Economic Value of Sustainable Brownfield Redevelopment

Transcription

1 Economic Value of Sustainable Brownfield Redevelopment Olesya Savchenko* University of Illinois at Urbana-Champaign 1301 W. Gregory Drive, Room 326 Urbana, IL USA John B. Braden University of Illinois at Urbana-Champaign 1301 W. Gregory Drive, Room 326 Urbana, IL USA *Corresponding author: P: ; F:

2 Abstract Though the community economic benefits of conventional modes of brownfield redevelopment have been studied extensively, little is known about sustainable forms that include exceptional measures to conserve resource inputs, energy and water use, and stormwater flows. In the U.S., federal agencies have incentivized sustainable methods of redevelopment, but it is not clear that these methods bestow additional value on the community at large. This study examines the values of nearby properties to determine whether sustainable and conventional modes of redevelopment differ in their neighborhood effects. The main hypothesis tested in this paper is whether sustainable redevelopment of brownfield sites increases nearby property values more than does traditional redevelopment. A sample of fifty-five redeveloped brownfield sites located in New York, NY is classified according to redevelopment type. House-level panel data for sale transactions and characteristics for allows for analysis at a fine geographic scale and high level of precision, controlling unobserved correlated heterogeneity. The results indicate that sustainable and conventional forms of brownfield redevelopment both have positive but statistically different impacts on nearby property values resulting, respectively, in a 8% and 1% appreciation in property values located within 0.5 miles of a brownfield site. The effects continue to be significant and statistically different, but their magnitudes decrease with radial distances up to 2 miles.

3 1. Introduction Redevelopment of abandoned, underused and potentially contaminated properties, called brownfields, is central to the revitalization efforts of many urban areas in the United States. With an estimated half a million brownfields across the country (U.S. Conference of Mayors 2010), redevelopment of such sites presents both challenges and potential opportunities for improving the living and environmental conditions of many American communities. Besides environmental contamination and potential health risks, brownfields are associated with declining property values, higher local unemployment rates, poor neighborhood aesthetics, and the opportunity cost of idle land. Many federal and state programs have been created in an effort to correct market failures that have resulted in insufficient brownfield regeneration. Over the past twenty years these programs have evolved from the enforcement of cleaning up of the most hazardous Superfund sites to the support of sustainable forms of brownfield redevelopment (U.S. Environmental Protection Agency 2010). Though the economic benefits of conventional modes of brownfield redevelopment have been studied extensively, little is known about sustainable forms. This study examines the values of nearby properties to determine whether sustainable and conventional modes of redevelopment differ in their neighborhood effects. The main hypothesis tested in this paper is whether sustainable redevelopment of brownfield sites increases nearby property values more than does traditional redevelopment. Sustainable redevelopment incorporates low-impact features such as efficient building design, recycled materials, water conservation and use of renewable energy that differentiate it from more conventional initiatives, which have no such elements. The results of this study indicate that sustainable and conventional forms of brownfield redevelopment both have positive impacts on nearby property values resulting in a 8% and 1%

4 appreciation in property values located within 0.5 mile of a brownfield site, respectively. The magnitude of the impact decreases for both forms of brownfield redevelopment with distance but remains statistically different within 2 miles of a redeveloped brownfield site. These results are timely and important in light of the strong interest the U.S. Environmental Protection Agency has shown in integrating sustainable practices into brownfield redevelopment. They also provide policy relevant information valuable for urban brownfield redevelopment efforts. Unlike the majority of brownfield studies that focus on large hazardous sites often associated with Superfund cleanups (Noonan et al. 2007; Kiel and Williams 2007, Gamper- Rabindran and Timmins 2013), this paper examines low risk brownfield sites enrolled in the New York (NY) State Brownfield Remediation Program. To our knowledge, this is the first study that identifies a differential impact of brownfield redevelopment type on nearby property values. A sample of fifty-five redeveloped brownfield sites located in New York, NY is classified according redevelopment type. Sustainable sites are those that were redeveloped to achieve US Council on Green Building LEED Certification. House-level panel data for sale transactions and characteristics spanning over twenty years allows for analysis at a fine geographic scale and a high level of precision. In addition, property fixed effects analysis is used to enable the recovery of housing price effects of brownfield redevelopment type, while controlling for time-invariant and time-varying unobserved heterogeneity that can bias results. The vast body of literature on the economic benefits of brownfield remediation ranges from the studies that focus on a single brownfield (Greenberg and Lewis 2000) to analyses that measure the average economic impacts of multiple sites (Greenstone and Gallagher 2008; Zabel and Guignet 2012; Gamper-Rabindran and Timmins 2013). Recent studies mostly focus on the 2

5 benefits of brownfield site remediation (Kiel and Williams 2007; Haninger et al. 2012; Mihaescu et al. 2013; Gamper-Rabindran and Timmins 2013), while earlier research examines price effects of proximity to a brownfield site (Kohlhase 1991; Kiel and Zabel 2001; Hite et al. 2001). In a meta-analysis of hazardous sites, Braden et al. (2011) provide a comprehensive review of these studies. Though with some exceptions (Greenstone and Gallagher 2008), there is a general consensus that brownfield site remediation leads to increases in nearby housing values. However, the estimated magnitude of the appreciation that results from brownfield cleanup varies widely across studies (Braden et al. 2011). To date, there are only a handful of studies that examine the links between sustainability and brownfield redevelopment. Wedding and Crawford-Brown (2007), for example, develop indicators of brownfield redevelopment success in achieving sustainability goals that include green building design as one of their success measures, while Adams and Watkins (2002) look at the importance of brownfield redevelopment in achieving overall sustainability goals. This paper adds to the existing brownfields literature by providing estimates of economic value that sustainable and conventional forms of redevelopment have on nearby property values. Though smaller brownfield sites are much more abundant in the U.S. than the severely contaminated Superfund sites, they have received little attention in the literature most likely due to their lower levels of contamination and lesser availability of government assistance for and information about their remediation. Ihlanfeldt and Taylor (2004) is one of the few exceptions. This paper also focuses on the less risky smaller sites that have been much less studied. If there is evidence of differential market effects by redevelopment form at these less visible sites, it would be reasonable to expect similar differences at the more prominent sites. Furthermore, few economic studies of brownfields look beyond the fact of remediation to 3

6 assess differences in the type of redevelopment. Mastromonaco (2013) is one of the few exceptions; he estimates the impact of the Construction Complete stage in Superfund site remediation process, which indicates construction of physical infrastructure associated with remediation activities. His findings indicate that designation of a site as Construction Complete by Environmental Protection Agency leads to a 2.8% increase in local property values in addition to the 1.6% appreciation resulting from the listing of a site on the Final National Priorities list. The present study goes beyond Mastromonaco s work by examining whether the benefits of redevelopment are different for projects that have sustainable features than for those redeveloped with conventional technologies. In addition to providing efficient use of resources and savings on-sight, such projects may also have positive effects on property values off-site by improving neighborhood amenities, decreasing neighborhood environmental foot print and attracting economic development. The remainder of the paper is organized as follows. Section 2 presents the methodology and empirical estimation strategy, while sections 3 and 4 describe data and report estimation results. Section 5 concludes the paper. 2. Method Economic value of non-market environmental goods, including remediation of contaminated land, can be estimated using a revealed preference approach such as the hedonic method. The hedonic price function recovers the implicit prices of product attributes by modeling the relationship between the observed price of the product (e.g. house) and the unobserved prices of its characteristics. Formalized by Rosen (1974), first-stage hedonic analysis 4

7 has been widely used in the literature to recover marginal willingness to pay (MWTP) for brownfield cleanup. Comprehensive theoretical treatment of the hedonic method can be found in Palmquist (2005), Freeman (1993) and Taylor (2003). Following Taylor (2003) a brief exposition of the theoretical foundations of the hedonic price model is as follows. In a housing market, utility of a consumer is given by:, (1) where is a vector of house characteristics and neighborhood amenities, is brownfield redevelopment type, is a numeraire good representing all other goods, and represents preference parameters. Each consumer maximizes utility function subject to the budget constraint:, (2) where is consumer s income. Utility maximization problem leads to the first-order conditions described in equation (3), where marginal rate of substitution between each house characteristic and the numeraire is equal to the marginal price of the characteristic: Rosen (1974) describes consumer choice with a bid function,, that represents the maximum price a consumer is willing to pay for a house with characteristic X given the (3) 5

8 expenditure on other goods. Using equation (2) to derive the bid function and substituting it into the consumer s utility leads to: (4) Formally, the bid function is given by: (5) and can be substituted into equation (4): (6) Maximizing consumer utility in (6) with respect to X yields: which in combination with equation (3) leads to: (7) Equation (8) implies that in equilibrium marginal price is equal to the marginal MWTP. (8) 2.1 Empirical Estimation The empirical specification of the first-stage hedonic price function relates property values to housing and neighborhood characteristics and to the amenity of interest (Freeman 1993):, (6) 6

9 where the dependent variable is log of sale price of the property in period t. Variables representing the type of brownfield redevelopment are included in, contains a vector property attributes, and is an error term. Using log of sale price as the dependent variable in equation (6) helps minimize heteroskedasticity (Wooldridge 2002), while the semi-log specification is robust in presence of omitted variable bias (Cropper et al. 1988)., where and are defined as count variables that represent the number of brownfields redeveloped in a sustainable or conventional manner a house is exposed to within a specified distance. Exploiting data on geographical location of houses and brownfields, transaction and brownfield redevelopment dates, counts of each brownfield redevelopment type a house is exposed to at the time of sale are constructed. Although proximity to the nearest contaminated site is commonly employed in the literature as a measure of exposure, recent studies rely on count variables to estimate housing market response to Superfund site remediation (Mastromonaco 2013) and drilling of shale gas wells (Muehlenbachs et al. 2012). Defining SUST and CONV as count variables accounts for the possibility that a property may be exposed to more than one redeveloped brownfield. If there is more than one redeveloped site in close proximity to a property, simply measuring the property s distance to the closest site may suffer from omitted variable bias. Additionally, distance to the nearest sites or dummy variables indicating exposure within a certain buffer such as used in Ihlanfeld and Taylor (2004) do not capture the density of sites, hence, are inappropriate for this study as some properties may be exposed to more than one site as evident from Figure 3. Count variables, however, reflect site clustering, while if a house is exposed to only one redeveloped site, they will act like distancebased measures. Additionally, the use of count variables is more appropriate in this study as 7

10 unlike the distance measure that remains constant over time, count variables provide the variation in brownfield site status necessary for the estimation of property level fixed effects (Mastromonaco 2013). Brownfield redevelopment dates are necessary to properly construct variables of interest and accurately estimate the impact of brownfield redevelopment type on nearby property values. Brownfield redevelopment dates (i.e. the exact date of construction completion on a former brownfield site) are not available for all sites. However, the date when Certificate of Completion (COC) that releases owners, developers and lessees from liability and certifies that the site is remediated is available for each brownfield site under analysis. COC is provided to the site once the final engineering report is approved or no further action decision document is issued. COC allows for redevelopment of the site based on applicable restrictions (e.g. residential use, industrial use). It is common for site redevelopment to be completed at the time of COC receipt. Hence, when the exact construction date on a former brownfield is not available, the date of COC issuance will be used to construct the variables of interest. Empirical estimation of the first-stage hedonic price function presents a number of challenges. Housing is a highly heterogeneous good in its structural and neighborhood characteristics. It is difficult to account for this heterogeneity in the model because many property attributes are not observable to the researcher. If any of the omitted variables are correlated with the independent covariates included in the model, the estimated price effect of brownfield redevelopment will be biased. Furthermore, endogeneity arises if one or more of the explanatory variables are codetermined with the dependent variable. For example, housing prices are influenced by the type of brownfield redevelopment, while redevelopment may also be partly determined by property values. 8

11 One alternative to the standard hedonic method is the property fixed-effects method (i.e. repeat sale estimator) (Palmquist 1982). This approach eliminates the need to control for property structural and neighborhood characteristics and demographics and local market conditions because the same properties that sold more than once at different points in time are used in the analysis. Taking advantage of the panel data set, property fixed-effects are estimated at very fine geographic scales controlling house specific unobserved heterogeneity that does not change over time. 2.2 Property Level Fixed Effects To address the issue of omitted variable bias in cross-sectional hedonic models, this study relies on the property level fixed effects method to recover the effect that the brownfield redevelopment type has on nearby property values. The estimated model adds property specific fixed effects to equation (6) as follows:, (7) where all variables are defined as in equation (6) with the exception of which represents a set of year fixed effects and that includes a set of property fixed effects. If in equation (6) unobserved heterogeneity was contained in the error term, equation (7) captures the timeinvariant component of the error that can be correlated with through fixed effects,. Using Monte Carlo simulations Kumibov et al. (2010) showed that spatial fixed effects substantially reduce bias in any function form used in hedonic estimation. Equation (7) will first be estimated using census tract fixed effects. The main specification, however, includes property level fixed effects in equation (7) to allow for the tightest control of time-invariant spatially correlated unobservables at a very fine scale. 9

12 The set of year fixed effects included in the model captures common macroeconomic shocks that influence property prices in each time period. This is particularly important as the data set used in this study includes periods of appreciation and decline in housing prices as evident from Figure 1. Year fixed effects effectively control for any temporal variation in the dependent variable, including the impact of a housing market collapse and other underlying impacts of market conditions on property values. Econometric identification in the property fixed effects method comes from differences in property prices, time-varying attributes and changes in the variable of interest over time, which in this study is brownfield redevelopment status. This model removes the time invariant component of the error term (e.g. local amenities or unobserved property characteristics that do not change over time) that could be correlated with the type of brownfield redevelopment. Though the property fixed effects approach improves upon the hedonic method by controlling for unobserved factors that can confound identification in cross-sectional analysis, its reliance on properties with repeat transactions reduces sample size and raises concerns whether such a sample is representative of the housing market. To alleviate these concerns, this study uses a large sample of repeat sale transactions over an extended period of time. Moreover, as evident from summary statistics in Table 1 and Figures 1 and 2, mean sale prices of properties that sold once have a similar trend as houses that sold multiple times. A number of recent studies use repeat sale data to estimate housing market impacts of Superfund remediation (Mastromonaco 2013), air pollution (Bajari et al. 2012) and proximity to shale gas wells (Muehlenbachs 2012). In addition to the omitted variable bias, it is important to control for reverse causality (Linn 2013). While reverse causality is commonly addressed in hedonic studies estimating the 10

13 effects of foreclosures on property values (Campbell et al.2011), with some exceptions (Linn 2013), the possibility that rising property prices lead to brownfield redevelopment has not been controlled for in the brownfield literature. It is particularly important to address reverse causality when it comes to sustainable brownfield redevelopment that may be induced by improved market conditions. For example, developers may be more likely to construct sustainable buildings on former brownfields in neighborhoods that experience rising property values. Assuming that property values include expectations of future trends (Bajari et al. 2012), changes in the rate of prices before brownfield redevelopment are included in the analysis to control for expectations of future trends and reverse causality. Using the property fixed effects model while also controlling for housing price trends, reverse causality and year fixed effects minimizes possible bias and leads to more precise estimates as compared to studies that use fixed effects at larger geographic scales. Spatial dependence and spatial autocorrelation are the other important issues of concern. Spatial dependence may arise if prices of properties located in close proximity affect each other. To account for spatial autocorrelation (spatial error dependence) error clustering at the property level is applied. The literature brownfield remediation generally reveals localized effects on nearby property values, but the spatial extent of the effects differs across studies. In a study of the effect of listing and delisting of hazardous waste sites on commercial and industrial properties in Atlanta, GA Ihlanfeldt and Taylor (2004) found that precision of the estimates declined for properties located further than 2 miles from a site. Mihaescu et al. (2013) report a reduction in property values within 2,000 feet of a brownfield, while Gamper-Rabindran and Timmins (2013) use 3 km (~1.9 miles) buffers in their analysis of the benefits of Superfund site remediation. 11

14 Similar to Gamper-Rabindran and Timmins (2013). This paper limits the off-site effects of redevelopment to properties located within a 3-mile buffer around a former brownfield site. Housing market impacts of the number of redeveloped brownfield by type is examined at 0.5, 1 and 2 mile buffers. 3. Data This study uses two main data sets: (1) proprietary transactions and characteristics data for residential properties in all boroughs of New York, New York purchased from Dataquick Information Systems, and (2) data on brownfield properties obtained from the publicly available New York State Department of Environmental Conservation (DEC) Environmental Site Database. Rich industrial past has left many brownfield sites throughout the city of New York (NYC). Today NYC is at the forefront of brownfield redevelopment, while New York State offers one of the best-maintained databases of brownfield sites in the country. Both the number of brownfield sites and the availability of information about them make NYC an attractive area to study Property Transactions and Characteristics The dataset obtained from Dataquick Information Systems includes 582,402 sale transactions that took place in NYC between 1989 and Along with property sale price, date and exact address, these data include names of property owners, sellers and buyers, loan amounts and housing characteristics such as square footage, year built, lot size, width and length, number of stories, number of units and number of structures on the lot. However, some of the desirable 12

15 house characteristics such as latitude and longitude coordinates for each property, number of bedrooms and bathrooms are not included in the data. Property characteristics provided in the dataset do not vary over time because Dataquick applies property structural characteristics information recorded at the most recent sale to all previous transactions. This means that in a property level fixed effects model housing characteristics that do not vary over time will drop out of the analysis. To alleviate concerns that properties with significant changes in their structural characteristics that are not observable in the dataset may be correlated with brownfield redevelopment, thus, potentially biasing the results, properties that appreciate more than 50% in one year are removed from the sample (Mastromonaco 2013). A number of recent studies also assume constant property characteristics as they rely on similar data sets that do not report changes in housing structural characteristics over time (Bajari 2012; Haninger et al. 2012; Mastromonaco 2013). Missing, duplicate or invalid observations (e.g. a property that the data report as sold before it was even built) on key variables such as property transaction price, date and address are removed from the dataset. Properties with more than one sale recorded on the same day or sold five or more times in one year are eliminated because they indicate non-arms-length transactions. Observations that report zero structures on the lot are also eliminated. Furthermore, to eliminate outliers sale transactions that fall below the 5 th and above the 95 th percentile are removed from the sample. These necessary cleaning operations reduce the initial data set to 212,778 property transactions. The final sample includes 163,217 repeat sales and 49,830 sales that took place only once during the study period. Geographic coordinates of each property are determined using ArcGIS software and the map obtained from the New York Department of City Planning. Using a unique identification 13

16 number attached to each property a house-level panel data set is created. Table 1 provides sample statistics for properties with single and multiple sale transactions. Mean sale prices for properties that sold more than once are slightly higher than that for homes with only one sale transaction during the study period. Properties with multiple sales also tend to have on average bigger lot size, higher number of stories but are also older, have less square feet, smaller lot width and depth than properties that sold once. Table 1: Summary Statistic for Properties Sold in New York, NY During Period Properties Sold Once Properties Sold Multiple Times Mean Std. Dev. Mean Std. Dev. Sale Price 516, , , , Square Feet Year Built Lot Size Lot Width Lot Depth Number of Stories Number of Units N Observations 51, ,955 Figure 1 depicts mean, seventy-fifth and twenty-fifth percentiles of sale prices by year for the properties that sold more than once during the study period. A period of stable price from 1989 to 1998 is followed by a period of slow appreciation from 1989 to Rapid increase in prices starts in 2001 and lasts till a sharp decline in 2007 followed by the slow recovery in This is consistent with the periods of the housing market bubble that preceded the housing market collapse of A set of year fixed effects used in property fixed effects model accounts for those common to all properties macroeconomic shocks that occurred in different years. 14

17 Price Figure 2 shows mean, seventy-fifth and twenty-fifth percentiles sale prices for properties that sold only once during the study period. They follow a very similar trend of slow and then rapid appreciation, steep decline in 2008 and slow recovery. Figure 1: Sale Prices in New York, NY by Year. Repeat Transactions. 1,000, , , , , , , , , , th percentile 25th percentile mean Year 15

18 Price Figure 2: Sale Prices in New York, NY by Year. Single Transactions 900, , , , , , ,000 75th percentile 25th percentile mean 200, ,000 0 Year 3.2. Sample of Brownfield Sites A sample of fifty-five brownfield sites located in NYC was obtained from the DEC s Environmental Site Remediation Database. This database contains 2,223 brownfields located in New York State that are being addressed under one of the DEC s remedial programs such as the Brownfield Cleanup Program (BCP), the State Superfund or the Environmental Restoration and Voluntary Cleanup programs. The database is updated nightly and provides information about each site s address and geographic coordinates, ownership, history, type of contamination, current use, environmental and health assessment as well as cleanup completion dates. The database is publicly available. Hence, NYC residents have access to information on site remediation. Furthermore, the DEC offers a free subscription based electronic newsletter which provides information on sites that enter the program, are in the process of remediation or have 16

19 completed the required cleanup activities and are ready for reuse. Of the ninety-three sites located in NYC, twenty-eight were cleaned up, at fifty-seven of them redevelopment work was completed, of which 5 were redeveloped sustainably. Location of all sites is presented in Figure 3. Two brownfield sites were dropped from the sample because their redevelopment dates did not fall into the period of study. 17

20 18

21 Redeveloped brownfield sites in the sample are similar in prior use, contamination and redevelopment type. The majority of sites served as gas stations or auto repair shops prior to their redevelopment into mixed-use residential apartment complexes. All sites are enrolled in the BCP program. The program is designed to provide incentives for brownfield cleanup by the private sector. Established under the New York State Environmental Conservation Law, the program encourages voluntary brownfield cleanup and redevelopment. The DEC reviews applications for entry into the program and accepts all eligible sites except those that have hazardous waste or petroleum contamination or are subject to any ongoing federal enforcement action (DEC, 2013). Sites participating in the BCP program receive Certificates of Completion (COC) once remedial activities are completed or will be completed under an approved Work Plan. COC relieves program participants from liability to the Sate for contamination at the sites and makes them eligible for tax credits that relate to the remediation and redevelopment of the site (DEC, 2013). The U.S. Green Building Council (USGBC) LEED certification is used as an indicator of sites that qualify as sustainable. LEED certification is awarded to those buildings that achieve excellence in environmental design. The program, well-known nationally and internationally known, is based on a point system. A building can receive credits based on its performance in the following categories: 1) site sustainability, which promotes strategies to decrease the impact on water resources and ecosystems, 2) water efficiency, 3) energy & atmosphere credits that encourage efficient energy use through innovative design, 4) materials & resources, which promotes waste reduction and use of sustainable building materials, and 5) indoor environmental quality, which also includes access to daylight and views (USGBC 2013). 19

22 The USGBS maintains a publicly available database of LEED certified buildings. The list of all LEED certified buildings in NYC was matched, by address, to the NYC list of fifty-five brownfield sites to determine which redevelopment projects received certification. This process helped to identify the brownfield sites, out of the total fifty-five redeveloped sites, five were classified as sustainably redeveloped. The buildings constructed on these five sites received LEED Gold and LEED Silver certifications. In addition to using USGBC LEED certification as a proxy for sustainable brownfield redevelopment, other programs and brownfield awards that distinguish superior environmental design have been explored. The programs include Environmental Protection Agency Energy Star certified buildings, New York City Brownfield Partnership (NYCBP) Big Apple Brownfield Awards and EPA Phoenix Awards. EPA Energy Star certification is provided to buildings that are energy-efficient. Starting in 2009 NYCBP distinguishes excellence in NYC brownfield redevelopment in various categories including green building design, while the EPA Phoenix Awards recognize outstanding redevelopment projects in each of the ten EPA regions. Using information on location, name and nature of redevelopment, each brownfield site in the sample was compared to the sites that were distinguished through one of these programs. Although only five redeveloped brownfield sites that are identified as sustainable, a sufficient number of sale transactions around these sites facilitate accurate estimation of their relative effect on property prices. The final dataset is formed by joining geocoded transactions and characteristics data for properties that sold more than once in NYC during period with the sample of brownfields classified by redevelopment type. These detailed panel data cover a long period of 20

23 time, which allows for better control of time trends and accurate property level fixed effects analysis that relies on properties that sold before and after brownfield redevelopment. 4. Estimation Results and Discussion Table 1 and Table 2 describe estimation results of two different specifications of equation (7). In both specifications variables of interest, SUST and CONV, represent, respectively, the number of sustainably and conventionally redeveloped brownfield sites that each property is exposed to within 0.5, 1, and 2 mile buffers. This assumes that brownfield redevelopment has no impact beyond the 2 mile buffer. Several prior studies have also used similar cut-off distance (Gamper- Rabindran and Timmins 2013; Mastromonaco 2013). Since there assumed to be no price effects of brownfield redevelopment type on property values beyond the 2 cut-off point, the analysis is limited to properties within a 2 mile buffer of a brownfield site. The impact of sustainable and conventional brownfield redevelopment is first estimated using census tract fixed effects. This specification includes property characteristics such as square feet, lot size, lot width, lot depth, year built and number of stories. Although this specification includes year dummies to control for any common temporal variation that can influence housing prices, it does not control for reverse causality. Results of this specification, described in Table 1, suggest that the estimated impact of the conventional brownfield redevelopment leads to a 3.2% increase in property values located within 0.5 miles of the site. The magnitude of the impact decreases with distance to 2% within 1 mile and 1.8% within 2 miles. SUST variable that represents the density of the sites redeveloped in a sustainable manner is not statistically significant at any distance. Most property characteristics are statistically significant and have the expected sign, although the magnitude of their coefficients is very small. 21

24 Lot Size and Lot Width, however, have an unexpected negative coefficient. Insignificance of the sustainable redevelopment and negative signs on Lot Size and Lot Width may result from lack of sufficient housing and neighborhood characteristics included in this specification. Moreover, reverse causality is not controlled for, while fixed effects at the census level do not account for any property specific time-invariant correlated heterogeneity. Table 1: Estimated Impacts of Sustainable and Conventional Forms Brownfield Redevelopment. Census Tract Fixed Effects Dependent variable: log sale price (1) (2) (3) SUST 0.5m (0.0541) CONV 0.5m *** (0.0103) SUST 1m (0.0393) CONV 1m *** ( ) SUST 2m (0.0183) CONV 2m *** ( ) Sqft *** *** *** ( ) ( ) ( ) Lot Size *** *** *** (9.58e-08) (9.57e-08) (9.58e-08) Year Built *** *** *** ( ) ( ) ( ) Lot Depth ** ** *** ( ) ( ) ( ) Lot Width *** *** *** 22

25 ( ) ( ) ( ) N of Stories *** *** *** ( ) ( ) ( ) Year FE Y Y Y FE Level Property Property Property _cons *** *** *** (0.145) (0.145) (0.145) N 68,953 68,953 68,953 R Notes:. * p < 0.1, ** p < 0.05, *** p < 0.0. Standard errors in parentheses clustered at property level. Results of the property level fixed effects specification of equation (7) are presented in Table 2. Fixed effects at the property level allow for control of the time-invariant unobservables that can be correlated with brownfield redevelopment type at the finest geographical scale. This leads to a more precise estimation of the impact that brownfield redevelopment form has on nearby property values. In addition to controlling for unobservables that do not change over time, this specification also includes rates of price change prior to brownfield redevelopment. This accounts for endogeneity associated with reverse causality. Furthermore, the model includes year fixed effects to account for any common time-varying shocks that could impact property values. Reliance on property level fixed effects eliminates the need to drop all observations reporting zero in housing characteristics. However, this is operation is necessary in estimating census tract fixed effect specification. This explains the differences in the number of observations reported in each table. The estimated impacts of sustainable and conventional brownfield redevelopment are statistically significant at each of the three distance specifications. Sustainable site redevelopment results in appreciation of 8%, 2.5% and 1.4%, respectively, of properties located 23

26 within 0.5, 1, and 2 mile buffers of a redeveloped site. The magnitude of the impact decreases with distance. This is consistent with studies that show that the estimated price effects of brownfield remediation, particularly Superfund sites, decreases quickly with distance (Kohlhase 1991; Kiel 1995; Ihlanfeldt and Talylor 2004).. The coefficients on CONV variable are also statistically significant at all distance specifications. Conventional brownfield redevelopment leads to a 1% appreciation in properties located 0.5 and 1 mile away from the site. The magnitude of this impact decreases to 0.4% as the distance between the site and nearby properties increases to 2 miles. The preferred buffer size is 0.5 miles away from a redeveloped brownfield site. This specification demonstrates the greatest impact on nearby property values. Furthermore, at this distance the impact of sustainable and conventional forms of brownfield redevelopment are statistically different as indicated by a p-value of of a t-test for coefficient equality. Although both redevelopment types lead to positive price effects on the nearby property values, the price effects they have on nearby are statistically different at 0.5 and 2 miles. Compared with the impact of sustainable and conventional brownfield redevelopment estimated using census tract fixed effects, property level fixed effects model produces smaller and more precise estimates. This result is not unexpected as coefficients in census tract fixed effects specification may be biased downward due to correlation between brownfield redevelopment type and neighborhood amenities that the model is not able to adequately control for. The estimates obtained from the property fixed effects specification are smaller than the price effects of Superfund site cleanups reported in the literature (Zabel and Guignet 2012; Gamper-Rabindran and Timmins 2013). This smaller magnitude can be explained by a few of 24

27 factors. First, brownfields under the analysis are smaller, less contaminated sites that present low risk to human health as compared to Superfund sites that pose imminent danger to human health. Second, property fixed effects model provides very tight controls for omitted variable bias and endogeneity that can be associated with the type of brownfield redevelopment, whereas many Superfund studies are not able to account for unobserved heterogeneity at the individual property level (Gamper-Rabindran and Timmins 2013). This reasoning is further supported by the evidence provided from the recent study by Mastromonaco (2013) that suggests that positive effects of Superfund cleanup decreases with the severity of contamination. 25

28 Table 2. Estimated Impacts of Sustainable and Conventional Forms Brownfield Redevelopment. Property Level Fixed Effects Dependent variable: log sale price (1) (2) (3) SUST 0.5m *** (0.0121) CONV 0.5m *** ( ) SUST 1m *** ( ) CONV 1m *** ( ) SUST 2m *** ( ) CONV 2m *** Controls ( ) Rate Change Y Y Y Year FE Y Y Y FE Level Property Property Property Cons *** *** *** (0.0426) (0.0427) (0.0427) N 114, , ,568 T-test p-value Notes:. * p < 0.1, ** p < 0.05, *** p < 0.0. Standard errors in parentheses clustered at property level. 5. Conclusion This paper investigates whether sustainable and conventional forms of brownfield redevelopment have different price effects on nearby property values. This study makes a 26

29 valuable contribution to the extensive brownfields literature that has paid little attention to the price effects that brownfield redevelopment forms have on the housing market. Unlike the vast majority of existing literature that examines the effects of cleanup of Superfund sites, this paper focuses on brownfields with low levels of contamination that are less studied in the literature. This study effectively controls for various sources of bias present in the hedonic analysis by using property fixed effects model and exploiting detailed panel data set of sale transactions. Results indicate that sustainable and conventional forms of brownfield redevelopment both have positive but statistically different impacts on nearby property values resulting, respectively, in a 8% and 1% appreciation in property values located within 0.5 miles of a brownfield site. The effects continue to be significant and statistically different, but their magnitudes decrease with radial distances up to 2 miles. The impact of sustainable forms of brownfield redevelopment on nearby property values is greater than that of conventional redevelopment at all distance specifications tested in the study These findings are timely and important as brownfield redevelopment projects that include sustainable elements such as green building design, efficient energy and water use, recycled materials, and green space have gained momentum over the past decade. As the number of such redevelopment projects grows throughout the country, it is becoming crucial, from a policy perspective, to understand the differences between economic benefits derived from sustainable brownfield redevelopment and those from more conventional efforts. 27

30 References Abbott, J. K., & Klaiber, H. A. (2011). An embarrassment of riches: Confronting omitted variable bias and multiscale capitalization in hedonic price models. Review of Economics & Statistics, 93(4), Adams, D. & Watkins, G. (2002). Greenfields, Brownfields and Housing Development. Wiley- Blackwell Bartsch, C.,& Anderson, C. (1998). Matrix of Brownfields Programs by State. Northeast- Midwest Institute, Washington, DC / Bajari, P., Fruehwirth, J. C., Kim, K. i., & Timmins, C. (2012). A rational expectations approach to hedonic price regressions with time-varying unobserved product attributes: The price of pollution. American Economic Review, 102(5), Braden, J. B., Feng, X., & Won, D. (2011). Waste sites and property values: A meta-analysis. Environmental and Resource Economics, 50(2), Cameron, T. A. (2006). Directional heterogeneity in distance profiles in hedonic property value models. Journal of Environmental Economics and Management, 51, Campbell, J. Y., Giglio, S., & Pathak, P. (2011). Forced sales and house prices. American Economic Review, 101(5), Cropper M., L. Deck, and K.E. McConnell. (1988). On the Choice of Functional Form for Hedonic Price Functions. Review of Economics and Statistics 70: Department of Environemntal Concervation of New York State (2013). Environmental Cleanup & Brownfields. Available at accessed March 3, 2013). De Sousa, C. A. (2003). Turning brownfields into green space in the city of toronto. Landscape and Urban Planning, 62(4),

31 Dorsey, J.W. (2003). Brownfields and greenfields: the intersection of sustainable development and environmental stewardship. Environmental Practice, 5 (1), Farber, S. (1998). Undesirable facilities and property values: a summary of empirical studies. Ecological Economics, 24:1{14. Freeman, A. M. III, (1993), The Measurement of Environmental and Resource Values (Resources for the Future, Washington DC). Gamper-Rabindran, S., & Timmins, C. (2013). Does cleanup of hazardous waste sites raise housing values? evidence of spatially localized benefits. Journal of Environmental Economics & Management, 65(3), Greenberg, M., & Lewis, M. J. (2000). Brownfields redevelopment, preferences and public involvement: A case study of an ethnically mixed neighbourhood. Urban Studies (Routledge), 37(13), Greenstone, M., & Gallagher, J. (2008). Does hazardous waste matter? evidence from the housing market and the superfund program. Quarterly Journal of Economics, 123(3), Gayer, T., Hamilton, J. T., & Viscusi, W. K. (2002). The Market Value of Reducing Cancer Risk: Hedonic Housing Prices with Changing Information. Southern Economic Journal, 69(2), Guignet, D. & Alberini, A. (2010). Voluntary Cleanup Programs and Redevelopment Potential: Lessons from Baltimore, Maryland. Journal of Policy Development and Research, 12(3), Haninger, K., Ma, L., & Timmins, C. (2012). Estimating the Impacts of Brownfield Remediation on Housing Property Values. Duke Environmental Economics Working Paper Series. Hite D, ChernW, Hitzhusen F, Randall A (2001) Property-value impacts of an environmental disamenity: the case of landfills. J Real Estate Financ Econ 22(2/3):

32 Ihlanfeldt and Talylor (2004). Externality Effects of Small-Scale Hazardous Waste Sites: Evidence from Urban Commercial Property Markets. Journal of Environmental Economics and Management, 47, Kohlhase, J.E. (1991). The Impact of Toxic Waste Sites on Housing Values. Journal of Urban Economics. 30: Kiel, K. A. (1995). Measuring the Impact of the Discovery and Cleaning of Identified Hazardous Waste Sites on Home Values. Land Economics. 74: Kuethe, T, Foster, K.A., & Florax R.J. (2008). A Spatial Hedonic Model with Time-Varying Parameters: A New Method Using Flexible Least Squares. Selected Paper prepared for presentation at the American Agricultural Economics Association Annual Meeting, Orlando, Florida, July 27-29,2008. Kuminoff, N. V., Parmeter, C. F., & Pope, J. C. (2010). Which hedonic models can we trust to recover the marginal willingness to pay for environmental amenities? Journal of Environmental Economics & Management, 60(3), Kuminov, N., C. F. Parmeter and J. C. Pope, (2009), Specification of Hedonic Price Functions: Guidance for Cross-sectional and Panel Data Applications," AAEC Working Paper , Virginia Tech. Linn, J. (2013). The effect of voluntary brownfields programs on nearby property values: Evidence from illinois. Journal of Urban Economics, 78(0), Mastromonaco, R. (2013). Hazardous Waste Hits Hollywood: Superfund and Housing Prices in Los Angeles. Environmental and Resource Economics. McCluskey, J. J., & Rausser, G. C. (2003). Hazardous Waste Sites and Housing Appreciation Rates. Journal of Environmental Economics and Management, 45(1),

33 Mihaescu, O., & Vom Hofe,,Rainer. (2013). Using spatial regression to estimate property tax discounts from proximity to brownfields: A tool for local policy-making. Journal of Environmental Assessment Policy & Management, 15(1), -1. Muehlenbachs, L., E. Spiller, and C. Timmins (2012). Shale gas development and property values: Differences across drinking water sources. Working Paper 18390, National Bureau of Economic Research. Nijkamp, P., Rodenburg, C. A., & Wagtendonk, A. J. (2002). Success factors for sustainable urban brownfield development: A comparative case study approach to polluted sites. Ecological Economics, 40(2), 235. Noonan DS, Krupka DJ, Baden BM (2007) Neighborhood dynamics and price effects of Superfund site clean-up. Journal of Regional Science 47(4): Palmquist, R. B., (2005), Property Value Models, in Handbook of Environmental Economics, edited by Karl-Gran Mler and Jeffrey R. Vincent. Amsterdam: North-Holland. Palmquist, R. B. (1982). Measuring environmental effects on property values without hedonic regressions. Journal of Urban Economics, 11(3), 333. Revitalizing contaminated sites--liability protections for owners and prospective purchasers. (2008). Hazardous Waste Consultant, 26(5), Rosen, S. (1974). Hedonic prices and implicit markets: Product differentiation in pure competition. Journal of Political Economy, 82(1), 34. Taylor, L., (2003), The Hedonic Method," in A Primer on the Economic Valuation of the Environment, eds. P. Champ, T. Brown, and K. Boyle, , Kluwer. Sigman, H. & Stafford,S. (2011).Management of Hazardous Waste and Contaminated Land. Annual Review of Resource Economics, 3(1), U.S. Environmental Protection Agency (2009). Challenges in Applying Property Value Studies to Assess the Benefits of the Superfund Program. Office of Superfund Remediation and Technology Innovation. 31

34 U.S. Conference of Mayors. (2003). Recycling America's Land: A National Report on Brownfields Redevelopment (Volume 4). Washington, DC. U.S. Environmental Protection Agency (2010). Partnership for Sustainable Communities Brownfield Pilots, EPA 560-F Available at accessed January 15, Wedding, C& Crawford-Brown, D. (2007). Measuring site-level success in brownfield redevelopments: a focus on sustainability and green building. Journal of Environmental Management, 85(2) Wooldridge, J. (2002). Econometric Analysis in Cross Section and Panel Data. MIT Press. Zabel, J. E., & Guignet, D. (2012). A hedonic analysis of the impact of LUST sites on house prices. Resource & Energy Economics, 34(4),