The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this...

22
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=teep20 Download by: [67.0.192.32] Date: 13 November 2017, At: 05:40 Journal of Environmental Economics and Policy ISSN: 2160-6544 (Print) 2160-6552 (Online) Journal homepage: http://www.tandfonline.com/loi/teep20 The case of the missing negative externality? Housing market effects of fracking in the Niobrara shale play, Colorado Xuanhao He, Na Lu & Robert P. Berrens To cite this article: Xuanhao He, Na Lu & Robert P. Berrens (2017): The case of the missing negative externality? Housing market effects of fracking in the Niobrara shale play, Colorado, Journal of Environmental Economics and Policy To link to this article: http://dx.doi.org/10.1080/21606544.2017.1398683 Published online: 13 Nov 2017. Submit your article to this journal View related articles View Crossmark data

Transcript of The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this...

Page 1: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=teep20

Download by: [67.0.192.32] Date: 13 November 2017, At: 05:40

Journal of Environmental Economics and Policy

ISSN: 2160-6544 (Print) 2160-6552 (Online) Journal homepage: http://www.tandfonline.com/loi/teep20

The case of the missing negative externality?Housing market effects of fracking in the Niobrarashale play, Colorado

Xuanhao He, Na Lu & Robert P. Berrens

To cite this article: Xuanhao He, Na Lu & Robert P. Berrens (2017): The case of the missingnegative externality? Housing market effects of fracking in the Niobrara shale play, Colorado,Journal of Environmental Economics and Policy

To link to this article: http://dx.doi.org/10.1080/21606544.2017.1398683

Published online: 13 Nov 2017.

Submit your article to this journal

View related articles

View Crossmark data

Page 2: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

The case of the missing negative externality? Housing marketeffects of fracking in the Niobrara shale play, Colorado

Xuanhao He, Na Lu and Robert P. Berrens

Department of Economics, University of New Mexico, Albuquerque, NM, USA

ARTICLE HISTORYReceived 26 April 2017Accepted 26 October 2017

ABSTRACTRecent rapid growth of shale gas exploration in the state of Colorado (CO)and elsewhere in the United States has caused considerable publicconcern over potential environmental costs to local communities,proximal to the location of energy development. In Weld County, CO,shale gas exploration has grown substantially since 2013. Both populationand new construction of houses also increased significantly after 2012.Combined, this increased the potential for negative externalities. Theobjective of the analysis is to apply the hedonic pricing method, usingsingle-family residential data from October 2014 to March 2017 and atemporal-spatial identification strategy, to estimate the environmentalcost of shale gas exploration on nearby house prices in Weld County, CO.However, results from spatial econometric models provide no evidence ofsignificant environmental impacts on housing values. Our policydiscussion explores a possible Coasian bargaining solution as the sourcefor this case of a missing negative externality. The energy and housingmarkets appear to be internalising externalities, where side paymentsfrom energy developers to homeowners are enough to compensate forany environmental impacts to housing.

KEYWORDSShale gas; hedonic pricing;weld county; bonuspayments; coase

1. Introduction

Shale gas is the most rapidly growing form of fossil fuel-based energy in the United States (US)(DOE 2013). The widespread introduction of new techniques, such as horizontal drilling and multi-stage hydraulic fracturing, has allowed shale gas to be explored on a much larger scale. Starting in2008, such development initially boomed in the Marcellus Shale Play in Pennsylvania (PA) andthen further expanded to 16 states in the US. By 2012, the US had the largest shale gas productionin the world, with 10,371 billion cubic feet (EIA 2013). Colorado (CO) is one of the US states withabundant shale gas reserves, and its production levels increased substantially from 18 billion cubicfeet in 2013 to 236 billion cubic feet in 2014 (EIA DATA1). In northeastern Colorado, Weld Countyis located within the Niobrara Shale Play and is home to the bulk of the shale gas permits issued(1789 in 2015). In 2015, 97% of the well permits approved in Weld County were shale gas wellpermits (COGCC 2016).

In terms of benefits, shale gas development can create additional jobs in the sector, and increasetax, lease, and royalty receipts for governments (Feyrer, Mansur, and Sacerdote 2017; Hartley et al.2015; Newell and Raimi 2015). It can also generate royalty and lease payments to private propertyowners (Weber, Brown, and Pender 2013; Hardy and Kelsey 2015; Brown, Fitzgerald, and Weber2016). However, shale gas extraction can bring environmental costs, including air pollution, possible

CONTACT Xuanhao He [email protected]

© 2017 Journal of Environmental Economics and Policy Ltd

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY, 2017https://doi.org/10.1080/21606544.2017.1398683

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 3: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

water contamination, increased truck traffic, noises and other dis-amenities (Olmstead et al. 2013;Adgate, Goldstein, and McKenzie 2014; Vengosh et al. 2014; Brown, Lewis, and Weinberger 2015).As local policy-makers and voters grapple with these potential trade-offs, several local, county-levelbans or moratoriums on fracking in Colorado have been debated (e.g. see review in Bennett andLoomis 2015).

As part of larger economic considerations, recent hedonic pricing method (HPM) studies havebegun to assess the environmental risks to proximal housing values from shale gas activities. Theseinclude HPM analyses focusing on the Marcellus Shale Play (Delgado, Guilfoos, and Boslett 2014;Gopalakrishnan and Klaiber 2014; Muehlenbachs, Spiller, and Timmins 2015), as well as for theNiobrara Shale Play, especially in the Denver basin in Colorado on the eastern front of the Rockies.Therein, Weld County, CO is a focal point of shale gas development, and several prior HPM studiesusing housing value data from 2012 and earlier have found small negative impacts (Bennett andLoomis 2015; James and James 2014). However, shale gas exploration boomed in Colorado after2012. Further, residential areas in the eastern portion of the Denver Basin in the Niobrara ShalePlay have developed substantially at the same time; that is, both new house construction and popula-tion increased substantially since 2012. In combination, all these point towards the much strongerpotential for negative externalities from shale gas activities.

With this background as motivation, the objective of the analysis is to apply the hedonic pricingmethod (HPM), using single-family residential data from 20 October 2014 to 1 March 2017 with spa-tial effects, to estimate the environmental cost of shale exploration activities on nearby housing pricesin Weld County, CO. Following Muehlenbachs, Spiller, and Timmins (2015) and Gopalakrishnan andKlaiber (2014), we utilise both spatial and temporal variations in shale activities, including potentiallyoil and gas companies paying bonus upon mineral right owners signing a lease contract and on-sitewell preparation after permitting, in our identification strategy. Surprisingly, results provide clear evi-dence that no negative impacts from these shale activities (as proxied by well permit approval) arebeing capitalised into the housing market. The impact of shale activities is insignificant regardless ofutilised spatial and temporal windows, using both full and subsample analyses. This case of a missingnegative externality is also robust to model specifications that include a variable measuring monthlyemployment counts in the oil and gas extraction sector in Weld County, CO.

We discuss a Coasian bargaining solution as the possible explanation for this specific case of anapparent ‘missing externality’. The ancillary evidence is consistent with homeowners receivingbonus payments or expecting to receive lease payments from oil and gas companies. In a kind ofinduced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such that side payments from polluters are more than enough to offset per-ceived environmental costs, including air pollution, noise, traffic congestion, and other related dis-amenities. The caution is that the absence of a negative environmental externality being capitalisedinto this case study housing market neither implies that this result will hold in other housing mar-kets. (e.g. which may be more dependent on groundwater) nor the absence of any environmentaleffects (e.g. where market perceptions may not match with actual environmental risks).

2. Background

2.1. Shale gas overview

Shale gas is defined as the natural gas found in shale formations (DOE 2013). Known concentrationsof shale gas remained undeveloped for decades, due to economic and technological limitations. Itwas not until the early 2000s that the newest techniques, such as multi-stage hydraulic fracturingcombined with horizontal drilling, made it economically feasible to extract natural gas from thesepreviously inaccessible geologic formations. Hydraulic fracturing (commonly known as fracking)involves the injection of high-pressure water mixed with sand and chemicals into horizontallydrilled wells to crack layers of rock and release trapped gas.

2 X. HE ET AL.

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 4: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

A shale gas boom started in 2008 in the Marcellus Shale Play in the north-eastern US. Nationally,this ongoing production boom is expected to continue through at least 2035, when contributingabout half of all US natural gas production (DOE 2013). Ceteris paribus, this shift in supply isexpected to exert downward pressure on the price of natural gas. Further, shale gas developmentalso contributes to employment, tax, lease and royalty receipts for state and federal governments.Direct benefits to local people come from royalty fees and lease payments to households who ownthe mineral rights underground. Because of the nature of horizontal drilling in shale gas production,benefits can be obtained if pipelines are lining across the areas underground of households’ lands,besides the case when shale gas wells are on their lands. It is estimated that local royalty income percapita on the Niobrara Shale Play in Colorado was $236 in 2014 (Brown, Fitzgerald, andWeber 2016).

2.2. Shale gas activities

According to Gopalakrishnan and Klaiber (2014), to extract shale gas from the subsurface area of aproperty, oil and gas extraction companies must gain access to the underground minerals, oftenthrough leasing mineral rights from the surface right owners, if mineral rights are not severed.Upon signing the lease, a bonus is paid to the lessor as an inducement for the contract, and a royaltyrate is agreed on as a portion of production revenue paid to the lessor (McMahon 2017). This con-tract allows the lessee to access part of the surface area for shale gas extraction. Colorado Oil & GasConservation Commission (COGCC) Rule 303 a requires that drilling companies must obtain per-mits from state governments for the well sites. After permit approval, on-site well preparationbegins. Most visible shale activities happen at this stage (e.g. tree and vegetation removal, well-pad,reserve pits and access road construction) (UA and Argonne 2017). Finally, drilling activities start.Based upon a well permit approval date, this analysis focuses on activities around permit approval,and on-site preparation for horizontal wells after permits are approved.

2.3. Environmental effects from shale gas activities

While it may generate significant economic development, the shale gas well preparation and extrac-tion process can also create significant concerns of negative external effects. We briefly review someof that literature here. Potential surface and groundwater contamination, potentially the most con-cerning environmental risk, happen after drilling activities begin. Lauer, Harkness, and Vengosh(2016) find that brine spills in North Dakota elevated concentrations of dissolved salts and othercontaminants in surface water. The treatment of shale gas waste or the presence of shale gaswells in a watershed may raise downstream chloride or total suspended solids (TSS) concentrations(Olmstead et al. 2013), and chloride, bromide, and iodide concentrations (Harkness et al. 2015).Since the flow-back fluid needs to be returned to the surface pit, the hazardous chemicals in thewastewater can lead to shallow groundwater contamination if not treated well. Also, injected waterin the producing wells could lead to deep groundwater contamination. For example, DiGiulio et al.(2011) find both contaminations above the Pavillion gas field in the Wind River Basin, Wyoming.Further, the impacts of water contamination to nearby residents can work through drinking-waterpollution. Osborn et al. (2011) found that average methane concentration was 17.45 times higher indrinking-water wells within a 1 km (0.62 miles) radius of a well head compared to those outside the1 km radius in northeastern Pennsylvania and upstate New York.

Air pollution comes from both traffic and well sites (Litovitz et al. 2013). It can be caused by theemissions of increased truck traffic and diesel-powered pumps (Considine Watson and Considine2011). During the preparation stage of wells, a large amount of truck trafficking happens near prop-erties (Considine Watson and Considine 2011). The estimated quantity of traffic necessary for wellcompletion is somewhere between 1,500 to over 2,000 truck trips (Hill 2013). It may also exist in thedrilling period. Both the number and concentration of non-methane hydrocarbons (NMHCs) are

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY 3

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 5: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

highest during the initial drilling phase (Colborn et al. 2014). Overall, the risk of benzene, as a majorcontributor to air pollution from unconventional natural gas extraction, is greater for most residentsliving within half-mile distance to wells, compared to those living outside this range (McKenzieet al. 2012). Greenhouse gases including benzene and methane increase low birth weight anddecrease term birth weight within 2.5 km (2.17 mile) of permitted wells (Hill 2013). Swarthout et al.(2015) find elevated concentrations of both methane and volatile organic compounds (VOCs)in areas with the highest density of shale gas wells in the southwestern Pennsylvania region of theMarcellus Shale. Selected polycyclic aromatic hydrocarbons (PAHs) have been found at a greatlyhigher concentration rates within 1.1 km (0.68 miles) of the well pad than those thresholds at whichprenatally exposed children in urban studies develop lower developmental and IQ scores (Colbornet al. 2014).

2.4. Shale gas development in weld County, CO

Colorado is one of the major natural gas producing states in the US, with natural gas the largest pro-duced energy in the state. Shale gas began to develop from 2009 in Colorado and boomed from2013. By 2014, 3775 billion cubic feet shale gas reserves had been proved compared to 136 billioncubic feet in 2013 (EIA DATA2). While there are several large shale plays (Niobrara, Piceance,Gothic, Pierre), production in the Niobrara has expanded most rapidly. Located over the NiobraraShale Play, Weld County issued 1,789 horizontal well permits in 2015 (89% of the total annual hori-zontal well permits issued in Colorado) (COGCC 2016).

3. The hedonic pricing method, and applications to energy development

HPM is a revealed preference technique used to value nonmarket effects, such as changes in envi-ronmental impacts. In HPM, variation in the price of a heterogeneous or differentiated good, suchas a house, is econometrically de-composed to reveal the effects of its attributes (Taylor 2003). Inequilibrium, the marginal implicit price of each attribute is equal to the consumers’ marginal will-ingness to pay for the attribute (Rosen 1974). Through the isolation of these marginal implicit pri-ces, HPM can be applied to measure the external effect (e.g. reduction in house price) of locallyundesirable facilities, e.g. the presence of power plants (Davis 2011), and oil and gas wells (Boxall,Chan, and McMillan 2005).

Residential property values within shale gas development areas may experience significant increasesor decreases, depending on the relative strength of the benefits and costs. Previous studies using HPMprovide mixed results. Some analyses find a significant negative effect of proximal shale wells on hous-ing prices (Gopalakrishnan and Klaiber 2014; James and James 2014; Bennett and Loomis 2015;Muehlenbachs, Spiller, and Timmins 2015). In contrast, Delgado, Guilfoos, and Boslett (2014) find nosignificant effects. In terms of study area, Delgado, Guilfoos, and Boslett (2014), Gopalakrishnan andKlaiber (2014), and Muehlenbachs, Spiller, and Timmins (2015) apply HPM to housing prices, andshale gas development in PA, in the Marcellus Shale Play, the largest shale play in the US and the ear-liest to be explored. Importantly for our analysis, Bennett and Loomis (2015) and James and James(2014) previously used HPM to examine shale activity effects, in the Niobrara Shale Play, on houseprices. Below, we briefly summarise some of these recent prior studies.

Delgado, Guilfoos, and Boslett (2014) focus on effects in Lycoming and Bradford counties, PA.Employing HPM with both parametric and semi-parametric approaches, they find little recent evi-dence of negative externalities from unconventional natural gas extraction; they do find some evi-dence that negative environmental effects may have been capitalised into property values in theearliest years of unconventional natural gas extraction.

Gopalakrishnan and Klaiber (2014) find evidence that shale gas exploration leads to a decline inhousing values in Washington County, PA. The decline is larger for houses relying on groundwaterwells. They also find that the impact of shale gas exploration on surrounding houses is largely short

4 X. HE ET AL.

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 6: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

term. It lasts mainly in the 6 months from the permit acquisition date and dies out by 12 monthslater. This is due to the majority of visible shale activities occurring within this time window. Also,they find that houses in proximity to major roads experience significant negative impacts due toheavy truck traffic and congestion.

Muehlenbachs, Spiller, and Timmins (2015) use HPM to identify several different effects onhouse prices connected to shale gas activities. An adjacency effect is measured by the number ofwells within eyesight. A groundwater contamination risk effect is measured using an indicator vari-able for houses that depend on groundwater wells instead of piped water. They find large negativeimpacts of shale gas exploration on groundwater-dependent houses with 1-1.5 km range. In addi-tion, they find small positive impacts for house using piped water adjacent to shale wells, whichmight be resulting from royalty payment and lease fees to property owners (Muehlenbachs, Spiller,and Timmins 2015, 19, 23).

As noted, the external cost of shale gas exploration in Weld County, CO has also been previouslystudied. Using a data set of 3557 single-family house sales in 2012, James and James (2014) find thatshale wells have a short- run negative effect on proximate house prices in Colorado. The linearHPM model predicts that housing value decreases by $15,000 (roughly 7% of average house price)as its distance to the nearest wells reduces by 1 kilometre.

Further, James and James (2014) discuss property rights for minerals in Colorado, which can beseparated from land rights (Radow 2011). If the mineral rights have not been severed from landrights, the landowner owns both rights to surface and subsurface areas. They are entitled to royaltiesfrom horizontal drilling that runs subsurface below their property if they lease out the mineralrights. James and James (2014) find that properties, with wells located on or intersected with thesubsurface area are sold at significantly higher prices, suggesting lease payments are capitalised intohousing values.

Bennett and Loomis (2015) find small negative effects of shale oil and gas wells on house prices inthe Weld County, CO. Using a dataset of single-family house sales for the period 2009–2012, sevenout of twelve econometric HPM models show that the number of wells has a one percentage pointsignificant negative effect on house prices within a half-mile range and during active drillingperiods in urban areas. This equates to a $1342 to $1936 reduction, in 2009 dollars in the sale price(0.6 – 0.9% of average house value).

Both prior HPM studies in Weld County are important precursors to our analysis, and appear todemonstrate some apparent negative effects of shale gas exploration on house prices. However, bothanalyses are based on data before the period of significant growth in population, residential homeconstruction, and shale gas development. Using more current data, a priori, we would expect to seesignificant growth in negative externalities on housing.

4. Data

Since 2013, Weld County CO has experienced the largest share of both annual horizontal drillingpermits (e.g. 1789 out of 2016 totals in 2015), and active horizontal wells (e.g. 4791 out of 5270 totalsas of 18 July 2016), of the state total in Colorado (COGCC 2016). Historically, Weld County hasbeen one of the richest agricultural counties in the US, east of the Rocky Mountains, because of itsleading role in producing cattle, grain and sugar beets. But more recently, it has seen significant pop-ulation and housing growth. We use data from Weld County to explore the effect of proximity toshale activity on house prices. The econometric specification requires data on housing transactions,well location and timing, as well as additional neighbourhood controls.

4.1. Housing transactions data

Weld County single-family residential house sales were taken from Weld County Assessor’s pub-licly-available data. Sale prices were inflation-adjusted to 2015 dollars. Matched housing

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY 5

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 7: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

characteristics include: built-up area size (SQFT); lot size (ACRE); number of baths (BATH); hous-ing structural binary variables indicating if a house has a garage (GARAGE), balcony (BALCONY),and finished basement (FINBASE); housing type binary variables indicating if a property is a 1-storyranch house (RSTORY1), a bi-level house (BILEVEL); housing exterior type binary variables indi-cating if a property is a frame house with hardboard exterior (HARDBD), vinyl exterior (VINYL),and masonry veneer exterior (MASONRY); age of property (AGE); and a binary variable indicatingif a property was sold in the same year as built (NEW). Houses with either missing values, zero bath-room or bedrooms were removed from the sample. Land sales were also dropped. To map the pre-cise locations of individual properties, we obtained GIS data from the Department of GIS for WeldCounty. In addition to housing characteristics, we formed a supplemental distance variable usingArcGIS: the inverse of the minimum distance from each house to either the nearest US highway orInterstate (INDISTRD). Census tract data were obtained from the Department of Local Affairs, CO.

We restricted housing data to the period from 20 October 2014 to 1 March 2017, based on ouridentification strategy. We collected data on well permits on beginning with 20 April 2016. COGCCmaintains the most recent two years’ permit data only. So, the earliest available permits wereapproved on 21 April 2014. The impact of shale activity on proximate housing values occurredmainly within six months after the permit approval date (Gopalakrishnan and Klaiber 2014). Basedon a 6-month time window and the permit approval dates, we limited housing data to houses sold6-months after first permitting date to fully capture the impact of shale activity on proximate houseprices. We selected the end date of housing data based on the availability of employment data in oiland gas sector (see below).

4.2. Shale gas activity data

Following Gopalakrishnan and Klaiber (2014), permit information is preferred to spud well infor-mation for HPM given significant visible shale gas activity occurring between permit approval dateand spud date. Further, wells may be important to consider, and thus preferred to use rather thanwell pads, since a single well pad may contain 8 or more horizontal wells due to horizontal drilling(Abdalla et al. 2012). The Colorado Department of Natural Resources and COGCC maintainrecords of all currently active well permits in the state, including pending permits and approved per-mits. As of 29 December 2016, there were 737 active, pending permits in Weld County, out of 1033in Colorado. The application reception dates of pending permits range from 25 February 2015 to 27July 2017. Once a pending permit is approved, it is valid for two years before expiration, upon whichthe permit can be refiled and approved for another two years.1 The permit data contains 7032 per-mits in Weld County (over the Niobrara Shale Play), approved between 21 April 2014 and 28 July2017. We were unable to obtain information on the length of time between permit date and spud(start to produce) date for a well.2 We assume there is a well to be constructed after the well permitis approved. Permits are therefore referred to hereafter as permitted wells.

Each permitted well was mapped into ArcGIS using its longitude and latitude. Different bufferzone distances of 0.5, 1 , 1.5, and 2 miles were created to examine the effect of permitted wells onthe housing prices across space. Moreover, time windows of 6 and 12 months between permittingdate and sales date were generated, allowing us to test the temporal variation of shale activityimpacts on housing prices.

4.3. Groundwater data

To capture the shale gas impact on housing prices through possible concerns of groundwater con-tamination, we download the map of Designated Basin and Ground Water Management Districtsfrom Colorado Ground Water Commission (CGWC) website and use it to identify the districts inwhich residential properties are located. Within Weld County, there are three designated groundwa-ter basins, in which households could primarily depend on groundwater as their main source of

6 X. HE ET AL.

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 8: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

water supply. They are Upper Crow Creek 8 in middle North Weld, Lost Creek 9 and North KiowaBijou 7 in Southeast Weld.

4.4. Employment data

We limited houses to those sold before 1 March 2017, 60 days after December 2016, the latestmonthly employment data in oil and gas extraction sector of Weld County, CO. This allowed collec-tion of employment data in the sector, and conducting HPM model specifications that controlled forpossible employment effects (e.g. Bennett and Loomis 2015). We collected monthly employmentcounts in the oil and gas extraction sector in Weld County from Quarterly Census of Employmentand Wages (QCEW), Bureau of Labor Statistics (BLS). This data is available through December2016. Following Bennett and Loomis (2015), we matched each month’s employment counts tohouses sold two months after.

5. Empirical strategy

5.1. Empirical model and hypothesis

The hedonic pricing method (HPM) assumes a representative consumer, whose choice over a differ-entiated good, such as a residential property, can be reflected in a standard utility maximisationproblem. The utility function is expressed in terms of a bundle of attributes capitalised into propertyvalues and a composite commodity representing all other goods:

Ukij ¼ U c;Xi; Eij;aj; d

k� �

(1)

Consumer k, with preferences, dk, chooses house i in neighbourhood j, based on characteristics ofthe property Xið Þ, location-specific characteristics aj

� �, environmental attributes Eij

� �, and a com-

posite numeraire commodity cð Þ. This consumer maximises her utility Ukij subject to a budget con-

straint and price of the house:

Pijt ¼ Pijt b;Xi; Eij;aj;’t

� �(2)

Assuming the housing market is in equilibrium and buyers are free to choose a property anywherein the market, the equilibrium price of the house provides information about the consumer’s will-ingness to pay for housing attributes. The partial derivative of the price function, also known as themarginal implicit price, with respect to any attribute (e.g. an environmental amenity or dis-amenity)equates to the marginal willingness to pay for that attribute, ceteris paribus.

To estimate the model econometrically, we use the following general functional form:

f Pijt� � ¼ b0 þ b1Xi þ b2aj � ’t þ gEij þ eijt (3)

Different transformations of price are used, including linear, semi-log, Box-Cox. Multicollinearitytests are carried out to select suitable control variables Xi. To control for both unobserved space andtime factors, we use census tract by year fixed effects (aj � ’t).

Further, we capture the impact of shale gas exploration (Eij) by defining total number of permit-ted wells within a specific distance and time window from each property and its sales date. Formally,shale gas impact gEij can be expressed as follows,

gEij ¼ gPERMITij (4)

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY 7

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 9: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

Since prior investigations provide mixed results, whether the sign of g is zero or not remains anempirical question. But in our Weld County case, there are prior HPM studies showing negativeeffects (albeit modest in size), and the subsequent growth in both shale gas development and resi-dential housing. Thus, against the null hypothesis of no effect, our alternative hypothesis is:

Ha: g < 0 (5)

That is, we expect to reject the null, and that the evidence will support Ha; as the number of permit-ted wells, within a specific spatial distance from the location of a residential property and time win-dow prior to its sales date, increases, the sales price will decrease. The magnitude of this decrease foran additional well would be the marginal implicit price or measure of the negative environmentalexternality.

5.2. Spatial autocorrelation of housing prices

We assume the error term eijt is independently and identically distributed within a same census tractand a same year in the hedonic price function 3ð Þ. However, this assumption fails when the price ofproperty i and that of property v are correlated with each other within census tract j. To account forspatial autocorrelation across housing prices because of adjacency effects, we follow Se Can andMegbolugbe (1997), and use an instrumental variable approach.3 This approach calculates theweighted average price of neighbouring properties (COMPARABLE) based on prior transactionsfor each property, formally represented as:

COMPARABLEi ¼Xn

v¼1wivPv ¼

Xn

v¼1

1divPnv¼1

1div

" #Pv (6)

where Pv is the vth neighbouring house price within a 1.8-mile distance from a subject property and6-months prior to the its sales date. Weight wiv is given to the vth neighboring house based on itsinverse distance to the ith target house. The weights are normalised to 1 for each subject property i.This approach accounts for both spatial and temporal autocorrelation.

5.3. Spatial and temporal variations of permitted wells

Following several recent studies (Gopalakrishnan and Klaiber 2014; Muehlenbachs, Spiller, andTimmins 2015), we include both buffer distance and time window to identify the permitted shalegas wells affecting the housing market. We explore buffer zones with radius of 0.5, 1, 1.5 and 2 milescentred at houses’ locations to examine the effect across space. According to Gopalakrishnan andKlaiber (2014), the impact of shale activities on housing price is concentrated in the 6 months afterpermit approval and dies out 12 months after that. We also create time windows of 6 and 12 monthsbetween permit approval dates and houses’ sales dates to explore the time periods in which shale gasdevelopment risk or benefit is apparent to nearby residents and prospective homebuyers. We countthe number of permitted wells both within a specific distance band of the location of a house andwithin a time window prior to the house’s sale date. This allows us to capture the variation of theimpact over time.

5.4. Subsample of properties around pending permits

If wells tended to locate in less desirable places, then comparing housing prices proximate to wellsand not would upward bias estimates of shale activity impacts (assuming the shale activity impactswere negative). Conversely, if wells tended to be located near more desirable amenities, then that

8 X. HE ET AL.

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 10: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

would downward bias the estimated coefficients on permits count. To account for the unobservedheterogeneity between properties near wells and not, either time-invariant (i.e. different landscapes)or time-varying (i.e. changed local economic activities), we restrict our sample to properties within2-miles of a pending permit for a subsample analysis. Since well sites within 2-miles of each propertyhad been selected for drilling in this subsample, the properties have more comparable locationalcharacteristics. In this subsample, all houses potentially have mineral reserves underground within a2-mile range. Houses proximate to approved permits may have well preparation activities nearby,whereas those with only a pending permit do not. In short, the subsample analysis provides betterestimates of the difference in the average housing values because of shale activities, using more com-parable houses within a 2-mile distance of shale gas reserve.

5.5. Employment of oil and gas extraction sector

Controlling for local employment effects is not commonly done in HPM studies of environmentalimpacts. However, for this fracking case, Bennett and Loomis (2015) proposed that rising jobs in oiland gas sector might increase the demand for housing in the local area, thus raising housing prices.After including employment of oil and gas sector of Weld County into their hedonic model, Bennettand Loomis (2015) found a small but positive impact of the employment in oil and gas sector onhouse prices in the whole county. If this was the case in our data, and if permits count and theemployment in oil and gas sector are positively correlated, then missing the control of the employ-ment effect would potentially bias upward the estimated coefficient on permits count.

We speculate that possible impacts from oil and gas sector employment, a noted boom and bustsector, may be more open with respect to the expected sign on residential housing market impacts.Bennett and Loomis (2015) reference Platt (2013) to support their argument for upward pressureon housing prices. Review of Platt (2013), however, allows that the impact of oil and gas sectoremployment on housing price remains an open empirical question. According to (Platt 2013), theinflux of migrant workers into an area is likely to increase local hotel and rent prices, which doesnot necessarily translate into increasing housing prices. Further, oil and gas sector workers may cre-ate negative impacts to nearby neighbourhoods in cases where field workers may live in camps, tem-porary field housing or even vehicles.

Whatever the possible direction of impact, we need to control for possible employment effects.Thus, as a robustness check, we include monthly employment counts for the oil and gas extractionsector into equation (3) to test any possible change to the estimated coefficient on well permitscount. If the result for our primary hypothesis of interest (5), the environmental impacts from frack-ing as proxied by well permits, is robust, then we would expect no change to the sign and signifi-cance of the estimated coefficient.

6. Results

6.1. Graphical evidence

Figure 1 depicts the location of properties, approved permits, and pending permits in Weld County,CO. Most transactions happen close to the city of Greeley, and the Southwest boundary of WeldCounty. While there are many approved permits located in the north and northeast of Weld County,where very few people live, a large number of permits are in close proximity to residential propertiesboth inside and outside the city of Greeley. Similarly, a high proportion of pending permits are prox-imate to residential properties. Further, by comparing Figure 1 to the map of Designated Basin andGround Water Management Districts from CGWC, we identified only 45 residential propertieslocate in designated groundwater basins. All of them are in Southeast Weld (in the town of Keenes-burg) near Interstate 76 and in the Lost Creek 9 designated groundwater area. This means 11,408

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY 9

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 11: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

(99.61%) of the sample are in public water providers’ coverage areas. This likely greatly reduces con-cerns in this area for groundwater contamination risks on housing values.

Figure 2 shows that the number of new spud wells increased gradually from 2008 to 2014 before itstarted to level out. The number of approved shale gas permits grew substantially in 2013, reached itshighest in 2014, and levelled out in 2015–2016 in Weld County, CO. This implies a significant increase

Figure 1. Map of house sales and permits in Weld County.

Figure 2. Building permits, shale gas permits, and newly spud wells in Weld County, CO.

10 X. HE ET AL.

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 12: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

of well preparation activities, in which most negative externalities might happen. Second, new construc-tion of single-family houses increased consistently until 2015, based on reported building permits.Third, Weld County population rose significantly from 264,117 in 2012 to 285,147 in 2015. Combined,this has only increased the potential for externalities from shale gas exploration inWeld County.

6.2. Descriptive statistics

Table 1 presents the mean, standard deviation, minimum and maximum values for each variableused in our analysis. The average house price is $290,892, with 1,735 square feet of built-up area,0.20 acres lot size, and 2.70 bathrooms. Ninety-seven percent of houses have a garage, nineteen per-cent have a balcony, and thirty-seven percent have a finished basement. Almost half (forty-eight per-cent) are one-story ranch houses, whereas only three-percent are bi-level houses. In terms of frameexterior, forty-three percent of houses use hardboard, and five percent each use vinyl or masonryveneer. Houses are relatively new, (average age of 17 years). Houses sold within a year of being builtrepresent twenty percent. The inverse of the minimum distance to U.S. highways or interstates is2.28 (1/miles). In addition, 1,108 employees work in the oil and gas extraction sector in WeldCounty, CO per month on average.

Table 2 depicts the spatial and temporal variations in approved permit counts. The averagenumber of permits increases as either distance or the time window increases. For a time window of

Table 1. Summary statistics (N = 11,453).

Variable Description Mean SD Min Max

PRICE Sale price (2015 dollars) 290,892 88,116 20,024 1,095,900COMPARABLE Weighted average property prices within 1.8-miles of the subject

property and 6-months before its transaction293,722 80,337 65,000 1,094,699

SQFT Property size (square feet) 1735 571 438 5781ACRE Lot size (acres) 0.20 0.23 0.01 9.35ACRESQ Squared Lot size 0.10 1.17 0.0001 87.42BATH Number of bathrooms 2.70 0.80 1 7GARAGE Equal to 1 if property has a garage, 0 otherwise 0.97 0.16 0 1BALCONY Equal to 1 if property has a balcony, 0 otherwise 0.19 0.39 0 1FINBASE Equal to 1 if property has a finished basement, 0 otherwise 0.37 0.48 0 1RSTORY1 Equal to 1 if property is a 1 story ranch, 0 otherwise 0.48 0.50 0 1BILEVEL Equal to 1 if property is a bi-level house, 0 otherwise 0.03 0.16 0 1HARDBD Equal to 1 if property is a frame house with hardboard exterior,

0 otherwise0.43 0.49 0 1

VINYL Equal to 1 if property is a frame house with vinyl exterior, 0 otherwise 0.05 0.22 0 1MASONRY Equal to 1 if property is a frame house with masonry veneer exterior,

0 otherwise0.05 0.21 0 1

AGE Age of property (years) 17.03 23.16 0 166NEW Equal to 1 if property is sold in the same year as being built, 0 otherwise 0.20 0.40 0 1INDISTRD Inverse of minimum distance to US highway or Interstate (1/miles) 2.28 69.14 0.04 7114.86EMPLOYMENT Monthly employee counts of oil and gas extraction sector in Weld

County1108 21 1081 1152

Table 2. Spatial and temporal variations in permitted wells.

PERMITDistance buffer

(miles)Time window(months) Mean SD Min Max

Permit_05_6 0.5 6 0.13 1.39 0 22Permit_05_12 0.5 12 0.27 1.95 0 22Permit_1_6 1 6 1.01 3.71 0 27Permit_1_12 1 12 2.03 5.52 0 41Permit_15_6 1.5 6 2.51 5.87 0 59Permit_15_12 1.5 12 5.13 9.13 0 59Permit_2_6 2 6 4.68 7.86 0 59Permit_2_12 2 12 9.41 11.98 0 62

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY 11

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 13: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

6-months, it is 0.13 within 0.5-mile, 1.01 within 1-mile, 2.51 within 1.5-miles, and 4.68 within 2-miles. These numbers more than double when the time window increase to 12-months.

6.3. Full sample estimation results

We use the linear functional form for the main results, because of its easier interpretation than Box-Cox results, and better performance compared to semi-log specifications4. All the econometric mod-els include the same housing characteristics, COMPARABLE variable, and census tract by year fixedeffects. In all cases, standard errors are clustered by census tract. Variations come from different lev-els of shale activities across space and time in the same census tract and year. Because of the possibil-ity5. that some proportion of permitted wells did not start well preparation, our estimate provides alower bound of the net shale activity impact on the housing market.

Eight different models reported in Table 3 variously employ four spatial cut-off radiuses of 0.5-mile, 1-mile, 1.5-miles, and 2-miles, and two time windows of 6-months and 12-months. All thehousing characteristics have expected signs and similar magnitude of estimated coefficients. Forinstance, an additional square foot of house size will increase housing value by approximate $58.The spatial term COMPARABLE is positive and significant at the 1% level.

Results across various spatial and temporal cut-offs reveal that the net impact of shale activities, asrepresented by well permit counts, is insignificantly different from zero. Thus, we cannot reject the

Table 3. Full sample analysis (N = 11,453, R2 = 0.92).(1) (2) (3) (4) (5) (6) (7) (8)

VARIABLES0.5 Mile6 Month

0.5 Mile12 Month

1 Mile6 Month

1 Mile12 Month VARIABLES

1.5 Mile6 Month

1.5 Mile12 Month

2 Mile6 Month

2 Mile12 Month

SQFT 57.60��� 57.60��� 57.59��� 57.61��� SQFT 57.63��� 57.67��� 57.60��� 57.61���

(2.124) (2.130) (2.127) (2.125) (2.108) (2.080) (2.120) (2.108)ACRE 33,539��� 33,515��� 33,525��� 33,446��� ACRE 33,575��� 33,448��� 33,592��� 33,505���

(7751) (7733) (7737) (7743) (7701) (7710) (7711) (7716)ACRESQ ¡4133��� ¡4130��� ¡4129��� ¡4120�� ACRESQ ¡4137��� ¡4122�� ¡4140��� ¡4131���

(1560) (1556) (1559) (1560) (1557) (1561) (1559) (1560)BATH 7372��� 7379��� 7395��� 7388��� BATH 7399��� 7382��� 7378��� 7369���

(697.7) (698.7) (699.8) (700.2) (701.2) (694.9) (699.2) (697.5)GARAGE 11,065��� 11,054��� 11,103��� 11,093��� GARAGE 11,148��� 11,126��� 11,096��� 11,089���

(2052) (2056) (2047) (2045) (2059) (2067) (2055) (2059)BALCONY 4149��� 4138��� 4119��� 4121��� BALCONY 4133��� 4154��� 4155��� 4169���

(1240) (1240) (1243) (1245) (1251) (1251) (1254) (1258)FINBASE 14,598��� 14,602��� 14,611��� 14,628��� FINBASE 14,593��� 14,620��� 14,592��� 14,617���

(1269) (1274) (1275) (1275) (1272) (1265) (1271) (1269)RSTORY1 23,316��� 23,320��� 23,305��� 23,306��� RSTORY1 23,350��� 23,344��� 23,333��� 23,315���

(1789) (1790) (1790) (1790) (1785) (1780) (1792) (1790)BILEVEL 20,653��� 20,675��� 20,716��� 20,720��� BILEVEL 20,734��� 20,707��� 20,681��� 20,639���

(2308) (2312) (2313) (2299) (2296) (2295) (2306) (2318)HARDBD ¡4435�� ¡4448�� ¡4431�� ¡4459��� HARDBD ¡4461��� ¡4498��� ¡4444��� ¡4451���

(1688) (1691) (1683) (1674) (1665) (1640) (1679) (1682)VINYL ¡5549��� ¡5557��� ¡5511��� ¡5479��� VINYL ¡5554��� ¡5579��� ¡5553��� ¡5562���

(1996) (1998) (1989) (1984) (1996) (1990) (1998) (2002)MASONRY ¡5686�� ¡5704�� ¡5635�� ¡5651�� MASONRY ¡5743�� ¡5733�� ¡5729�� ¡5801���

(2217) (2233) (2202) (2208) (2196) (2182) (2194) (2194)AGE ¡333.9��� ¡334.0��� ¡333.8��� ¡333.4��� AGE ¡333.0��� ¡332.4��� ¡333.3��� ¡332.6���

(46.64) (46.78) (46.90) (47.02) (46.73) (46.94) (46.90) (47.11)NEW 3854 3877 3939 3949 NEW 3910 3962 3861 3894

(2887) (2878) (2844) (2827) (2859) (2832) (2888) (2862)INDISTRD 2.871��� 2.869��� 2.874��� 2.867��� INDISTRD 2.874��� 2.877��� 2.868��� 2.871���

(0.444) (0.443) (0.441) (0.439) (0.441) (0.442) (0.441) (0.443)COMPARABLE 0.659��� 0.659��� 0.659��� 0.659��� COMPARABLE 0.659��� 0.659��� 0.659��� 0.659���

(0.0276) (0.0276) (0.0276) (0.0277) (0.0274) (0.0274) (0.0273) (0.0275)PERMIT ¡133.5 ¡171.1 ¡189.1 ¡140.7 PERMIT ¡120.9 ¡116.3 ¡45.72 ¡55.91

(379.0) (300.8) (194.2) (125.9) (97.42) (87.48) (64.12) (51.44)

Note: All the regressions use linear sale prices as dependent variables. They include year by census tract fixed effects. Robuststandard errors clustered by census tract. ��� p < 0.01, �� p< 0.05, � p < 0.1.

12 X. HE ET AL.

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 14: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

null, g ¼ 0. The evidence clearly does not support the alternative hypothesis (Ha in [5]) of a negativeenvironmental externality. In contradiction to Gopalakrishnan and Klaiber (2014), our results suggestno negative impact of shale activities during the well preparation period after permitting. Rather, ifthere are no negative impacts, then some type of compensation may be occurring, which could beupfront one-time bonus payments upon signing a lease contract between property owners6 and oil andgas companies, or monthly lease payments related to the amount of production.

6.4. Full sample temporal variations in the impact of shale gas activities

We further explore the potential impacts of shale gas development activities within a 2-mile radiusacross possible temporal windows. Table 4 presents a wide range of time windows, which vary from6-months to 26-months from well permitting, each combined with the 2-mile spatial cut-off. Theestimated coefficients of permits count remain insignificant regardless of the size of time windows.This is consistent with the finding of Table 1 that no negative environmental impact on housing val-ues exists during the preparation period after permitting.

6.5. Subsample estimation results

Next, we restrict our sample to houses within 2-miles of a pending permit only (presented inTable 5). Restricting the sample this way controls for unobserved heterogeneity (either time

Table 4. Full sample temporal variations at 2-mile (N = 11,453, R2 = 0.92).(1) (2) (3) (4) (5) (6) (7) (8)

VARIABLES2 Mile6 Month

2 Mile9 Month

2 Mile12 Month

2 Mile15 Month VARIABLES

2 Mile18 Month

2 Mile21 Month

2 Mile24 Month

2 Mile26 Month

SQFT 57.60��� 57.60��� 57.61��� 57.60��� SQFT 57.59��� 57.60��� 57.59��� 57.59���

(2.120) (2.115) (2.108) (2.113) (2.112) (2.112) (2.117) (2.119)ACRE 33,592��� 33,576��� 33,505��� 33,532��� ACRE 33,565��� 33,525��� 33,567��� 33,552���

(7711) (7723) (7716) (7750) (7765) (7769) (7799) (7805)ACRESQ ¡4140��� ¡4139��� ¡4131��� ¡4133�� ACRESQ ¡4138�� ¡4134�� ¡4138�� ¡4137��

(1559) (1560) (1560) (1565) (1567) (1566) (1571) (1574)BATH 7378��� 7374��� 7369��� 7368��� BATH 7373��� 7365��� 7372��� 7371���

(699.2) (698.7) (697.5) (696.0) (692.1) (692.3) (693.4) (694.4)GARAGE 11,096��� 11,091��� 11,089��� 11,089��� GARAGE 11,088��� 11,085��� 11,088��� 11,088���

(2055) (2055) (2059) (2054) (2049) (2053) (2050) (2050)BALCONY 4155��� 4,160��� 4,169��� 4,160��� BALCONY 4,162��� 4,161��� 4,162��� 4,161���

(1,254) (1,254) (1,258) (1,252) (1,251) (1,253) (1,251) (1,251)FINBASE 14,592��� 14,592��� 14,617��� 14,595��� FINBASE 14,588��� 14,597��� 14,589��� 14,590���

(1,271) (1,270) (1,269) (1,270) (1,272) (1,271) (1,270) (1,271)RSTORY1 23,333��� 23,323��� 23,315��� 23,318��� RSTORY1 23,322��� 23,316��� 23,322��� 23,320���

(1,792) (1,791) (1,790) (1,789) (1,790) (1,789) (1,790) (1,790)BILEVEL 20,681��� 20,666��� 20,639��� 20,657��� BILEVEL 20,659��� 20,649��� 20,658��� 20,655���

(2,306) (2,310) (2,318) (2,312) (2,316) (2,318) (2,319) (2,320)HARDBD ¡4,444��� ¡4,439�� ¡4,451��� ¡4,435�� HARDBD ¡4,422�� ¡4,437�� ¡4,422�� ¡4,426��

(1,679) (1,682) (1,682) (1,680) (1,679) (1,678) (1,678) (1,677)VINYL ¡5,553��� ¡5,537��� ¡5,562��� ¡5,549��� VINYL ¡5,533��� ¡5,544��� ¡5,534��� ¡5,536���

(1,998) (1,996) (2,002) (1,999) (1,995) (1,995) (1,993) (1,993)MASONRY ¡5,729�� ¡5,730�� ¡5,801��� ¡5,718�� MASONRY ¡5,658�� ¡5,727�� ¡5,659�� ¡5,674��

(2,194) (2,191) (2,194) (2,193) (2,189) (2,190) (2,180) (2,183)AGE ¡333.3��� ¡333.3��� ¡332.6��� ¡333.1��� AGE ¡333.5��� ¡333.1��� ¡333.5��� ¡333.4���

(46.90) (46.85) (47.11) (46.75) (46.52) (46.80) (46.47) (46.48)NEW 3,861 3,867 3,894 3,882 NEW 3,872 3,891 3,872 3,878

(2,888) (2,881) (2,862) (2,865) (2,859) (2,846) (2,848) (2,834)INDISTRD 2.868��� 2.868��� 2.871��� 2.869��� INDISTRD 2.870��� 2.871��� 2.869��� 2.870���

(0.441) (0.442) (0.443) (0.442) (0.444) (0.444) (0.446) (0.447)COMPARABLE 0.659��� 0.659��� 0.659��� 0.659��� COMPARABLE 0.659��� 0.659��� 0.659��� 0.659���

(0.0273) (0.0274) (0.0275) (0.0274) (0.0273) (0.0274) (0.0276) (0.0276)PERMIT ¡45.72 ¡30.47 ¡55.91 ¡17.45 PERMIT 1.676 ¡16.71 1.376 ¡2.595

(64.12) (43.47) (51.44) (53.19) (55.67) (57.30) (57.55) (57.21)

Note: All the regressions use linear sale prices as dependent variables. They include year by census tract fixed effects. Robuststandard errors clustered by census tract. ��� p < 0.01, �� p< 0.05, � p < 0.1.

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY 13

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 15: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

invariant or time-varying) between houses with and without oil and gas reserves in nearby under-ground areas. Results show that again permits count is an insignificant determinant of house pricesno matter what spatial and temporal cut-offs are used. Again, we find no impact from shale gasactivities, failing to reject the null (i.e. the evidence does not support Ha).

6.6. Subsample temporal variations in shale gas development activities

We also explore the potential time window of well permit impacts under the sample restriction.Table 6 employs temporal variations for a 2-mile cut-off. After excluding incomparable properties,results still show that being located within 2-miles of permitted wells is associated with insignificanteffects (again counter to the hypothesis Ha). The net impact of an additional well within a 2-miledistance, or marginal implicit price, is statistically insignificant from zero for all the time windows.

6.7. The impact of oil and gas employment on housing values

As noted, Bennett and Loomis (2015) found small and positive significant impact of oil and gas sec-tor employment on housing values in the whole county. Table 7 presents the results of full sampleanalysis, for model specifications that include the monthly employment counts in the oil and gasextraction sector in Weld County. In contrast to the result of Bennett and Loomis (2015), our resultsindicate a consistent, negative, and highly significant impact of employment in oil and gas extraction

Table 5. Subsample Analysis (N = 5,060, R2 = 0.90).(1) (2) (3) (4) (5) (6) (7) (8)

VARIABLES0.5 Mile6 Month

0.5 Mile12 Month

1 Mile6 Month

1 Mile12 Month VARIABLES

1.5 Mile6 Month

1.5 Mile12 Month

2 Mile6 Month

2 Mile12 Month

SQFT 59.22��� 59.21��� 59.18��� 59.21��� SQFT 59.30��� 59.39��� 59.25��� 59.26���

(3.562) (3.555) (3.576) (3.568) (3.520) (3.446) (3.538) (3.518)ACRE 39,916��� 39,901��� 39,634��� 39,591��� ACRE 39,567��� 39,268��� 39,845��� 39,761���

(11,119) (11,068) (11,134) (11,159) (11,061) (11,040) (11,048) (11,034)ACRESQ ¡4,361�� ¡4,360�� ¡4,320�� ¡4,323�� ACRESQ ¡4,319�� ¡4,291�� ¡4,353�� ¡4,346��

(1,677) (1,671) (1,672) (1,684) (1,681) (1,696) (1,676) (1,680)BATH 9,439��� 9,451��� 9,521��� 9,500��� BATH 9,469��� 9,436��� 9,446��� 9,440���

(1,009) (999.5) (990.5) (997.4) (1,009) (1,018) (1,004) (1,009)GARAGE 9,480��� 9,492��� 9,479��� 9,444��� GARAGE 9,582��� 9,459��� 9,454��� 9,415���

(2,724) (2,720) (2,731) (2,732) (2,712) (2,720) (2,756) (2,776)BALCONY 3,745� 3,746� 3,631� 3,659� BALCONY 3,688� 3,685� 3,754� 3,772�

(1,939) (1,959) (1,966) (1,961) (1,982) (1,990) (2,000) (2,002)FINBASE 13,768��� 13,768��� 13,779��� 13,816��� FINBASE 13,793��� 13,839��� 13,742��� 13,768���

(2,343) (2,356) (2,343) (2,359) (2,352) (2,328) (2,339) (2,344)RSTORY1 24,711��� 24,716��� 24,787��� 24,761��� RSTORY1 24,756��� 24,762��� 24,722��� 24,707���

(2,812) (2,815) (2,819) (2,810) (2,802) (2,783) (2,819) (2,823)BILEVEL 20,960��� 20,942��� 20,822��� 20,926��� BILEVEL 21,100��� 21,121��� 20,911��� 20,918���

(4,001) (4,002) (3,999) (4,008) (3,888) (3,856) (3,995) (4,011)HARDBD ¡2,828 ¡2,834 ¡2,866 ¡2,955 HARDBD ¡2,957 ¡3,122 ¡2,869 ¡2,882

(2,623) (2,621) (2,615) (2,575) (2,492) (2,353) (2,576) (2,568)VINYL ¡3,708 ¡3,711 ¡3,815 ¡3,797 VINYL ¡3,735 ¡3,799 ¡3,688 ¡3,725

(3,244) (3,233) (3,235) (3,232) (3,276) (3,271) (3,258) (3,250)MASONRY ¡4,064 ¡4,041 ¡4,063 ¡4,058 MASONRY ¡4,178 ¡4,017 ¡4,042 ¡4,111

(4,460) (4,454) (4,472) (4,501) (4,403) (4,444) (4,422) (4,410)AGE ¡280.9��� ¡280.6��� ¡280.9��� ¡279.6��� AGE ¡280.2��� ¡279.8��� ¡280.3��� ¡279.9���

(72.86) (72.95) (73.41) (73.55) (73.09) (74.06) (73.58) (73.71)NEW 1,542 1,579 1,705 1,712 NEW 1,641 1,751 1,507 1,564

(3,694) (3,658) (3,609) (3,566) (3,661) (3,589) (3,732) (3,679)INDISTRD ¡123.4 ¡122.9 ¡102.9 ¡115.6 INDISTRD ¡107.4 ¡116.6 ¡117.0 ¡117.9

(168.3) (168.6) (178.8) (175.4) (172.8) (169.5) (169.0) (168.8)COMPARABLE 0.630��� 0.630��� 0.631��� 0.631��� COMPARABLE 0.632��� 0.633��� 0.629��� 0.630���

(0.0489) (0.0487) (0.0494) (0.0493) (0.0489) (0.0483) (0.0484) (0.0484)PERMIT ¡113.9 ¡87.48 ¡363.4 ¡177.5 PERMIT ¡269.5 ¡230.8 ¡92.05 ¡65.17

(475.9) (397.0) (299.3) (196.2) (176.4) (193.3) (115.3) (82.14)

Note: All the regressions use linear sale prices as dependent variables. They all include year by census tract fixed effects. Robuststandard errors clustered by census tract. ��� p < 0.01, �� p< 0.05, � p < 0.1.

14 X. HE ET AL.

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 16: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

sector on housing values. The magnitude of impacts is between $64 – 65; on average; that is, theprice of a single-family house will decrease by $65 for one more worker in the oil and gas extractionsector in Weld County. This result implies that the temporary influx workers in the oil and gas sec-tor presented net negative impacts on nearby housing prices. While this is an interesting result inand of itself, for our primary hypothesis of interest (5), the estimated coefficients on fracking wellpermits count stayed insignificant after including the employment variable in all model specifica-tions examined. Also, the magnitudes for the estimated coefficients on well permits do not changemuch with and without including the employment counts of oil and gas extraction sector.

7. Summary

Using data for the period 20 October 2014 to 1 March 2017, we find consistent insignificant esti-mated impacts of permitted well counts (g) on residential house prices located within 2-miles ofpermitted wells in Weld County, CO. The estimated marginal implicit price of a well approved fordrilling within 2-miles of an average property is not statistically different from zero. What to makeof the result, in a setting where a priori we might have thought negative externalities in the housingmarket would be apparent? We speculate that this result is more consistent with evidence that prop-erty owners within 2-miles of permitted wells receiving bonus payments or expecting to receive leasepayments. It could also be compensation received from lessors accessing subsurface minerals from

Table 6. Subsample temporal variations at 2-mile (N = 5,060, R2 = 0.90).(1) (2) (3) (4) (5) (6) (7) (8)

VARIABLES2 Mile6 Month

2 Mile9 Month

2 Mile12 Month

2 Mile15 Month VARIABLES

2 Mile18 Month

2 Mile21 Month

2 Mile24 Month

2 Mile26 Month

SQFT 59.25��� 59.24��� 59.26��� 59.25��� SQFT 59.33��� 59.37��� 59.29��� 59.26���

(3.538) (3.522) (3.518) (3.498) (3.473) (3.472) (3.475) (3.490)ACRE 39,845��� 39,900��� 39,761��� 39,804��� ACRE 39,680��� 39,642��� 39,734��� 39,744���

(11,048) (11,024) (11,034) (11,086) (11,115) (11,094) (11,111) (11,107)ACRESQ ¡4,353�� ¡4,361�� ¡4,346�� ¡4,352�� ACRESQ ¡4,343�� ¡4,344�� ¡4,351�� ¡4,348��

(1,676) (1,676) (1,680) (1,687) (1,687) (1,681) (1,689) (1,696)BATH 9,446��� 9,430��� 9,440��� 9,435��� BATH 9,421��� 9,408��� 9,430��� 9,431���

(1,004) (1,006) (1,009) (1,006) (1,014) (1,014) (1,009) (1,009)GARAGE 9,454��� 9,450��� 9,415��� 9,459��� GARAGE 9,442��� 9,432��� 9,467��� 9,477���

(2,756) (2,746) (2,776) (2,761) (2,775) (2,775) (2,759) (2,752)BALCONY 3,754� 3,764� 3,772� 3,757� BALCONY 3,750� 3,748� 3,761� 3,765�

(2,000) (1,997) (2,002) (1,993) (1,997) (1,999) (1,995) (1,993)FINBASE 13,742��� 13,763��� 13,768��� 13,751��� FINBASE 13,743��� 13,732��� 13,738��� 13,743���

(2,339) (2,337) (2,344) (2,336) (2,328) (2,321) (2,329) (2,326)RSTORY1 24,722��� 24,706��� 24,707��� 24,715��� RSTORY1 24,735��� 24,731��� 24,726��� 24,721���

(2,819) (2,829) (2,823) (2,812) (2,800) (2,798) (2,795) (2,803)BILEVEL 20,911��� 20,944��� 20,918��� 20,981��� BILEVEL 20,992��� 21,065��� 21,017��� 21,009���

(3,995) (4,002) (4,011) (3,962) (3,967) (3,930) (3,912) (3,905)HARDBD ¡2,869 ¡2,844 ¡2,882 ¡2,862 HARDBD ¡2,907 ¡2,936 ¡2,887 ¡2,874

(2,576) (2,585) (2,568) (2,562) (2,547) (2,540) (2,545) (2,550)VINYL ¡3,688 ¡3,664 ¡3,725 ¡3,696 VINYL ¡3,733 ¡3,752 ¡3,711 ¡3,710

(3,258) (3,239) (3,250) (3,233) (3,246) (3,252) (3,235) (3,234)MASONRY ¡4,042 ¡4,081 ¡4,111 ¡4,099 MASONRY ¡4,162 ¡4,208 ¡4,121 ¡4,101

(4,422) (4,402) (4,410) (4,403) (4,394) (4,389) (4,383) (4,394)AGE ¡280.3��� ¡280.8��� ¡279.9��� ¡279.9��� AGE ¡279.2��� ¡279.1��� ¡279.6��� ¡279.8���

(73.58) (73.38) (73.71) (73.00) (73.85) (74.27) (73.32) (73.15)NEW 1,507 1,539 1,564 1,566 NEW 1,580 1,603 1,592 1,603

(3,732) (3,709) (3,679) (3,684) (3,679) (3,659) (3,662) (3,630)INDISTRD ¡117.0 ¡121.9 ¡117.9 ¡120.4 INDISTRD ¡118.4 ¡117.5 ¡119.5 ¡120.0

(169.0) (168.4) (168.8) (168.5) (169.0) (168.9) (168.2) (168.2)COMPARABLE 0.629��� 0.629��� 0.630��� 0.630��� COMPARABLE 0.630��� 0.630��� 0.630��� 0.630���

(0.0484) (0.0484) (0.0484) (0.0485) (0.0484) (0.0483) (0.0486) (0.0487)PERMIT ¡92.05 ¡37.14 ¡65.17 ¡35.28 PERMIT ¡61.10 ¡71.63 ¡39.29 ¡31.58

(115.3) (71.65) (82.14) (88.15) (89.20) (83.74) (85.93) (87.36)

Notes: All the regressions use linear sale prices as dependent variables. They all include year by census tract fixed effects.Robust standard errors clustered by census tract. ��� p<0.01, �� p<0.05, � p<0.1.

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY 15

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 17: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

property owners’ surface area, and for damage caused by accessing. If so, such payments appear tobe large enough to offset any negative housing externalities associated with shale well preparationactivities, irrespective of temporal cut-offs, for properties located within 2-miles of permitted wells.As a note of caution, the externalities considered in this Weld County analysis likely do not involverisk of groundwater contamination to household drinking water supplies, since there are very fewproperties (approximate 0.47%) located in groundwater management districts.

8. Discussion and conclusions

With the combination of much higher levels of shale gas production, well-preparation activities,population growth and new housing construction in Weld County, the expectation was a substantialincrease in negative externalities from shale exploration. However, our HPM modelling does notsupport this hypothesis, with the effect of well-permitting showing no evidence of negative external-ities on home prices, regardless of the utilised spatial and temporal cut-offs. Further, our primaryresult held when we investigated HPM model specification that controlled for the effects of oil andgas extraction sector employment on housing prices in Weld County. The finding of a negativeeffect from oil and gas extraction sector employment, in contrast to the positive effect found by

Table 7. The impact of oil and gas employment (N = 11,453, R2 = 0.92).(1) (2) (3) (4) (5) (6) (7) (8)

VARIABLES0.5 Mile6 Month

0.5 Mile12 Month

1 Mile6 Month

1 Mile12 Month VARIABLES

1.5 Mile6 Month

1.5 Mile12 Month

2 Mile6 Month

2 Mile12 Month

SQFT 57.80��� 57.81��� 57.80��� 57.82��� SQFT 57.83��� 57.88��� 57.81��� 57.82���

(2.123) (2.128) (2.126) (2.124) (2.109) (2.078) (2.120) (2.107)ACRE 34,000��� 33,977��� 33,981��� 33,904��� ACRE 34,023��� 33,907��� 34,036��� 33,963���

(7,749) (7,732) (7,738) (7,741) (7,709) (7,706) (7,721) (7,712)ACRESQ ¡4,170��� ¡4,167��� ¡4,165��� ¡4,157��� ACRESQ ¡4,173��� ¡4,159��� ¡4,176��� ¡4,167���

(1,559) (1,555) (1,557) (1,558) (1,556) (1,559) (1,558) (1,557)BATH 7,384��� 7,390��� 7,406��� 7,400��� BATH 7,408��� 7,395��� 7,388��� 7,381���

(694.8) (695.7) (696.6) (697.3) (698.0) (691.9) (696.1) (694.7)GARAGE 11,136��� 11,125��� 11,169��� 11,160��� GARAGE 11,208��� 11,195��� 11,160��� 11,157���

(2,072) (2,076) (2,066) (2,065) (2,078) (2,087) (2,073) (2,079)BALCONY 4,147��� 4,136��� 4,117��� 4,118��� BALCONY 4,132��� 4,150��� 4,153��� 4,165���

(1,234) (1,234) (1,236) (1,238) (1,244) (1,244) (1,246) (1,251)FINBASE 14,602��� 14,606��� 14,616��� 14,634��� FINBASE 14,599��� 14,626��� 14,597��� 14,625���

(1,267) (1,272) (1,273) (1,272) (1,269) (1,263) (1,269) (1,267)RSTORY1 23,414��� 23,416��� 23,401��� 23,403��� RSTORY1 23,442��� 23,442��� 23,425��� 23,412���

(1,806) (1,806) (1,806) (1,806) (1,802) (1,796) (1,808) (1,807)BILEVEL 20,698��� 20,717��� 20,757��� 20,763��� BILEVEL 20,770��� 20,753��� 20,717��� 20,682���

(2,302) (2,306) (2,305) (2,292) (2,289) (2,289) (2,299) (2,311)HARDBD ¡4,462��� ¡4,474��� ¡4,459��� ¡4,488��� HARDBD ¡4,485��� ¡4,528��� ¡4,466��� ¡4,482���

(1,686) (1,689) (1,681) (1,673) (1,664) (1,638) (1,679) (1,681)VINYL ¡5,589��� ¡5,597��� ¡5,555��� ¡5,522��� VINYL ¡5,594��� ¡5,624��� ¡5,590��� ¡5,607���

(2,009) (2,011) (2,002) (1,996) (2,008) (2,003) (2,009) (2,015)MASONRY ¡5,739�� ¡5,756�� ¡5,693�� ¡5,708�� MASONRY ¡5,790�� ¡5,792�� ¡5,766�� ¡5,864���

(2,230) (2,245) (2,216) (2,222) (2,210) (2,195) (2,207) (2,208)AGE ¡336.9��� ¡337.1��� ¡336.9��� ¡336.6��� AGE ¡336.1��� ¡335.6��� ¡336.5��� ¡335.7���

(47.38) (47.50) (47.62) (47.75) (47.44) (47.68) (47.55) (47.87)NEW 3,716 3,736 3,795 3,807 NEW 3,767 3,821 3,725 3,752

(2,909) (2,898) (2,865) (2,849) (2,879) (2,853) (2,907) (2,884)INDISTRD 3.308��� 3.306��� 3.308��� 3.305��� INDISTRD 3.304��� 3.319��� 3.303��� 3.312���

(0.443) (0.442) (0.441) (0.440) (0.439) (0.443) (0.438) (0.444)EMPLOYMENT ¡65.28��� ¡65.06��� ¡64.85��� ¡65.33��� EMPLOYMENT ¡64.23��� ¡65.83��� ¡64.78��� ¡65.79���

(16.67) (16.67) (16.78) (16.83) (16.96) (16.88) (16.73) (16.77)COMPARABLE 0.653��� 0.653��� 0.653��� 0.653��� COMPARABLE 0.653��� 0.653��� 0.653��� 0.653���

(0.0275) (0.0275) (0.0275) (0.0275) (0.0273) (0.0272) (0.0272) (0.0273)PERMIT ¡108.9 ¡149.8 ¡180.4 ¡139.3 PERMIT ¡108.4 ¡118.6 ¡32.03 ¡58.48

(355.9) (289.1) (191.0) (122.0) (94.38) (86.86) (59.99) (51.80)

Note: All the regressions use linear sale prices as dependent variables. They all include year by census tract fixed effects. Robuststandard errors clustered by census tract. ��� p < 0.01, �� p< 0.05, � p <0.1.

16 X. HE ET AL.

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 18: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

Bennett and Loomis (2015), highlights a possible social/employment external effect attached to theindustry (but not necessarily specific to fracking).

For the specific absence of any negative impact from well permitting for fracking on house prices,we call this a ‘case of the missing externality.’ What to make of this absence of evidence for a nega-tive environmental externality from fracking in the Weld County housing market? Are there rationalreasons, theoretical support, or corollary empirical evidence for the HPM evidence in this case? Onespeculative possibility is that some of the congestion impacts might be better reflected in in the nega-tive extraction sector employment effects we find in this HPM case study. This result merits furtherinvestigation in future HPM fracking studies.

Going deeper, economists have long advocated for market and exchange mechanisms to correctfor the presence of negative externalities (e.g. pollution) (Coase 1960); this policy advice can rangefrom the creation of formal markets for pollution permit (Baumol and Oates 1971) to more informalbargaining and negotiations. On the latter, side payments have long been recognised as a vehicle forreplacing the missing market (if transaction costs can be minimised) and reaching a deal betweenparties. Such deals have been referred to as ‘induced equilibrium,’ capable of weakly Pareto-domi-nating the original Nash equilibrium between parties without side payments (Bigelow 1993). Asargued by Coase (1960, 15), in what is now popularly recognised as the Coase Theorem: ‘It is alwayspossible to modify by transactions on the market the initial legal delimitation of rights. And, ofcourse, if such market transactions are costless, such a rearrangement of rights will always take placeif it will lead to an increase in the value of production’.

Essentially, the Coase Theorem asserts that under certain conditions bargaining can replace themissing market. One corollary might be that when we expect to find a missing market and insteadfind a missing externality, we might then expect to find evidence of bargaining to internalise theexternality (e.g. mitigating behaviours in the production process, or compensation for the expectedlosing party). Thus, in the case where we expect but fail to find evidence of pollution dis-amenityeffects in behavioural trails (say, being negatively capitalised into affected housing market) giventheir known physical presence, we might expect to find the presence of significant side payments. Isthere evidence of such payments creating or approaching an induced equilibrium in hydraulic frack-ing for Weld County?

Although we do not have housing-level transaction data to model this, side payments to propertyowners appear to regularly occur in Weld County and surrounding areas in Colorado. This is con-sistent with the missing negative impact of well permitting found in our HPM results; the implica-tion is that property owners are either receiving bonus payments or expect to receive leasepayments, and such side payments are large enough to offset the perceived negative environmentalexternalities involved with shale well site preparation activities. Available descriptive evidence indi-cated that local landowners could receive from several dollars to thousands of dollars per acre bonuspayments, upon signing a lease with oil and gas companies (Brasier et al. 2011; Muehlenbachs,Spiller, and Timmins 2012).

Further, if there is such a missing environmental externality in house price, then we might expectit to at least partially mute public support against fracking. To wit, several attempts to implementbans or restrictions over oil and gas development in Colorado, especially in the city of Greeley,Weld County, have failed despite vocal concerns over environmental and health costs from fracking.As one example, in March 2016, the Greeley City Council reversed the Planning Commission’sdenial to a permit to drill 22 wells West of the city; this reversal essentially protected the privateproperty rights of 1,800 citizens in Greeley who have mineral rights (Aguilar 2016). At county level,as of November 2016, Weld County commissioners created a county permit for oil and gas projectsin the unincorporated Weld County, which no longer requires landowners to sign off, or hold publichearing in most cases, even if they are near homes or schools (Sweeney 2016). State-wide, therecently proposed Ballot Initiative #78 would shift control of development of Colorado’s oil and gasnatural resources to local government, and add a mandatory setback zone of 2,500 feet around occu-pied buildings and in open public spaces, making roughly 90% of surface acreage in Colorado (85%

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY 17

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 19: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

of surface acreage in Weld County) unavailable for future oil and gas facility development (Alford2016a). As of August 2016, however, this initiative could not make the November 2016 state ballot(Alford 2016b). Also at a more general level for oil and gas development in Colorado, as of April2016, House Bill 1355, which would give local governments the ability to exercise land use authorityover fracking, was voted down by the Colorado House of Representatives (Alford 2016c).

In closing, appealing to the Coase Theorem, we argue that our results are consistent with a kindof induced equilibrium, where side payment may be explaining the ‘missing externality’ from frack-ing in the housing market. Further, this may help partially explain failed public attempts to stopshale gas development in Colorado, especially in Weld County. Rather than asserting that this closesthe case, this opens further questions, such as what impacts or affected populations may be left out?For instance, households replying upon groundwater might experience much more significant nega-tive externalities from shale activities. Note that with less than one percent (0.39%) of properties inour Weld County sample located in groundwater-designated areas, groundwater contamination riskfrom shale activities are likely absent from our shale impact estimates. Thus, we caution that our pri-mary result may not apply to areas with significant groundwater concerns (presuming that negativeenvironmental risks are fully recognised by the public and then capitalised into the housing market).Finally, how compensation may be working in these markets has attracted some initial research, butcould use additional investigation. Thus, we hope this analysis spurs further research delving intothe distribution, extent, and magnitude of side payments relating to shale gas extraction, and howsuch payments link to housing markets and other private or public areas.

Notes

1. Different from other states, Colorado does not charge a permit fee for an application. This provides an incentivefor operators to keep a number of permits to drill and wait until the oil price to rise or sell to another operatorwho is ready to drill. That is why some wells go through several re-fillings before they get drilled or they neverget drilled.

2. After wells spud, they are eliminated from the permit data-set. The only way to match spud date and permitapproval date for the same well is to look at each individual scout card, one well at a time.

3. Maximum likelihood estimation with spatial lag or spatial error models is often carried out (Anselin 1988). Thisapproach requires calculation of an n� n spatial weighting matrix, which is not feasible in the context of large num-ber of varying properties transacted over time. In such cases, an instrumental variable method may be preferred.

4. These three alternative functional forms for the models have no substantial difference in estimated results. Fullresults are available upon request from the lead author.

5. This possibility is higher for Colorado compared to those states with permit application fees.6. In Weld County, surface property rights can be severed from subsurface mineral rights. It is possible that individ-

ual property owners do not have claims to either bonus or lease payments underground. Results represent anaverage effect across properties, whose owners own mineral rights and not.

Acknowledgments

We thank Dennis Ahlstrand, Colorado Oil & Gas Conservation Commission for answering questions concerning GISdata, Courtney Anaya, Weld County Assessor’s Office for help on housing data, and Ric Wise, Ashlie Koehn from theU.S. Bureau of Labor Statistics for detailed answers to our questions about oil and gas employment data.

Disclosure statement

No potential conflict of interest was reported by the authors.

References

Abdalla, Charles, Joy Drohan, Brian Rahm, Jeffrey Jacquet, John Becker, Alan Collins, H. Allen Klaiber, Gregory Poe,and Deb Grantham. 2012. Water's Journey Through the Shale Gas Drilling and Production Processes in the Mid-Atlantic Region. Review of Penn State Extension. University Park, PA: College of Agricultural Sciences.

18 X. HE ET AL.

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 20: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

Adgate, John L., Bernard D. Goldstein, and Lisa M. McKenzie. 2014. “Potential Public Health Hazards, Exposures andHealth Effects from Unconventional Natural Gas Development.” Review of Environmental Science & Technology48 (15): 8307–8320.

Aguilar, J. 2016. “Greeley City Council Reverses Planning Commission’s Oil and Gas Denial Verdict.” DenverPost.com. http://www.denverpost.com/2016/03/08/greeley-city-council-reverses-planning-commissions-oil-and-gas-denial-verdict/ (accessed Apr 26, 2017).

Alford, E. 2016a. “Colorado’s Setback Rule is a Game Changer.”MineralWise.com. http://www.mineralweb.com/news/colorados-setback-rule-is-a-game-changer/ (accessed Apr 26, 2017).

Alford, E. 2016b. “Anti-Fracking Measures Fall Short in Colorado.” MineralWise.com. http://www.mineralweb.com/news/anti-fracking-measures-fall-short-in-colorado/ (accessed Apr 26, 2017).

Alford, E. 2016c. “Colorado Says No to Local Control of Oil & Gas.” MineralWise.com. http://www.mineralweb.com/news/colorado-says-no-to-local-control-of-oil-gas/ (accessed Apr 26, 2017).

Anselin, Luc. 1988. Spatial Econometrics: Methods and Models. Dordrecht: Springer.Baumol, William J., and Wallace E. Oates. 1971. “The Use of Standards and Prices for Protection of the Environment.”

Review of The Swedish Journal of Economics 73: 42–54.Bennett, Ashley, and John Loomis. 2015. “Are Housing Prices Pulled Down or Pushed Up by Fracked Oil and Gas

Wells? A Hedonic Price Analysis of Housing Values in Weld County, Colorado.” Review of Society & NaturalResources 28 (11): 1168–1186.

Bigelow, John Payne. 1993. “Inducing Efficiency: Externalities, Missing Markets, and the Coase Theorem.” Review ofInternational Economic Review 34 (2): 335–346.

Boxall, Peter C., Wing H. Chan, and Melville L. McMillan. 2005. “The Impact of Oil and Natural Gas Facilities onRural Residential Property Values: A Spatial Hedonic Analysis.” Review of Resource and Energy Economics 27 (3):248–269.

Brasier, Kathryn J., Matthew R. Filteau, Diane K. McLaughlin, Jeffrey Jacquet, Richard C. Stedman, Timothy W. Kel-sey, and Stephan J. Goetz. 2011. “Residents’ Perceptions of Community and Environmental Impacts from Devel-opment of Natural Gas in the Marcellus Shale: A Comparison of Pennsylvania and New York Cases.” Review ofJournal of Rural Social Sciences 26 (1): 32.

Brown, David R., Celia Lewis, and Beth I. Weinberger. 2015. “Human Exposure to Unconventional Natural GasDevelopment: A Public Health Demonstration of Periodic High Exposure to Chemical Mixtures in Ambient Air.”Review of Journal of Environmental Science and Health, Part A 50 (5): 460–472.

Brown, Jason P., Timothy Fitzgerald, and Jeremy G. Weber. 2016. “Capturing Rents from Natural Resource Abundance:Private Royalties from US Onshore Oil & Gas Production.” Review of Resource and Energy Economics 46: 23–38.

Coase, Ronald H. 1960. “The Problem of Social Cost.” Review of Journal of Law and Economics 3: 1–44.COGCC (Colorado Oil & Gas Conservation Commission). 2016. Staff Report for 07/18/2016. http://cogcc.state.co.us/

library.html#/staffreports (accessed Sep 10, 2017).Colborn, Theo, Kim Schultz, Lucille Herrick, and Carol Kwiatkowski. 2014. “An Exploratory Study of Air Quality near

Natural Gas Operations.” Review of Human and Ecological Risk Assessment: An International Journal 20 (1): 86–105.Considine, Timothy J., Robert W. Watson, and Nicholas B. Considine. 2011. Energy Policy and Environment Report:

The Economic Opportunities of Shale Energy Development. New York, NY: The Manhattan Institute.Davis, Lucas W. 2011. “The Effect of Power Plants on Local Housing Values and Rents.” Review of Review of Econom-

ics and Statistics 93 (4): 1391–1402.Delgado, Michael S., Todd Guilfoos, and Andrew Boslett. 2014. “The Cost of Hydraulic Fracturing: A Hedonic

Analysis.” Working Paper. http://www.webmeets.com/files/papers/wcere/2014/890/Shale%20Feb%206%202014.pdf (accessed Sep 10, 2017).

DiGiulio, Dominic C., Richard T. Wilkin, Carlyle Miller, and Gregory Oberley. 2011. Investigation of Ground WaterContamination Near Pavillion. Wyoming, IL/Ada, OK: Office of Research and Development, National Risk Man-agement Research Laboratory.

DOE (Department of Energy). 2013. “Natural gas from shale: questions and answers.” https://energy.gov/fe/downloads/natural-gas-shale-questions-and-answers (accessed Sep 10, 2017).

EIA (U.S. Energy Information Administration). 2013. “North America leads the world in production of shale gas.”https://www.eia.gov/todayinenergy/detail.php?id=13491 (accessed Sep 10, 2017).

EIA (U.S. Energy Information Administration). DATA1. (Shale Gas Production data released on Dec 14, 2016).https://www.eia.gov/dnav/ng/ng_prod_shalegas_s1_a.htm (accessed on Apr 16, 2017).

EIA (U.S. Energy Information Administration). DATA2. (Shale Gas Reserve data released on Dec 14, 2016);. https://www.eia.gov/dnav/ng/ng_enr_shalegas_dcu_SCO_a.htm (accessed on Apr 16, 2017).

Feyrer, James, Erin T. Mansur, and Bruce Sacerdote. 2017. “Geographic Dispersion of Economic Shocks: Evidencefrom the Fracking Revolution.” Review of The American Economic Review 107 (4): 1313–1334.

Gopalakrishnan, Sathya, and H. Allen Klaiber. 2014. “Is the Shale Energy Boom a Bust for Nearby Residents? Evidencefrom Housing Values in Pennsylvania.” Review of American Journal of Agricultural Economics 96 (1): 43–66.

Hardy, Kirsten, and Timothy W. Kelsey. 2015. “Local Income Related to Marcellus Shale Activity in Pennsylvania.”Review of Community Development 46 (4): 329–340.

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY 19

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 21: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

Harkness, Jennifer S., Gary S. Dwyer, Nathaniel R. Warner, Kimberly M. Parker, William A. Mitch, and Avner Ven-gosh. 2015. “Iodide, Bromide, and Ammonium in Hydraulic Fracturing and Oil and Gas Wastewaters: Environ-mental Implications.” Review of Environmental Science & Technology 49: 1955–1963.

Hartley, Peter R., Kenneth B. Medlock, Ted Temzelides, and Xinya Zhang. 2015. “Local Employment Impact fromCompeting Energy Sources: Shale Gas versus Wind Generation in Texas.” Review of Energy Economics 49: 610–619.

Hill, Elaine L. 2013. “Shale Gas Development and Infant Health: Evidence from Pennsylvania.” http://www.damascuscitizensforsustainability.org/wp-content/uploads/2014/10/Shale-Gas-Development-and-Infant-Health-Elaine-Hill-Aug-2014.pdf (accessed Sep 10, 2017).

James, Alexander, and Jasmine James. 2014. “A Canary near a Gas Well: Gas Booms and Housing Market Busts inColorado.” Working Paper. http://alexandergjames.weebly.com/uploads/1/4/2/1/14215137/hedonic.weld.manuscript.jeem.pdf (accessed Sep 10, 2017).

Lauer, Nancy E., Jennifer S. Harkness, and Avner Vengosh. 2016. “Brine Spills Associated with Unconventional OilDevelopment in North Dakota.” Review of Environmental Science & Technology 50: 5389–5397.

Litovitz, Aviva, Aimee Curtright, Shmuel Abramzon, Nicholas Burger, and Constantine Samaras. 2013. “Estimation ofRegional Air-quality Damages from Marcellus Shale Natural Gas Extraction in Pennsylvania.” Review of Environ-mental Research Letters 8 (1): 014017.

McKenzie, Lisa M., Roxana Z. Witter, Lee S. Newman, and John L. Adgate. 2012. “Human Health Risk Assessment ofAir Emissions from Development of Unconventional Natural Gas Resources.” Review of Science of the TotalEnvironment 424: 79–87.

McMahon, M. Brian. 2017. “A Mineral Owner’s Introduction to Oil and Gas Leases.” MBM Law Offices. http://brianmcmahonlaw.com/CM/Client-Bulletin/Mineral-Owners-Intro-to-Leases.html (accessed Sep 10, 2017).

Muehlenbachs, Lucija, Elisheba Spiller, and Christopher Timmins. 2012. “Shale Gas Development and Property Val-ues: Differences across Drinking Water Sources.” NBER Working Paper 18390. http://www.nber.org/papers/w18390 (accessed Sep 10, 2017).

Muehlenbachs, Lucija, Elisheba Spiller, and Christopher Timmins. 2015. “The Housing Market Impacts of Shale GasDevelopment.” Review of The American Economic Review 105 (12): 3633–3659.

Newell, Richard G., and Daniel Raimi. 2015. “Oil and Gas Revenue Allocation to Local Governments in Eight States.”NBER Working Paper 21615. http://www.nber.org/papers/w21615 (accessed Sep 10, 2017).

Olmstead, Sheila M., Lucija A. Muehlenbachs, Jhih-Shyang Shih, Ziyan Chu, and Alan J. Krupnick. 2013. “Shale GasDevelopment Impacts on Surface Water Quality in Pennsylvania.” Review of Proceedings of the National Academyof Sciences 110 (13): 4962–4967.

Osborn, Stephen G., Avner Vengosh, Nathaniel R. Warner, and Robert B. Jackson. 2011. “Methane Contamination ofDrinking Water Accompanying Gas-well Drilling and Hydraulic Fracturing.” Review of Proceedings of theNational Academy of Sciences 108 (20): 8172–8176.

Platt, J. 2013. “Oil and Fracking Booms Creating Housing Busts.” MNN-Mother Nature Network. https://www.mnn.com/earth-matters/energy/stories/oil-and-fracking-booms-creating-housing-busts (accessed Aug 28, 2017)

Radow, Elisabeth N. 2011. “Homeowners and Gas Drilling Leases: Boon or Bust?” Review of New York State Bar Asso-ciation Journal 83 (9): 10–21.

Rosen, Sherwin. 1974. “Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition.” Review ofJournal of Political Economy 82 (1): 34–55.

Se Can, A.y., and Isaac Megbolugbe. 1997. “Spatial Dependence and House Price Index Construction.” Review of TheJournal of Real Estate Finance and Economics 14 (1–2): 203–222.

Swarthout, Robert F., Rachel S. Russo, Yong Zhou, Brandon M. Miller, Brittney Mitchell, Emily Horsman, Eric Lipsky,David C. McCabe, Ellen Baum, and Barkley C. Sive. 2015. “Impact of Marcellus Shale Natural Gas Development inSouthwest Pennsylvania on Volatile Organic Compound Emissions and Regional Air Quality.” Review of Environ-mental Science & Technology 49: 3175–3184.

Sweeney, C. 2016. “Weld County Officials Adopt New Oil and Gas Regulations.” GreeleyTribune.com. http://www.greeleytribune.com/news/local/weld-county-officials-adopt-new-oil-and-gas-regulations/ (accessed Sep 10, 2017).

Taylor, Laura O. 2003. “The Hedonic Method.” In A Primer on Nonmarket Valuation. The Economics of Non-MarketGoods and Resources, edited by P.A. Champ, K.J. Boyle, and T.C. Brown, Vol. 3, 331–393. Dordrecht: Springer.

UA and Argonne (University of Arkansas and Argonne National Laboratory). 2017. “Fayetteville Shale: Reducing theEnvironmental Impact of Natural Gas Development.” http://lingo.cast.uark.edu/lingopublic/index.htm (accessedApr 26, 2017).

Vengosh, Avner, Robert B. Jackson, Nathaniel Warner, Thomas H. Darrah, and Andrew Kondash. 2014. “A CriticalReview of the Risks to Water Resources from Unconventional Shale Gas Development and Hydraulic Fracturingin the United States.” Review of Environmental Science & Technology 48 (15): 8334–8348.

Weber, Jeremy Glenn, Jason P. Brown, and John Pender. 2013. “Rural Wealth Creation and Emerging Energy Indus-tries: Lease and Royalty Payments to Farm Households and Businesses.” Federal Reserve Bank of Kansas CityWorking Paper No. 13-07. https://ssrn.com/abstract=2307667 or http://dx.doi.org/10.2139/ssrn.2307667 (accessedSep 10, 2017).

20 X. HE ET AL.

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17

Page 22: The case of the missing negative externality? …...induced equilibrium (Bigelow 1993), this localised energy market appears to be internalising hous-ing-related externalities, such

Appendix

Table A.1. The impact of oil and gas employment using subsample (N = 5060, R-squared = 0.90).

(1) (2) (3) (4) (5) (6) (7) (8)

VARIABLES0.5 Mile6 Month

0.5 Mile12 Month

1 Mile6 Month

1 Mile12 Month

1.5 Mile6 Month

1.5 Mile12 Month

2 Mile6 Month

2 Mile12 Month

SQFT 59.44��� 59.43��� 59.40��� 59.43��� 59.50��� 59.59��� 59.46��� 59.47���

(3.530) (3.522) (3.545) (3.536) (3.498) (3.422) (3.515) (3.493)ACRE 40,076��� 40,072��� 39,785��� 39,762��� 39,745��� 39,465��� 40,022��� 39,951���

(11,078) (11,027) (11,098) (11,119) (11,034) (11,009) (11,017) (10,995)ACRESQ -4,361�� -4,361�� -4,318�� -4,324�� ¡4,322�� ¡4,295�� ¡4,356�� ¡4,349��

(1,671) (1,664) (1,665) (1,677) (1,675) (1,690) (1,670) (1,672)BATH 9,464��� 9,470��� 9,546��� 9,521��� 9,491��� 9,461��� 9,469��� 9,465���

(1,017) (1,007) (998.8) (1,005) (1,017) (1,027) (1,012) (1,017)GARAGE 9,486��� 9,494��� 9,482��� 9,451��� 9,575��� 9,464��� 9,468��� 9,436���

(2,735) (2,732) (2,742) (2,741) (2,725) (2,730) (2,757) (2,779)BALCONY 3,756� 3,759� 3,635� 3,669� 3,698� 3,694� 3,761� 3,773�

(1,940) (1,959) (1,966) (1,959) (1,982) (1,991) (1,997) (1,999)FINBASE 13,740��� 13,739��� 13,754��� 13,788��� 13,766��� 13,809��� 13,723��� 13,741���

(2,363) (2,377) (2,362) (2,377) (2,370) (2,348) (2,358) (2,363)RSTORY1 24,821��� 24,824��� 24,892��� 24,864��� 24,853��� 24,860��� 24,825��� 24,816���

(2,854) (2,855) (2,861) (2,851) (2,844) (2,826) (2,860) (2,864)BILEVEL 21,129��� 21,119��� 20,996��� 21,098��� 21,247��� 21,271��� 21,092��� 21,097���

(3,927) (3,928) (3,932) (3,940) (3,831) (3,793) (3,937) (3,945)HARDBD ¡2,869 ¡2,871 ¡2,912 ¡2,991 ¡2,986 ¡3,141 ¡2,897 ¡2,911

(2,625) (2,623) (2,619) (2,580) (2,503) (2,368) (2,586) (2,580)VINYL ¡3,695 ¡3,693 ¡3,814 ¡3,788 ¡3,729 ¡3,790 ¡3,684 ¡3,713

(3,298) (3,287) (3,288) (3,286) (3,322) (3,319) (3,299) (3,294)MASONRY ¡4,171 ¡4,156 ¡4,179 ¡4,172 ¡4,273 ¡4,130 ¡4,155 ¡4,209

(4,438) (4,425) (4,453) (4,477) (4,386) (4,427) (4,406) (4,394)AGE ¡284.7��� ¡284.6��� ¡284.8��� ¡283.6��� ¡284.0��� ¡283.6��� ¡284.2��� ¡283.9���

(75.42) (75.49) (75.95) (76.02) (75.49) (76.38) (75.83) (75.95)NEW 1,250 1,271 1,405 1,405 1,351 1,450 1,232 1,268

(3,732) (3,689) (3,642) (3,598) (3,684) (3,622) (3,751) (3,710)INDISTRD ¡116.1 ¡115.8 ¡96.27 ¡109.2 ¡102.4 ¡110.4 ¡112.0 ¡112.1

(169.9) (170.3) (180.1) (176.6) (173.8) (170.9) (170.3) (170.3)EMPLOYMENT ¡80.42��� ¡80.37��� ¡79.72��� ¡79.10�� ¡75.55�� ¡76.54�� ¡78.37�� ¡79.31��

(28.79) (28.89) (29.10) (29.28) (30.01) (30.58) (29.27) (29.85)COMPARABLE 0.625��� 0.625��� 0.626��� 0.625��� 0.627��� 0.628��� 0.625��� 0.625���

(0.0484) (0.0483) (0.0490) (0.0489) (0.0486) (0.0479) (0.0481) (0.0480)PERMIT ¡70.34 ¡46.20 ¡355.9 ¡163.6 ¡240.7 ¡210.9 ¡62.69 ¡49.36

(444.8) (383.4) (295.9) (190.9) (173.9) (194.9) (109.6) (85.14)

Notes: All the regressions use linear sale prices as dependent variables. They all include year by census tract fixed effects.Robust standard errors clustered by census tract. ��� p < 0.01, �� p < 0.05, � p< 0.1.

JOURNAL OF ENVIRONMENTAL ECONOMICS AND POLICY 21

Dow

nloa

ded

by [

67.0

.192

.32]

at 0

5:40

13

Nov

embe

r 20

17