 154-173. Gipe, P. (1995) Wind Energy Comes of Age. Wi ley Press. New York, NY. 560 pages. ISBN 978-0471109242. Haurin, D. R. and Brasington, D. (1996) School Quality and Real House Prices: Inter-and Intrametropolitan Effects. Journal of Housing Economics. 5(4): 351-368. Heintzelman, M. D. and Tuttle, C. (2011) Values in the Wind: A Hedonic Analysis of Wind Power Facilities. Working Paper : 39. Heintzelman, M. D. and Tuttle, C. (2012) Values in the Wind: A Hedonic Analysis of Wind Power Facilities. Land Economics. August (88): 571-588. Hinman, J. L. (2010) Wind Farm Proximity a nd Property Values: A Pooled Hedonic Regression Analysis of Property Values in Central Illinois. Thesis Prepared for Masters Degree in iversity, Normal. May, 2010. 143 pages. Applied Economics. Illinois State Un 40

48 Hoen, B., Wiser, R., Cappers, P., Thayer, M. a nd Sethi, G. (2009) The Impact of Wind Power in the United States: A Multi-Site Hedonic Projects on Residential Property Values boratory, Berkeley, CA. December, 2009. 146 Analysis. Lawrence Berkeley National La pages. LBNL-2829E. nd Sethi, G. (2011) Wind Energy Facilities and Hoen, B., Wiser, R., Cappers, P., Thayer, M. a Proximity and View on Sales Prices. Journal of Residential Properties: The Effect of Real Estate Research. 33(3): 279-316. Hubbard, H. H. and Shepherd, K. P. (1991) Aeroacoustics of Large Wind Turbines. The Journal of the Acoustical Society of America. 89(6): 2495-2508. Intergovernmental Panel on Climate Change (IPCC) (2011) Special Report on Renewable Energy Sources and Climate Change Mitigation . Cambridge University Press. Cambridge, United Kingdom and New York, NY, US A. 1076 pages. ISBN 978-1-107-02340-6. Jackson, T. O. (2001) The Effects of Environmen tal Contamination on Real Estate: A Literature Review. Journal of Real Estate Research. 9(2): 93-116. Jackson, T. O. (2003) Methods and Technique s for Contaminated Property Valuation. The Appraisal Journal. 71(4): 311-320. Kane, T. J., Riegg, S. K. and Staiger, D. O. (2006) School Quality, Neighborhoods, and Housing 8(2): 183-212. Prices. American Law and Economics Review. Kelejian, H. H. and Prucha, I. R. (1998) A Generalized Spatial Two-Stage Least Squares gressive Model with Autoregressive Procedure for Estimating a Spatial Autore Disturbances. The Journal of Real Estate Finance and Economics. 17(1): 99-121. Kelejian, H. H. and Prucha, I. R. (2010) Speci fication and Estimation of Spatial Autoregressive nd Heteroskedastic Disturbances. Journal of Econometrics. Models with Autoregressive a 157(1): 53-67. Kroll, C. A. and Priestley, T. (1992) The Eff ects of Overhead Transmission Lines on Property Values: A Review and Analysis of the Literatu re. Prepared for Edison Electric Institute, Washington, DC. July, 1992. 99 pages. Kuethe, T. H. (2012) Spatial Fragmentati on and the Value of Residential Housing. Land Economics. 88(1): 16-27. Lantz, E. and Tegen, S. (2009) Economic Develo pment Impacts of Commun ity Wind Projects: A Review and Empirical Evaluation. Prepared for National Renewable Energy Laboratory, Golden, CO. Conference Paper, NREL/CP-500-45555. Laposa, S. P. and Mueller, A. (2010) Wind Farm Announcements and Rural Home Prices: Journal of Sustainable Real Estate. Maxwell Ranch and Rural Northern Colorado. 2(1): 383-402. 41

49 Loomis, D. and Aldeman, M. (2011) Wind Farm Implications for School District Revenue. nter for Renewable Energy,, Normal, IL. July Prepared for Illinois State University's Ce 48 pages. 2011. onomic Impact of Wind Energy Development in Loomis, D., Hayden, J. and Noll, S. (2012) Ec Illinois. Prepared for Illinois State Univer sity's Center for Renewable Energy,, Normal, IL. June 2012. 36 pages. Malpezzi, S. (2003) Hedonic Pricing Models: A Selective and Applied Review. Section in Housing Economics and Public Policy: Essa ys in Honor of Dun can Maclennan. Wiley- Blackwell. Hoboken, NJ. pp. 67-85. ISBN 978-0-632-06461-8. Palmer, J. (1997) Public Acceptance Study of the Searsburg Wind Power Project - One Year Post Construction. Prepared for Vermont E nvironmental Research Associates, Inc., Waterbury Center, VT. December 1997. 58 pages. Ready, R. C. (2010) Do Landfills Always Depress Nearby Property Values? Journal of Real Estate Research. 32(3): 321-339. Rogers, W. H. (2006) A Market for Institutions: Assessing the Impact of Restrictive Covenants on Housing. Land Economics. 82(4): 500-512. Markets: Product Differentiation in Pure Rosen, S. (1974) Hedonic Prices and Implicit Competition. 82(1): 34-55. Journal of Political Economy. Simons, R. A. and Saginor, J. D. (2006) A Me ta-Analysis of the Effect of Environmental Contamination and Positive Amenities on Residential Real Estate Values. Journal of Real Estate Research. 28(1): 71-104. Sims, S. and Dent, P. (2007) Property Stigma : Wind Farms Are Just the Latest Fashion. Journal of Property Investment & Finance. 25(6): 626-651. Sims, S., Dent, P. and Oskrochi, G. R. (2008) Modeling the Impact of Wind Farms on House International Journal of St Prices in the Uk. 12(4): 251- rategic Property Management. 269. E. N. (2005) The Compos ition of Hedonic Pricing Sirmans, G. S., Macpherson, D. A. and Zietz, Models. Journal of Real Estate Literature. 13(1): 3-42. Slattery, M. C., Lantz, E. and Johnson, B. L. (2011) State and Local Economic Impacts from Wind Energy Projects: A Texas Case Study. Energy Policy. 39(12): 7930-7940. Sunak, Y. and Madlener, R. (2012) The Im pact of Wind Farms on Property Values: A Geographically Weighted Hedonic Pricing Mode l. Prepared for Institute for Future Energy Consumer Needs and Behavior (ACN ), RWTH Aachen University. May, 2012 27 pages. FCN Working Paper No. 3/2012. (revised March 2013). 42

50 Tiebout, C. M. (1956) A Pure Theory of Local Expenditures. The Journal of Political Economy. 64(5): 416-424. White, H. (1980) A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity. Econometrica. 48(4): 817-838. ive and Fair Decision-Making Renewables Schemes: Deliberat Wolsink, M. (2007) Planning of achful Accusations of Non-Cooperation. Energy on Landscape Issues Instead of Repro 35(5): 2692-2704. Policy. Zabel, J. E. and Guignet, D. (2012) A Hedonic An alysis of the Impact of Lust Sites on House Prices. Resource and Energy Economics. 34(4): 549-564. 43

51 8. Appendix – Full Results OneMile SEM HalfMile SEM HalfMile OLS OneMile OLS coef s e s e coef s e coef s e Variables coef 11.330* ** (0.058) 11.292* * * (0.090) In tercep t 11.332** * (0.058) 11.292* * * (0.090) 0.002 -0.017 fd p3td is 3_ 11 (0.031) (0.024) (0.016) fd p3td is 3_ 12 -0.015 (0.011) 0.008 -0.035 (0.029) -0.038 (0.033) fd p3td is 3_ 21 (0.017) fd p3td is 3_ 22 -0.001 (0.014) -0.033* -0.033* * * (0.009) (0.008) fd p3td is 3_ 23 -0.006 fd p3td is 3_ 31 0.019 (0.026) -0.022 (0.031) -0.001 (0.018) fd p3td is 3_ 32 0.044* * * (0.014) -0.031* * * (0.012) fd p3td is 3_ 33 -0.005 (0.010) 0.053 fd p3td is 4_ 10 (0.045) 0.001 (0.039) fd p3td is 4_ 11 -0.023 (0.027) -0.018 (0.035) 0.008 (0.016) fd p3td is 4_ 12 -0.015 (0.011) -0.065 (0.056) (0.049) -0.028 fd p3td is 4_ 20 fd p3td is 4_ 21 -0.038 (0.033) -0.027 (0.036) (0.014) -0.034* (0.017) fd p3td is 4_ 22 -0.001 -0.006 (0.008) fd p3td is 4_ 23 -0.033* * * (0.009) (0.041) fd p3td is 4_ 30 -0.036 -0.016 (0.046) -0.016 0.032 fd p3td is 4_ 31 (0.035) (0.031) (0.018) fd p3td is 4_ 32 -0.001 (0.014) 0.044** * fd p3td is 4_ 33 (0.010) -0.031* * * (0.012) -0.005 ls fla1000_ia_ car (0.042) 0.723* ** (0.045) 0.722* ** (0.045) 0.750* * * (0.042) 0.749** * (0.054) 0.900** * (0.054) 0.879* ** (0.060) 0.88* * * (0.060) 0.899* * * ls fla1000_ia_ flo (0.077) ls fla1000_ia_ fra (0.077) 0.932* ** (0.083) 0.934* ** (0.083) 0.980* * * 0.980** * 0.683* * * 0.683** * (0.061) 0.633* ** (0.065) 0.633* ** (0.064) ls fla1000_ia_ s ac (0.061) (0.037) (0.037) 0.382* ** (0.040) 0.38* * * (0.040) 0.442* * * 0.441** * ls fla1000_il_ dek (0.030) 0.641** * (0.030) ls fla1000_il_ liv 0.643* ** (0.046) 0.641* * * 0.643* ** (0.046) 0.512* * * (0.019) 0.512** * (0.019) 0.428* ** (0.029) 0.428* ** (0.029) ls fla1000_il_ mcl 0.800* * * (0.052) 0.800** * ls fla1000_mn _ co t 0.787* ** (0.077) 0.787* ** (0.077) (0.052) ls fla1000_mn _ fre (0.028) 0.595** * (0.028) 0.539* ** (0.031) 0.539* ** (0.031) 0.594* * * 0.587* * * (0.101) (0.101) 0.551* ** (0.102) 0.55* * * (0.102) ls fla1000_mn _ jac 0.587** * (0.025) ls fla1000_mn _ mar 0.603* ** (0.029) 0.603* ** (0.029) 0.643** * 0.643* * * (0.025) 0.421** * 0.389* ** (0.014) 0.389* ** (0.014) (0.012) (0.012) 0.421* * * ls fla1000_n j_ atl (0.044) ls fla1000_ (0. 044) 0.606* ** (0.045) 0.606* ** (0.045) ny_cli 0.635*** 0.635*** 0.373* * * (0.092) 0.375** * (0.092) 0.433* ** (0.094) 0.436* ** (0.094) ls fla1000_ n y _ fra 0.520*** ls fla1000_ 0.520*** (0. 034) 0.559* ** (0.035) 0.559* ** (0.035) ny_her (0.034) 0.556*** 054) 0.556*** (0. ny_lew 0.518* ** (0.057) 0.518* ** (0.057) ls fla1000_ (0.054) ny_mad (0.025) 0.503*** (0. 025) 0.502* ** (0.025) 0.502* ** (0.025) ls fla1000_ 0.503*** ny_ste 0.564*** (0.032) 0.564*** (0. 032) 0.534* ** (0.034) 0.534* ** (0.034) ls fla1000_ ny_wyo ls fla1000_ (0.034) 0.589*** (0. 034) 0.566* ** (0.034) 0.566* ** (0.034) 0.589*** oh_pau 0.625*** 0.624*** (0. 080) 0.567* ** (0.090) 0.565* ** (0.090) ls fla1000_ (0.080) 0.529*** (0. 0.529*** oh_woo 030) 0.487* ** (0.035) 0.487* ** (0.035) ls fla1000_ (0.030) (0.037) 0.838** * (0.037) 0.794* ** (0.046) 0.793* ** (0.046) ls fla1000_o k_cus 0.838* * * 0.750* * * (0.063) 0.750** * (0.063) 0.706* ** (0.072) 0.706* ** (0.072) ls fla1000_o k_g ra ls fla1000_p a_fay 0.332* * * 0.332** * (0.111) 0.335* ** (0.118) 0.334* ** (0.118) (0.111) 0.564* * * 0.564** * (0.025) 0.548* ** (0.031) 0.548* ** (0.031) ls fla1000_p a_s om (0.025) 0.486* * * (0.056) 0.486** * (0.056) 0.44* * * (0.063) ls fla1000_p a_way (0.063) 0.44* * * ls fla1000_wa_kit (0.073) 0.540** * (0.073) 0.540* * * 0.494* ** (0.078) 0.494* ** (0.078) 44

52 HalfMile SEM OneMile SEM HalfMile OLS OneMile OLS Variables coef coef s e coef s e coef s e s e 0.033 (0.030) (0.030) 0.013 (0.032) 0.013 (0.032) acres _ ia_car 0.033 (0.014) 0.050* ** 0.044** * (0.014) 0.044* ** (0.014) acres _ ia_flo 0.050* ** (0.014) -0.008 -0.009 -0.009 (0.022) (0.022) (0.022) acres _ ia_fra -0.008 (0.022) 0.064* ** (0.014) 0.054** * (0.015) 0.054* ** (0.015) 0.064* ** acres _ ia_s ac (0.014) (0.027) 0.064* * (0.027) 0.055* (0.029) 0.048* (0.029) acres _ il_dek 0.068* * 0.023 (0.014) (0.014) 0.014 (0.018) 0.014 (0.018) acres _ il_liv 0.023 (0.010) (0.010) 0.092** * (0.011) 0.092* ** (0.011) 0.091* ** 0.091* ** acres _ il_mcl (0.011) -0.030** * (0.011) -0.024* (0.013) -0.024* acres _ mn_ cot -0.030* ** (0.013) -0.002 (0.007) -0.002 (0.007) 0.002 (0.008) 0.002 (0.008) acres _ mn_ fre 0.019 (0.016) 0.020 (0.016) 0.03* (0.016) 0.03* (0.016) acres _ mn_ jac acres _ mn_ mar 0.020* * 0.020* * (0.008) 0.017* (0.009) 0.017* (0.009) (0.008) -0.041 (0.031) (0.031) -0.013 (0.026) -0.013 (0.026) acres _ nj_ atl -0.041 (0.007) (0.007) 0.022** * (0.007) 0.022* ** (0.007) 0.019* ** 0.019* ** acres _ ny_ cli (0.010) 0.009 (0.010) 0.014 (0.011) acres _ ny_ fra (0.011) 0.009 0.014 -0.004 (0.008) -0.004 (0.008) 0.012 (0.008) 0.012 (0.008) acres _ ny_ her 0.014* (0.008) 0.014* acres _ ny_ lew 0.014 (0.009) 0.014 (0.009) (0.008) acres _ ny_ mad (0.003) 0.021* ** (0.003) 0.021** * (0.004) 0.021* ** (0.004) 0.021* ** 0.009* 0.009* (0.005) 0.007 (0.005) 0.007 (0.005) acres _ ny_ s te (0.005) 0.016* ** (0.004) 0.016* ** acres _ ny_ wy o 0.019** * (0.004) 0.019* ** (0.004) (0.004) acres _ oh_ pau -0.010 (0.020) -0.010 (0.020) 0.01 (0.024) 0.009 (0.024) acres _ oh_ wo o -0.007 (0.010) -0.007 (0.010) 0.002 (0.010) 0.002 (0.010) acres _ ok_cu s (0.019) -0.037* (0.019) -0.034 (0.022) -0.034 (0.022) -0.037* 0.014 (0.010) (0.010) 0.019* (0.011) 0.019* (0.011) acres _ ok_g ra 0.014 (0.023) (0.023) 0.01 (0.023) 0.01 (0.023) -0.006 -0.006 acres _ pa_fay (0.009) 0.004 (0.009) acres _ pa_s o m (0.010) 0.009 (0.010) 0.003 0.009 0.017* * (0.007) 0.017* * (0.007) 0.024** * (0.007) 0.024* ** (0.007) acres _ pa_way 0.009 (0.010) 0.009 (0.010) 0.014 (0.011) 0.014 acres _ wa_kit (0.011) acres lt1_ia_ car (0.136) 0.448* ** (0.136) 0.559** * (0.144) 0.56*** (0.143) 0.446* ** 0.436* ** 0.435* ** (0.112) 0.384** * (0.118) 0.383* ** (0.118) acres lt1_ia_ flo (0.112) 0.670* ** (0.124) 0.668* ** (0.124) 0.684** * (0.139) 0.68*** (0.139) acres lt1_ia_ fra acres lt1_ia_ s ac 0.159 0.160 (0.115) 0.222* (0.123) 0.221* (0.123) (0.115) 0.278* ** (0.066) 0.285* ** (0.066) 0.282** * (0.073) 0.294* ** (0.073) acres lt1_il_d ek acres lt1_il_liv 0.278* ** (0.063) 0.276* ** (0.063) 0.383** * (0.088) 0.38*** (0.088) acres lt1_il_mcl -0.069* ** -0.070** * (0.021) -0.007 (0.032) -0.007 (0.032) (0.021) 0.529* ** 0.529* ** (0.093) 0.466** * (0.120) 0.465* ** (0.120) acres lt1_mn _co t (0.093) 0.314* ** (0.053) 0.314* ** (0.053) 0.294** * (0.061) 0.293* ** (0.061) acres lt1_mn _fre 0.250* (0.144) 0.247* (0.145) 0.169 (0.146) 0.162 (0.146) acres lt1_mn _jac 0.452* ** (0.062) 0.452* ** (0.062) 0.461** * (0.069) 0.462* ** (0.069) acres lt1_mn _mar acres lt1_nj_atl 0.135* ** (0.048) 0.135* ** (0.048) 0.044 (0.047) 0.043 (0.047) acres lt1_ny _cli 0.115* ** 0.115* ** (0.044) 0.108** (0.047) 0.108* * (0.047) (0.044) acres lt1_ny _fra (0.100) 0.118 (0.100) 0.113 (0.115) 0.113 (0.115) 0.118 acres lt1_ny _h er 0.364* ** (0.047) 0.364* ** (0.047) 0.331** * (0.050) 0.332* ** (0.050) (0.067) acres lt1_ny _lew (0.061) 0.120* * (0.061) 0.117* 0.119* 0.117* (0.067) 45

53 HalfMile SEM OneMile SEM HalfMile OLS OneMile OLS coef s e s e coef s e coef s e Variables coef (0.031) acres lt1_ny _mad 0.043 (0.032) 0.043 (0.032) 0.018 0.017 (0.031) 0.100* * 0.100** (0.047) 0.18* ** (0.047) acres lt1_ny _s te (0.042) (0.042) 0.18* ** (0.035) 0.137** * (0.039) 0.144* ** 0.137** * (0.039) acres lt1_ny _wyo 0.144** * (0.035) (0.087) 0.507** * (0.120) 0.507** * (0.120) acres lt1_oh _p au 0.426** * (0.087) 0.425* ** (0.034) 0.114** * (0.041) 0.114** * (0.041) 0.124* ** (0.034) 0.124** * acres lt1_oh _woo 0.104 (0.070) 0.091 (0.092) 0.093 (0.092) acres lt1_ok_ cu s 0.103 (0.070) -0.038 acres lt1_ok_ gra -0.065 (0.066) -0.065 (0.066) -0.038 (0.054) (0.054) 0.403* ** 0.42* * (0.165) 0.42* * (0.164) (0.153) (0.153) acres lt1_pa_ fay 0.403** * 0.243* ** (0.039) 0.223** * (0.047) acres lt1_pa_ s o m 0.243** * (0.039) 0.223** * (0.047) (0.062) 0.138* * (0.062) 0.108 (0.077) 0.109 (0.077) 0.138** acres lt1_pa_ way (0.134) 0.335* * acres lt1_wa_kit 0.342** (0.164) 0.342** (0.164) 0.335** (0.134) -0.013** * (0.001) -0.011* ** (0.001) -0.011** * (0.001) ag e_ ia_car -0.013** * (0.001) -0.013** * (0.002) ag e_ ia_flo -0.013* ** (0.002) -0.013** * (0.002) -0.013** * (0.002) -0.011* ** (0.003) -0.012** * (0.003) ag e_ ia_fra -0.011** * (0.003) -0.012** * (0.003) -0.011** * (0.003) ag e_ ia_s ac -0.011* ** (0.003) -0.013** * (0.003) -0.013** * (0.003) ag e_ il_ dek -0.004** * (0.001) -0.004* ** (0.001) -0.004** * (0.001) -0.004** * (0.001) -0.002 (0.001) -0.003 (0.002) -0.003 (0.002) -0.001 ag e_ il_ liv (0.001) -0.004** * (0.001) -0.006* ** (0.001) -0.006** * (0.001) -0.004** * (0.001) ag e_ il_ mcl -0.021** * (0.003) -0.013* ** (0.005) ag e_ mn_ co t -0.021** * (0.003) -0.013** * (0.005) -0.013** * (0.001) -0.012* ** (0.002) -0.012** * (0.002) ag e_ mn_ fre -0.013** * (0.001) -0.018** * (0.005) -0.018** * (0.005) -0.018** * (0.005) ag e_ mn_ jac -0.018* ** (0.005) ag e_ mn_ mar -0.010** * (0.001) -0.010** * (0.001) -0.009* ** (0.002) -0.009** * (0.002) -0.004** * (0.000) -0.003* ** (0.001) -0.003** * (0.001) ag e_ nj_atl -0.004** * (0.000) -0.005** * (0.001) -0.005** * (0.001) -0.005** * (0.001) ag e_ ny _cli -0.005* ** (0.001) (0.003) -0.005 -0.004 -0.005* (0.003) -0.005* (0.003) ag e_ ny _fra (0.003) -0.008** * (0.001) -0.008** * (0.001) -0.008* ** (0.001) -0.008** * (0.001) ag e_ ny _h er -0.008** * (0.001) ag e_ ny _lew -0.009* ** (0.001) -0.009** * (0.001) -0.008** * (0.001) -0.006** * (0.001) -0.006** * (0.001) -0.006** * (0.001) ag e_ ny _mad -0.006* ** (0.001) -0.006** * (0.001) -0.007* ** (0.001) ag e_ ny _s te -0.006** * (0.001) -0.007** * (0.001) -0.006* ** (0.001) -0.006** * (0.001) -0.006** * (0.001) ag e_ ny _wy o -0.006** * (0.001) 0.003 (0.003) 0.003 (0.004) 0.003 (0.004) ag e_ oh _p au 0.003 (0.003) 0.008* ** (0.001) 0.01* ** (0.001) 0.01* ** (0.001) (0.001) 0.008** * ag e_ oh _wo o (0.002) -0.000 (0.002) 0.002 (0.003) 0.002 (0.003) ag e_ ok_cu s -0.000 -0.000 (0.002) (0.002) 0.001 (0.002) 0.001 (0.002) ag e_ ok_g ra -0.000 (0.004) (0.004) 0.01* * (0.005) 0.01* * (0.005) 0.010** ag e_ pa_fay 0.010* * -0.006** * (0.001) -0.008* ** (0.001) -0.008** * (0.001) ag e_ pa_s o m -0.006** * (0.001) 0.006** * (0.002) (0.002) 0.007** * (0.002) 0.007** * (0.002) ag e_ pa_way 0.006* ** 0.010** * 0.010* ** (0.003) 0.014** * (0.003) 0.014** * (0.003) ag e_ wa_kit (0.003) 0.034** * (0.011) 0.034* ** (0.000) 0.022* (0.012) 0.022* (0.012) ag es q_ ia_car ag es q_ ia_flo 0.040** * 0.040* * (0.016) 0.044** * (0.016) 0.044** * (0.016) (0.016) 0.025 (0.022) (0.022) 0.02 (0.023) 0.021 (0.023) ag es q_ ia_fra 0.025 (0.022) (0.022) 0.025 (0.023) 0.025 (0.023) 0.032 0.032 ag es q_ ia_s ac (0.010) 0.008 (0.010) 0.013 ag es q_ il_d ek 0.013 (0.011) 0.008 (0.012) -0.023** (0.009) -0.023** (0.009) -0.011 (0.014) -0.011 (0.014) ag es q_ il_liv 0.005 (0.007) 0.005 (0.007) ag es q_ il_mcl (0.011) 0.021* (0.011) 0.021* ag es q_ mn_ cot (0.043) 0.109* * (0.043) 0.032 (0.069) 0.033 (0.069) 0.109** 0.046** * 0.045* ** (0.010) 0.044** * (0.012) 0.044** * (0.012) ag es q_ mn_ fre (0.010) 0.103** * (0.035) 0.104* ** (0.035) 0.1** * (0.034) ag es q_ mn_ jac 0.101** * (0.034) ag es q_ mn_ mar (0.012) 0.012 (0.012) 0.006 0.012 0.006 (0.014) (0.014) 46

54 HalfMile SEM OneMile SEM HalfMile OLS OneMile OLS coef s e s e coef s e coef s e Variables coef (0.003) ag es q _nj_ atl 0.003 (0.005) 0.003 (0.005) 0.010** * 0.010** * (0.003) 0.011* (0.006) (0.006) 0.011* (0.006) 0.011* ag es q _ n y _ cli (0.006) 0.011* (0.022) (0.020) -0.002 (0.020) -0.011 -0.002 ag es q _ n y _ fra -0.011 (0.022) (0.005) 0.022* * * (0.006) ag es q _ n y _ h er 0.022* * * (0.005) 0.022* * * 0.022* * * (0.006) 0.031* * * (0.006) 0.032* * * (0.007) 0.032* * * (0.007) (0.006) 0.031* * * ag es q _ n y _ lew 0.017* * * (0.003) 0.023* * * (0.003) 0.023* * * (0.003) ag es q _ n y _ mad 0.017* * * (0.003) 0.013** (0.005) 0.018** * (0.005) 0.018* ** (0.005) ag es q _ny _s te 0.013** (0.005) 0.016* * * 0.017* * * (0.005) 0.017* * * (0.005) (0.005) ag es q _ n y _ wy o 0.016* * * (0.005) -0.045* * (0.022) -0.043 (0.028) -0.043 (0.028) ag es q _ o h _ p au -0.044* * (0.022) -0.091* ** (0.009) -0.074* ** (0.007) ag es q _oh _woo -0.074* ** (0.007) -0.091* ** (0.009) -0.113* ** (0.026) -0.113* ** (0.026) -0.091* ** (0.019) -0.091* ** (0.019) ag es q _ok_ cu s -0.081* ** (0.023) -0.081* ** (0.023) -0.097* ** (0.029) -0.097* ** (0.029) ag es q _ok_ gra -0.112* ** (0.032) -0.105* ** (0.034) -0.106* ** (0.034) ag es q _pa_ fay -0.112* ** (0.032) (0.008) ag es q _pa_ s o m (0.008) 0.016* (0.009) 0.016* (0.009) 0.000 0.002 -0.052* ** (0.012) ag es q _pa_ way -0.053* ** (0.014) -0.000* ** (0.012) -0.053* ** (0.014) -0.132* ** (0.031) -0.000* ** (0.027) ag es q _wa_kit -0.132* ** (0.031) -0.097* ** (0.027) (0.073) -0.082 b ath s im_ ia_ s ac -0.081 (0.077) -0.050 (0.073) -0.050 (0.077) -0.005 baths im_il_dek 0.001 (0.018) 0.001 (0.018) -0.005 (0.015) (0.015) (0.025) 0.090* * * (0.025) 0.087* * * (0.024) 0.087* * * (0.024) 0.090* * * b ath s im_ n y _ cli (0.062) 0.245* * * (0.062) 0.213* * * (0.064) b ath s im_ n y _ fra 0.246* * * 0.212* * * (0.064) 0.099* * * 0.099* * * (0.022) 0.079* * * (0.022) 0.079* * * (0.022) b ath s im_ n y _ h er (0.022) (0.030) (0.030) 0.142* * * (0.031) 0.142* * * (0.031) 0.168* * * b ath s im_ n y _ lew 0.167* * * (0.014) 0.180* * * (0.014) 0.157* * * (0.013) 0.157* * * (0.013) b ath s im_ n y _ mad 0.180* * * (0.019) 0.189* * * (0.019) 0.166* * * (0.020) 0.166* * * (0.020) b ath s im_ n y _ s te 0.189* * * (0.021) 0.107* * * (0.021) 0.1* * * (0.021) 0.1* * * (0.021) b ath s im_ n y _ wy o 0.107* * * 0.095* (0.051) 0.095* (0.051) 0.149* * * (0.057) 0.149* * * (0.057) b ath s im_ o h _ p au 0.094* * * b ath s im_ o h _ wo o 0.094* * * (0.017) 0.092* * * (0.019) 0.092* * * (0.019) (0.017) 0.367* * * (0.077) (0.077) 0.301* * * (0.082) 0.302* * * (0.082) b ath s im_ p a_ fay 0.367* * * (0.036) baths im_p a_ way 0.081** (0.041) 0.081* * (0.041) 0.082** 0.082** (0.036) -2.521* -2.011 (1.936) -2.019 (1.937) (1.467) (1.468) pctvacant_ia_car -2.515* 0.921 (1.152) 1.358 (1.409) p ctv acan t_ ia_ flo (1.410) 0.903 (1.152) 1.339 (3.521) 8.928** (3.518) -2.596 (1.703) -2.6 (1.703) 8.887** pctvacant_ia_fra (0.527) 0.673 (0.527) pctvacant_ia_s ac 1.266* ** (0.377) 0.672 1.267** * (0.377) 0.052 0.062 (0.638) 0.037 (0.964) 0.069 (0.961) pctvacant_il_ dek (0.639) (0.474) (0.474) -0.699 (0.872) -0.701 (0.872) -0.475 -0.476 p ctv acan t_ il_ liv (0.397) -0.366 (0.397) 0.445 pctvacant_il_mcl 0.442 (0.670) -0.365 (0.670) 1.072* (0.592) 1.072* (0.592) 0.272 (1.039) 0.273 (1.039) pctvacant_mn_ co t -1.782* * (0.703) -1.787* * (0.703) -1.372 (0.965) -1.384 p ctv acan t_ mn _ fre (0.965) p ctv acan t_ mn _ jac (0.883) -1.318 (0.884) -1.285 (1.084) -1.313 (1.084) -1.345 (0.502) 2.175* * * 1.53* * (0.622) 1.528* * (0.622) p ctv acan t_ mn _ mar 2.178* * * (0.502) (0.062) (0.062) 0.096 (0.085) 0.095 (0.085) -0.054 -0.054 pctvacant_nj_atl (0.224) 0.709* * * (0.224) p ctv acan t_ n y _ cli 0.841* * * (0.251) 0.709* * * 0.842* * * (0.251) 6.173** * (2.110) 6.104** * (2.113) 0.519 (0.710) 0.499 (0.709) pctvacant_n y_ fra -1.226* ** (0.247) -1.226* ** (0.247) -1.347* ** (0.288) -1.347* ** (0.288) pctvacant_n y_ her -0.125 (0.127) -0.125 pctvacant_n y_ lew -0.266* (0.159) -0.266* (0.159) (0.127) (0.196) (0.196) 0.767* * * (0.246) 0.765* * * (0.246) p ctv acan t_ n y _ mad 0.750* * * 0.752* * * 0.280 (0.190) 0.281 (0.190) 0.039 (0.242) 0.04 (0.242) p ctv acan t_ n y _ s te (0.101) (0.101) p ctv acan t_ n y _ wy o 0.179* 0.225* (0.119) 0.224* (0.119) 0.178* -1.256 -1.473 p ctv acan t_ o h _ p au (1.499) -1.341 (1.951) (1.498) (1.952) -1.473 47

55 HalfMile SEM OneMile SEM HalfMile OLS OneMile OLS coef s e s e coef s e coef s e Variables coef -0.565 pctvacant_oh_woo -0.565 (0.563) -0.306 (0.563) (0.400) (0.400) -0.304 -0.140 -0.127 (0.521) -0.189 (0.521) pctvacant_ok_cus (0.359) (0.358) -0.167 (0.777) (1.045) 0.536 (1.045) 1.414* pctvacant_ok_gra 1.413* (0.777) 0.537 (0.596) 0.232 (0.807) 0.235 (0.807) pctvacant_pa_fay 0.227 (0.596) 0.229 0.562*** (0.138) (0.098) 562** * (0.138) pctvacant_pa_som 0.517*** (0.098) 0.516*** 0. 0.446** (0.156) 0. 446** (0.175) 0.444*** (0.156) pctvacant_pa_way 0.445*** (0.175) -0.075 pctvacant_wa_kit -0.377 (0.282) -0.377 (0.281) -0.076 (0.546) (0.546) (0.244) (0.244) -0.156 (0.324) -0.156 (0.324) pcto wn er_ ia_ car -0.225 -0.225 0.578* * 0.75** * (0.290) 0.75** * (0.290) (0.238) (0.238) pcto wn er_ ia_ flo 0.579* * 0.206 (0.310) 0.172 (0.393) pcto wn er_ ia_ fra (0.393) 0.207 (0.310) 0.169 (0.585) 0.261 (0.586) -0.34 (0.545) -0.345 (0.545) 0.274 pcto wn er_ ia_ s ac (0.088) 0.073 (0.087) 0.032 (0.123) pcto wn er_ il_d ek (0.123) 0.075 0.028 0.176 0.176 (0.140) 0.265 (0.200) 0.264 (0.200) pcto wn er_ il_liv (0.140) (0.051) 0.388* * * 0.331** * (0.101) 0.331** * (0.101) pcto wn er_ il_mcl 0.389* ** (0.051) 0.375* * * 0.609** (0.254) 0.609** (0.254) (0.138) (0.138) pcto wn er_ mn _co t 0.375* ** -0.120 (0.090) -0.072 (0.124) pcto wn er_ mn _fre (0.124) -0.119 (0.090) -0.073 (0.474) -0.205 (0.474) -0.175 (0.569) -0.185 (0.570) -0.206 pcto wn er_ mn _jac (0.076) 0.262* * * (0.076) 0.151 (0.103) 0.151 (0.103) pcto wn er_ mn _mar 0.262* ** (0.037) -0.087** (0.037) -0.036 (0.052) -0.037 (0.052) pcto wn er_ nj_atl -0.087* * (0.171) (0.171) -0.305 (0.199) -0.303 (0.199) -0.229 pcto wn er_ ny _cli -0.229 (1.500) 2.693* (1.505) -0.315 (1.447) -0.398 (1.442) pcto wn er_ ny _fra 2.743* 0.246* ** (0.095) (0.095) 0.213* (0.109) 0.213* (0.109) pcto wn er_ ny _h er 0.246* * * -0.034 -0.034 (0.185) -0.126 (0.219) -0.126 (0.219) pcto wn er_ ny _lew (0.185) 0.750* ** (0.075) 0.750* * * (0.075) 0.723** * (0.084) 0.723** * (0.084) pcto wn er_ ny _mad 0.192 pcto wn er_ ny _s te 0.191 (0.128) -0.083 (0.162) -0.084 (0.162) (0.128) -0.089 (0.111) (0.111) -0.109 (0.138) -0.108 (0.138) pcto wn er_ ny _wyo -0.089 (0.347) -0.185 -1.245** * (0.473) -1.249** * (0.474) pcto wn er_ oh _p au -0.187 (0.348) 0.264* * * 0.274** (0.136) 0.274** (0.136) (0.092) (0.092) pcto wn er_ oh _woo 0.263* ** 0.068 (0.104) -0.041 (0.146) pcto wn er_ ok_ cus (0.146) 0.068 (0.104) -0.043 (0.159) 0.271* (0.159) 0.253 (0.217) 0.253 (0.217) 0.271* pcto wn er_ ok_ gra (1.736) -0.420 (1.736) -0.15 (2.037) -0.165 (2.037) pcto wn er_ pa_ fay -0.413 (0.114) 0.170 (0.114) 0.098 (0.173) 0.098 (0.173) pcto wn er_ pa_ s om 0.171 (0.441) (0.441) -0.251 (0.345) -0.252 (0.345) -0.351 pcto wn er_ pa_ way -0.348 (2.139) 0.259 (2.139) -0.358 (1.889) -0.361 (1.890) pcto wn er_ wa_ kit 0.257 0.002 (0.002) (0.002) 0.003 (0.003) 0.003 (0.003) med _ag e_ ia_ car 0.002 0.003 0.003 (0.002) 0.004 (0.003) 0.004 (0.003) med _ag e_ ia_ flo (0.002) 0.066* ** (0.015) 0.066* * * (0.015) 0.014** (0.006) med _ag e_ ia_ fra (0.006) 0.014** med _ag e_ ia_ s ac (0.014) 0.028* * (0.014) 0.012 (0.010) 0.012 (0.010) 0.028* * -0.001 (0.002) (0.002) -0.001 (0.003) -0.001 (0.003) med _ag e_ il_d ek -0.001 (0.004) (0.004) -0.005 (0.005) -0.005 (0.005) -0.004 med _ag e_ il_liv -0.004 -0.006** * (0.002) -0.006** (0.003) -0.006** (0.003) med _ag e_ il_mcl -0.006* * * (0.002) 0.017* ** (0.005) (0.005) 0.018** (0.008) 0.018** (0.008) med _ag e_ mn _co t 0.017* * * (0.002) 0.012* * * (0.002) 0.013** * (0.002) 0.013** * (0.002) med _ag e_ mn _fre 0.012* ** 0.013 (0.008) 0.013 (0.008) 0.012 med _ag e_ mn _jac 0.012 (0.010) (0.010) med _ag e_ mn _mar (0.003) 0.013* * * (0.003) 0.012** * (0.003) 0.012** * (0.003) 0.013* ** 0.010* ** 0.010* * * (0.001) 0.016** * (0.002) 0.016** * (0.002) med _ag e_ nj_atl (0.001) 0.020* ** (0.004) 0.020* * * (0.004) 0.02** * (0.004) 0.02** * (0.004) med _ag e_ ny _cli -0.517* * * (0.198) -0.511** * (0.198) 0.008 (0.040) 0.01 (0.039) med _ag e_ ny _fra 0.005 0.007* med _ag e_ ny _h er (0.003) 0.005 (0.003) (0.003) (0.003) 0.007* 48

56 HalfMile SEM OneMile SEM HalfMile OLS OneMile OLS coefsecoefsecoefsecoefse Variables (0.005) 0.013*** (0.005) 0.008 (0.005) 0.008 (0.005) med_age_ny_lew 0.013*** (0.002) 0.004** 0.004* (0.002) 0.004* (0.002) med_age_ny_mad 0.004** (0.002) 0.012*** 0.012*** (0.004) 0.001 (0.004) med_age_ny_ste (0.003) (0.003) 0.001 (0.005) (0.006) (0.006) 0.008 0.007 med_age_ny_wyo 0.008 (0.005) 0.008 0.034*** 0.019 (0.012) 0.019 (0.012) (0.013) 0.034*** med_age_oh_pau (0.013) -0.004 (0.003) -0.004 (0.004) med_age_oh_woo (0.004) -0.004 (0.003) -0.004 (0.002) (0.002) 0.008* * (0.004) 0.008* * (0.004) med _ ag e_ o k_ cu s 0.004 0.004 0.011 0 0 (0.006) (0.009) (0.006) med_age_ok_gra 0.011 (0.009) (0.073) (0.073) 0.052 (0.095) 0.052 (0.095) 0.049 med _ ag e_ p a_ fay 0.049 (0.002) 0.008*** (0.002) 0.012*** (0.004) 0.012*** (0.004) med_age_pa_som 0.008*** -0.005 (0.012) (0.012) 0.002 (0.007) 0.002 (0.007) med _ ag e_ p a_ way -0.005 (0.095) med _ ag e_ wa_ kit 0.025 (0.034) 0.025 (0.034) -0.015 -0.015 (0.095) -0.034* * -0.039* * * (0.015) -0.039* * * (0.015) (0.015) (0.015) s win ter_ ia -0.034* * -0.020* * (0.008) -0.013 (0.012) s win ter_ il (0.012) -0.020* * (0.008) -0.013 (0.009) -0.053* * * (0.009) -0.057* * * (0.011) -0.057* * * (0.011) -0.053* * * s win ter_ mn (0.006) -0.007 (0.006) -0.008 s win ter_ n j -0.008 (0.007) -0.007 (0.007) -0.030* * * -0.030* * * (0.007) -0.026* * * (0.007) -0.026* * * (0.007) s win ter_ n y (0.007) (0.012) -0.055* * * (0.014) -0.055* * * (0.014) -0.048* * * -0.048* * * (0.012) s win ter_ o h (0.015) -0.039* * (0.015) -0.024 (0.018) -0.024 s win ter_ o k -0.039* * (0.018) -0.025* (0.015) -0.025* (0.015) -0.02 (0.017) -0.02 (0.017) s win ter_ p a -0.004 (0.046) -0.004 (0.046) 0.014 (0.051) 0.013 (0.051) s win ter_ wa s y _ 1996_ ia -0.436* * * -0.433* * * (0.137) -0.493* * * (0.157) -0.489* * * (0.157) (0.137) -0.267* * * (0.037) -0.344* * * (0.061) -0.344* * * (0.061) s y _ 1996_ il -0.267* * * (0.037) (0.058) s y _ 1996_ mn -0.585* * * (0.065) -0.521* * * (0.059) -0.521* * * -0.585* * * (0.065) -0.820* * * (0.022) -0.820* * * s y _ 1996_ n j -0.717* * * (0.038) (0.022) -0.717* * * (0.038) -0.43* * * -0.298* * * (0.042) -0.43* * * (0.053) (0.042) -0.298* * * s y _ 1996_ o h (0.053) -0.444* * * (0.073) -0.846* * * (0.079) s y _ 1996_ o k -0.444* * * (0.073) -0.846* * * (0.079) (0.060) -0.604* * * (0.067) -0.604* * * (0.067) s y _ 1996_ p a -0.584* * * -0.584* * * (0.060) -0.242* * * (0.036) -0.234* * * (0.052) -0.232* * * (0.052) -0.242* * * s y _ 1997_ il (0.036) (0.055) s y _ 1997_ mn -0.535* * * (0.060) -0.535* * * (0.060) -0.445* * * -0.445* * * (0.055) -0.791* * * -0.791* * * (0.021) -0.686* * * (0.038) -0.686* * * (0.038) s y _ 1997_ n j (0.021) (0.043) -0.39* * * (0.053) -0.39* * * (0.053) -0.302* * * -0.302* * * (0.043) s y _ 1997_ o h (0.057) -0.458* * * (0.057) -0.51* * * (0.066) -0.51* * * (0.066) s y _ 1997_ p a -0.458* * * (0.078) s y _ 1998_ ia -0.633* * * (0.099) -0.634* * * (0.099) -0.442* * * -0.441* * * (0.078) -0.156* * * -0.156* * * (0.031) -0.175* * * (0.048) -0.175* * * (0.048) s y _ 1998_ il (0.031) -0.391* * * (0.054) s y _ 1998_ mn -0.484* * * (0.059) -0.484* * * (0.059) -0.391* * * (0.054) s y _ 1998_ n j (0.020) -0.723* * * (0.021) -0.633* * * (0.037) -0.633* * * (0.037) -0.723* * * -0.217* * * (0.040) -0.302* * * (0.047) -0.302* * * (0.047) s y _ 1998_ o h -0.217* * * (0.040) (0.048) s y _ 1998_ o k -0.818* * * (0.059) -0.395* * * (0.048) -0.394* * * -0.816* * * (0.059) (0.059) -0.554* * * (0.068) -0.552* * * (0.067) -0.481* * * s y _ 1998_ p a -0.480* * * (0.059) (0.115) s y _ 1998_ wa -0.356* * (0.161) -0.356* * (0.161) -0.433* * * -0.433* * * (0.115) -0.347* * * (0.085) -0.345* * * (0.086) -0.568* * * (0.117) -0.565* * * (0.117) s y _ 1999_ ia -0.155* * * (0.031) -0.156* * * (0.031) -0.215* * * (0.046) -0.214* * * (0.046) s y _ 1999_ il s y _ 1999_ mn -0.302* * * -0.303* * * (0.055) -0.367* * * (0.059) -0.368* * * (0.059) (0.055) -0.679* * * (0.020) -0.583* * * (0.036) -0.583* * * (0.036) s y _ 1999_ n j -0.679* * * (0.020) -0.161* * * -0.161* * * (0.040) -0.243* * * (0.047) -0.243* * * (0.047) s y _ 1999_ o h (0.040) -0.347* * * (0.044) -0.348* * * (0.044) -0.743* * * (0.050) s y _ 1999_ o k -0.743* * * (0.050) s y _ 1999_ p a (0.058) -0.452* * * (0.058) -0.515* * * (0.066) -0.515* * * (0.066) -0.452* * * -0.432* * * s y _ 1999_ wa -0.432* * * (0.114) -0.454* * * (0.166) -0.453* * * (0.165) (0.114) 49

57 HalfMile SEM OneMile SEM HalfMile OLS OneMile OLS coef s e s e coef s e coef s e Variables coef (0.145) s y_ 2000_ia -0.246 (0.183) -0.246 (0.183) -0.164 -0.165 (0.146) -0.172** * (0.045) -0.171* ** (0.045) -0.088* ** (0.031) s y_ 2000_il -0.088** * (0.031) -0.224* ** (0.053) s y_ 2000_mn -0.148* ** (0.051) -0.149** * (0.051) -0.224** * (0.053) -0.565* ** (0.020) -0.565** * (0.020) -0.461** * (0.036) -0.462* ** (0.036) s y_ 2000_n j (0.041) s y_ 2000_o h -0.16** * (0.047) -0.098* * (0.041) -0.098** -0.161** * (0.047) -0.748** * (0.059) -0.749* ** (0.059) -0.331** * (0.050) -0.330* ** (0.050) s y_ 2000_o k -0.478** * (0.067) -0.478* ** (0.067) s y_ 2000_p a -0.394* ** (0.057) -0.395** * (0.057) -0.463** * (0.115) -0.403** -0.402* * (0.160) s y_ 2000_wa -0.463* ** (0.115) (0.160) -0.435** * (0.066) -0.332** * (0.065) s y_ 2001_ia -0.334* ** (0.065) -0.433* ** (0.066) -0.080** * (0.031) -0.101** (0.048) -0.101* * (0.048) s y_ 2001_il -0.080* * (0.031) -0.119** (0.050) -0.204** * (0.051) -0.204* ** (0.052) s y_ 2001_mn -0.119* * (0.050) -0.333** * (0.034) -0.333* ** (0.034) -0.438* ** (0.018) s y_ 2001_n j -0.438** * (0.018) (0.036) -0.033 s y_ 2001_o h -0.078** (0.040) -0.078* * (0.040) -0.033 (0.036) -0.250* ** (0.041) -0.648** * (0.044) -0.648* ** (0.044) s y_ 2001_o k -0.251** * (0.041) -0.402** * (0.055) s y_ 2001_p a -0.446** * (0.063) -0.402* ** (0.055) -0.447* ** (0.063) -0.275* -0.275* (0.163) -0.378** * (0.122) (0.163) s y_ 2001_wa -0.378* ** (0.122) -0.128** (0.059) -0.264** * (0.064) -0.261* ** (0.064) s y_ 2002_ia -0.130* * (0.059) 0.007 (0.030) -0.013 (0.043) -0.013 (0.043) (0.030) 0.008 s y_ 2002_il (0.050) -0.072 (0.050) -0.138** * (0.051) s y_ 2002_mn -0.072 -0.139* ** (0.051) -0.330* ** (0.019) -0.195** * (0.035) -0.195* ** (0.035) s y_ 2002_n j -0.330** * (0.019) -0.307** * (0.020) -0.342* ** (0.020) -0.307* ** (0.020) -0.342** * (0.020) s y_ 2002_n y (0.038) -0.022 (0.038) -0.053 (0.042) -0.053 (0.042) s y_ 2002_o h -0.022 -0.249** * (0.045) -0.649** * (0.052) -0.649* ** (0.052) -0.249* ** (0.045) s y_ 2002_o k s y_ 2002_p a -0.313* ** (0.053) -0.313** * (0.053) -0.355** * (0.059) -0.354* ** (0.059) -0.241* * (0.123) -0.241** (0.123) -0.216 (0.166) -0.216 (0.166) s y_ 2002_wa -0.195* * (0.081) -0.194** (0.081) -0.311** * (0.085) -0.314* ** (0.084) s y_ 2003_ia 0.034 (0.030) (0.030) 0.021 (0.040) 0.021 (0.040) s y_ 2003_il 0.034 (0.049) 0.034 -0.026 (0.049) -0.026 (0.049) s y_ 2003_mn 0.034 (0.049) 0.023 0.023 (0.033) -0.119** * (0.017) s y_ 2003_n j (0.033) -0.119* ** (0.017) -0.276** * (0.020) -0.276* ** (0.020) s y_ 2003_n y -0.247* ** (0.020) -0.247** * (0.020) 0.005 (0.036) -0.019 (0.039) -0.019 (0.039) (0.036) s y_ 2003_o h 0.005 -0.229** * (0.046) -0.632** * (0.053) -0.632* ** (0.053) s y_ 2003_o k -0.229* ** (0.046) -0.191** * (0.052) -0.191* ** (0.052) -0.213* ** (0.054) s y_ 2003_p a -0.213** * (0.054) -0.326** * (0.114) s y_ 2003_wa -0.337* * (0.159) -0.335** -0.326* ** (0.114) (0.159) -0.307** * (0.087) -0.308* ** (0.087) -0.209* ** (0.076) s y_ 2004_ia -0.208** * (0.076) (0.029) 0.087*** s y_ 2004_il 0.105** * (0.034) 0.105* ** (0.034) 0.087* ** (0.029) 0.082* (0.049) 0.081* (0.049) 0.036 (0.049) 0.036 (0.049) s y_ 2004_mn -0.179* ** (0.019) -0.179** * (0.019) -0.2* ** (0.020) -0.2** * s y_ 2004_n y (0.020) s y_ 2004_o h (0.037) 0.059 (0.037) 0.067* (0.039) 0.067* (0.039) 0.059 -0.143* ** (0.041) -0.143** * (0.041) -0.511* ** (0.044) s y_ 2004_o k -0.511** * (0.044) -0.146** * (0.052) -0.145* ** (0.053) -0.146* ** (0.052) -0.145** * (0.053) s y_ 2004_p a (0.113) -0.144 (0.113) -0.082 (0.152) s y_ 2004_wa (0.152) -0.144 -0.081 -0.074* * (0.037) -0.075** (0.037) -0.151** * (0.040) -0.151* ** (0.040) s y_ 2005_ia 0.125* ** (0.027) 0.125*** (0.027) 0.139** * (0.032) 0.138* ** (0.032) s y_ 2005_il 0.163* ** 0.162*** s y_ 2005_mn (0.048) 0.12** (0.048) 0.119* * (0.048) (0.048) 0.278* ** 0.278*** (0.018) 0.453** * (0.034) 0.453* ** (0.034) s y_ 2005_n j (0.018) -0.110* ** (0.019) -0.111** * (0.019) -0.122** * (0.019) -0.122* ** (0.019) s y_ 2005_n y 0.112* ** 0.112*** s y_ 2005_o h (0.036) 0.099** * (0.037) 0.098* ** (0.037) (0.036) -0.018 s y_ 2005_o k -0.018 (0.038) -0.354** * (0.038) -0.354* ** (0.038) (0.038) 50

58 OneMile OLS OneMile SEM HalfMile OLS HalfMile SEM Variables coefse coefse coefse coefse s y_ 2005_ p a (0.051) -0.058 (0.053) -0.058 (0.053) -0.060 -0.060 (0.051) -0.070 (0.111) (0.153) 0.025 (0.153) -0.070 s y_ 2005_ wa (0.111) 0.025 (0.028) -0.106* ** (0.028) -0.051* -0.106* * * (0.028) s y_ 2006_ ia -0.050* (0.028) (0.026) 0.215* * * (0.030) 0.215* ** (0.030) s y_ 2006_ il 0.192* * * (0.026) 0.192* * * (0.049) 0.164* * * (0.049) 0.164* ** (0.049) 0.206* * * (0.049) 0.206* * * s y_ 2006_ mn 0.340* * * (0.017) 0.514* * * (0.032) 0.514* ** (0.032) s y_ 2006_ n j 0.340* * * (0.017) -0.073* * * (0.019) -0.073* ** (0.019) s y_ 2006_ n y -0.066* * * (0.019) -0.066* * * (0.019) 0.147* * * 0.147* * * 0.144* ** (0.035) s y_ 2006_ o h (0.034) (0.034) 0.144* * * (0.035) (0.039) 0.026 (0.037) -0.3* ** (0.037) (0.039) 0.025 s y_ 2006_ o k -0.3* ** 0.008 s y_ 2006_ p a -0.001 (0.052) -0.001 (0.052) 0.008 (0.051) (0.051) (0.131) -0.066 (0.131) 0.02 (0.160) 0.021 (0.160) -0.066 s y_ 2006_ wa (0.028) 0.012 (0.028) -0.019 (0.028) -0.019 (0.028) s y_ 2007_ ia 0.013 0.218* * * (0.025) (0.025) 0.251* * * (0.028) 0.251* ** (0.028) s y_ 2007_ il 0.218* * * (0.049) s y_ 2007_ mn 0.145* * * (0.048) 0.144* ** (0.048) 0.177* * * 0.177* * * (0.049) 0.297* * * 0.459* * * (0.031) 0.459* ** (0.031) (0.017) s y_ 2007_ n j 0.297* * * (0.017) -0.020 (0.019) -0.022 (0.019) -0.022 (0.019) s y_ 2007_ n y -0.020 (0.019) 0.143* * * (0.035) 0.138* * * (0.036) 0.138* ** (0.036) s y_ 2007_ o h 0.144* * * (0.035) (0.037) (0.037) -0.154* * * (0.034) -0.154* ** (0.034) 0.149* * * s y_ 2007_ o k 0.150* * * (0.051) 0.030 (0.051) 0.067 (0.052) 0.067 (0.052) s y_ 2007_ p a 0.030 (0.110) s y_ 2007_ wa (0.110) 0.209 (0.147) 0.209 (0.147) 0.189* 0.189* (0.029) (0.029) -0.029 (0.029) -0.029 (0.029) 0.011 s y_ 2008_ ia 0.010 (0.026) 0.218* * * (0.026) 0.217* * * (0.029) 0.217* ** (0.029) s y_ 2008_ il 0.219* * * (0.050) 0.149* * * (0.050) 0.108* * (0.049) 0.108* * (0.049) s y_ 2008_ mn 0.149* * * 0.195* * * 0.195* * * (0.018) 0.35* ** (0.032) 0.35* ** (0.032) s y_ 2008_ n j (0.018) -0.000 (0.019) -0.000 (0.019) -0.008 (0.019) -0.008 (0.019) s y_ 2008_ n y 0.084* * s y_ 2008_ o h (0.036) 0.061* (0.037) 0.061* (0.037) (0.036) 0.084* * (0.039) 0.153* * * -0.145* * * (0.035) -0.145* ** (0.035) s y_ 2008_ o k 0.154* * * (0.039) 0.044 0.055 (0.053) 0.056 (0.053) (0.053) (0.053) s y_ 2008_ p a 0.044 0.179 (0.117) 0.326* * s y_ 2008_ wa 0.325* * (0.148) 0.178 (0.117) (0.148) (0.036) -0.057 (0.036) -0.102* * * (0.036) -0.102* ** (0.036) -0.056 s y_ 2009_ ia (0.026) 0.158* * * (0.026) 0.176* * * (0.028) 0.176* ** (0.028) s y_ 2009_ il 0.158* * * (0.051) 0.104* * (0.051) 0.089* (0.050) 0.089* (0.050) s y_ 2009_ mn 0.104* * (0.019) (0.019) 0.238* * * (0.032) 0.238* ** (0.032) 0.071* * * s y_ 2009_ n j 0.071* * * (0.019) -0.005 (0.019) -0.013 (0.019) -0.013 (0.019) s y_ 2009_ n y -0.005 0.036 (0.035) (0.035) 0.028 (0.036) 0.028 (0.036) s y_ 2009_ o h 0.036 (0.038) 0.219* * * (0.038) -0.102* * * (0.034) -0.101* ** (0.034) s y_ 2009_ o k 0.219* * * 0.009 (0.053) 0.010 (0.053) 0.0003 (0.054) 0.0004 (0.054) s y_ 2009_ p a 0.018 s y_ 2010_ ia (0.029) -0.004 (0.028) -0.004 (0.028) (0.029) 0.017 (0.028) (0.028) 0.104* * * (0.029) 0.104* ** (0.029) 0.105* * * s y_ 2010_ il 0.105* * * (0.050) 0.180* * * (0.050) 0.137* * * (0.049) 0.137* ** (0.049) s y_ 2010_ mn 0.181* * * (0.019) 0.010 (0.019) 0.177* * * (0.032) 0.178* ** (0.032) s y_ 2010_ n j 0.010 (0.021) 0.003 (0.021) -0.006 (0.020) -0.006 (0.020) s y_ 2010_ n y 0.003 -0.017 (0.036) -0.017 (0.036) -0.024 (0.036) -0.024 (0.036) s y_ 2010_ o h 0.231* * * s y_ 2010_ o k (0.038) -0.074* * (0.033) -0.074* * (0.033) (0.038) 0.231* * * 0.013 0.013 (0.057) 0.013 (0.057) 0.013 (0.057) s y_ 2010_ p a (0.057) 0.207 (0.127) 0.207 (0.127) 0.305* (0.165) 0.305* (0.165) s y_ 2010_ wa no te: ** * p <0.01, * * p <0.05, * p <0.1 38,407 38,407 51,276 51,276 N 2 0.64 0.66 0.64 0.66 Adjusted R 51

Related documents

Modern Control Design for Flexible Wind Turbines

Modern Control Design for Flexible Wind Turbines

July 2004 • NREL/TP-500-35816 Modern Control Design for Flexible Wind Turbines A lan D. Wright National Renewable Energy Laboratory 1617 Cole Boulevard, Golden, Colorado 80401-3393 303-275-3000 • www....

More info »
Monographie BIPM 5 Tables Vol3

Monographie BIPM 5 Tables Vol3

Monographie BIPM-5 Table of Radionuclides (Vol. 3 – A = 3 to 244) M.-M. Bé, V. Chisté, C. Dulieu E. Browne, C. Baglin V. Chechev, N. Kuzmenco R. Helmer F.Kondev D. MacMahon K.B. Lee 2006 BUREAU INTERN...

More info »
lbnl section h

lbnl section h

C n tr a c t N o . DE - A C 02 - 0 5 C H 1 1 2 31 o M f i c a t i o n i N o . 1 05 6 od S e c t i o n H PART II SECTION H CONTRACT CLAUSES TABLE OF CONTENTS PAGE NO. - LABORATORY FACILITIE S ... ... H...

More info »
LBNL Electrification of Buildings 2018

LBNL Electrification of Buildings 2018

Electrification of buildings and industry in the United States Drivers, barriers, prospects, and policy approaches : Authors Jeff Deason, Max Wei, Greg Leventis, Sarah Smith and Lisa Schwartz Energy A...

More info »
Small  and Medium Sized Commercial Building Monitoring and Controls Needs: A Scoping Study

Small and Medium Sized Commercial Building Monitoring and Controls Needs: A Scoping Study

PNNL -22169 Prepared for the U.S. Department of Energy Contract DE -AC05- 76RL01830 under and Medium Small- -Sized Commercial Building Monitoring and Controls Needs: A Scoping Study S Katipamula MA Pi...

More info »
Smart2020Report

Smart2020Report

SMART 2020: Enabling the low carbon economy in the information age

More info »
U.S. Department of Energy Commercial Reference Building Models of the National Building Stock

U.S. Department of Energy Commercial Reference Building Models of the National Building Stock

U.S. Department of Energy Commercial Reference Building Models of the National Building Stock Michael Deru, Kristin Field, Daniel Studer, Kyle Benne, Brent Griffith, and Paul Torcellini National Renew...

More info »
The Best Value for America’s Energy Dollar:

The Best Value for America’s Energy Dollar:

The Best Value for America’s Energy Dollar: National Review of Cost of Utility Energy A the Efficiency Programs Maggie Molina March 2014 Report Number U1402 © American Council for an Energy - Efficien...

More info »
lbnl 1003671

lbnl 1003671

LBNL-1003671 E RNEST O RLANDO L AWRENCE ERKELEY N ATIONAL L ABORATORY B Benefits of Leapfrogging to Superefficiency and Low Global Warming Potential Refrigerants in Room Air Conditioning Nihar Shah, M...

More info »
Energy Efficiency in Local Government Operations

Energy Efficiency in Local Government Operations

LOCAL GOVERNMENT CLIMATE AND ENERGY STRATEGY SERIES Energy Efficiency in Local Government Operations A Guide to Developing and Implementing Greenhouse Gas Reduction Programs Energy Efficiency U.S. ENV...

More info »
Primary Frequency Response and Control of Power System Frequency

Primary Frequency Response and Control of Power System Frequency

LBNL- 2001105 Primary Response and Frequency Control of Power System Frequency Author: John Undrill John Undrill, LLC Energy Analysis and Environmental Impacts Division Lawrence Berkeley National Labo...

More info »
A Retrospective Analysis of the Benefits and Impacts of U.S. Renewable Portfolio Standards

A Retrospective Analysis of the Benefits and Impacts of U.S. Renewable Portfolio Standards

A Retrospective Analysis of the Benefits and Impacts of U.S. Renewable Portfolio Standards

More info »
Benchmarking Non Hardware Balance of System (Soft) Costs for U.S. Photovoltaic Systems, Using a Bottom Up Approach and Installer Survey – Second Edition

Benchmarking Non Hardware Balance of System (Soft) Costs for U.S. Photovoltaic Systems, Using a Bottom Up Approach and Installer Survey – Second Edition

Hardware - Benchmarking Non Balance -of -System (Soft) Costs for U.S. Photovoltaic Systems, Using a Bottom -Up Approach – Second and Installer Survey Edition Barry Friedman, Kristen Ardani, David Feld...

More info »
Pennsylvania Climate Action Plan

Pennsylvania Climate Action Plan

Pennsylvania Climate Action Plan Pennsylvania Climate Action Plan , 2018 (DRAFT) November 20 Prepared for: Prepared by: A cknowledgements and Disclaimer 0

More info »
Best Practices Handbook for the Collection and Use of Solar Resource Data for Solar Energy Applications

Best Practices Handbook for the Collection and Use of Solar Resource Data for Solar Energy Applications

Best Practices Handbook for the Collection and Use of Solar Resource Data for Solar Energy Applications 1 1 1 1 2 A. Dobos, S. , Wilbert, A. Habte, S. Kurtz, M. Sengupta 3 6 4 7 5 T. ard, D. Myers, C....

More info »
Renewable power generation costs in 2017

Renewable power generation costs in 2017

Renewable Power Generation Costs in 2017

More info »
2016 Renewable Energy Grid Integration Data Book

2016 Renewable Energy Grid Integration Data Book

2016 Renewable Energy Grid Integration Data Book

More info »
The Costs of Decarbonisation: System Costs with High Shares of Nuclear and Renewables

The Costs of Decarbonisation: System Costs with High Shares of Nuclear and Renewables

The Costs of Decarbonisation The Costs of Decarbonisation: System Costs with High : System Costs with High Shares of Nuclear and Renewables Shares of Nuclear and Renewables

More info »