Abstract
This paper contributes to the debate over public benefits and costs of state-funded voluntary cleanup programs, using evidence from property values in New York City. We value site redevelopment separately from cleanup and examine time to capitalization. Using property fixed effects and controlling for time-varying shocks, New York’s Brownfield Cleanup Program added 4% to property values. Off-site gains averaged 5.6% for properties with three units or less and 1.2% for multifamily residences, producing a $579.3 million tax gain that does not exceed the $667.9 million in program spending. Benefits stem from program participation and cleanup, but not from site redevelopment. (JEL Q51, Q58)
1. Introduction
Policy makers rely on state voluntary cleanup programs (VCPs) to encourage revitalization of contaminated land in the United States. Established in every state, these programs offer economic incentives and liability relief in return for voluntary cleanup and redevelopment of contaminated properties. An important justification for VCPs is the off-site benefits generated by site cleanup, in addition to the on-site gains, that are captured by the community at large and other neighborhood properties (Wernstedt et al. 2013; McCarthy 2002, 2009). Therefore, although most of the contaminated sites are privately owned, public money is used to assist property owners in bringing contaminated land back to productive use. The U.S. Environmental Protection Agency (EPA) (USEPA 2004) has estimated that cleanup costs for 355,000 properties in the United States will reach $250 billion by 2033, with more than half of these contaminated sites cleaned up under state VCPs. Despite these substantial costs and the central role of VCPs in U.S. land revitalization policy, the off-site benefits induced by these programs are not well understood.
In this paper, we measure the public economic benefits of New York’s Brownfield1 Cleanup Program (BCP), as they are capitalized into local residential property values in New York City, and compare those benefits to program costs. New York’s BCP is among the largest VCPs in the country and, unlike most other state programs, offers incentive structures that differentiate between cleanup and redevelopment. This study advances the literature on benefits of contaminated site cleanup in several ways. First, to the best of our knowledge, this paper provides the first evaluation of public costs and benefits of a state VCP. Because the cost of site cleanup through state VCPs is funded in part by the public sector, it is important to understand public cost-benefit trade-offs of such programs. Since its inception, New York’s BCP has provided over $1 billion to incentivize site cleanup and redevelopment. Furthermore, the external benefits induced by state VCPs are poorly understood, with only a handful of prior empirical studies that provide conflicting evidence of the neighborhood effects (Lang and Cavanagh 2018; Linn 2013; Alberini 2007). Second, our analysis distinguishes the external effects of site cleanup from those of site redevelopment on property prices; to date, with a few exceptions (Mastromonaco 2014), the majority of the brownfields literature has focused on the value of site cleanup alone. Given that some state VCPs promote redevelopment in addition to cleanup, understanding whether these efforts produce greater off-site benefits is useful for policy makers in prioritizing state funding for these activities. Finally, we also examine how quickly the off-site benefits of cleanup affect the prices of nearby properties and the heterogeneity of these effects across different housing markets.
The vast majority of studies that seek to measure the economic benefits of contaminated land cleanup have focused on highly hazardous sites addressed under the federal Superfund program. Braden, Feng, and Won (2011) and Sigman and Stafford (2011) provide extensive reviews of this literature. Among the few exceptions is the study by Haninger, Ma, and Timmins (2017) that measures the value of brownfield site cleanup under the EPA Brownfields Program. The authors find property value appreciation due to cleanup ranges from 5.0% to 11.5% when exploiting variation in panel data and as high as 15.2% when using a double-difference matching estimator that does not rely on temporal variation. Using these property value increases due to site cleanup under the EPA Brownfield Program estimated by Haninger, Ma, and Timmins (2017), Sullivan (2017) calculates residential property tax revenue resulting from cleanup of 48 sites under the EPA Brownfield Program for one single tax year to be between $29 and $97 million (2014 dollars). These tax gains exceed the $12.4 million that the EPA invested in these sites.
Among the studies that focus specifically on the external neighborhood effects of VCPs, Linn (2013) finds that the Illinois Site Remediation Program leads to a 1% appreciation in property values 0.25 mile away from a brownfield enrolled in the program. Lang and Cavanagh (2018), however, show that brownfield remediation through Rhode Island’s VCP leads to a decline in nearby residential property values due to incomplete information about the presence of risk in the market. The authors further find that brownfield cleanup leads to decreases in property prices in low-value neighborhoods and to increases in property prices in high-value neighborhoods. This demonstrates considerable heterogeneity in availability of information about contamination prior to cleanup across different market segments. Alberini (2007) shows site participation in Colorado’s VCP has no significant effect on property values.
Unlike these previous studies on state VCPs, our analysis exploits the panel structure of sale transaction data, which allows tighter controls for time-invariant correlated unobservables at the house level. Further, we include additional controls for potential time-varying correlated heterogeneity at the neighborhood level. Brownfield cleanup and redevelopment is not randomly sited, and property or neighborhood characteristics may be correlated with both house prices and proximity to cleaned and redeveloped sites. To minimize the potential for omitted variable bias from correlated time-invariant and time-varying unobservables, we use a property-level fixed effects model and control for neighborhood-level unobservables that vary over time and space.
Our analysis employs 123,852 residential property repeat sale transactions between 1996 and 2011 in New York City. We choose to focus the analysis on New York City; more than half of the BCP’s spending has been devoted to projects located there.2 Fine geographic scale and high temporal frequency of New York City real estate transaction data enable us to measure the effects of site cleanup and site redevelopment with high levels of precision over space and time. A sample of 185 contaminated sites located in New York City and addressed through the BCP are classified as cleaned up and redeveloped. Those sites that were redeveloped are further characterized by redevelopment type, such as apartment buildings, offices, schools, and retail. After identifying residential property price effects of cleanup and redevelopment supported by the BCP, we compare the aggregate value of these effects in the form of property tax gains to the associated public investments. Unlike prior studies that rely on estimated average site cleanup costs in their cost-benefit analyses (Haninger, Ma, and Timmins 2017), we use the actual amounts of public spending received by each BCP site in the form of cleanup and redevelopment tax credits, which we obtained from the New York State Department of Taxation and Finance. Finally, we supplement our analysis with Google Maps Street View data to verify the redevelopment status of each site.
During the period of our analysis, we find that the BCP had substantial positive external neighborhood effects. Site entry into the BCP and subsequent cleanup through the program on average led to a 4% appreciation in residential property values 0.25 mile away from a site. These off-site gains averaged 5.6% for residential properties with three units or less, and 1.2% for multifamily residencies with more than three units. We also find heterogeneous effects of BCP site entry and cleanup across the five boroughs of New York City. Cleanup had a significant effect on nearby property values in the Bronx, the poorest borough of New York City, while site entry into the program captured most of the effect of the program in the rest of the boroughs. In total, BCP site entry, cleanup, and redevelopment led to a $3.45 billion increase in the value of properties within 0.25 mile of these sites and a corresponding total tax gain of $579.3 million, computed as a stream of annual tax benefits since the start of the program until 2011.3 Though substantial, this tax gain does not exceed the $667.9 million4 that the BCP expended toward these projects.5 Our results also suggest that site redevelopment did not add to the off-site benefits of the program. This is an interesting finding considering that over 85% of BCP public spending is focused on redevelopment incentives. From the perspective of property values and assuming that entry into the program is not significantly affected by redevelopment incentives, our findings suggest that targeting program spending on cleanup incentives with reduced or no added inducements for redevelopment may result in the highest net public economic off-site benefits. Finally, we also find that site entry and site cleanup did not immediately capitalize into nearby property values and were valued by the housing market with a lag of approximately three years.
2. Background
State VCPs began in the 1990s in response to concerns that contaminated site owners’ fears of potential liability for damages and cleanup costs under the federal Superfund program had deterred sales, cleanup, and redevelopment of contaminated land. The Comprehensive Environmental Response, Compensation and Liability Act (CERCLA), commonly known as Superfund, addresses contaminated sites that pose severe health and environmental risks. CERCLA authorizes the EPA to initiate urgent and long-term remediation at such sites. However, properties that have lower levels of contamination are excluded from Superfund. In 2002, an amendment to CERCLA created the Small Business Liability Relief and Brownfields Revitalization Act, often referred to as the Brownfields Law, which provides conditional environmental liability relief for property owners and buyers of potentially contaminated land. The Brownfields Law also established the federal EPA Brownfields Program, which supports cleanup activities by providing assessment and cleanup grants to states and local communities. In addition, states also created VCPs to address contamination at brownfield sites. Today, every state has a VCP that, in exchange for voluntary cleanup, provides site owners with liability protection, technical support, and some financial assistance with cleanup costs and, in some cases, redevelopment costs in the form of grants, tax credits, and loans.
New York State established its VCP in 1994 and replaced it with the Brownfield Cleanup Program in 2003. The BCP encourages voluntary cleanup and redevelopment of thousands of contaminated sites in the state (DiNapoli 2013). It is administered by New York State’s Department of Environmental Conservation (NYSDEC). NYSDEC reviews applications and accepts all eligible sites except those that have hazardous wastes, petroleum contamination, or are subject to any ongoing federal enforcement action or other regulatory program. The BCP offers technical support, legal relief, and financial incentives to program participants. When site cleanup is finished, the BCP issues a certificate of completion that provides site owner with liability protection against any future claims for further remediation. A certificate of completion eliminates a constraint to sources of financing, private as well as public, to support redevelopment.
A rich set of financial incentives offered through the BCP includes tax credits for a percentage of cleanup cost, groundwater remediation, and tangible site redevelopment costs. The base percentage for credits each site can claim ranges between 10% and 12% of site cleanup and 10% to 12% of site redevelopment costs. An additional tax credit of 8% is available for sites in environmental zones (En-Zone).6 Sites can earn another 2% in tax credits if cleanup achieves unrestricted use of the site. Due to the high potential cost of the program, in 2008, individual site credits were capped at the lesser of $35 million or three times the site preparation and on-site groundwater remediation costs and $45 million or six times the site preparation and on-site groundwater remediation costs for those projects that are redeveloped for manufacturing. Only sites that have been issued a certificate of completion can claim applicable tax credits (NYSDEC 2016).
At its inception, the BCP was expected to provide tax credits of $135 million annually; however, the average annual cost of the program for fiscal years 2008 to 2012 was on average $188 million. As of 2013, the cumulative cost of the program has exceeded $1 billion, with an estimated potential outstanding tax credit liability to the state of $3.3 billion for the sites currently in the program (DiNapoli 2013).
Information about sites addressed under the BCP is publicly available through NYSDEC’s website, and communication with the public is required by the Department at several milestones during the cleanup process. Public comments and participation are required after a site application is completed and before the remedial investigation work plan is approved. Fact sheets describing the remedial investigation report, final engineering report, and certificate of completion are sent to the relevant public contact lists developed at the time of a site’s entry into the program. Therefore, the public has easy access to information on the sites that enter the program, site contamination levels, and actions taken to clean up and redevelop these properties.
3. Method and Identification
Economic value of nonmarket environmental goods, including those flowing from cleanup and redevelopment of contaminated land, is commonly estimated using preferences for complimentary goods. Such is the case with the hedonic method applied to housing values to gain insights into the capitalized value of urban amenities (or disamenities) (Palmquist 2005; Taylor 2003). The hedonic price function recovers the implicit prices of product attributes by modeling the relationship between the observed price of the properties and the unobserved prices of their characteristics. Rosen (1974) developed a two-stage procedure to recover marginal willingness-to-pay (MWP) functions for heterogeneous individuals and to capture welfare effects of nonmarginal changes in amenities. However, due to serious endogeneity problems associated with the estimation of the second stage, hedonic studies rely on the first stage to estimate the price effect of marginal changes in amenities (Bishop and Timmins 2015; Bartik 1987; Epple 1987).
First-stage hedonic analysis has been widely used to estimate MWP for brownfield cleanup (Lang and Cavanagh 2018; Haninger, Ma, and Timmins 2017; Linn 2013) and to value hazardous site remediation (Taylor, Phaneuf, and Liu 2016; Gamper-Rabindran and Timmins 2013; Zabel and Guignet 2012; Kiel and Williams 2007; Ihlanfeldt and Taylor 2004; Kiel and Zabel 2001). However, the many studies that rely on variation in panel data, such as fixed effects or difference-in-difference specifications, can recover MWP for amenities under several restrictive assumptions (Kuminoff and Pope 2014). One such assumption is the stability of the hedonic price gradient over time. If this assumption does not hold (e.g., neighborhood sociodemographic characteristics change in response to site cleanup), these methods recover capitalization effects that are not necessarily equal to MWP. Haninger, Ma, and Timmins (2017) suggest using a double-difference matching estimator that does not rely on temporal variation to recover estimates that have a welfare interpretation. However, the effectiveness of this method largely depends on the quality of the matches used in the analysis. Further, while it alleviates the concerns over temporal stability of the hedonic price gradient, it assumes that the gradient is constant across space. However, despite the concerns over intertemporal stability of a hedonic price gradient over time, econometric models using temporal variation in amenities are widely used to estimate capitalization effects because of their ability to eliminate time-invariant unobservables (Locke and Blomquist 2016; Livy and Klaiber 2016; Muehlenbachs, Spiller, and Timmins 2015; Mastromonaco 2014).
Building on this literature, we estimate the nonmarket benefits of site cleanup and site redevelopment by measuring their capitalization into local property values. A major challenge to accurate estimation of the impact of contaminated site cleanup and redevelopment on nearby property values is the presence of correlated unobserved heterogeneity that may confound identification (Mastromonaco 2014; Greenstone and Gallagher 2008). In an ideal research setup, site cleanup and site redevelopment would be assigned randomly, allowing for unbiased identification of the effects of cleanup and redevelopment. However, the decision to clean up or redevelop a site through the BCP is not random and could be strategic on the part of site owners. Therefore, we adopt several approaches to control for potential confounding unobservables. We begin by demonstrating the bias introduced by a cross-sectional specification of the hedonic regression. We then describe the steps we take to minimize the potential for omitted variable bias caused by time-invariant and time-varying correlated unobservables. Finally, we explain the choice of variables of interest and site proximity measures.
Estimation Strategy
Our goal is to measure localized public economic off-site benefits of contaminated site cleanup and redevelopment through the BCP by estimating their capitalization into property values. A standard hedonic price function, described in equation [1], models the effects of site cleanup and site redevelopment through the BCP on local property values:
[1]
where the dependent variable is log of sale price of the property i that transacts at time t, Xit is a vector of variables of interest that include sites entering the BCP to undergo cleanup and redevelopment, Zi is a vector of observable house characteristics, Yt is a set of temporal fixed effects, and εit is an error term.
Estimation of equation [1], however, will produce biased results if any of the housing characteristics or neighborhood amenities not observable to the researcher are correlated with property prices and site cleanup or redevelopment. Therefore, we modify equation [1] in several important ways to control for time-invariant and time-varying correlated unobservables.
A common approach to control for correlated heterogeneity that does not change over time is to use econometric models with spatial fixed effects (Livy and Klaiber 2016; Buck, Auffhammer, and Sunding 2014; Kuminoff, Parmeter, and Pope 2010). Spatial fixed effects alone, however, may not fully account for those unobservables that are specific to properties surrounding sites participating in the BCP. To address this concern, we control for correlated unobservables at the property level, removing any potential confounding time-invariant characteristics specific to each property (gardens, hardwood floors, historical significance, school district, proximity to an urban center or a water body, etc.). Property-level fixed effects also control for unobservable factors that may influence buyers’ location choices and location of sites participating in the BCP.
Another source of potential omitted variable bias in hedonic estimation is time-varying amenities or changing property value trends that may correlate with property prices and decisions to clean up and redevelop a site through the BCP. To control for spatially varying time trends throughout our sample period, our main specification includes census tract by year fixed effects that flexibly control for any housing market changes that vary over time within each neighborhood. We test the robustness of this approach by including a number of other controls to capture time-varying correlated trends, as discussed in the next sections.
An alternative approach to using year by census tract fixed effects is to include fixed effects for houses sold in the same census tract and the same sets of time periods (Mastromonaco 2014). The main limitation of this approach is that it greatly reduces the number of observations—by 60% in our sample. We, therefore, chose not to follow it.
Property-Level Fixed Effects Model
Our main specification estimates a repeat sales property-level fixed effects model, an alternative to the standard hedonic model in equation [1], which tightly controls for property-specific correlated unobservables and for time-varying heterogeneity at the neighborhood level:
[2]
where lnPit is the natural log of the price of property i that transacts at time t; Xit is a vector of variables of interest that relate to site cleanup and redevelopment through the BCP, defined below, and α includes coefficients of interest; θt is a temporal fixed effect indicating the quarter of the year that accounts for common seasonal shocks to the housing market (Ngai and Tenreyro 2014); τct is a census tract by year fixed effect that controls flexibly for spatially varying housing market conditions that change over time and could be correlated with site participation in the BCP. This includes general neighborhood improvement or deterioration, addition of neighborhood amenities such as a playground, increased crime or traffic, and other changes over time that could be correlated with site cleanup and redevelopment. The property-level fixed effect, μi, captures any confounding observed and unobserved property characteristics that did not change between the repeat sales. Examples include housing characteristics such as number of bedrooms, square footage, general quality of the house and local neighborhood, among others. Finally, vit is the random error term.
Identification in the repeat sales property fixed effects model comes from the changes in property sale prices and in the variables of interest over time, which, in this study, are site cleanup and redevelopment through the BCP. The identification relies on the panel aspect of the data, which include only properties that sold at least twice in the period analyzed. The estimation process compares sale prices of properties before and after cleanup or redevelopment, relying on the geographically fine spatial fixed effects at the property level. The strategy of using changes in entry and cleanup over time allows for a difference-in-difference interpretation. Our approach improves on the methodology used in prior studies on external benefits of VCPs that control for time-invariant unobservables at a more aggregate spatial scale (Linn 2013). In addition, the model identifies the effects at a fine temporal scale using date, month, and year of property sales, site entry, and cleanup and redevelopment through the BCP.
Several recent studies employ a difference-in-difference methodology that compares properties in close proximity to a contaminated site (treatment group) with properties located farther away (control group) (Taylor, Phaneuf, and Liu 2016; Zabel and Guignet 2012). Spatial clustering of sites, as evident from Figure 1, makes this approach inappropriate in our case because properties in the control group would be exposed to other sites participating in the BCP.
Geographic Distribution of the New York City Sites Participating in the Brownfield Cleanup Program
Variables of Interest and Proximity Definition
Though our primary interest lies in measuring the impacts of site cleanup and site redevelopment through the BCP, site entry into the program may also affect property values. Depending on the information available to buyers and sellers, site entry into the BCP may provide a positive signal of upcoming site cleanup that can capitalize in nearby property values. Because site entry can influence nearby property values and is correlated with site cleanup, omission of this variable will lead to biased estimates. Therefore, our analysis measures the effect of entry, cleanup, and redevelopment of sites through the BCP, and the capitalized off-site benefit of the program is the sum of these effects. We further distinguish between the types of redevelopments such as apartment buildings, schools, retail, or office to determine whether the type of redevelopment leads to different off-site effects.
Construction of the variables of interest requires decisions about (1) the definition of property exposure to sites in the BCP, and (2) the length of this exposure. Proximity to the nearest contaminated site is commonly employed in the literature as a measure of exposure (Haninger, Ma, and Timmins 2017; Taylor, Phaneuf, and Liu 2016; Ihlanfeldt and Taylor 2004; Kiel and Williams 2007). Distance to the nearest site or dummy variables indicating exposure within a certain buffer, however, do not capture the density of sites and may introduce omitted variable bias if some properties are exposed to more than one site, as is the case in our sample (Figure 1). Several recent hedonic studies rely on exposure measures that account for site density to estimate housing market response of Superfund site remediation (Mastromonaco 2014), drilling of shale gas wells (Muehlenbachs, Spiller, and Timmins 2015), impact of foreclosures (Bak and Hewings 2017), and brownfield cleanup (Lang and Cavanagh 2018; Linn 2013). One way to capture density is to count the number of sites within a certain buffer (Mastromonaco 2014), implicitly assuming that each site within the buffer has the same influence on housing values. Studies have shown, however, that the effects of site contamination are localized and diminish with distance (Gamper-Rabindran and Timmins 2013; Ihlanfeldt and Taylor 2004). To account for spatial clustering of sites that can expose a property to more than one site as well as the diminishing effects with increasing distance, we construct density indices of site entry, site cleanup, and site redevelopment through the BCP (Lang and Cavanagh 2018; Linn 2013). Specifically, our density index of site cleanup through the BCP sums the inverse distances of all sites that are cleaned up through the program and are located within 1 mile of a property7 such that
[3]
where 
indicates the inverse distance between property i and site k that is cleaned up through the BCP; Ikt equals 1 if cleanup of site k occurs by time t. Variables indicating site entry into the BCP, Eit, and site redevelopment through the BCP, Rit, are constructed analogously. Therefore, in equation [2], Xit ≡ {Eit,Cit,Rit}, where each term is a density variable representing site entry, site cleanup, and site redevelopment. Further, to explore whether the effects of site redevelopment differ by redevelopment type, we interact the site redevelopment variable, Rit, with an indicator variable equal to 1 if the site was redeveloped into an apartment building, school, office, or retail.
Using an inverse relationship between contaminated sites and nearby properties, we can infer the estimated impact of site entry, site cleanup, and site redevelopment at distances smaller than 1 mile. However, the use of a density index as a measure of exposure requires several implicit assumptions. First, the imposed inverse relationship between the impact of a site and its proximity to the property implies that sites located closer to a property are weighted more heavily than those that are farther away (Linn 2013; Ihlanfeldt and Taylor 2004). Second, in equation [3] the summation is taken over all sites within a 1 mile radius from a property, which effectively assumes that sites beyond 1 mile have no impact. To verify that 1 mile is a valid cut off, we rely on a data-driven approach and allow cleanup treatment effect to depend on distance (Taylor, Phaneuf, and Liu 2016; Dröes and Koster 2016). Specifically, we estimate the following equation:
[4]
where lnPit is the natural log of the sale price of property i that transacts at time t; Sitk equals 1 if the nearest cleaned up BCP site is located within a distance band k of property i transacted at time t. We use 0.25 mile intervals to define eight distance bands that extend to 2 miles. Therefore, the coefficient αk compares the price difference relative to properties outside of the 2 mile distance band. Finally, the remaining variables in equation [4] include a vector of housing characteristics listed in Table 1, Zi, a temporal fixed effect indicating the quarter of the year, θt, and a census tract by year fixed effect, τct. Appendix Figure A1 plots the estimated coefficients from equation [4] and their confidence intervals for each of the distance bands. As evident from the figure, the impact of site cleanup gradually decreases with distance. The effects become very small and are only marginally significant beyond the (0.75–1] mile distance band and are insignificant beyond the (1.5–1.75] distance band. This is consistent with the inverse relationship between the impact of site cleanup and its proximity to a property imposed by the density measures used in our analysis. While these results do not suggest that there is no effect beyond 1 mile, the choice of a 1 mile cutoff facilitates the identification of economically important results. Limiting the range of impact to 1 mile is generally consistent with prior studies of contaminated sites that revealed localized effects (Haninger, Ma, and Timmins 2017; Linn 2013; Mihaescu and vom Hofe 2012). The effects appear to be the strongest within 0.25 mile. Therefore, we exploit the properties of the density indices to explore localized effects at 0.25 mile, which is about four blocks in New York City. This is intuitively plausible given the density of the study area, while also consistent with prior literature that has interpreted property capitalization effects at fine geographic scale (Linn 2013; Schwartz et al. 2006).
Summary Statistics for Residential Properties Sold in New York City More Than Once over 1996–2011 and 2004–2011 Periods; Property Proximity to Sites Participating in the Brownfield Cleanup Program (BCP)
Any measure of exposure to sites participating in the BCP also requires an implicit assumption about the timing of effects. Although some effects of the BCP may capitalize into property values immediately due to the availability of public information on sites entering the program, other effects may occur with a lag. Our variable specification in equation [3] allows for the effects of entry or cleanup to capitalize into property values with a lag. We also examine several specific lags to understand how long it takes for the effects to capitalize in nearby property values.
4. Data
This study uses two main datasets: (1) data on brownfield properties obtained from the publicly available NYSDEC Environmental Site Database,8 and (2) proprietary transactions and characteristics data for residential properties in all boroughs of New York City purchased from DataQuick Information Systems9 for the years 1996–2011.
BCP Sites
Our brownfield sample includes 185 sites participating in the BCP. By 2011, 85 sites in our sample were cleaned up, while the remaining sites were in the process of being addressed through the program. Prior to cleanup and redevelopment, BCP sites’ uses ranged from former industrial parks, manufacturing companies, and dry cleaners to electric substations, abandoned buildings, and parking lots. Of the cleaned sites, 33 were redeveloped with conventional structures that include 21 apartment buildings with shops on the lower level, 6 shopping stores/malls, 5 schools, and 1 office building. Figure 1 shows the spatial distribution of all sites; it is evident that many sites are clustered closely to each other.
The NYSDEC Environmental Site Database as well as the weekly Environmental Notice Bulletin, an official publication of the NYSDEC,10 provide the exact date, month, and year each site enters the BCP and when it receives a certificate of completion that certifies site cleanup. Site redevelopment years are obtained from the annual census of all buildings in the city, their locations, and construction years provided by the New York City Department of Planning. Because the exact day and month of redevelopment is not available, we rely on Google Maps Street View images to obtain this information. Specifically, Google Maps Street View provides snapshots of the same properties over time (typically between 2007 and 2016) with the month and year these snapshots are recorded. This provides us with the month site redevelopment is completed. Finally, those properties for which Google Street View is not available (nine sites), we assign December 1 as the month and date of redevelopment. We check the sensitivity of these assumptions in our analysis.
We also use annual reports provided by the New York State Department of Taxation and Finance to obtain data on the public cost of the BCP. All program cost data are available and cover the period from 2005 to 2014.11 The information provided in each annual report includes the exact amounts of tax credits received by each individual BCP site. Appendix Table A1 demonstrates the distribution of total nominal BCP public spending on sites located in New York City and the state of New York by tax credit type. The program spends the most on tangible site redevelopment incentives, compared to cleanup and groundwater remediation spending.
Property Data
We started out with 788,758 observations of property sale transactions and characteristics spanning 1996 to 2011. The data include property sale price, sale date, transaction type (arm’s-length, quit claim, etc.), and address. Structural characteristics include property square footage, total lot size, lot width and length, number of properties on site, number of stories in each property, type of property (single home, condominium, apartment, etc.), and the year each property was built. Housing characteristic data are reported as they were recorded at the time of the most recent transaction and do not vary over time.
We performed a series of cleaning operations to remove missing records and outliers as well as nonrepeating transactions, which reduced our sample to a panel of 123,852 transactions of 49,769 properties that sold more than once (refer to the Appendix for a description of data cleaning steps). The final dataset was formed by joining geocoded transactions and characteristics data for properties that sold more than once in New York City during the 1996–2011 period with the sample of brownfield sites enrolled in the BCP. Panel A of Table 1 shows summary statistics of property sales and characteristics data we use, while Panel B shows summary statistics of site entry, cleanup, and redevelopment density variables used in the analysis. The property sample is broken down into two periods because the period between 2004 and 2011 is used in estimation of the effects of site redevelopment. The average price of properties is slightly higher for the 2004–2011 period than for the full sample, which includes property sale values all the way to 1996.
In addition to the property transaction data obtained from DataQuick, we rely on an annual census of all properties located in New York City provided by the New York City Department of City Planning. The census includes extensive data on all residential and nonresidential properties in New York City for the period between 2002 and 2016. Address, latitude and longitude, zoning, census tract and block, as well as structural characteristics such as the year property was built, lot size, and square footage are provided for each property in the dataset. We use these data in the cost-benefit analysis to identify all residential properties (sold or unsold) around each BCP site.
5. Results and Discussion
This section presents estimation results. In all models, the dependent variable is the natural log of property sale price. Variables of interest are defined as density measures that sum inverse distances between all sites that enter, are cleaned up or redeveloped within 1 mile of a property prior to its sale transaction. Such variable construction facilitates interpretation of all effects at 0.25 mile by multiplying the estimated coefficients by 4 (1/0.25). Standard errors are clustered at the respective fixed effect to control for correlation in the error terms.
Property-Level Fixed Effects Model
Table 2 reports estimation results of the BCP site entry and cleanup. All estimates are statistically significant across various specifications and are of the expected positive sign. As we move from column (1) to column (3), we demonstrate the impact of adding fixed effects at different spatial scales. In column (1) we start by estimating a naive hedonic regression from equation [1] that includes housing characteristics and accounts only for temporal year-quarter fixed effects. Column (2) adds spatial fixed effects at the property level, while column (3) estimates the effect of site entry and site cleanup according to the model specified in equation [2]. In addition to property-level fixed effects, column (3) includes an interaction of year and census tract fixed effects that controls flexibly for changing housing market conditions over time common to properties within the same census tract. Column (3) represents our main specification.
Capitalization Effects of Brownfield Cleanup Program Site Entry and Cleanup on Local Property Values
Interestingly, as we move from column (1) to column (3), where we add more precise fixed effects, the estimated coefficient on site cleanup decreases in magnitude by 50% compared to the estimate in column (1). This suggests the importance of controlling for correlated unobservables at a fine geographic scale. In column (4), we also include brownfield site fixed effects to control for any time-invariant brownfield characteristics. Addition of the brownfield fixed effect does not alter our results, therefore, column (3) that estimates equation [2] remains our main specification.
Together, site entry and cleanup increase property values 0.25 mile away by 4%. Figure 2 demonstrates the decay of the combined effects of site entry and cleanup on nearby property values with distance. The effects are strongest at smaller distances between properties and BCP sites and slowly decline as the distance increases to 1 mile. The magnitude of the estimated coefficients on entry and cleanup is a bit higher but generally consistent with prior work that has documented relatively small increases in nearby property values due to site participation in a cleanup program (Linn 2013).
Decay of Combined Capitalization Effects of Brownfield Cleanup Program Site Entry and Cleanup over Distance
Table 3 reports estimates of equation [2] that includes the effect of site redevelopment in addition to site entry and site cleanup. Because the first site redevelopment we observe occurred in 2003, we restrict our property transaction sample to the period between 2004 and 2011. This avoids any endogeneity that might be introduced by estimating these effects over the full sample. Column (1) of Table 3 includes site entry, site cleanup and site redevelopment. In column (2), we differentiate between the various types of site redevelopment such as apartment buildings, office, retail and schools, using office redevelopment as the omitted category. We do not find any statistically significant effects of redevelopment nor different types of redevelopment. Insignificant effect of site redevelopment on nearby property values may suggest that expectations about redevelopment have already been capitalized into property values after cleanup. This may be due to the fact that cleanup provides a strong signal of future redevelopment, particularly in a place such as New York City, where land for redevelopment is scarce. It may also suggest that site participation and cleanup through the BCP alone without further redevelopment can alleviate buyers’ concerns about the contamination of a site and, therefore, are sufficient for property values to rebound from the negative effects of being near a contaminated site (Dale et al. 1999). However, the relatively small sample of redeveloped sites warrants further exploration in future studies. In particular, future research should consider studying the effect of redevelopment on a regional or a national scale that would allow for analysis of a larger number of redeveloped sites, therefore increasing the power to detect redevelopment effects.
Capitalization Effects of Site Redevelopment through the Brownfield Cleanup Program on Local Property Values
Heterogeneity of Effects across Housing Submarkets
We explore the potential heterogeneous property effects of the BCP on different housing submarkets within New York City. We first divide the New York City housing market into submarkets defined by the geographic boundaries of the five New York City boroughs. For each borough, we estimate the model specified in equation [2]. Variables of interest are defined according to the equation [3] as the sums of inverse distances of all sites that entered, are cleaned up or redeveloped through the BCP, respectively, within 1 mile of a property. Appendix Figure A2 summarizes the effect of site entry, cleanup and redevelopment on property values 1 mile away in each borough of New York City. The graph depicts considerable heterogeneity in the program effects across these five submarkets. Site entry into the BCP appears to have significant effect on nearby property values in most boroughs. Specifically, in Manhattan, Queens and Brooklyn, site entry is positive and statistically significant, while cleanup does not significantly affect property values. This may suggest that entry provides a strong signal to the market about future cleanup (and redevelopment), therefore, the expectations about cleanup may be capitalized into the housing market after site entry into the program. Interpreting these effects for properties located 0.25 mile away from BCP sites suggests that site entry leads to appreciation of property values by more than 10% in Manhattan and 4% in Queens and Brooklyn, respectively. Interestingly, in the Bronx, the poorest borough of New York City, it is the cleanup that has a large positive and statistically significant effect on the housing market, while site entry does not have a significant effect. This result may suggest that site entry into the program alone does not provide a strong signal of future cleanup to home buyers in this market, therefore, property values respond only to the actual cleanup. This can also suggest that home buyers in the Bronx may have incomplete information about the BCP and become aware of the program only after the cleanup is complete. We do not find any statistically significant effect of the program in Staten Island, which may be due to the fact that in our sample only a handful of BCP sites are located there. Finally, the external effect of site redevelopment is consistently insignificant in all boroughs where redeveloped sites are located.
We also explore the potential heterogeneity of BCP capitalization effects on different property types. Specifically, we subdivide our housing data into Class 1 and Class 2 property tax classes designated by New York City. Class 1 includes one-to three-unit residencies and Class 2 consists of residential properties with more than three units. Appendix Figure A3 depicts the effects of entry, cleanup and redevelopment on Class 1 and Class 2 properties 1 mile away from a BCP site. For Class 1 properties, both entry and cleanup have positive and statistically significant effects on nearby property values. Interestingly, the values of Class 2 properties decrease in response to site entry into the program, while site cleanup has a significant positive effect. The effect of site entry on nearby property values can depend on the information available to the housing market. If information about contamination is not generally known by buyers and sellers, then site entry into the program that reveals this information, may produce a negative effect on property values. Lang and Cavanagh (2018) also find that incomplete information about risk may lead to declines in property values when this information is disclosed by remediation.
The combined effects of site entry and cleanup are positive for both Class 1 and Class 2 properties. Interpreting the combined effects for properties located 0.25 mile away from a site suggests that, together, site entry and cleanup increase Class 1 and Class 2 property values by 5.6%, and 1.2%, respectively. Interestingly, Class 2 properties experience a smaller increase in value from entry and cleanup compared to Class 1 properties. This may be due to the fact that Class 2 properties include mostly rented units and owners of these properties may be less likely to live in them themselves. The effect of site redevelopment is insignificant for both Class 1 and Class 2 properties.
Time to Capitalization Effects
While prior literature has documented positive effects of brownfield and Superfund site cleanup on nearby property values (Haninger, Ma, and Timmins 2017; Mastromonaco 2014; Linn 2013; Gamper-Rabindran and Timmins 2013), with few exceptions (Lang and Cavanagh 2018; Linn 2013), most studies that value contaminated land cleanup do not consider the time it takes for the effects to capitalize into property values. Table 4 shows estimates of the time to capitalization effects of site entry and site cleanup with various time periods between site participation in the program and property sale. We estimate the effects using several time bins to determine how quickly the effects are capitalized into nearby property values. Appendix Figures A4 and A5 plot the estimated results and demonstrate that property capitalization effects of site entry and site cleanup do not appear immediately, but take effect with at least a 3 year lag. The effects of both site entry and site cleanup are small and not statistically significant at first but grow steadily over time. The estimated results from our main specification that do not restrict the length of the lag between site participation in the program and sale of a property is robust to other time specifications presented in Table 4.
Time to Capitalization of Effects of Brownfield Cleanup Program Site Entry and Cleanup on Local Property Values
Robustness Checks
One way to evaluate the robustness of an identification strategy is to use a placebo (falsification) test by applying the model in the context where no effect should be found. We do this by artificially and randomly varying entry and cleanup dates. Specifically, we move entry and cleanup dates 100, 200, 300, and 400 days before the actual dates and reestimate, our main specification using the sample of properties that transacted before entry or cleanup i.e. pretreatment period (Bak and Hewings 2017; Galiani, Gertler, and Schargrodsky 2005). If the effect of entry and cleanup is found using falsified dates, this would suggest that neighborhoods near BCP sites trend differently than those that are far from such sites, bringing the identification strategy into question. The results of the falsification test are presented in Appendix Table A2. None of the entry or cleanup variables are significant, supporting our identification strategy and suggesting that the neighborhoods close to BCP sites do not trend differently from the neighborhoods located far from these sites.
Measurement error in the month of a site’s redevelopment could potentially introduce some degree of attenuation bias into our estimates of redevelopment effect. Therefore, we check the sensitivity of our results to using Google Maps and the assumptions to identify the month of redevelopment. To do this we move the month of redevelopment by three and six months before and after the month used in our analysis. While it would be valuable to measure whether the period when redevelopment begins and the period during the construction could affect nearby property values, the information on these dates is not available. When we move the month of redevelopment, we may also capture some of these effects if they exist. Appendix Table A3 presents these results. Moving the month of redevelopment does not influence our main results, presented in column (1) of Table 3. The estimated capitalization effect of site redevelopment remains statistically insignificant.
To further address concerns over potential bias caused by time-varying correlated unobservables and to check the robustness of our base model, we estimate equation [2] with several additional controls. First, we control for construction activity in close proximity to properties that are exposed to sites participating in the BCP. New York City experiences constant construction activity. Major construction projects nearby might influence sale values of existing properties. Moreover, increased construction activity may be associated with favorable improvements to the neighborhood that can make it more desirable and increase property values. This, in turn, may influence a site owner’s decision to enter the BCP as the value of site cleanup and redevelopment increases with improved neighborhood desirability and local property prices. Using data from the New York City Department of Planning on the exact location of all buildings (both residential and nonresidential) in New York City and the year each building is constructed or altered, we define a variable that counts the number of buildings constructed one year before the sales of properties in our data and located within 0.3 mile of these transacting properties. We also construct a similar variable that counts the number of buildings that were altered one year prior to the sales of properties located within 0.3 mile. To vary the time lag with which construction or alterations to the existing buildings may influence nearby property prices, we also estimate the equation with construction and alteration activity that took place two years prior to property sales. The results of adding these controls are presented in columns (1) and (2) of Table 5. The estimated coefficients on site entry and cleanup remain positive and statistically significant demonstrating that our model specification is robust to these additional controls.
Robustness Checks
Second, rising property prices may also induce site cleanup and redevelopment, making entry into the BCP more attractive if a site is located in the neighborhood that is experiencing increasing property values. Assuming property values include expectations of future trends (Bajari et al. 2012), we control for the potential confounding effects of property value trends by including in equation [2] an interaction term of a set of year fixed effects and the average price of properties sold before any site enters the BCP or receives certification (before year 1997) and located within 0.3 mile of transacting properties (Linn 2013). Column (3) of Table 5 demonstrates that this control does not change our main results. This implies our main model as specified in equation [2] already accounts for the potential time-varying unobservables that could be captured by this interaction term.
Moreover, to account for changing neighborhood amenities associated with industrial activity we add a control for proximity to CERCLIS12 sites. Under the Superfund program, potentially hazardous sites are added to the CERCLIS database and investigated by the EPA. CERCLIS sites that are determined by EPA to present imminent danger to human health and the environment based on its hazard ranking system score are placed on the National Priorities List (NPL) and are commonly referred to as Superfund sites. Superfund sites, can have significant impacts on nearby property values (Matromonaco 2014; Keil and Williams 2007; Kiel and Zabel 2001) and are typically located in less desirable neighborhoods with a heavy industrial past (Mastromonaco 2014). However, those CERCLIS sites that do not achieve the necessary score to be placed on the NPL may also have serious levels of contamination or potential contamination. These sites receive a non-NPL status code in the CERCLIS database, which indicates that the EPA has determined that they will not be placed on the NPL list. We, therefore, use density of CERCLIS sites as a proxy for the presence of unfavorable neighborhood amenities associated with industrial activity that can affect property values at different points in time. We construct a density measure of CERCLIS sites analogously to our main variables of interest as described in equation [3]. That is, we sum the inverse distances of all CERCLIS sites within 1 mile of a transacting property that received non-NPL status in the program prior to the transaction.13 Our main results remain robust to the addition of the CERCLIS variable as evident from column (4) in Table 5.14
Third, to alleviate concerns that the assumption of constant housing characteristics and their coefficients may bias estimation results (Linn 2013; McMillen 2008), we divide our sample into five-year time periods between 1996 and 2011 and interact property characteristics with period fixed effects. Column (5) of Table 5 shows that our main model is generally robust to this specification as the coefficients remain positive and significant.
Finally, we add fixed effects for interactions of the year of the sale by the year of prior sale, which limits identification to properties that sold in the same time period (i.e., have the same sale year and the same year of prior sale). Our results are also generally robust to this specification, presented in column (6) of Table 5.
Comparison of Local Public Economic Benefits and Program Costs
To date, little is known about the public costs and benefits of state VCPs. Although existing literature has shown positive neighborhood effects of site entry and cleanup through some state VCPs (Linn 2013), it is unclear how these benefits compare to program costs. Relying on unique program cost data obtained from the New York State Department of Taxation and Finance, we quantify the public cost benefit comparison for the BCP. Our analysis focuses on the local external benefits generated by increases in nearby property values due to site entry and cleanup through the BCP in New York City and the corresponding tax gains they generate. It excludes private costs and benefits realized at BCP sites, employment effects, impacts on nearby commercial properties, and inconvenience costs experienced by neighbors during cleanup and redevelopment. We compare residential property tax gains to the public program spending in the form of tax credits provided by the BCP to incentivize voluntary cleanup and redevelopment.
We follow several steps to compute aggregate increases in local residential property values due to site entry and cleanup. First, we identify all residential properties (sold or unsold) exposed to each BCP site. To do this, we use an annual census of all New York City properties and their geographical information obtained from the New York City Department of City Planning. Using GIS methods, we determine which of these properties are located within 0.25 mile of each site participating in the BCP (0.25 mile is the distance at which we interpret all effects). To ensure that we capture the aggregate off-site residential benefits due to site entry and site cleanup, respectively, we differentiate between the properties that are exposed only to the sites that have entered the program from those that are also exposed to the sites that are cleaned up through the BCP.
Second, we assign a “preentry” and “precleanup” price to each of the properties exposed to BCP sites. That is, we calculate an average sale price of exposed properties sold before the year each site enters the BCP or is cleaned up. This average price is then assigned to each single-family residential property and each unit in multiresidential properties (sold or unsold) within 0.25 mile of a site, based on the sites they are exposed to.
Finally, we use the preentry and precleanup prices, the number of properties within 0.25 mile of each site, and the coefficients on site entry and site cleanup for Class 1 and Class 2 properties, presented in Appendix Figure A3, to compute the increases in property values following each site entry and site cleanup. We use the estimated effects of site entry and cleanup for Class 1 and Class 2 properties because these properties are valued and assessed differently in property tax calculations by New York City. The aggregate neighborhood off-site benefits due to the BCP is the sum of these property value increases.
Following the above method, the local appreciation in residential property values within 0.25 mile of BCP sites led to a $3.45 billion increase in property values. The city and the state may recover some of its investment through increases in property taxes. We follow three steps in calculating the tax proceeds from these property value increases. First, we discount the estimated increased value of each property due to site entry and cleanup by 0.7 because appraised market values in New York City are on average 70% lower than sale prices (Schwartz et al. 2006). Second, we separate properties by property type because assessment ratios15 and tax rates vary by property type16 in New York City. We then apply the relevant assessment ratios and tax rates to these discounted property values by property type to obtain the tax gains for each property.17 In this calculation, we continue to differentiate between properties that are exposed only to sites that enter the BCP and those that are also exposed to sites that are cleaned up through the program. Finally, to capture the flow of tax benefits over time, we compute the total tax gains as the sum of the stream of annual tax receipts from the start of the program until 2011.18
Using this three-step method, we estimate that the total flow of tax gains resulting from increases in property values exposed to BCP sites within 0.25 mile between the start of the BCP and 2011 is $579.3 million. This amount is a conservative lower bound estimate on the tax gains that the BCP produces for the city for two reasons. First, while we interpret all our effects at 0.25 mile, our estimations suggest properties at a closer proximity to brownfields can experience greater value increases and the program’s benefits can extend up to 1 mile. Second, we limit our benefit calculation to 2011, the end of our study period. However, the flows of tax benefits extend into the future and will increase as more sites enter and get cleaned up through the program.
All BCP cost data were obtained from the New York State Department of Taxation and Finance and cover the period from 2005 to 2014. Because the site cleanup component of tax credits is available for 5 years and the site redevelopment component is available for 10 years, we use tax credit data available for all years even though our study period ends in 2011. The BCP financed 40 projects located in New York City, with public cost in the form of tax credits provided to these sites totaling $667.9 million.19 These costs also represent a lower bound on the total credits received by the sites in our analysis. First, these sites may receive additional financial assistance from the BCP in the coming years due to the long periods of time that are allowed for claiming cleanup and redevelopment credits. Second, in the 2010, 2011, and 2012 tax years, taxpayers were required to defer credits exceeding $2 million in aggregate. Excess deferred brownfield credits for sites under our analysis were received in 2013, 2014, and 2015.20
Our results suggest that the BCP had substantial external benefits over the period of our analysis. The increases in property values due to site participation in the BCP led to significant tax benefits that, however, fall short of outweighing the public spending on the program. Tax gains from properties exposed to sites within 0.25 mile are estimated at $579.3 million through 2011, while program costs totaled $667.9 million.
The BCP has heavily incentivized site redevelopment following site cleanup, with over 85% of the program’s spending focused on tangible redevelopment incentives. However, we do not find that site redevelopment leads to additional external property benefits. Site entry and cleanup through the BCP alone produce substantial positive effects on property values despite the disproportionally smaller spending on cleanup incentives. This implies that higher spending on site cleanup with smaller or no added inducements for redevelopment may increase the net public economic off-site benefits. One important caveat to this conclusion is the possibility that entry into the BCP might be influenced by the availability of incentives for redevelopment in addition to the incentives for cleanup. During the period of our analysis, more than 40% of New York City sites that claimed cleanup credits also claimed redevelopment credits. This number is over 50% for the sites located in the rest of New York State.
Although this analysis is among the first attempts to understand the public cost-benefit trade-off of a state VCP, data limitations preclude a more comprehensive analysis. For example, we do not have information on such potential positive impacts of the BCP as job creation, increases in commercial property values, benefits received by developers, or any additional services and amenities provided through site redevelopment. Similarly, costs and potential negative externalities such as homeowner displacement and increased noise, crime, and traffic congestion due to redevelopments are also not captured in our analysis because of lack of data. Future research should address these concerns as well as potential heterogeneous housing market response to brownfield sites’ size, the extent and type of contamination, and contaminated media such as soil, groundwater, or air.
6. Conclusion
This paper contributes to the debate over the public economic benefits and costs of state-funded VCPs. Established in every state, VCPs are the primary mechanisms used by policy makers to address low-risk land contamination throughout the United States. Given the central role these programs play in U.S. land revitalization efforts and the substantial public costs of VCPs, it is important from a policy perspective to quantify the public economic benefits and costs of such programs. This paper fills this gap by estimating the public economic benefits of New York’s BCP, as its projects are capitalized into property values in New York City, and comparing those benefits to the public costs of the program. New York’s BCP is one of the largest VCPs in the country, having spent over $1 billion dollars on contaminated land cleanup through financial incentives for site cleanup and site redevelopment. We focus our analysis on New York City, where more than half of the program’s public spending has been concentrated. Our analysis relies on property-level fixed effects to control for time-invariant correlated unobservables, while also controlling for time-varying sources of omitted variable bias addressed in only a few existing studies.
Our findings suggest that the BCP results in significant external neighborhood benefits. During the period we analyzed, contaminated site entry into the BCP and site cleanup altogether led on average to a 4% appreciation in residential property values 0.25 mile away from the sites in the program. These increases were 5.6% for residences with three units or less and 1.2% for multifamily properties. This translated into a $3.45 billion increase in residential property values with a corresponding $579.3 million tax gain between program inception and 2011. The property tax gains fall short of the $667.9 million public program spending during the same period. This analysis provides one important data point for assessing the public cost-benefit trade-off of state-funded VCPs. However, data limitations preclude a more comprehensive and conclusive analysis. Our analysis does not include other potential positive impacts of the BCP such as job creation, sales tax generation, increases in commercial property values and residential property values more than 0.25 mile from the BCP sites, and developer benefits. Likewise, the cost calculation does not account for some negative effects the program could generate, such as homeowner displacement and other private costs, increased noise, traffic congestion, and any future tax credits that brownfields included in our analysis may claim through the program.
Unlike prior literature, we estimate values of site entry and cleanup separately from the value of redevelopment. Our results suggest that the local housing market does not place a premium on redevelopment or different types of redevelopment. Further, our analysis shows that the effects of site entry into the BCP and subsequent cleanup are valued by the housing market with a lag.
The BCP has heavily incentivized tangible site redevelopment through the program, but our results suggest that site entry and cleanup provide the bulk of the benefits to nearby residential properties. Unless entry into the program is significantly affected by redevelopment incentives, targeting program spending on cleanup incentives and reducing inducements for redevelopment might increase the net public economic benefits.
Acknowledgments
This work was supported in part by the U.S. Environmental Protection Agency through K6 grant TR-83418401, by project ILLU-470-320 of the Agricultural Experiment Station, University of Illinois, and a data acquisition grant from the Library of the University of Illinois at Urbana-Champaign. However, the contents of this paper are solely the responsibility of the authors and do not necessarily represent the official views of the sponsors. The authors are grateful to Amy Ando, Kathy Baylis, and members of the UIUC Program in Environmental and Resource Economics seminar group for helpful comments. We also thank the anonymous referees for their helpful comments.
Footnotes
↵1 The EPA defines a brownfield as a “real property, the expansion, redevelopment, or reuse of which may be complicated by the presence or potential presence of hazardous substance, pollutant, or contaminant.” (http://epa.gov/brownfields/.)
↵2 External effects of the BCP in New York City may be different from those in the rest of New York State. While it would be informative to estimate these effects, lack of data on property sale transactions in the rest of the state precludes such analysis.
↵3 The calculation of total tax gain accounts for the flows of tax receipts in every year following the initial increase until 2011. The discount rate used is 5% for benefit and cost calculations.
↵4 BCP spending on cleanup and redevelopment incentives for sites located outside of New York City totaled $624.1 million. To make costs comparable to benefits, we bring the annual public costs incurred in years 2007 to 2014 to 2011 using future value and discounting calculations. The discount rate used is 5%. For more information on program costs refer to Section 5. Appendix Table A1 provides a detailed breakdown of the nominal amounts of credits by type claimed by sites located in New York City and the rest of New York State.
↵5 We compare off-site benefits of the program in the form of tax gains accumulated to residential properties within 0.25 mile of BCP sites and the costs that the program expended to these brownfields in the form of cleanup and redevelopment tax credits. Therefore, our analysis excludes other potential positive or negative impacts of the program such as property value increases beyond 0.25 mile, private costs and benefits, job creation, sales tax generation, increases in commercial property values, and so forth. Furthermore, our calculations include only the costs and benefits realized during the period of our analysis. This implies that any financial assistance that the BCP may provide to the sites in the future due to the long period of time allowed for claiming cleanup and redevelopment credits is not part of our calculations. Therefore, the cost and benefit amounts can be considered lower bounds on the total benefits and costs of the program.
↵6 En-Zones are areas that as of the 2000 census have a poverty rate of at least 20% and an unemployment rate of at least 1.25 times the statewide unemployment rate.
↵7 While it would be informative to explore the variation in BCP sites on observable characteristics such as size, contamination type, and severity, these data are not available for all brownfield sites in our sample. To control for time-invariant brownfield characteristics we include brownfield fixed effects in our model presented in column (4) of Table 2. Addition of brownfield fixed effects does not change the results.
↵9 DataQuick subsequently was acquired by CoreLogic (https://www.corelogic.com).
↵11 The first tax credits were claimed in 2007.
↵12 The Comprehensive Environmental Response, Compensation, and Liability Information System (CERCLIS) is the EPA database that contains information on activities conducted under the Superfund program. The EPA is currently transitioning to the Superfund Enterprise Management System (SEMS), which replaces the CERCLIS database. SEMS provides the same content as CERCLIS.
↵13 Among the CERCLIS sites located in New York City, 39 received non-NPL status and 2 were placed on NPL. Because NPL sites are more severely contaminated than those that receive non-NPL status, the CERCLIS variable does not include the two NPL sites located in New York City. As a robustness check, we control for the two NPL sites using their NPL listing dates. Inclusion of NPL sites does not change our main results.
↵14 In addition, 61 brownfield sites enrolled in the EPA Brownfields Program were identified in New York City. Cleanup dates at 18 of these sites fell outside of the time period of this study. Of the remaining 43 sites, cleanup has begun at two brownfield sites and was completed at one of them during the time frame of our study. As a robustness check we control for these sites by creating two indicator variables (one indicating cleanup start and the other one indicating cleanup completion) equal to 1 if a site is within 0.25 mile of a property and its cleanup start or cleanup completion occurs before property sale. Inclusion of these controls does not change the results.
↵15 New York City applies and assessment ratio of 6% to Class 1 residential properties and 45% to Class 2 residential properties.
↵16 For tax purposes New York City classifies one- to three-unit residential properties as Class 1 properties and residential properties with more than three units as Class 2 properties.
↵17 We account for tax rate increases over time by applying the relevant tax rates for each year obtained from the New York City Department of Finance in the tax gain calculation for each exposed property. Our calculation does not include exemptions such as for veterans or elderly.
↵18 We use a 5% discount rate in a future value formula to bring cash flows that occurred in the past to 2011 and include the flows of tax receipts in every year following the initial increase until 2011. Although we do not have information on the exact interest rate specific to the time frame of our analysis, we believe a 5% rate is a reasonable assumption given that in 2006 the average rate for New York City’s general obligation issues was 5.35% Schwartz et al. (2006), while in 2016 it was 5.1% (NYCIBO 2017).
↵19 To make costs comparable to benefits, we bring the annual public costs incurred in years 2007 to 2014 to 2011 using future value and discounting calculations. The discount rate used is 5%.
↵20 Fifty percent of total deferred credits allowed in 2013; 75% of remaining credit allowed in 2014; all remaining credit allowed in 2015.
