NCEE0 NATIONAL CENTER FOR ENVIRONMENTAL ECONOMICS A Hedonic Analysis of the Impact of LUST Sites on House Prices in Frederick, Baltimore, and Baltimore City Counties Jeffrey Zabel and Dennis Guignet Working Paper Series Working Paper# 10-01 January, 2010 stA}.^ U.S. Environmental Protection Agency National Center for Environmental Economics 1200 Pennsylvania Avenue, NW(MC 1809) JQL , |^L $ Washington, DC 20460 K. S? http://www.epa.gov/economics i* o \ ------- A Hedonic Analysis of the Impact of LUST Sites on House Prices in Frederick, Baltimore, and Baltimore City Counties Jeffrey Zabel and Dennis Guignet NCEE Working Paper Series Working Paper # 10-01 January, 2010 DISCLAIMER The views expressed in this paper are those of the author(s) and do not necessarily represent those of the U.S. Environmental Protection Agency. In addition, although the research described in this paper may have been funded entirely or in part by the U.S. Environmental Protection Agency, it has not been subjected to the Agency's required peer and policy review. No official Agency endorsement should be inferred. ------- A Hedonic Analysis of the Impact of LUST Sites on House Prices in Frederick, Baltimore, and Baltimore City Counties Jeffrey Zabel Economics Department Tufts University and Dennis Guignet Department of Agricultural and Resource Economics University of Maryland KEYWORDS: LUST, Hedonic Analysis, Groundwater Contamination, Remediation Benefits Abstract Petroleum from leaking underground storage tanks (LUSTs) can contaminate local soil, and surface and groundwater. In some cases this can pose health risks to the surrounding population. Focusing on single family home sales from 1996-2007 in three Maryland counties, we use a hedonic house price model to estimate the willingness to pay to live father away from LUST sites. Particular attention is given to how property values are affected by leak and cleanup activity at a LUST site, the severity of contamination, the presence of a primary exposure path (i.e., private groundwater wells), and publicity surrounding a LUST site. The results suggest that although the typical LUST site may not significantly affect nearby property values, more publicized (and more contaminated sites) can impact surrounding home values by more than 10%. We would like to thank the National Center for Smart Growth at the University of Maryland for providing us with the housing transaction data. We also would like to thank Anna Alberini, Chip Paterson, and particularly Kelly Maguire for useful comments. 1 ------- 1. INTRODUCTION Petroleum products are used in many industrial activities, and some products (such as motor fuel) are sold to consumers at commercial facilities. Such facilities are widespread, and often store petroleum products onsite in underground storage tanks (USTs). For example, USTs are commonly used at gas stations to store gasoline, diesel, and other petroleum products. Over time leaks may occur as a result of corrosion and rusting, cracks, defective piping, and because of spills during refilling and maintenance activities. Petroleum from leaking underground storage tanks (LUSTs) contaminates the surrounding soil and can percolate into local groundwater aquifers. Oil contamination can migrate via surface run- off or local groundwater flows, and could potentially contaminate the surrounding environment and nearby water bodies. As of March 2009, there were over 482,166 known UST releases throughout the United States.1 In addition to environmental contamination, LUSTs can pose potential adverse health risks. Vapors can travel upwards into nearby homes and other structures. This poses several acute health risks such as headaches, nausea, and even potential explosions.2 Exposure to petroleum products over long periods of time increases the risk of some chronic diseases. Consumption of contaminated groundwater is the primary exposure path of concern. Petroleum products break down to several carcinogens and other contaminants that can affect the kidneys, liver, and nervous system. As a result, 1 US Environmental Protection Agency (EPA) htto://www. epa. gov/oust/faas/faa9a.htm. Accessed July 16, 2009. 2 Maryland Department of Environment (MDE), http://www.mde.state.md.us/assets/document/LRP%20Vapor%20Intrusion%20Guidance(6').pdf. accessed July 16, 2009. 2 ------- concentration levels of these petroleum constituents in drinking water are regulated by the US Environmental Protection Agency (EPA).3 Due to the potential environmental costs and health risks, LUSTs may adversely affect the welfare of nearby residents. If so, the cleanup of contamination from a LUST should result in some benefit to residents. We attempt to measure the benefits of cleaning up LUSTs, as reflected in residential property values. A hedonic property value model is estimated using single family home sales from 1996-2007 in three Maryland counties: Baltimore City, Baltimore, and Frederick. Careful attention is given to how property values are affected by leak and cleanup activity at LUST sites, the severity of contamination, the presence of a primary exposure path (i.e., private groundwater wells), and publicity surrounding a LUST site. The main conclusion from this analysis is that the average LUST site is unlikely to have a significant impact on house prices. However, the results suggest that the most publicized (and more contaminated sites) LUST sites can significantly impact nearby property values by more than 10%. This paper proceeds as follows. In Section 2, we provide a literature review. In Section 3, we discuss the data that we will use to estimate the hedonic model. This includes the property transaction data we were able to obtain from the National Center for Smart Growth and the LUST data that we obtained from the Maryland Department of the Environment. In Section 4, we lay out the framework for using the hedonic model. We pay particular attention to measuring the baseline impact of living near the UST site prior to discovery of the leak so that the impact of the LUST site is measured with respect to this baseline. In Section 5, we present the results, followed by some concluding remarks in Section 6. 3 US EPA, http://www.epa.gOv/safewater/contaminants/index.html#listmcl. accessed July 16, 2009. 3 ------- 2. LITERATURE REVIEW There is a large literature that provides evidence that hazardous waste sites adversely affect the prices of nearby residences. Boyle and Kiel (2001) provide a recent survey of the literature. A significant portion of the hedonic residential property value studies focus on Superfund sites. Comparing across these studies Farber (1998) finds that surrounding residential property values increase, on average, by $3,500 for each additional mile away from a hazardous site. Boyle and Kiel (2001) find significant variation in this premium across studies ranging from $190 to $11,450. The relatively small literature on the impact of contamination on the value of non- residential properties is surveyed in Jackson (2001). This includes the impact on the values of nearby commercial and industrial properties and on the contaminated property itself. Jackson analyzes seven studies (Dotzour (1997), Guntermann (1995), Page and Rabinowitz (1993), Patchin (1994), Sementelli and Simons (1997), Simons and Sementelli (1997), and Simons, Bowen and Sementelli (1999). Jackson reports that all of these studies that estimate the impact of contamination on the sales prices of commercial and industrial properties find significantly negative effects. Some studies focus on the value of the contaminated property itself. McGrath (2000) estimates a hedonic equation of sales prices of industrial properties in Chicago that includes the probability of contamination, PROBCON. The estimated coefficient for PROBCON is negative and significant. The impact is a 76% median unit discount or approximately a one million dollar ($1995) per acre decrease in parcel value. This is a particularly large impact and McGrath speculates that "investors are perhaps either 4 ------- overestimating the financial liability or that the discounts incorporate the present value of required legal costs certain to be part of any site redevelopment." (page 440). Jackson (2002) estimates the impact of current or previous contamination on prices using a hedonic equation applied to sales of industrial properties in Southern California. He finds that contaminated industrial properties sell for a discount of approximately thirty percent, on average. Alberini (2007) examines the "contamination discount" of selected contaminated properties in Colorado and finds that the contaminated property appreciates in value after participation in the Colorado Voluntary Cleanup Program. In the remainder of this literature review, we first focus on studies relevant to our analysis of the impact of LUST sites/groundwater contamination on property values. Initially we consider impacts on nearby properties. We then review the literature on the impacts on the LUST sites, themselves. Second, we consider the impact of property contamination on the likelihood that such properties will be redeveloped. Third, we look at three studies that use data from our study area, Maryland. 2.1 The Impact of LUST Sites/Groundwater Contamination on Property Values; Nearby Sites One study that looked specifically at LUST sites is Simons, Bowen and Sementelli (1997; henceforth SBS97). SBS97 analyze the impact of USTs on 16,990 residential sales in Cuyahoga County, Ohio in 1992 (this includes the city of Cleveland). They considered three types of USTs: non-leaking tanks registered with the State of Ohio, and registered and unregistered LUSTs. There were 2,513 tank sites; 1,151 non- leaking, 835 leaking but unregistered, and 527 leaking and registered. SBS97 cite a study 5 ------- by Bowen, Sailing, Haynes, and Cyran (1995) that developed a ranking of the toxicity of noxious environmental releases. Based on their analysis, LUSTs are expected to have a very localized impact. SBS97 interpreted this to mean being within sight distance or within a city block (300 feet). Hence, they generated indicator variables for units within this distance of the three types of USTs. There were 83 sales within the required distance of an UST; 42 near non-leaking USTs, 24 near leaking but unregistered USTs, and 17 near leaking and registered USTs. The only indicator that was marginally significant (at 5% but not at 1%) was for leaking and registered USTs. The estimated coefficient indicated that houses near a registered UST that is known to have leaked sold for a discount of $15,152 or 17% of the average sales price in 1992. This result should be viewed with caution since it is based on a small number of sales (17) and the model does not control for other potential LULUs (locally undesirable land uses) that could bias the result. Page and Rabinowitz (1993) analyze the impact of groundwater contamination on residential and non-residential properties. They note that the nature of groundwater flows complicates the analysis; "Neither the direction nor the rate of movement of plumes of toxic chemicals in ground water is predictable without a thorough and costly hydrogeological investigation." (page 473) The analysis of non-residential properties is based on a few case studies of abandoned industrial properties. The authors find that groundwater contamination significantly negatively affected the value of these properties (though it is not clear how they did this). The residential analysis considers properties in seven rural towns or small cities in Wisconsin that depend on private groundwater wells. The authors compare units with groundwater contaminated with toxic chemicals to 6 ------- similar nearby properties with wells with no identified contamination. They find no difference in the prices across these two groups of properties. Dotzour (1997) looks at the impact on sales prices of residential properties in an area of Wichita Kansas where groundwater contamination had been discovered. However, few of the properties in the contaminated area used the groundwater as drinking water. Dotzour compared the change in average sales price of units in the contaminated area during the year before and after the contamination announcement to comparable changes in two control areas. The results showed no significant differences across the three study areas. 2.2 The Impact of LUST/Groundwater Contamination on the Property Values and Transaction Rates of LUST Sites Simons, Bowen and Sementelli (1999; henceforth SBS99), Simons and Sementelli (1997) and Sementelli and Simons (1997) compare property values and transaction rates of LUST sites versus non-LUST sites. All three studies use data from the same location and hence cannot be considered to provide independent information. SBS99 analyze residential and commercial properties in Cuyahoga County, Ohio. Using the residential properties, SBS conducted a limited hedonic analysis. They found that residential properties near and/or with actual contamination from a LUST sold for a 14- 16% discount (consistent with SBS97). They also estimated that commercial LUST sites sold at an annual rate of 2.7% whereas the annual transaction rate for uncontaminated sites was 4.0%. Hence, the transaction rate for the LUST sites was 33% lower than for those sites without contamination. These results are suggestive at best since they are 7 ------- based on a very small sample of contaminated properties. In the case of the commercial analysis, it is likely that the difference in transaction rates between contaminated and uncontaminated properties is not statistically different from zero. Further, this analysis does not control for the characteristics of the sites, so it is unclear if this difference is solely driven by differences in contamination levels. Using a similar dataset, Sementelli and Simons (1997) find that a No Further Action (NFA) letter has no impact on the transaction rates of LUST sites. Simons and Sementelli (1997) compare the transaction rates of LUST and registered nonleaking tank (RUST) commercial sites. They note that in Cuyahoga County, most of the drinking water comes from Lake Erie and is provided by the City of Cleveland Water Department. Hence, 98% of the LUST sites use municipal drinking water and hence the health risks are minimal. But it is expected that LUST sites will be slower to sell. Results show that the transaction rates for LUST sites over a four year period was 3.8% versus 10.4% for comparable, uncontaminated (non-RUST or LUST) sites. Further, the transaction rate for RUST sites was only 4.9%. Relative to sites with no USTs present, buyers may be reluctant to purchase properties with RUSTs in fear of future liability, and remediation and removal costs. RUST and LUST sites were also found to be less likely to obtain secured mortgage financing and loan-to-value ratios were lower than for other commercial properties. 2.3 The Impact of Contamination on Redevelopment Many observers suggest that contamination—whether actual or merely suspected—is likely to impair the redevelopment of properties. Three studies focus on the 8 ------- impact of contamination on the redevelopment of such properties. Sigman (2005) estimates the impact of CERCLA liability laws on the redevelopment rates of industrial sites in the U.S. The data are annual city-level observations from 1990 to 2000. The data are from surveys of realtors and are not transaction data. The dependent variable is the vacancy rate of industrial space. Sigman uses fixed effects to capture unobserved city- specific factors that can affect vacancy rates. The presence of CERCLA joint and several liability laws implies a 40% increase in vacancy rates in city centers. There is suggestive evidence that joint and several liability has a bigger impact in cities with a higher risk of contamination. Strict liability does not significantly affect vacancy rates. The impact of joint and several liability on vacancy rates in suburban areas is negative but not significant. Sigman also finds similar results using a data set of brownfield sites; the presence of joint and several liability in a city is associated with 67% more brownfield sites. These results are not as strong as the previous ones since the data are cross- sectional and hence it is not possible to use fixed effects to capture unobserved city-level factors that are correlated with liability laws. Also, the definition of a brownfield is not standardized across cities. McGrath (2000) also analyzes the impact of contamination on the likelihood of redevelopment for 195 industrial properties in Chicago that sold between August 1983 and November 1993; 95 of which were redeveloped. Individual property contamination levels are not known, so McGrath uses a list of contamination probabilities for 25 industrial and commercial land-uses to generate the probability of contamination variable, PROBCON. McGrath estimates a probit model where the dependent variable is whether or not a property that sold is redeveloped. The estimated coefficient for 9 ------- PROBCON is negative but not significant. Hence, there is no evidence that redevelopment of a purchased site is affected by the probability that a site is contaminated. Lange and MacNeil (2004) estimate a logit model where the dependent variable is whether or not the redevelopment of a brownfield site was "successful" or "not-so- successful" (the authors do not state was it means for redevelopment to be successful). The data on 26 successful and 26 not-so-successful sites were obtained from surveys sent to 228 representatives of EPA brownfield assessment pilots (the response rate was 24%). Four factors were found to significantly affect successful redevelopment: an index of political support (financial incentives and limitations on developer liability) and the willingness of the lending institution to cooperate on project financing, adequacy of infrastructure, the fraction of the site redeveloped as office or commercial use, and the fraction devoted to greenspace (the latter two are relative to the fraction redeveloped for residential use). 2.4 Three Studies using Data from Maryland Thayer, Albers and Rahmatian (1992) estimate the impacts of hazardous and non- hazardous waste sites on house prices in Baltimore from 1985 -1986. Results show a strong positive relationship between distance to hazardous waste site and price; prices increase by approximately 2% per mile further from the site. This positive relationship seems to level off with increased distance, but remains for at least four miles away from the site. They also found a significant positive relationship between air quality and price; a 6% increase in air quality led to an approximate 4% increase in price. 10 ------- Howland (2000) focuses on parcels in an industrial area of Baltimore, finding that contamination reduces the sale price, but does not slow down transactions. Schoenbaum (2002) examines values, and vacancy and turnover rates for another industrial area in Baltimore, and reports no evidence of significant differences across brownfields and non- brownfield properties. In summary, there have been numerous studies on the effects of hazardous waste sites on surrounding residential property values. In contrast, based on the literature review above, there are few studies of the effects on residential property values from groundwater contamination and specifically from LUSTs. Research on LUSTs and surrounding residential property values have been confined to just one geographic area (Cuyahoga, Ohio), and are limited in reliability due to few sales in close proximity of a LUST site. Further, the analysis of the impact of environmental contamination on non- residential properties is relatively small and not well developed from a statistical standpoint. 3. DATA The hedonic analyses will focus on three counties in Maryland: Baltimore City, Baltimore County, and Frederick. First we give a description of the UST sites in these three counties and then provide details of the housing data. 11 ------- 3.1 UST Sites Description Data on the 640 "Remediation Cases" in the study area were obtained from the Maryland Department of Environment's (MDE) Oil Control Program. We focus on the 387 cases where a leak was discovered between 1996 and 2007. This corresponds to the period of available home sales data. Out of the 387 cases, 180 were in Baltimore County, 123 in Baltimore City County, and 84 in Frederick County. We exclude cases with invalid coordinates, cases that are simply a residential location with a contaminated groundwater investigation and not linked to a specific LUST, when the 'leaking' event was minimal and resulted in nothing that could conceivably affect house prices, and if contamination was the result of something other than a leaking tank. This leaves 219 cases: 110 in Baltimore County, 66 in Baltimore City County, and 43 in Frederick County. Figures 1 and 2 display the LUST sites in the three counties. Table 1 shows the breakdown of case openings and closings by year. A case is open when an investigation regarding a potential leak is warranted, which may occur for several reasons, including: odor or water taste complaints from nearby residents, issues regarding routine onsite groundwater testing or UST system compliance checks, discrepancies in product inventory records, and if an UST owner reports an issue. Once a case is opened MDE investigates the situation and determines the best course of action, which may or may not include active cleanup. Petroleum products naturally degrade over time, so if there is no public or environmental threat, then ongoing monitoring and natural attenuation is sometimes deemed the best course of action (US EPA, 2004; Khan et al., 2004). 12 ------- A case is closed when MDE is satisfied that there is no contamination, or there may be contamination but no exposure, or, if undertaken, cleanup is well underway or complete. Overall a case is closed once the LUST is no longer considered an environmental or health threat. Of these 219 sites, 149 were closed by 2008. A few sites were open and closed on the same day. It is likely that this may happen when the results of a relatively small investigation that turned up little to worry about are entered (date open) at the same time when MDE enters their conclusion (little to worry about; date closed). Some of these cases are merely investigations in response to a complaint MDE receives. When the inspector gets to the site they may find nothing and just close the case right away. This seems to happen often with vapor investigations. Also, surface spill cases are sometimes minor and cleaned up right away with kitty litter, so these cases are usually closed right away also.4 Considering the 149 cases that were closed by 2008, the average leak case was open for 1.53 years, the median is 0.57 years, and the maximum is just under 10.5 years. Regarding the leak cases that remained open as of 2008, the average case is open for 3.10 years (the median duration is 4.68 years). There is information on relative risk categories (1-4; 1 is riskiest) but these do not appear to provide relevant information about the health risks associated with each LUST site. Instead, we use information on groundwater testing for petroleum concentration. We use these data because groundwater is the primary exposure path of concern and testing is done much more often than vapor and soil testing. We focus on concentration values for BTEX; the summation of benzene, toluene, ethylbenzene, and xylene. This 4 We have information on cleanup dates but they are reported only semi-annually. Therefore, we do not use this information in this analysis. 13 ------- aggregate measure of pollution is commonly reported, though only the individual components (benzene, toluene, ethylbenzene, and xylene) are regulated. The variable we use is btexmax; the maximum of the btex summation at any single time and testing location, including both on and offsite testing associated with a case. Testing is only carried out at 148 of the 219 LUST sites so we include a testing indicator in the hedonic model.5 The mean and median values for btex max are 17,818.82 and 280.75, respectively so the distribution is severely skewed right (concentrations are in micrograms/liter, which is equivalent to ppb). There are 24 LUSTS where the btex max concentration is zero. 3.2 Sales Data The data come from the MDProperty View CAMA (Computer Assisted Mass Appraisal) Database. This database is created on a yearly basis using data obtained from the State Department of Assessments and Taxation (SDAT). We have data from the 1996 - 2007 editions of this database. Each year provides information on the most recent sale for each unit in Frederick County, Baltimore County, and Baltimore City County so our dataset includes all sales between 1996 and 2007. Although much of the stock of housing in Baltimore consists of townhomes (attached and semi-attached homes) and condominium apartments, we will restrict attention to single-family homes. We do so for the sake of comparability with Baltimore 5 There does not appear to be an explicit testing criterion. Still, testing is more common at sites where there is a potential exposure path (groundwater being used) and if individual homes are nearby that could potentially be exposed. Further, the severity of the LUST event is also a factor in determining whether testing takes place. 14 ------- and Frederick Counties, where single family homes are prevalent, and with previous hedonic studies, which have largely focused on single-family homes. For each home, we have the exact address, latitude and longitude, and the names and the address of the owner. The latter information can be used to determine whether a home is owner-occupied. We also have the size of the lot, the square footage of the home, the age of the home, the quality of the structure (fair, average, good, very good), the type of heating and whether air conditioning is present, the number of bedrooms, the number of baths, the number and type of fireplaces, the presence, type and size of a porch, the presence and size of a garage, and the type of construction (e.g., brick, stucco). We have a general description of the dwelling (e.g., "1 story with basement") but we do not know the style of the home (e.g., Cape Cod, Federal style, etc.). Because we have the coordinates of most homes, we also know which census tract and block group these homes fall in. There are sufficient sales to allow us to include block group fixed effects in Baltimore and Frederick Counties and census tract fixed effects in Baltimore City County. These fixed effects allow us to control for all local amenities and disamenities that are common to all units in the block group (or census tract) and are constant over the time period of our analysis; 1996-2007. We believe that local public goods such as school quality and safety are constant over this time period so we do not have to include these variables in our model. We also do not include accessibility (in terms of distances) to the city center, downtown Washington DC and downtown Baltimore (employment centers), and tunnels (Harbor and Fort McHenry Tunnels) since these are essentially constant within block groups and census tracts. We do include distances to local amenities such as lakes, open spaces, commercial districts, 15 ------- and major roads. We also have calculated the number of UST facilities (leaking or not) within a 500 meter radius of each housing unit. We also know whether each house is within the public water service area, or outside this area and presumably reliant on private groundwater wells. For Baltimore City County, this is not an issue because all homes are served by city water. Units were excluded if lot size is greater than 10 acres (or recorded as zero), if the house was built prior to 1800, or was larger than 8,000 (enclosed) square feet. Units were also excluded if there were zero full baths or more than ten full baths and if ten half baths were recorded. Sales that were not arms length and prices that were less than twenty thousand dollars or more than five million dollars were dropped. Finally, we exclude cases with missing geographic coordinates. The final dataset includes 35,552 sales from Frederick County, 76,968 sales from Baltimore County, and 24,296 sales from Baltimore City County. Summary statistics for these three jurisdictions are given in Table 2. 4. MODEL DEVELOPMENT We now develop the framework for using the hedonic method (as applied to property values) to calculate the benefits from the cleanup of a nearby LUST site. For this analysis, we focus on measuring the benefits that accrue to residential units, though this can easily be generalized to include commercial and industrial sites. Assume that the price for house i in block group g at time t (Pigt) is a log-linear function of house characteristics (Hlt), neighborhood characteristics (Nigt), and a LUST site (LUST). Given the prevalence of LUST sites, we allow for the possibility that price can be affected by 16 ------- + p + p )+v- j= p p p 0 ------- 0 0 0 0 0 ------- ------- =p +p +p +p +p ------- =p +p +p +p +p +p +p +p +p +p +p +p + + ------- ------- = p +p +p +p +p +p +p +p +p +p +p +p +p + + +(- )• ------- '( (P "P )- ) •( (P "P )" ) •( (P -P )- ) ------- (p +p -p -p )-) (p +p -p -p )-) (p +p -p -p )-) (p +p +p -p -p -p )-) (p +p +p -p -p -p )-) (p +p +p -p -p -p )-) ------- ------- ------- ------- ------- ------- ------- ------- ------- ------- ------- ------- ------- ------- ------- Policy Association of American Geographers Literature Appraisal Journal Ecological Economics Environment and Planning A Journal of Real Estate Research Contemporary Economic Annals of the Journal of Real Estate The ------- Journal of the American Planning Association Journal of Environmental Economics and Management Journal of Real Estate Literature Journal of Real Estate Research Journal of Environmental Management Journal of Urban Economics The Journal of Real Estate Finance and Economics Journal of Urban Planning and Development Journal of Environmental Planning and Management, ------- Journal of Urban Economics Journal of the American Planning Association, The Appraisal Journal, Journal of Political Economy, Land Economics Economic Development Quarterly Turning Brownfields into Greenbacks The Appraisal Journal ------- Journal of Real Estate Research The Appraisal Journal Estate Research The Journal of Real ------- Table 1 Dates of Opened and Closed LUST sites Year Opened Closed 1996 21 6 1997 12 11 1998 15 8 1999 14 9 2000 12 11 2001 13 10 2002 11 4 2003 22 13 2004 25 19 2005 41 32 2006 25 19 2007 8 16 2008 0 4 Total 219 149 ------- Table 2 Summary Statistics for Housing Data Mean Std Dev Minimum Maximum Variable Baltimore City Count; / Nominal House Price (in $1,000s) 158.304 147.815 20.06 2520 Real House Price (in $1,000s, base is 2000) 147.037 133.042 17.441 2306.848 Lot Size (Acres) 0.205 0.148 0.003 5.280 Living Area (1000's of sq ft) 1.719 0.779 0.104 7.911 Age of House 71.186 20.389 0 206 Number of Full Bathrooms 1.579 0.800 1 10 Number of Half Bathrooms 0.294 0.510 0 5 1 if split foyer 2 levels of living area 0.006 0.080 0 1 1 if split level 3 or more levels of living area 0.007 0.086 0 1 1 if Attic or Attached Garage 0.080 0.271 0 1 1 if dwelling grade is low cost, economy, or fair 0.764 0.424 0 1 Nearest open space in 1,000s meters 0.457 0.291 0 1.454 Nearest surface water body in 1,000s meters 2.263 1.233 0.027 5.592 Nearest major road in 1,000s meters 2.524 1.111 0.017 5.132 Nearest commercial zone in 1,000s meters 0.368 0.250 0 1.281 Number of registered tanks within 500 meters 2.537 2.771 0 21 Baltimore County Nominal House Price (in $1,000s) 241.483 182.734 22.575 3300 Real House Price (in $1,000s, base is 2000) 226.677 161.197 20.280 2740.689 Lot Size (Acres) 0.512 0.874 0.002 10 Living Area (1000's of sq ft) 1.789 0.852 0 7.976 Age of House 38.217 26.056 0 206 Number of Full Bathrooms 1.711 0.738 1 8 Number of Half Bathrooms 0.550 0.549 0 5 1 if split foyer 2 levels of living area 0.068 0.251 0 1 1 if split level 3 or more levels of living area 0.093 0.290 0 1 1 if Attic or Attached Garage 0.401 0.490 0 1 1 if dwelling grade is low cost, economy, or fair 0.340 0.474 0 1 Nearest open space in 1,000s meters 0.540 0.597 0 7.296 Nearest surface water body in 1,000s meters 2.470 1.669 0 14.656 Nearest major road in 1,000s meters 1.949 1.772 0.001 12.139 Nearest commercial land use in 1,000s meters 0.663 0.676 0 6.775 Number of registered tanks within 500 meters 1.174 2.013 0 18 Frederick County Nominal House Price (in $1,000s) 270.828 143.943 25 2901.8 Real House Price (in $1,000s, base is 2000) 258.237 120.710 26.411 2901.8 Lot Size (Acres) 0.700 1.129 0.016 10 Living Area (1000's of sq ft) 1.997 0.801 0.348 7.929 Age of House 20.698 27.249 0 207 Number of Full Bathrooms 1.962 0.661 1 7 Number of Half Bathrooms 0.644 0.514 0 5 1 if split foyer 2 levels of living area 0.078 0.269 0 1 1 if split level 3 or more levels of living area 0.053 0.224 0 1 1 if Attic or Attached Garage 0.463 0.499 0 1 ------- Table 2 Summary Statistics for Housing Data Mean Std Dev Minimum Maximum 1 if dwelling grade is low cost, economy, or fair 0.098 0.297 0 1 Nearest open space in 1,000s meters 1.700 1.660 0 10.744 Nearest surface water body in 1,000s meters 3.977 2.330 0 12.664 Nearest major road in 1,000s meters 2.545 2.450 0.004 17.760 Nearest commercial zone in 1,000s meters 0.947 0.987 0 9.697 Number of registered tanks within 500 meters 0.644 1.772 0 16 ------- Table 3 Buffer Counts public water non-public water Buffer all tested cont>0 all tested cont>0 all tested cont>0 Frederick and Baltimore Counties PRE 100 155 126 111 138 109 98 17 17 13 OPEN 100 76 72 70 60 56 55 16 16 15 CLOSED 100 77 27 22 74 26 22 3 1 0 PRE 200 720 512 464 634 426 395 86 86 69 OPEN 200 308 262 255 264 218 215 44 44 40 CLOSED 200 421 152 141 402 144 137 19 8 4 PRE 100 200 573 392 359 504 323 303 69 69 56 OPEN 100 200 233 191 186 204 162 160 29 29 26 CLOSED 100 200 344 125 119 328 118 115 16 7 4 PRE 200 500 4190 3372 3038 3724 2926 2681 466 446 357 OPEN 200 500 1696 1380 1359 1549 1233 1224 147 147 135 CLOSED 200 500 2424 1225 1122 2303 1165 1071 121 60 51 Baltimore City County PRE 100 34 26 26 OPEN 100 11 9 9 CLOSED 100 32 2 2 PRE 200 179 122 122 OPEN 200 76 57 57 CLOSED 200 291 39 39 PRE 100 200 145 96 96 OPEN 100 200 65 48 48 CLOSED 100 200 260 37 37 PRE 200 500 1245 938 935 OPEN 200 500 457 302 302 CLOSED 200 500 2054 538 532 ------- Table 4 Results for Model 1 Bait/Fred Counties Bait City (1) (2) (3) Variable/Impact 100 Meter Buffer PRE -0.072* OPEN -0.026 CLOSED -0.037 p-value for joint sig 0.067 200 Meter Buffer PRE -0.029** 0.048 OPEN -0.015 0.024 CLOSED -0.024 -0.056 p-value for joint sig 0.025 0.239 100-200 Meter Buffer PRE -0.019 OPEN -0.012 CLOSED -0.021 p-value for joint sig 0.159 200-500 Meter Buffer PRE 0.001 0.001 0.028 OPEN 0.016 0.016 0.016 CLOSED -0.008 -0.008 -0.004 p-value for joint sig 0.243 0.239 0.518 Percent Impacts for 100 meter buffer OPEN PRE 4.713 CLOSED PRE 3.613 CLOSED OPEN -1.051 Percent Impacts for 200 meter buffer OPEN PRE 1.396 -2.378 CLOSED PRE 0.514 -9.836** CLOSED OPEN -0.870 -7.639 Percent Impacts for 100-200 meter buffer OPEN PRE 0.704 CLOSED PRE -0.250 CLOSED OPEN -0.947 Percent Impacts for 200-500 meter buffer OPEN PRE 1.523* 1.523* -1.126 CLOSED PRE -0.838 -0.858 -3.112 CLOSED OPEN -2.326 -2.346* -2.009 Observations 112502 112502 24296 Number of bg/tract 602 602 128 Adj R-squared 0.788 0.787 0.442 SER 0.205 0.205 0.410 ** p<0.01, * p<0.05 ------- Table 5 Regression Results for Model 2 Baltimore/Fred Counties Baltimore City County 200 200-500 200 200-500 Variable/Impact Without Testing PRE -0.011 -0.005 -0.010 0.051* OPEN -0.026 0.054 -0.049 -0.063 CLOSED -0.019 0.005 -0.065* 0.004 p-values for joint sig 0.706 0.119 0.284 0.198 With Testinc , Contamination = 0 PRE -0.046*** -0.002 0.133*** 0.008 OPEN -0.004 0.013 0.006 0.044 CLOSED -0.034 -0.016 0.125 -0.046 p-values for joint sig 0.032 0.208 0.000 0.426 With Testing, Contamination = 10,000 (around 90th pctile) PRE -0.043*** -0.001 0.036 0.040 OPEN -0.007 0.012 0.110 0.088*** CLOSED -0.029 -0.018 0.020 -0.024 p-values for joint sig 0.033 0.197 0.398 0.006 Percent Impact, Without Testing OPEN - PRE -1.446 6.050** -3.913 -10.777* CLOSED - PRE -0.784 0.951 -5.386 -4.675* CLOSED-OPEN 0.672 -4.808 -1.534 6.839 Percent Impact, With Testing, Contamination = 0 OPEN - PRE 4.252* 1.586 -11.896*** 3.721 CLOSED - PRE 1.191 -1.376 -0.800 -5.231 CLOSED-OPEN -2.936 -2.915 12.594 -8.631 Pet Impact, With Testing, Contam = 10,000 (around 90th pctile) OPEN - PRE 3.652* 1.262 -7.454 4.910 CLOSED - PRE 1.366 -1.735 -1.553 -6.288* CLOSED-OPEN -2.206 -2.960** 6.376 -10.674** *** p<0.01, ** p<0.05, * p<0.10 ------- Table 6 Regression Results for Model 6 Baltimore and Frederick Counties Public Water Source Non-Public Water Source 200 200-500 200 200-500 Variable/Impact Wthout Testing PRE -0.010 -0.002 OPEN -0.025 0.055* CLOSED -0.018 0.004 -0.082 0.019 p-values for joint sig 0.752 0.176 With Testing, Contamination = 0 PRE -0.042** -0.009 -0.046* 0.040** OPEN 0.004 0.017 -0.018 0.002 CLOSED -0.036 -0.016 -0.065 -0.044** p-values for joint sig 0.132 0.101 0.340 0.003 With Testing, Contamination = 10,000 (around 90th pctile)+ PRE -0.044** -0.010 -0.046* 0.040** OPEN -0.002 0.013 -0.018 0.002 CLOSED -0.033 -0.019 -0.060 0.294** p-values for joint sig 0.085 0.077 0.000 0.000 Percent Impact, Wthout Testing OPEN - PRE -1.504 5.866** CLOSED - PRE -0.830 0.651 CLOSED-OPEN 0.684 -4.926 Percent Impact, With Testing, Contamination = 0 OPEN - PRE 4.771* -3.067 2.910 -3.692 CLOSED - PRE 0.578 -1.340 -1.811 8.015*** CLOSED-OPEN -4.002 1.782 -4.588 -4.488 Pet Impact, With Testing, Contam = 10,000 (around 90th pctile)+ OPEN - PRE 4.271* -0.322 2.906 -3.692 CLOSED - PRE 1.084 -0.230 -1.407 7.704*** CLOSED-OPEN -3.057 0.092 -4.191 -4.166 *** p<0.01, ** p<0.05, * p<0.10 + The contamination level at which the LUST impact is calculated is 1,000 for the private water source since this is the highest contamination level recorded for these cases ------- Table 7 Regression Results Dependent Variable: Number of Transaction Rate Baltimore/Fred Counties Baltimore City County 200 200-500 200 200-500 Variable/Impact Without Testing PRE 0.253*** 0.245*** 0.259* 0.366*** OPEN 0.116 0.513*** -0.133 0.536*** CLOSED 0.279** 0.463*** 0.362*** 0.648*** p-values for joint sig 0.001 0.000 0.000 0.000 With Testinc , Contamination = 0 PRE 0.430*** 0.384*** 0.422*** 0.120*** OPEN 0.389*** 0.434*** 0.267 0.135*** CLOSED 0.827*** 0.498*** 0.321 0.267** p-values for joint sig 0.000 0.000 0.000 0.000 With Testing, Contamination = 10,000 (around 90th pctile) PRE 0.402*** 0.371*** 0.571*** 0.496*** OPEN 0.371*** 0.424*** 0.396** 0.221 CLOSED 0.796*** 0.465*** 0.319* 0.552*** p-values for joint sig 0.000 0.000 0.000 0.000 Partial Elasticity, Without Testing OPEN - PRE -6.825 13.329** -15.436 6.705 CLOSED - PRE 1.264 10.850** 4.074 11.119** CLOSED-OPEN 8.089 -2.479 19.510* 4.415 Partial Elasticity, With Testing, Contamination = 0 OPEN - PRE -2.011 2.464 -6.087 5.592 CLOSED - PRE 19.746* 5.675* -3.986 8.880 CLOSED-OPEN 21.757 3.211 2.102 10.952 Partial Elasticity With Testing, Contam = 10,000 (90th pctile) OPEN - PRE -1.529 2.621 -6.916 -10.838* CLOSED - PRE 19.576** 4.662 -9.913 2.211 CLOSED-OPEN 21.106* 2.041 -2.997 13.050 ***p<0.01,** p<0.05 * p<0.10 ------- Table 8 Publicized Cases in Baltimore and Frederick Counties Case no Spill Location City Date Opened # of sales before/after opening 96-2047FR GRESHAM STORE/FLINTHILL GROCY ADAMSTOWN 30-Sep-96 97-0257FR HAHN TRANSPORT NEW MARKET 12-Aug-96 47/321* 97-0646FR BARNES STORE FREDERICK 8-Oct-96 3/38 00-0575FR CARL CLINGAN LIBERTYTOWN 27-Sep-99 20/37 00-1125FR SHELL MT. AIRY 28-Dec-99 16/18 00-1183FR FARMERS & MECHANICS BANK UNION BRIDGE 11-Jan-00 54/42* 00-1301FR GREEN VALLEY GARAGE MONROVIA 9-Feb-00 100/102* 00-1332FR 7-ELEVEN STORE 28961 LIBERTYTOWN 15-Feb-00 21/36 03-1335BA2 FORMER STEBBINS BURNHAM OWINGS MILLS 10-Mar-03 59/36* 03-1758FR SHEETZ STORE #176 KNOXVILLE 7-May-03 43/17 04-2121BA4 CROWN MD-81 JOPPA 23-Jun-04 4/2 05-0326 BA2 AMOCO STATION #3033 PHOENIX 9-Sep-04 127/38* 05-0522BA3 CHEVRON/EXXON STATION HEREFORD 25-Oct-04 21/5 05-0834FR GREEN VALLEY CITGO MONROVIA 19-Jan-05 214/47* 05-0856BA2 JACKSONVILLE CITGO JACKSONVILLE 24-Jan-05 129/33* 06-0239FR MT. PLEASANT CITGO FREDERICK 21-Sep-05 49/13 06-0245FR EXXON #26463 FREDERICK 22-Sep-05 116/9 06-0303BA2 EXXON SERVICE STA 2-8077 PHOENIX 6-Oct-05 141/23* 06-0317FR CIFCO #1 6/10 GAS MART CLARKSBURG 13-Oct-05 40/9 06-0675FR JEFFERSON BP JEFFERSON 9-Feb-06 239/14 06-0825BA2 FORK CITGO #23 KINGSVILLE 31-Mar-06 46/11 06-0826BA2 MARYLAND LINE GARAGE MD LINE 31-Mar-06 23/4 07-0593FR GAS MART OF FREDERICK FREDERICK 16-Feb-07 2/0 Note: * - chosen for individual analysis ------- Table 9 Results for Publicized LUST Sites; Baltimore and Frederick Counties 500 Meter Buffer Variable/Impact PRE 0.020 OPEN 0.000 p-values for joint sig 0.425 OPEN - PRE -1.935 500-1000 Meter Buffer PRE 0.013 OPEN -0.041** p-values for joint sig 0.067 OPEN - PRE -5.257** 1000 Meter Buffer PRE 0.015 0.015 OPEN 1 0.035** OPEN 2 -0.005 OPEN 3G -0.025* OPEN 13 0.058** OPEN 46 0.026 OPEN G6 -0.070** p-values for joint sig 0.000 0.000 Percent Impacts OPEN 1 - PRE 2.037 OPEN 2-PRE -1.900 OPEN 3-PRE -3.872** OPEN 2-OPEN 1 -3.858*** OPEN 3G - OPEN 1 -5.791*** OPEN 3G - OPEN 2 -2.010 OPEN 13-PRE 4.467** OPEN 46-PRE 1.100 OPEN G6 - PRE -8.151*** OPEN 46-OPEN 13 -3.224* OPEN G6 - OPEN 13 -12.079*** OPEN G6 - OPEN 46 -9.150*** Observations 112502 112502 112502 Number of bg/tract 602 602 602 Adj R-squared 0.787 0.788 0.788 SER 0.205 0.205 0.205 *** p<0.01, ** p<0.05, * p<0.10 ------- Table 10 Results for Individual Publicized LUST Sites; Baltimore and Frederick Counties 1000 Meter Buffer 1000-2000 Meter Buffer LUST SITES PRE OPEN IMPACT PRE OPEN IMPACT (1) (2) (3) (4) (5) (6) Jacksonville Exxon 0.087 -0.045 -12.358** 0.003 -0.033 -3.537 (0.076) (0.038) (0.022) (0.023) Green Valley -0.023 0.036** 6.015** Garage/CITGO (0.018) (0.011) Hahn Transport -0.173** -0.164** 0.903 (0.027) (0.026) Farmers & Mechanics -0.046** -0.102** -5.438** Bank (0.010) (0.015) Former Stebbins -0.062 0.070 14.180** Burnham (0.167) (0.154) Observations 112502 Adj R-squared 0.788 SER 0.205 Robust standard errors in parentheses ** p<0.01, * p<0.05 ------- Baltimore City & County area 0 3,7507,500 15,000 22,500 30,000 I Yards Major Roads Public water service Subset of LUSTs for Hedonics Original 640 LUST Cases ------- Frederick County 0 4,100 8,200 A ~ 16,400 24,600 Major Roads Public water service area Subset of LUSTs for Hedonics Original 640 LUST Cases 32,800 l Yards ------- |