ML12334A756
| ML12334A756 | |
| Person / Time | |
|---|---|
| Site: | Indian Point  |
| Issue date: | 05/14/2009 |
| From: | Williams College |
| To: | Atomic Safety and Licensing Board Panel |
| SECY RAS | |
| References | |
| RAS 21573, 50-247-LR, 50-286-LR, ASLBP 07-858-03-LR-BD01 | |
| Download: ML12334A756 (94) | |
Text
United States Nuclear Regulatory Commission Official Hearing Exhibit NYS000232 Entergy Nuclear Operations, Inc. Submitted: December 17, 2011 In the Matter of:
(Indian Point Nuclear Generating Units 2 and 3)
V\..~P.R REG(J<.q)<o ASLBP #: 07-858-03-LR-BD01 l~\
Docket #: 05000247 l 05000286
" 0'!:
Exhibit #: NYS000232-00-BD01 Identified: 10/15/2012
'" ~ Admitted: 10/15/2012 Withdrawn:
~
'1"'....1-: 0' i Rejected: Stricken:
" .*** .. " Other:
WILLIAMS COLLEGE LIBRARIES Your unpublished thesis, submitted for a degree at Williams College and administered by the Williams College Libraries, will be made available for research use. You may, through this form, provide instructions regarding copyright, access, dissemination and reproduction of your thesis.
_ The faculty advisor to the student writing the thesis wishes to claim joint authorship in this work.
In each section, please check the ONE statement that reflects your wishes.
- 1. PUBLICATION AND QUOTATION: LITERARY PROPERTY RIGHTS A student author automatically owns the copyright to his/her work, whether or not a copyright symbol and date are placed on the piece. The duration of U.S. copyright on a manuscript--and Williams theses are considered manuscripts--is the life of the author plus 70 years.
_ I1we do not choose to retain literary property rights to the thesis, and I wish to assign them immediately to Williams College.
Selecting this option will assign copyright to the Collegc. This in no way precludes it student author from later publishing his/her work; the student would, however. need to contact the Archives for a permission f{xlTI. The Archives would be free in this case to also grant permission to another researcher to publish small sections from the thesis. Rarely would there be any reason for the Archives to grant permission to another party to publish the thesis in its entirely; if such a situation arose, the Archives would be in touch with the author to let them. know that slich a request had been made.
_I1we wish to retain literary property rights to the thesis for a period of three years, at which time the literary property rights shall be assigned to Williams College.
Seieeting this option gives the Huthor a it;w years to make exdusive usc of the thesis in up-coming "v I/W::::: ::l::::;~':;~~::::o::rty rights to the ilicsis for aperiod of ~ years, or until my death, whichever is the later, at which time the literary property rights shall be assigned to Williams College.
Selecting this option allows the author great flexibility in extending or shortening the time of his/her automatic copyright period. Some students arc interested in using tbeir thesis in graduate school work. Tn this case, it would make sense for them to enter a number such as '10 years' in the blank, and line out the .'ords 'or until my death, whichever is the later.' Tn any event it is easier for the Archi ves to administer copyright on a manuscript if the period ends with the individual's deatfl-"ollr staff won't have to search for estate executors in this case***but this is entirely up to each student.
II. ACCESS The Williams College Libraries arc investigating the posting of theses online, as well as their retention in hardcopy.
~in hardcopy Williams College is granted permission to maintain and provide access to my thesis and via the Web both on and off campus.
Selecting this option allows researchers around the world to access the digital version of your work.
OAGI0001301_00001
_ Williams College is granted permission to maintain and provide access to my thesis in hardcopy and via the Web for on-campus use only.
Selecting this option allows access to the digital version of your work from tbe on-cmnpus network only.
_ The thesis is to be maintained and made available in hardcopy form only.
Selecting this option allows access to your work only from the hardcopy you submit. Such access pertains to lhe entirety of your work, including any media that it comprises or includes.
III. COPYING AND DISSEMINATION Because theses are listed on FRANCIS, the Libraries receive numerous requests every year for copies of works. IfIwhen a hardcopy thesis is duplicated for a researcher, a copy of the release form always
" \ accompanies the copy. Any digital version of your thesis will include the release form.
~ Copies of the thesis may be provided to any researcher.
Selecting this option all.ows any researcber to request a copy from the Williams College Libraries, or to make one from an electronic version.
_ Copying of the thesis is restricted for _ years, at which time copies may be provided to any researcher.
Th.is option allows the author to set a time limit on copying restrictions. During this period, an electronic version of tbe thesis wil! be protected against duplication.
_ Copying of the thesis or portions thereof, except as needed to maintain an adequate number of research copies available in the Williams College Libraries, is expressly prohibited. The electronic version of the thesis will be protected against duplication.
Selecting this option allows no reproductions to be made for researchers. The electronic version of the thesis will be protected against duplication. 'This opti.on does not dis**allow researchers from reading/viewing tbe work in either hardcopy or digital form.
Signed (student author) Signat:ure RerTloved Signed (faculty advis~ Sig nat:u re RerTloved Thesis title Ae4M:"j & ~f;i0 .( ;/;;dltJ? ~ A!I~.(; ;;+,,1' Date AaV I 'fl ;).001 I /
Accepted for the LibrariesSig nat:u re RerTloved Date accepted .,), . ! lj, (,/ q OAGI0001301 00002
Measuring the Externalities of Nuclear ower: A Hedonic Study by Brian Prest Professor Stephen Sheppard, Advisor A thesis submitted i.n partial fulfillment of the requirements for the Degree of Bachelor of Arts with Honors in Economics WILLIAMS COLLEGE Williamstown, Massachusetts May 14,2009 OAGI0001301 00003
Abstract In this paper, I use house price data to measure the extent to which the negative externalities posed by the Pilgrim Nuclear Generation Station in Plymouth, Massachusetts are capitalized into house prices. I find that homebuyers are willing to pay a significant premium to live farther away from the nuclear plant, finding a statistically significant price-distance elasticity estimate of 0.09.
This effect is estimated to extend to about 8 km. This elasticity is found to fall over time and with the median age of the community. Under reasonable parameters, the aggregate damages attributable to the nuclear plant, and hence the estimated external benefits of its closure, are estimated at $7,940 per affected house on average, aggregating to about $53 million (in year 2000 dollars), or
$1,024 per Plymouth resident. To conduct a full cost-benefit analysis the plant's continued operation, these external benefits must be weighed with the costs of replacing the energy produced by Pilgrim Station. Entergy Corporation, which owns and operates Pilgrim Station, also owns a large wooded green space that surrounds it, creating a buffer between the plant and the community as well as generating pleasing views and public recreational benefits. The price-distance elasticity of the green space is estimated to be a statistically significant -0.05, extending to about 5 km. This estimate implies external benefits of about $11,758 per affected house on average, or $79 million in aggregate. The effect of media coverage of the risks of nuclear power on house prices is also examined, finding the anomalous and counter-intuitive result that more media coverage is associated with higher house prices, especially near the nuclear plant.
OAGI0001301 00004
Table of Contents I. Introduction ................................................................................................... 1 II. The Economic Theory of Externalities ........................................................ 6 III. Literature Review .......................................................................... " ... " .... 11 IV. Methodology .............................................................................................. 18 IV. 1. Hedonic Model Estimation ... ... ............ .......... ...................................... 19 IV. 2. Damage Estimation ........ ... ............. ..................................................... 21 V. Data Description .................................................................................. " .. ".32 VI. Results ........................................................................................................ 42 VI. 1. Regression Estimation ........................ .................... ............................. 42 VI. 2. Damage Estimation .......................... ............... ... ........ ......................... 49 VI. 3. Benefit Estimation: The Green Space .................... ........ ....... ........ ....... 58 Results Summary .......................................................................................... 62 VII. Discussion ................................................................................................. 64 VII. 1. Potential Problems and Biases ............ .................................... ........... 64 VII. 2. Inaccurate Public Perception of Risk ...... ............................. .............. 71 VIII. Conclusion .............................................................................................. 73 Appendix .......................................................................................................... 75 A.l. Media-Distance Interaction ........... .......................... .............................. 75 A.2. Lagged and Cumulative Media Effects ........ ............ .............. ............. .... 77 A.3. Cumulative Media Effects................................................ ...................... 79 A.4. Time-varying Price-Distance Relationships .... ....................................... 82 A.5. Age-Distance Interactions ..................................................................... 84 References ...................................................................................... ,.. """ .. " ..... 87 11 OAGI0001301 00005
Acknowledgements I would like to thank my advisor, Stephen Sheppard, for his help, guidance, and comments throughout the thesis writing process. I would also like to thank Lara Shore-Sheppard and Y oshi Konishi for their helpful comments on an earlier draft of this paper. Finally, I thank my friends and family for always supporting me.
111 OAGI0001301 00006
I. Introduction In the face of the threat of fossil-fuel-induced global climate change, the expansion of environmentally cleaner means of producing energy has been a subject of much public discussion. One solution that has been proposed is the expanded use of nuclear power. When working properly, it produces little air pollution and consumes little fuel, making it an appealing alternative to dirty fossil fuels such as coal. However, there are clearly problems with nuclear power as well. The catastrophe at Chernobyl and the accident at Three Mile Island are clear examples of the dangers of nuclear power when essential precautionary systems fail. The issues of constant exposure to radiation and the safe storage of nuclear waste are costs of nuclear power as well. Further, the risk of accidents is not the only safety issue; the revelation that nuclear power plants were considered as potential targets for the September 11, 2001 terrorist attacks makes the threat of disaster all the more real.
In this paper, I use house sale data to examine the magnitude of the negative externality associated with Pilgrim Nuclear Generating Station in Plymouth, Massachusetts. By observing the extent to which a house's proximity to the nuclear plant is associated with its sale price, I estimate the price that individuals are willing to pay to avoid the risks of nuclear power, turning this price into an estimate of the aggregate damages associated with the plant. Pilgrim Station is currently under consideration for a renewal of its operating license, which is set to expire in 2012. The impact of the Pilgrim plant on the local OAGI0001301 00007
community is certainly in the public eye, as evidenced by the numerous newspaper articles that are frequently published on the subject, and the issue of the plant's relicensing has sparked several activist groups and protests. When deciding whether to relicense the nuclear plant, the magnitude of the negative externality it generates is an important cost that must be estimated and considered.
Over the years, Pilgrim has been fraught with problems. There have been several instances of missing radioactive waste, breaking valves, and fires, and workers have been repeatedly found asleep or intoxicated on the job (see Boston Globe, 5/24/06, 1129/89, 10/31108, 10114/04,2/22/07, and 12111/91). While its record today is much better than it has been in the past, it has historically received low safety ratings from the Nuclear Regulatory Commission (NRC), once even being included on the list of the ten worst run nuclear plants in the nation (The Boston Globe, 5/24/96).
The series of problems at the plant have contributed to public awareness, along with a highly publicized 1990 study of the health impact ofliving near the plant, in which the author studied leukemia rates near Pilgrim Station before and after the release of radioactive material. They found that individuals living downwind of the plant were four times as likely to suffer from leukemia after radioactive releases compared with the same local population before the release (Clapp et. al 1990). Another study found similarly detrimental effects of accidental leaks of radioactive material from Pilgrim Station on infant mortality (Sternglass 1986). While there has been controversy in the medical community 2
OAGI0001301 00008
about the validity of these studies, they are commonly referred to in newspaper articles on Pilgrim Station, suggesting that this information is easily accessible to the public. These facts suggest that the risks of living near Pilgrim Station have invaded the public consciousness. Media coverage of these studies and more generally coverage of the dangers associated with the plant (such as newspaper articles that discuss accidents at Pilgrim) could increase public awareness of the risks associated with living near the nuclear plant. This could feasibly result in lower property values, especially with regard to properties near the plant, as homebuyers gain information (or fear) regarding the dangers of living near a nuclear plant.
The above discussion leads to the following two hypotheses, which I test.
First, that house sale prices will be positively associated with distance from the plant, holding other factors constant. Second, media coverage of the risks posed by the nuclear plant reduces house prices, especially near the plant. I test these hypotheses and quantify the value placed on these nuclear risks by using a hedonic model.
The hedonic method involves modeling property values as a function of various housing amenities, including proximity to Pilgrim Station. By considering a house as a bundle of many different goods and services (like size, style, location, etc.) which consumers can value independently, one can see the price of the entire "bundle" (the house) as made up of the prices of its component goods and services. One can model house prices as a function of these amenities to 3
OAGI0001301 00009
isolate the price paid for each particular factor, such as the price of distance from a nuclear power plant. Measuring this price of distance by using data on home sales then allows the estimation of the dollar-value that individuals place on the risks of nuclear power. Once the impact of proximity to the nuclear plant on prices is estimated, I use this to estimate the value that individuals place on the risks of nuclear power by comparing the predicted price of each house as a function of its distance from the plant to the predicted price of a hypothetical identical house at a "safe" distance from the plant. To test the second hypothesis, I by testing for a statistically significant relationship between the sale price of the house and the number of Boston Globe articles in the year of the house's sale.
Ultimately, I find strong evidence that house prices are adversely affected by proximity to the nuclear plant, with a one percent increase in distance from the plant increasing house prices by between 0.026 to 0.104 percent, depending on the model specification. With an average real sale price of$198,077.90 for Plymouth houses in my sample, this implies that doubling the distance of a house from the nuclear plant would be associated with an increase in house price of roughly $5,150 to $20,600 (in year 2000 dollars), a substantial sum. Most specifications produce estimated price-distance elasticities that are significant at the 99% level. These elasticities are used to generate estimates of the damages attributable to the nuclear plant. A range of estimates is presented in the results section, with a choice estimate of $52.9 million.
Entergy Corporation, the company that owns and operates Pilgrim Station, 4
OAGI0001301 00010
owns a large undeveloped green space that surrounds the plant. I find substantial benefits to living near the green space capitalized into house prices. In fact, on aggregate these benefits are of roughly the same size as or somewhat larger than the damages attributable to the plant itself, with a choice estimate of $78.8 million in benefits. This suggests that the undeveloped green space acts as a buffer to shield the community from the worst of the damages associated with the nuclear plant and generating some benefits of its own.
While the bundle of the nuclear plant and green space together seem to generate net benefits to the community, I still conclude that closing the plant and preserving the green space can generate significant external benefits, I on the order of$52.9 million. This is because the policy decisions regarding the status of the nuclear plant and green space can be made independently, so there is no necessary link between them. For example, the plant could be closed and the green space could be zoned to protect it from development. In order to make a policy decision regarding the plant's operation, the $52.9 million in damages would have to be compared with the value of the energy generated by Pilgrim Station. Similarly, development of the green space would require comparing the $78.8 million in benefits with the net benefits of an alternative scenario.
In the following sections I describe the methodology, data, and results from my study of the impact of Pilgrim Station on local property values and hence I Throughout this paper, I use the terms external benefits and external damages to refer to the effect of the nuclear plant and/or green space on the local community. These measures should not be confused with net economic damages benefits, which would account for all relevant costs and benefits.
5 OAGI0001301 00011
the value placed on the risks of nuclear power. In section II, I briefly review the economic theory behind externalities. In section III, I discuss the relevant literature. In section IV I outline my methodology. In section V, I describe the data, including some descriptive statistics and maps. In section VI, I describe my regression results and damage estimates. I discuss my results in section VII, focusing on interpretation and possible biases. Finally, I conclude by summarizing my findings.
II. The Economic Theory of Externalities In economics, an externality arises whenever "the actions of one party make another party worse or better off, yet the first party neither bears the costs nor receives the benefits of doing so" (Gruber 2007). In the case of an unregulated negative production externality, the private costs of production diverge from true social costs, leading to inappropriate production choices which cause inefficient deadweight loss. A simple model includes polluting firms in a competitive market. In this model, each firm's production of a good (Q) is associated with the production of a negative externality (e.g. air pollution) that harms other members of the community. The firm is not legally liable for this harm, so it does not take pollution levels into account when making production decisions. Instead, the firm makes production decisions based on its own private costs, maximizing profits at the point where the price of its production good equals private marginal cost (PMC). This produces the upward sloping supply curve seen in Figure 1.
6 OAGI0001301 00012
However, the true social costs of production are larger than this, since social costs must include the damages from pollution as well as production costs.
Thus, the social cost is equal to the sum of private costs and the damages. Social marginal costs (SMC) then are equal to private marginal costs plus the marginal damages (MD) of pollution (SMC = PMC + MD), so that SMC is an upward sloping curve located above the firm's supply curve.
Finally, since the example under consideration is a negative production externality, the demand curve is normal, meaning that the social marginal benefits of consumption equals the private marginal benefits.
Social welfare is maximized at the point where SMC = 5MB (point A, where Q = QJ, but market equilibrium is at the point where PMC = PMB (point B, where Q = QJ. This means that in a competitive market there is overproduction of the good and hence too much pollution, leading to deadweight loss D, represented by the shaded triangle in Figure 1.
Imposing a tax equal to t, the vertical distance between the SMC and PMC curves at Q, imposes larger marginal costs on the firm, shifting the PMC curve upward. With the tax, the new market equilibrium will be at Q, the socially optimal level. The tax shifted the burden of the externality onto the firms, "internalizing" it. This leads the firm to produce and pollute at the socially optimal level, eliminating the deadweight loss.
7 OAGI0001301 00013
Figure 1. Supply and Dema.nd with a. Negative Production Externality Price .:..
of Good Supply'" PMC
/
/
t
/ !.
. Demand "". PMB '" 5MB Qs Q" Quantity of Good Produced 8
OAGI0001301 00014
Turning to the subject of this paper, nuclear power arguably imposes a negative externality on the residents of the local community. The negative externality takes the form of the risk of accidents that would put residents in danger, the most obvious example being a nuclear meltdown. Other negative externalities could exist too, such as displeasure from pure unsightliness of the nuclear plant.
With regard to regulation, the primary methods of internalizing this externality in practice are legal liability and safety inspections. Placing legal liability on the nuclear plant for damages in the case of an accident. This liability does exist in practice, but it is not clear that it is sufficient. If the plant were liable for all damages in the event of an accident (and the only negative externality is due to the risk of these damages), then the risks would be internalized. Of course, there is the possibility that compensation gained in the event of an accident would be inadequate, in which case the externality would not be internalized. Indeed, in practice various factors limit the legal liability of the nuclear plant. In addition to corporate limited liability and bankruptcy, the Price-Anderson Nuclear Industries Indemnity Act is a federal law that limits the liability of nuclear plant operators, while requiring them to purchase insurance against the chance of an accident.
Another method of controlling the externalities associated with nuclear power is the use of safety regulations which directly impose requirements regarding the safety of a nuclear plant. These regulations are intended to reduce the risk of accidents occurring, and hence reduce the expected damages. However, 9
OAGI0001301 00015
only to the extent that there exist un-internalized risks due to the limitations of regulation and liability should I observe a people willing to pay a premium to live far away from the plant. If risks were fully internalized, there would be no reason to pay to avoid living near the plant because anyone harmed in the event of a nuclear accident would be fully compensated for their damages, so there would exist no net risk to the community.
In general, the goal of regulation is to price the externalities so that the emitter adjusts its operations to reach the socially optimal level of production and the externality. The efficient amount of regulation then should reflect the magnitude of harm caused by the negative externality. In order to find that magnitude, one must determine how much the affected individuals value the negative externality. As I have discussed, this value is exactly what is being reflected in differential house prices that are found using hedonic models. By estimating the price-distance relationship using a hedonic model, one can detern1ine how strictly to regulate Pilgrim Station, or whether it should be allowed to operate at all.
10 OAGI0001301 00016
III. Literature Review Hedonic models have been used in the past to measure the impact of nuclear power plants, but the results have been mixed. Some studies have shown plants negatively affecting property values, some have found no statistically significant effect, and some have even found positive effects.
The first major study of nuclear power's effect on house prices was presented by Jon Nelson, in which he examined the impact of the accident at Three Mile Island in March of 1979 (Nelson 1981). Using home sale data, he investigated the impact of two independent variables on prices for houses within four miles of the plant: a dummy variable representing whether the sale was before or after the accident and a variable that interacted this dummy with the date of sale (effectively allowing the accident's impact to vary over time). After controlling for characteristics of the house, he found that the accident did not cause house prices to fall or even slow their rate of appreciation. Nelson hypothesized two possible reasons for this: (1) the damage was perceived a temporary, short-term shock, and so was not reflected in a market of farsighted buyers and sellers and/or (2) expected state and federal aid acted as insurance policies that offset any potential losses.
Gamble and Downing looked at the Three Mile Island incident as well, but they also analyzed other plants that did not have accidents (1982). In their study of the other plants, they used home sale data to examine the impacts of two variables: the nuclear plant being visible from the house and the house's distance 11 OAGI0001301 00017
from the plant. Under several different specifications, the coefficients on each variable were statistically insignificant after controlling for characteristics of the house. They applied this same model to the Three Mile Island plant, with the addition of a dummy variable indicating if the sale occurred after the accident, as well as a variable that interacted this dummy with the house's distance from the plant. Their results indicated that the coefficient on distance from the plant was significant and positive before the accident, but insignificant afterwards. Gamble and Downing attributed this change to the fact that the Three Mile Island incident was followed by the large-scale hiring of clean-up workers and nuclear technicians, potentially increasing the housing demand near the plant as the new hires moved in.
In 1986, Galster critiqued this study for its implicit long-run focus.
Pointing out that the housing market collapsed for a brief period following the incident, he argued that the finding that there were little to no long-run effects of the Three Mile Island incident on house prices ignores the important short-run impacts (Galster 1986).
In 1997, Rephann used a quasi-experimental method to analyze the impact of various "LULUs" (locally unwanted land uses), including nuclear power (Rephann 1997), finding that nuclear plants significantly increase local tax revenue. He argued that the common finding that nuclear plants do not significantly affect house prices is due to the negative effects being swamped by the positive effects on the community of massive local property tax revenues.
12 OAGI0001301 00018
Large amounts of tax revenues from nuclear plants can increase the amount of public services offered by the local government or by decreasing the tax rate needed to achieve a given level of public services. Both of these effects increase the value ofliving in that community. As a result, tax revenue from nuclear plants can increase house prices. This positive impact could hide the negative effect on property values of the risks posed by nuclear plants.
In reference to the employment effect, Rephann finds evidence that nuclear power plants do not significantly stimulate local employment because they generally perform national searches for a few highly trained specialists and because they are generally a high-capital, low-labor business. This finding is helpful for the interpretation of observed price-distance relationships in this paper, since it implies that estimates for the house-price impact of the risks of nuclear power are not strongly biased by employment effects.
A 2006 study by Bezdek and Wendling examined the areas around seven nuclear plants across the United States. They investigated the effect of the plant on house prices and employment, but they also considered the effect of increased tax revenue from the plant on local incomes, community development, and schools. They found that in each case the creation of the plant was associated with large net benefits for the community. Property values in regions with nuclear plants tended to appreciate on par with the state average, and economic growth in these areas following the construction of the plant exceeded the state-level average. Bezdek and Wendling attributed much of this growth to the nuclear 13 OAGI0001301 00019
power plant: for example, the tax revenue from each plant in their sample funded 50% or more of the budget for its county and school district. However, they made some caveats to this claim, such as the fact that the areas where the plants were built were initially in poor economic shape, so their disproportionately high growth could simply be the region catching up to the mean.
The issues emphasized in Bezdek and Wendling's study should not threaten to bias this paper's results. First, Plymouth is not an economically depressed area: its median household income of $50,838 in 2007 was remarkably similar to this figure for the United States as a whole: $50,740, and a smaller proportion of the Plymouth population falls under the poverty line, 7.4%,
compared to the Massachusetts state average of 9.3% (2000 Census, City-Data).
Second, the potential positive effect on property values due to tax revenues should not bias this paper's results, since the positive effect should increase house prices evenly across the entire town. This implies that the observed price-distance relationship should not be biased, since the tax revenue from the nuclear plant should affect all houses equally.
In contrast to the above findings of insignificant or even positive impacts of nuclear power plants on house prices, Folland and Hough's 1991 cross-region study found that regions with nuclear power plants had significantly lower house prices than those without. This finding was reinforced by another cross-region study by Clark and Nieves in 1994 that looked at a variety of noxious facilities including nuclear power plants. They found that significant negative impacts on 14 OAGI0001301 00020
house prices for each nuclear plant within a set distance of the property. An objection to these findings is that when searching for a location to build a nuclear power plant, facility builders tend to seek out cheap land for construction. As a result, this negative interregional association between house prices and the existence of a power plant could simply reflect the tendency of plants to be built in areas with initially cheap land. In response to this criticism, Folland and Hough (2000) followed up their initial study by controlling for this tendency of builders to seek out cheap land by comparing house prices before and after the announcement and installation of each power plant, and the negative impact of nuclear plants on house prices persisted. They estimated a nuclear plant "installation effect" of about -10% on house value. Further, the authors criticized the studies of Three Mile Island by pointing out that the area was actually safer after the accident, since the reactor went offline, so the findings of no effect on house prices is not necessarily inconsistent with the existence of a nuclear-based negative externality in this instance.
In another application of the hedonic model to nuclear power plants, a study by Clark, Michelbrink, Allison, and Metz (1997) examined two California nuclear power plants: one in Diablo Canyon and one in Rancho Seco. In each case, Clark et al. found that house prices actually increased with proximity to the plant, suggesting that any detrimental impact of the plants could not have been large, if it existed at all. In a related study, Metz and Clark (1997) also examined the impact of media coverage of these two plants on house prices. They analyzed 15 OAGI0001301 00021
the level of newspaper coverage regarding plans to construct a new dry storage nuclear waste facility at these plants, finding that this coverage had only minor impacts on prices. Interestingly, the new dry storage facility was not housing more waste; it only changed the type of storage of pre-existing nuclear waste.
Although small, these price impacts could simply reflect changes in the salience of this nuclear waste storage among the public, not actual changes in risk. In this study, I similarly test the relevance of this kind of "salience effect" by testing for a relationship between media coverage and sale price.
Clark and Allison continued to analyze the impact of newspaper coverage of local nuclear power plant issues on housing prices in their 1999 paper. This study focused on the Rancho Seco plant examined in their previous papers. After including a large number of controls, they found that reminders of the plant in the forms of visual cues, proximity, and media attention had statistically significant negative impacts on house prices, although these impacts were generally small (Clark and Allison 1999).
Gawande and Jenkins-Smith performed a study in 200 I that examined the impact of a series of highly publicized shipments of nuclear waste in South Carolina on house prices. This study differentially considered the price impacts in different kinds of areas: those with low risk perception and those with high risk perception. Areas that had more experience dealing with nuclear waste tended to have a lower perception of the risk of the shipment, and correspondingly the house-price impact of these shipments was not statistically significant. However, 16 OAGI0001301 00022
urban areas considered these shipments more risky, and there was a significant negative impact on house prices there. They argue that "it may not matter whether public perceptions of the risk [on which price impacts are based] are accurate" since the impacts on house prices are nevertheless real. In terms of legal cases, they also point out that negative price impacts based on public risk perceptions are sti11legally considered grounds for a lawsuit, "even if the public perceptions of risk are not deemed 'reasonable'" (Gawande and Jenkins-Smith 200 1, p. 230).
In sum, the evidence on the impact of nuclear power plants on local property values is mixed, and the issue remains contentious. Many studies find no effect of nuclear power plants on local house prices, while others argue that there does exist a significant negative effect. Still others argue that nuclear power plants benefit the local community through various channels: one channel being the large amounts of tax revenue they generate, and a more controversial one being the claim that physical proximity to nuclear plants actually directly increases property values. Further research is needed to settle this contentious issue, and in this paper I hope to aid in the progress toward that goal by examining Pilgrim Station, keeping in mind the problems and biases discussed in past studies on the Issue.
17 OAGI0001301 00023
IV. Methodology In order to estimate the impact of the Pilgrim nuclear plant on house values, I use a hedonic model following Rosen (1974). These models generally assume that houses contain a bundle of services. Since homebuyers take into account the array of different services that homes provide, their willingness to pay for a house can then be broken down into a different willingness to pay for each individual service. By estimating the magnitude of the effects of each housing amenity on the price of a home, I can estimate consumer willingness to pay for certain environmental goods. For example, when comparing two otherwise identical houses, save for the fact that one is located near a nuclear power plant, one would expect this house to fetch a lower price than its twin. The difference between the prices of these houses can then be considered a rough measure of the value that individuals place on avoiding the risks of living near a nuclear power plant. In this paper, I use this type of method to examine the value that individuals place on the risks and unsightly imagery of nuclear power in Plymouth, MA.
The idea laid out above explains the core intuition behind the hedonic price method, but further elaboration is required. In this section, I formally layout the methodology that I use to estimate the value placed on avoiding living near Pilgrim Station.
18 OAGI0001301 00024
IV 1. Hedonic Model Estimation In my analysis, I assume that house prices can be modeled as a function of various housing amenities of the following fonn:
11 Log(Real House Price)j = f30 + f31Log(Pilgrim Distance)j + ~f3 jX jj +E (1) j~2 where each x ji variable represents an amenity of property i such as number of acres, house style, and local demographic statistics, with} indexing the n different housing amenities included in the model. These variables act as controls, so that I can isolate the effect of the primary variable of interest: the distance of the property from Pilgrim Station ("Log(Pilgrim Distance)" in equation (1>>. In the absence of bias, estimating f31 will then provide a measure of the "price" of reducing the risks of nuclear power.
Note that the functional fonn assumed above is a log-log model; this model has many advantages over a basic linear model. The log-log model has been used extensively in the hedonic studies, in which it has performed reasonably well at fitting the data and measuring marginal prices (Sheppard 1999). One advantage is that the coefficients on the log variables represent elasticities. Thus, f31 represents the percentage change in house price from a one percent increase in the distance from the nuclear plant. This makes interpreting the results from the log-log functional form intuitive and easy.
Second, the logarithmic functional form is one that allows the magnitude of the dollar-denominated effect of distance on price to diminish with distance, as intuition would imply. Clearly a one-kilometer increase in distance from the 19 OAGI0001301 00025
nuclear plant should matter more near the plant than it does further away because the damages of a nuclear accident would dissipate with distance, and the logarithmic model incorporates this idea. Third, the logarithmic form also "pulls in" outliers, so that properties with extreme prices or distances do not single-handedly drive the results.
I also include a variable representing the extent of media coverage of the dangers of nuclear power: the logarithm of the number of Boston Globe articles in the year of the house's sale2 that prominently feature Pilgrim Station and the risks to the community associated with the plant. The inclusion of this variable allows me to test for the existence of an effect of media coverage on the public perception of nuclear risks and hence on house prices. The coefficient of this variable represents the relationship between the extent of media coverage and house prices.
In this paper, I estimate equation (1) with several variations. I estimate the basic model outlined above, using two different sets of house style indicator variables to find the most appropriate model. I then estimate the two models including year fixed effects to control for year-to-year shocks (such as the burst of the housing bubble) and to gain a better sense of the effect of the media variable, which could produce anomalous results if it is associated with year-based shocks.
I use the results of these models to find the damages associated with the plant.
2 The main body of this paper uses the number of Boston Globe articles in the year of sale, but other time periods and specifications are explored in the appendix.
20 OAGI0001301 00026
IV 2. Damage Estimation To measure the economic efficiency of policy decisions regarding the operation of the negative externality-producing power plant, I must estimate costs and benefits. In the case of a nuclear power plant, the benefits of closing it are simply equal to the negative of the damages. These damages can be estimated as the marginal willingness to pay to avoid pollution (MWTP) that is reflected in house prices. However, as Figure 2 demonstrates, the price of avoidance and the marginal willingness to pay/unctions are not the same. MWTPI represents the marginal willingness to pay curve derived from the utility function of a hypothetical consumer 1, which should feasibly decline with distance. The price of increasing a house's distance from the plant represents the house price premium placed on increasing distance by one unit from the equilibrium hedonic price function. This is assumed to decline with distance in Figure 2, reflecting the declining value placed on distance father away from the plant, where exposure to radiation would be less sensitive to marginal changes in distance.
Given this price function, consumer 1 will "purchase" distance until its price is equal to the marginal willingness to pay for an additional unit. The same applies to consumer 2, who is more averse to living near the nuclear plant and hence buys a house farther away than consumer 1. Ideally, one would like to directly observe the marginal willingness to pay functions. Unfortunately, one only observes the prices paid for houses, although the hedonic method allows one to separate out the price paid for individual characteristics such as distance from the nuclear plant. Still, the observed prices provide a clue about each consumer's 21 OAGI0001301 00027
underlying willingness to pay. For both consumers, the value placed on the initial units of distance exceed the price paid for them,3 so the true willingness to pay to reduce the risks of nuclear power (and hence the damages of these risks) are larger than the price indicates. Thus, the price of distance (or, equivalently, the measured damages to a house) understates the true willingness to pay to reduce the risks of nuclear power. This result is analogous to the market for any good, in which the price of the good is less than or equal to consumers' willingness to pay.
If this were not the case, consumers would not purchase it.
3 This assumes that the marginal willingness to pay curve is steeper than the price curve. This is a reasonable assumption when one considers the alternative, in which the consumer would want to purchase a house infinitely far from the nuclear plant.
22 OAGI0001301 00028
Figure 2. The Price of Distance and lXIW'TP Unit of Distance While the "price of distance" from the nuclear plant understates the willingness to pay for a house farther away from the nuclear plant, the observed price-distance relationship can still be used to estimate damages attributable to the plant. Consider Figure 3, which shows the hypothetical supply and demand for housing in a region. D2 and D 1 represent the demand functions for properties that are identical in every \-vay except for their exposure to the negative externality, with <<D2 houses" being adversely affected and "D1 houses" being "safe," or unaffected. If housing supply is fixed and invariant to distance iTom the plant (a case represented by Sl), then the difference in marginal willingness to pay from increased exposure (D2-Dl) \vill be fully ret1ected In difference inpTice (PI-P2).
23 OAGI0001301 00029
Since an individual buys a house at this price, the observed difference in price corresponds to an individual's intersection point between his or her marginal willingness to pay curve and the price line in Figure 2. In this case, the price differential between the houses is an accurate measure of the damages from the negative externality.
Figure 3. Supply and Demand fo.r Ho.uses 51 52 (l)
U PI I....
0-(l)
Vl
- l o
- r:
Quantity 24 OAGI0001301 00030
Housing supply can reasonably be considered to be fixed in the short-run, when houses cannot be built quickly in response to changes in prices, but what if housing supply is upward sloping? This case is represented by 82 in Figure 3, showing that the difference in price (PI-P3) will not reflect the full difference in marginal willingness to pay. Thus, directly measuring the difference in average house prices gives a lower bound for the difference in marginal willingness to pay due to the increased exposure to the negative externality. While recognizing the fact that the observed difference in house prices based on distance to the plant will likely understate the true willingness to pay to avoid pollution and hence understate economic damages, estimating such price differences can still provide a useful lower bound measure of the magnitude of these damages. Once an estimate of the damage for each house is calculated, I can sum this estimate over all houses to give an estimate ofthe total damages of the negative externality.
The process described above works well when one has a natural experiment, but unfortunately I do not have one. For example, ifI had data on house prices in Plymouth before and after the plant was announced and built in 1972, then I could compare house prices before and after this time to accurately estimate the effect of thenuc1ear plant. It becomes more difficult when one has data for only non-experimental situations such as this. In such a situation, one must assume that the houses that are sufficiently far away from the plant are "safe," meaning that the price is unaffected by the negative externality associated with the nuclear plant. Then, I can assume that the price level of these houses 25 OAGI0001301 00031
'\vould prevail (ceteris paribus) for properties near the plant as welL In other words, 1 choose properties that are significantly far ,1\vay from the source of the negative externality as "safe" houses to which propelties near the plant cali be cornpared. The difference in prices between these sate houses and houses near the nuclear plant (holding other differences constant) can then be used as a measure of the damages of the negative externality. To elucidate the damage calculation mathematically, recall equation (1):
Log(Real House Price)i /30 + j31Log(Pilgrim Distance)i + 2f3jx jf + £ Raising e to each side of the equation gives the real house prices as a function of1og~distance and the other independent variables:
Note that the left side of this equation simplifies to the real hOllse price.
Once equation (1) is estin1ated. I usc the estimated }1 coefficients to predict real house prices for each house:
Predicted Real House Price; ':::: e Denote the distance that 1 assume to be the point beyond which houses are considered in the "safe" group as Dc' Comparing the real house price of house i that is distance D, from the power plant to the predicted real house price of a house that is Dc away results in equation (2), \vhich represents estirnated damages:
.:.b OAGI0001301 00032
_,. ._ t:J _. ~ _ .0 ~.
Po +t>lLag(D:)+ ~ fljXji 13u+t>JLl)g(Dc }+ 2;/'3;X j:"
Estimated Damages;(DJ:::: e FJ ~e j<, (2)
Estimating the damages using equation (2) for each house in effect compares each house's predicted price to a hypothetical identical house that is Dc kilometers aW'ay, attributing the price difference to the damage from proximity to the plant. Graphically, this estimation technique is similar to computing the difference ahove the "predicted price" line and helow the "distance cutoffline" in Figure 4 ata given distance. 4 Figure 4.
Damage Calculation Illustration 140000 130000 +.....*....... " .............*......*....... " ..* " .* "" .*.... "" ....*..*..*....*. :=;>.-""""'....................... .
,20000 +..................................... :""""~:c .......................................*.................................................*
110000 Col
- ila. 10DODD
- - Prf~dj(;ted Pr;Cf~
<!l
.......:.: .... Distance C:..:tnff 1: 10 km o~ 90000
- t:
SODOO :1 .*..*." ,- ,. ... ,-" ..* ' .. ............. *.*. _ ...*....*..*.*.... .......*..*....*....
lOOOQ l***************************************************** ...... .
60000 , ....................................... " ...................................................... ".
50000 .,..- ..--.--..-....-.-....*... -..- -..".-........-.-....- .......... - -................................................**
o s 10 is 20 Distance {km) 4 I say ":-;imilar" because it is not quite identical to this procedure. The difference is that the predicted price curve in Figure 4 is based on the control variables taken at their sample means, whereas the damage calculation us cd uses the actual values of sample variables for each observation.
27 OAGI0001301 00033
Since the price-distance relationship should be zero at distances beyond Dc, direct damages should theoretically be zero for houses where D; 2: Dc.
However, as Figure 4 illustrates, the logarithmic functional form is strictly increasing in Dc, so damages for these houses are measured as negative.
These '~negative damages" are excluded from my analysis 1~Jr several reasons. First, the distance cutoff should theoretically be chosen to be a point beyond vv'hich houses are no longer aHeeled by the nuclear plant. Otherwise, this is not a proper choice of comparison group, Second, if prices are believed to continue to rise beyond Dc, then these estimated "negative damages" would be unbounded \-vith distance, For cxample, houses 1000 kilometers away would be predicted to have exceedingly high value because of the assumed logarithmic functional foml, which is strictly increasing. Of course, this relationship is not likely true at extreme distances. In my calculations, "negative damages'" are only finite because the sample is finite and limited to the town of Plymouth, \vith the sample's maximum value ofthe distance from the nuclear plant 0[20.49 km.
Finany . the higher prices for houses furthcr away may he appreciated by their residents, but they still represent a willingness to pay to avoid a negative extemality, so they actuany ret1ect diminishing damages, For hOLlses within Dc of the plant, damages will be positive, decreasing in D;, and increasing .111 Dc- The first two ofthese three effects aIe dearly illustrated in Figure 4. Figure 5 demonstrates the fact that darnages are increasing in Dr" Cleclrly, the choice of a 20 km cutoffrather than a 10 kIn cutoff results in larger 28 OAGI0001301 00034
damages for every house, as the difference between the cutoff line and the predicted price line becomes larger with the high(,'f cutoff choice.
Figure 5. 5 Damage Calculation Illustration 140000 130000 120000 1.iOnnO 1:: 100VQQ
~ - Predkteo Pdce
- J
~.'~.'. Distance Cutoff I: l() km Ofstance Cutuf't 1: 20 km 0 90000 z:
SOOOO 70000 6aona 50000 0 10 15 20 Olsta"<<, {km)
The iJ1tuition for this relationship is simple: increasing the distance cut-off places more houses into the "damaged" group, as well as increasing the estimated hypothetical <<unaffected" price. As a result, Dc must be chosen carefully. In the results section, I estimate dan1ages for several values of Dc to illustrate the leve1 of sensitivity of the results to this choice and to provide a sense of the range of the damage estimates.
Once damages are estimated for each house, the total damages due to the nuclear plant can be calculated as the sum across an houses in Plymouth. Since 5 For the graphs illustrating the damage calculation technique, I have USi:.ed a hypothetical relationship wher.; Ihe predicted value of a hOllse is S 100,000 at 1 kilometer away f:rom the plant, with un elasticity of price to distance of 0.1 {ie. + 1'i{. distance implies +0,1 % house price).
29 OAGI0001301 00035
my sample does not contain every house in Plyrnouth, smnming damages across the sample will understate dam.ages. To (*on:ect for this, these damages are summed for each Census Blockj and scaled bya factor Ct) according to the sample's misrepresentation of the actual number of houses in that block. For example, if there arc 1000 houses in a block as measured by the year 2000 Census, but the sample only has 500 houses in that block, then the damages are summed across all 500 observations and multiohed by a:. == 1000 = 2 in order to
. ~.1 500 replicate the damages for the unobserved houses In that block. Then, estiuHlted total damages arc given by:
Total Damages 2: a j
J LDamages{ D j )
IE}
On average, each block contains about 1.88 as many households as observed in my sample according to the Census, so in effect this correction roughly scaks damage estimates by 1.88. A house's distance fTOm Pilgrim Station is negatively associated with ai' with a 1 km increase in distance being associated with a 0.023 decrease in a} (significant at the 99'% 1evel). This nleans that the sample was drawn more heavily fl"om areas neat the plant, but the magnitude ofthis relationship seems smail. This could suggest some sample selection hias,. but itis not immediately clear which direction this would hiasthe estimates.
As I have argued, the damage estimate would then be biased downward to the extent that prices understate true wiHingness to pay. However, there is also the 30 OAGI0001301 00036
possibility that prices for houses in the "safe" gmup are artificially high. This could be so if the presence of the negative externaLity shifts housing demand further away from the plant, pushing up hOllse prices in the "safe" group . This
\vouid bias estimates of damages upward, so the true net direction of these biases is unclear. Unfortunately, this is a potential bias that cannot be controlled fOT in the absence of an experimental situation.
31 OAGI0001301 00037
v . .Data Description The bulk of the data was obtained from the Plymouth Tax Assessor's offIce. After discarding ohviously problematic observations (e.g. houses with a real sale price under $5,000 in year lDDD doUars}, there exist well over ]3,000 observations, with theb111k ofthcm dating back to 1992, as well as a few observations from the 19708 and 1980s. The data include many variables describing the house, including most importantly sales price, sales date, street address, and number of acres.
I t1)cllsed the analysis on residential house sales that were sold at arms length as a single lot (i.e. not tied to the purchase of another lot). I focus on residential houses because commercial property prices are unlikely to rdlect the negative externalities of living near the nuclear plant, biasing estimates of the true impact on Plymouth residents. I do not include properties that \'\-'cre sold as part of a bundle package of many lots because the only price that was recorded for these sales was the aggregate price for the entire bundle. As a result, the saleprice of such houses is unlikely to reflect the true value of the single lot of land. Finally, I excluded houses that were not sold at "arms length!" A sale that was not done at "arms length" is defined as a sale within a famUy or corpora.tion,or a sale due to bankruptcy or foreclosure. In these cases, the sale was likely done at a price well below the tRW market value. Indeed, including these observations in the model along with indicators for each of these characteristic results in large, significant, 32 OAGI0001301 00038
and negative coeJ11cients on each indkator.6 Even after excluding houses that are not residential, single lot, and sold at arms length, the sample stilI has over 10,000 observed sales.
In order to examine the hypothesis that media coverage has an effect on the damages of the nuclear plant to property values, I searched the Lexis-Nexis database of The Boston Globe, tallying the number of articles from each year between 1992 and 2008 that significantly discuss the risks of nuclear power.
Articles that only tangentially referred to the plant (e.g. as a frarne of reference when describing the energy output of another plant) are not included in this taUy.
There \Vi.:Te only a few articles that reported positive aspects about the plant, and these articles \vere not included in the ta11y. The mean number of articles over the 17 -year period is 18.7 articles per year, \vith g years having fewer 1l@1 18 articles, and 9 years having more. Figure 6 S110\V5 the distribution ofthese articles over time.
6 Including these indicators in the model docs not signifkantlyaffect the estimated impact of proximity to tlwnLlclem plant, but it reduces the R-squared value.
33 OAGI0001301 00039
Figure 6. Boston Globe articles over time 60 SQ 20,---------'c------ ---<-------------------------------.------------------c_--------------.c-------------------------------
lQ -:c----------------------------------------------------._c-c_-___________ c_________cc ___ _
~.-------------- -------------------------.------.------------------------------------------------------------~--------'".
1£190 1994 1996 1998 2000 ~002 2GD4 2UQ6 200B Yeal' I also collected the monthly Housing Price Index (HPJ) for the Boston m.etro area from the OUIce of Federal Housing Enterprise Oversight (OFHEO). I use the Boston HPI to denate the observed nominal sales prices in the raw data.
This denation servest\?,Io purposes: first, it contTols for general inflation in the economy, and second, it controls for regional trends in house prices. The latter adjustment is especially important when considering the recent boom and bust in house prices in the United States, Boston included. However, even with this deflation of nominal prices, the real price data for Plymouth appears stilt to sho"\v a pattern of boom and bust in recent years, as can be seen in Figure 7. The persistence of this pattern indicates the fact that Plymouth was hit harder than the general Boston area by both the rise and fall of house prices.
34 OAGI0001301 00040
Figlu-e 7. Annual Average House Prices Deflated by Boston HPI, Year 2000 dollars 23{)OC{) ..................................................................................................................................................
220000 210000
~
~'" J.OOO(;U , ................................................................................. ,., ............................................... 00 *. , . , . . . . . .
~
c::
~ 19GaoO 5
b 180000
.~
=-
- << 17000C
~.
~5QQO<l .:...... ............. .. .......................................................................................................................
!400[Rl .l.. __ . ________ .. _........... ---.. --- ...............-..- ...- ...........- - -.. --- .. ---.**----..-- .. -... ~---~.....- ........- ..;
1992 1994 1996 1998 :woo 2.002 2004 2006 2006 Yeal' I used the program ArcMap (a Geographic Information System program) to measure 1he distance of each house to the nuclear plant, as well as the distance to several other relevant locations, by using the house's recorded street address.
For each house, I calculated a distance to each ofthc fonowing locations: the Pilgrim nuclear plant, the state parle, the shore, the highway, and the Pilgrim green space (the wooded area surrounding the nuclear plant). The means and standard deviatie}l1s of the independent variables induded in my models are reported in Tables 1 and 2.
35 OAGI0001301 00041
TablcL Summary Statistics n = 10,428 Variable Mean Std, Dev.
Real Sa Ie Price (HPt Jel1ej,d, YGa,. ZOOG duU",,) $198,077.90 $94,734.44 Adjusted TotHl Acres 0.589 acres 0.970 acres Distance from PilgrilTl Nuclear Plant 9.615 kilometers (km) 4.1% km Distance from Slate Park 5.282km 1.977 km Distance from Shore :U28 km 2.ll1 km Distance from Pilgrim Green Space 5.258 km 3.205 km Distance from Higb\vay 2.259 km ].631 km HouseAge 29.88 years 36.64 years Finished .Area 1860.843 sq. ft 2029.M sq. ft.
No. of Stories 1.69 0.55 No. of Rooms 6.57 2.19 No. Qfl3edrooms 2.96 1.08 No, of Full Baths 1.66 0.76 No, of Ha If Baths U.49 054 Percent of House Air Conditioned 44.6m<<
80S/Wi C; lobe articles in year of sale 20.{)6 13,Cil1J - - - - 1
.c-.
Sample Description Proportion of
~__-_--~-_-~---~--fat~g~-!:1~-----~-~-_~- ____ Ch~!,~~~!~_~isti~_____-----_ Pr-~.P-~E.t!_~~_
Yes 84.26%
Owner Occupied?
No 15.74%
A,vcrage 7302.%
Condition Good 26.22%
Ranch/Raised Ranch 19.65%,
Colonial 19.35t~1, Cape lS.53i %>
Condo 11.72t}~}
Building Type Contemporary 6.86%
Garrison 6.74%
Gambrel 5.29,}o Townhouse 3.51%
Cottage 1.64%
Regular 88.60%
Roof Type Metal 6.]5';";'
I:-~___________________________________________ j:4_~~~!"tE~~~ ___ ~ "--- 4.47% .~~"----l Residential- Medium [_,ot 37,57:;"c)
Residential- Rural 25.58'}~)
Znne Residential~" Small Lot 19.49%
._---- Residential- i\,iixed Density 11.86%
None 82J)5%
Changed Assessment 5.00%
Bank Sale 2.84'Yo Court/Probate 2.25'Y(,
Sale Information Other 1.93%
Changed Sale L84'%
Ptlrtial {nIs. [ .2! 'J,.-i,
'--_ _ _ _ _ _ _ _ _ _ _ _ _ _ _- L_ _ _...:(:::.~(:::H:.:.1V~*t;::'n~i.::en:.:.c::::e::.,' _________ ______U~_~~~______
36 OAGI0001301 00042
Average Household Size 2,74 tU6 PerccntCaucasian Percent of Properties Vacool 0.34% 7.94";,;
Percent of Properties Owner Occu pied 28.84%, 5.45%
- >!o'nK'. dcnlogr£lphjcslal:stj~s presented h"::l"(:: dnnot accuratety rq11"($Cnt the dC1110graphic m~lh:~up bfPlymouth a.s a.wlioJc. Rather, the statistics here arc;iv~rilg,~d across the s3mpic~ m~anlng f!~lat they represent the avCt,lgc demographics of the cDJlunuuitymnong aHhouscs, Tn (lthcr v.. ofdr~, ti1c.dC:nOlnl11ator \\"h~navcraglng.is the nunlbcr o1chousc*(Jbsl;.~tvati{)n::;.
A fe\v notes nmst be made on the control variables listed in Tahle 1. The reader may notice that many ofthe continuous "logged" variables are labeled as
~'Adjusted," such as "Log(Adjuskd .House Age)." For each ofthese, the underlying variahle was equal to zero in the original data for some homes, For some of these variables, this is logical (forex3mple, it is possible and common for a house to have an age of zero at the time of sale), but the logarithm of zero is undefined, Foran ofthese variables except "Log(Adjusted Total Acres)," I have added one to the underlying variable., and then taken the logarithm, so that these variahles can he included in the model while preserving the lmderlying relationship between the variables and house price. For the total acreage, some entries were recorded as having zero total acres. To help correct for these missing values, I adjusted the total acres for the observations to equal the total fillished area of the house in order to approximate thcminimnm size of the plot of land that the honse must be built upon.
OAGI0001301 00043
The reader may notice that several other location-based independent variables are included in the model. One may worry that this would pose problems of collineanty, which would manifest itself in the fonn of large standard errors. Fortunately, there appears to be sufficient independent variation in the data to provide reasonably precise estimates, as most estimates of the elasticity of house ptice to the .distallce from the nuclear plant are significant at the 99% level.
Figure 8. Sale Date 20GO Z(:05 Sale Year 38 OAGI0001301 00044
The span ofyeaJs fbr which I have sales data is fairly broad. Most sales 7
faU between 1992 and 2008, with only five sales occurring before 1992. Figure 8 illustrates the distribution of sales over time, showing that I have a broad distribution of observations over time. There is a trend of an inHeasing volume of house sales leading up to 2005, and falling thereafter, but this is unsurprising given the national housing boom and bust corresponding to this time frarne.
Further, the level of media coverage is tlnJeiated to the number of sales in each year. The correlation between the number 01'.8051011 Globe articles and the number of house sales in a year is ,,"0.002. There dOeS seem to be a significa.nt relationship between the location of sales and the amount of med1a coverage. /\
univariate regressionofthe annual average distance from the nuclear plant among all sales on the number of Globe articles found a negative relationship that ,vas signiikant at the 95~0 level. Ho\vever, the magnitude ofthis relationship is smalL with an additional article being assoc-iated with a reduction of 8 meters in average distance from the nuclear plant. This seems small in comparison, when considering that the average distance from the nnclear plant among all houses is 9.6 km.
Figure 9 presents a map of Plymouth, Massachuselts, created using ArcMap. This map incLudes a large red dot indicating the location of the Pilgrim nuclear plant, and smaller blue dots representing hOllse sale observations that are included in the model. This map provides a sense of the layout ofthe area of
, Because there is nnindlcation ofwby these five old sales 'A;ere induded in the data :;et. whereas other old sales were not, I have excluded them from the analy&is.
39 OAGI0001301 00045
study.
40 OAGI0001301 00046
Figure 9. Plymouth Map s
10 Kiiometers 41 OAGI0001301 00047
VI. Results In this section, I present the estimates of the relationship between the price of a house and its distance from the nuclear plant. I present a range of estimates, from a low elasticity of 0.026 to a high of 0. 104, using four different variations of the model. With these estimates in hand, I then proceed to calculate estimated damages associated with the nuclear plant. I further calculate these estimated damages using several different cutoff distances (Dc) as described in section III. I also estimate the benefits of the Pilgrim green space using a similar method and compare them with the estimated damages of the power plant.
VI. 1. Regression Estimation The initial model (model 1) includes the control variables listed in Tables 1 and 2. Column 1 of Table 3 shows the results of this regression. For purposes of space, the results for most control variables have been omitted, and only the following variables have been presented (all in logarithmic form): number of relevant Boston Globe articles in year of sale, 8 and the distances to the Pilgrim nuclear plant, the shore, the state park, and the green space.
Since this is a log-log model, these coefficients can be interpreted as elasticities. The results from this model imply that a 1% increase in distance from the nuclear plant would increase the price of a house by 0.0442% (ceteris paribus). This coefficient is significant at the 95% level. The sign of this 8 While all of the models presented in this paper include a media coverage independent variable, removing this variable from the models does not substantially change results.
42 OAGI0001301 00048
coefficient is sensible-increasing distance from the plant reduces the exposure to the negative externality of the nuclear plant, and so people should be willing to pay a premium for it. Further, the magnitude seems roughly appropriate as well; to put this in perspective, the compare this with the coefficient of the control variable for Log(Finished Area in Square Feet) of 0.3123 or the coefficient of the Log(Shore distance) variable of -0.0588. Thus, proximity to the nuclear plant appears to have a somewhat weaker impact on house prices than proximity to the shore does, but the effect is much weaker than that of having a larger house (in finished area).
43 OAGI0001301 00049
Table 3. Regression Results Dependent Variable: Log(Real House Price)
Coefficient (1) (2) (3) (4)
Log(Pilgrim Distance) 0.0442** 0.0260 0.1040*** 0.0861 ***
(2.39) (1.45) (5.65) (4.82)
Log( Globe articles) 0.0179*** 0.0163*** 0.0188*** 0.0195***
(4.96) (3.19) (5.12) (3.74)
Log(Shore Distance) -0.0588*** -0.0641 *** -0.0685*** -0.074***
(-14.03) (-15.80) (-16.31) (-18.22)
Log(Pilgrim Green Space Distance) -0.0327*** -0.0239*** -0.0564*** -0.0476***
(-3.89) (-2.93) (-6.76) (-5.87)
Constant 9.32*** 9.52*** 7.71*** 7.94***
(23.63) (24.91) (19.38) (20.57)
No. of House Style Indicators 48 48 9 9 Year Fixed Effects? No Yes No Yes R-squared 0.6138 0.6392 0.5976 0.6232 Adjusted R-squared 0.6096 0.6348 0.5947 0.6199 No. observations 10428 10428 10428 10428 t-statistics in parentheses
- , **, *** indicates significance at the 90%, 95%, and 99% level, respectively 0
G>
a0 44 0
w 0
I 0
0 0
(J1 0
The other coefficients reported are interesting as well. The coefficient on (the log of) distance from the shore has the expected sign and a sensible magnitude: a 1% increase in distance from the shore decreases house prices by 0.0588%. The coefficient on the distance from the Pilgrim green space similarly has the expected sign and magnitude.
Of these, the anomalous coefficient is the one on the logarithm of Boston Globe articles. According to this model, a I % increase in Globe articles discussing the dangers of the Pilgrim nuclear plant in the year of a house sale is associated with an increase in the sale price of a house by 0.0179% (significant at the 99% level). The sign on this coefficient is certainly counterintuitive, as one would expect that media coverage of the risks of nuclear power would increase the salience of nuclear risks and hence drive down house prices in Plymouth. One possible explanation for this result is that the Globe article variable is highly correlated with year-specific price shocks. In order to control for this possibility, I add year fixed effects to this initial model. The results of this model are presented in column 2 of Table 3.
As the table shows, the inclusion of year fixed effects does not substantially change the coefficient on the media coverage variable. The sign and significance of this coefficient are still troubling, with a 1% increase in media coverage being associated with a 0.0163% increase in house sale price. This anomalous result is explored more in the Appendix but put aside for the moment because of the lack of clear explanation, even under models 3 and 4.
45 OAGI0001301 00051
The inclusion of year fixed effects also reduces the estimated magnitude and significance of the coefficient on the Log(Pilgrim Distance) variable relative to model 1. Modell shows an estimated elasticity of price to distance of 0.0442, significant at the 95% level, whereas in model 2 this estimated elasticity falls to a statistically insignificant 0.0260. As models 3 and 4 illustrate, this is the least significant and smallest estimated elasticity of all reasonable model specifications, and so I use it as a lower bound for the elasticity of price to distance when estimating damages.
Another issue with model 1 is the inclusion of a large number of House Style indicators. Forty-eight house style indicators (e.g. "Contemporary Colonial" or "Raised Ranch") are used in the initial model, but most homebuyers could not likely identify the differences between more than a handful of different build styles. Many of these style variables unnecessarily differentiate between house types that are nearly identical by most reasonable judgments. As an extreme example, these style variables differentiate between thirteen different types of townhouses. Since there are a small number of observations for many of these finely differentiated variables, correlations picked up in the model may not reflect true homebuyer preferences. Condensing these thirteen different Townhouse variables into a single Townhouse indicator variable arguably better represents the level of detail that consumers take into account when purchasing a house. In model 3 I use the control variables of modell, replacing the 48 house style indicators with nine condensed style dummies: Cape, Colonial, Condo, 46 OAGI0001301 00052
Contemporary, Cottage, Gambrel, Garrison, Ranch, and Townhouse. As 90.3% of the sample fits into one of these nine categories, broad style effects would still be captured with these variables. I estimate this model both without and with year fixed effects in columns 3 and 4, respectively.
The estimated elasticity of price to distance is sensitive to the parameterization of house style, increasing substantially when the condensed style indicators are used. This model implies that a 1% increase in distance from the nuclear plant increases house price by 0.104% (significant at the 99% level).
Since this model does not attempt to control for fine differences in house styles, it finds stronger statistical significance of the variables of interest. It is somewhat worrying that the magnitude of this estimated elasticity is so sensitive to the choice of this controlling for house style. However, I have chosen to present these models to demonstrate the largest amount of volatility seen from these kinds of choices. The other variables of interest seem to be little affected by the inclusion of the nine house style indicators.
Including year fixed effects (column 4) results in a lower estimated elasticity of price to distance (0.0861) relative to model 3. However, this elasticity is still higher than the estimates from models 1 and 2.
The four models presented in Table 3 show that the estimated elasticity of price to distance is quite sensitive to model specification. While the sign on the coefficient for each distance variable was in the expected direction, the magnitude varied significantly. The four estimated elasticities of price to distance were 47 OAGI0001301 00053
0.0260,0.0442,0.0861, and 0.1040. I use each of these elasticities in the damage calculations to give a sense of the range of possible damages, depending on which specification one deems most reasonable. These figures cannot be compared directly to the literature, since past studies have not generally used the same log-log functional form. However, most of these coefficients are statistically significant, in contrast to the findings of some past authors such as Nelson (1981) who found no significant effect, but consistent with others like Folland & Hough (2000) who found a 10% average drop in house prices near nuclear plants. The magnitudes are roughly in line with other studies that estimate the effect of power plants on house prices. For instance, Blomquist (1974) found that the elasticity of house price to distance from electrical power plants (not just nuclear plants) was about 0.09.
Since model 4 controls for year fixed effects and uses "common sense" controls for house style, I argue that it is the most reasonable approach. While the 48 different house style indicators may more accurately control for fine differences in houses, the group of nine condensed style indicators more accurately reflect the degree of detail that home buyers generally take into account when considering house style. Consequently, I choose to use the elasticity from model 4, 0.0861, when focusing on specific estimates. However, I also calculate damages for the other three models to provide a range of estimates in case the mode14' s specification is deemed incorrect.
48 OAGI0001301 00054
VI. 2. Damage Estimation Using these results and the damage calculation methodology outlined in section III, I find the predicted price-distance relationships graphed in Figure lO.
This figure demonstrates graphically the large difference that small changes in estimated elasticities can make in the predicted price-distance relationship, especially at small distances. The three horizontal lines represent different cutoffs for the assumed "safe" group of houses. Clearly, damage estimates to the choice of distance cutoff for "safe" groups.
Cutoff 1 seems inappropriately low. Indeed, this would be using houses that are only 0.14 km, 0.77 km, 2.48 km, or 3.07 km away from the nuclear plant (depending on the elasticity chosen) as a "safe" group, while the data suggests that the price-distance relationship appears to remain strong beyond these distances. The other two distance cutoffs seem more appropriate but still produce very different damage estimates. Keeping in mind that the vertical distance between a point on the price line and the horizontal control cutoff line roughly represents the estimated damages to a house due to the nuclear plant, Figure 10 illustrates how the choices of elasticity and distance cutoff (and the interaction of the two) affect estimated damages.
49 OAGI0001301 00055
Figure 10.
Predicted House Prices with Varying Elasticities 190000 180000 170000
-e=0.026
., ........wo=0.0442
" 160000
";:: 0=0.0861 p...,
....... e=0.104
- s ...... Distance Cutoff 1 0 150000
== - Distance Cutoff 3
- - Distance Cutoff 2 140000 130000 120000 0 2 4 6 8 10 12 Distance (kIn)
Figure 11 illustrates the relationship between damages and distance at a 10 km cutoff distance. This graph is constructed by solving for the predicted price of each house at 10 km within each model and subtracting from this the predicted price of each house as a function of its distance from the plant. Notice in Figure 11 that for larger estimated elasticities (as opposed to smaller ones), the estimated damages for houses within 10 km of the plant are larger.
50 OAGI0001301 00056
Figure 11.
Damages with Varying Elasticies. Cutoff Distance: 10 km tOOOOO seono + ......................................... .
6()OOG :~ ........................*.........*..........................................................................*..............
'" - . Damage, 13.:;::0.026
~
E 40000" *'*c* ". . ............... ... ... . .......... .
, ;.......;...;. [)am<:lge~ (:'-'-0.0442 Damage, e~O.()861
<'l
~ """'nal11'\gc, e~{U04 20000
- OOOG ,. . .*. . .
Distance (km)
Table 4 presents the estimated damages for a variety of elasticities and distance cutoffs. These estimates are sensitive to the choiee of distance cutoff',
and narTowing dm*vn'one's choice of estimate necessitates detennining "reasonable" distance cutoffs. I frnd the most reasonable cutoffs by using the following method. I replace the distance variable in the model with a series of dummy variables representing distance fTOm the corresponding location (for example,PilgrimDistancc1 through PHgrimDistance21 ), PilgrlmDistancci is set equal to one ifthe observation's distance from the Pilgrim nuclear plant is between i and i ~ 1, and zero otherwise. The distance i at which the coefficients on PilgrimDistanceAnd PilgrimDistancc i + l are not signjficantly different from each other is considered the reasonable cutoff distance. This method results in an 51 OAGI0001301 00057
estimated appropriate distance cutoff of 8 km for the nuclear plant. As such, the most reliable the damage estimates should be the ones using roughly these cutoffs. The columns v>'ith "reasonable" distance cutoffs are highHghted in yellow, where reasonahle cutoffs are defined as those within a certain rangcofthe 8km figure.
This 8 km cutoff seems somewhat reasonable. During the Three Mile Island accident, children and pregnant or nursing mothers within a 5 mile (8 km) radius of the plant were evacuated, and others within 10 miles (16 km) were cautioned to stay indoors. The NRC has two ""emergency planning zones" (EPZs) around nuclear plants that specify courses of action in case of a nuclear accident.
The smaller of these t\:vo zones is the "Plume Exposure Pathvv'ay EPZ" (in which the exposure to radioactive materials through direct inhalation is a signit1cant risk), which ext(;nds to about 10 miles (16 km) from the nuclear plant. A larger zone beyond this 10 mile radius designates the area where inhalation is not a significant risk, but exposure through ingestion of contaminated food or water is a threat. At the Chernobyl catastrophe, individuals within a maximum of 32 km were evacuated.
A comparison of the 8 km cutoff with the numbers presented above suggests that 8 km may be a Imver bound ih!" the true effector the nuclear plant.
There are l\NO explanations for the generally higher distances used by the NRC and in the event of accidents, First, individuals may underestimate the geog1'aphical extent of the nuclear risks, causing the price of distance capitalized 52 OAGI0001301 00058
into housepdccs to diminish before the risk hIlly disappears. A second explanation is that the NRC and disaster re1iefworkers overestimate or overstate the extent of the risks. If they were to understate these risks, they would likely be blamed after an accident for failing to sufticiently prepare for an accident so overstating them may be preferable to reduce the chance of being accused of negligence.
The table presents various estimated based on different parameters for distance cutoffs and price-distance elasticities. VyThich measure is lk:emed conect depends on one's interpretation ofthe appropriate values of these parameters. The 10',:vest estimate, using the parameters Dc = 4km and e == 0.0260, is an average of
$2,123 per house and $3.6 million total damages. The highest estimate presented, using Dc = 12knl and e =0.1040, is an average ofSll,937 per house and 8l58.4 million total damages, These are rdatively extreme estimates; more reasonable parameter choices result in more moderate estimates, The parameters of Dc = 8km and e "".0.0861., which as previously argued appear to be the rnost reasonable, results in an average damage estimate of $7,940 per house and $52.9 million in aggregate. To give a sense of the size of these damages, this would be roughly eqnivalent to 81,024 in damages per person, based on Plymouth's population count from the 2000 CensllS of 51 ,701 people.
A comparison of these figures to Folland & Hough (2000) is instructive:
\vhi1e they estimate that the installation ora nuclear plant reduces house prices by about 1O~Q., my estimates amonntto only about 4~!6 of average house prices (40;[) =
53 OAGI0001301 00059
$7 ,940/S 198,078, since $198,078 is the mean sample house price). Even the most extreme damage estimateofSll,937 pCThouse only amounts to 6% of average house prices.
54 OAGI0001301 00060
Table 4. Damages Year 2000 Dollars Distance cutoff
'Price-Distance . 4 km Mean Damage per House $2,122.99 e ;:; 0.026 No. of' oils. damaged 924 (Model!)
$3,587,697.04
$3,646.42 e ""c O'(l442 92~
(Modell)
$6,172,990.82
$7,176.83 0
G>
a0 55 0
w 0
I 0
0 0
(j)
Figure 12. Heat Map of Pilgrim Damages N
s fiI Property Sales
@ Pilgrim Station Lo~,' . -144159:1 o 2.5 5 JU OAGI0001301 00062
Figure 12 shows the spatial distribution of these damages in the fonn of a heat map using model 4's elasticity estimate, e = 0.0861. This map illustrates visually the rate at which damages dissipate with distance. "Redder" areas correspond to higher damages due to the nuclear plant. With increasing distance from the plant, these damages fall and the color correspondingly becomes greener. Different elasticity choices would show the same basic trend, except that the damages would taper off more or less quickly. Also note that beyond a certain distance, houses are considered "negatively damaged" due to the choice of distance cutoff: and are represented as green. As previously noted, these. "negative damages" are not included in Table 4, and a map that corresponded to this choice would not have a continually "greening" trend beyond that distance beyond the cutoff, as Figure 12 does. Figure 12 includes this greening trend to show the degree to which estimated damages fall quickly near the plant but slowly farther away, as intuition would imply.
With these damage estimates in mind, I tum to the question of interpreting these numbers in the context of the costs associated with the risks of nuclear power. These external costs can be l11.easured as the willingness to pay to avoid the risks of nuclear power. As argued in section III, if consumers are rational, the price of avoiding the risks ofnuclear power (as represented by the estimates presented above) should be less than or equal to the representative consumer's willingness to pay. Since the per-house damages represent the degree to which houses are more expensive farther from the nuclear plant, these per-house 57 OAGI0001301 00063
damages retlect the price of avoiding nnclear power. Thus, the per-honse damages provide an estimate of the wiHingness to pay of a household unitto avoidnuc1ear risks. Aggregating these damages among an houses (and scaling them appropriately, as described in section Ill) then provides a lower bound of the aggregate willing11css to pay of affected households to avoid the risks posed by Pilgrim Station. As a result, the estimated aggregate positive damages can be considered lower bounds for the aggregate disvalue placed on the risks associated with the nuclear plant. Thus, the external benefits from the complete elimination of nuclear risk would be in the range of$7.9miUion to $108.8 minion, using reasonable parameters, with the choice estiumte being $52.9 million.
However, caution should be used \vhen applying this estimate. This number does not represent net economic benefits, since the nuclear plant presumahly creates value in its electricity production, and if shut down it would have to be replaced w'ith another source of energy, an issue discussed further in section VII. If this assumption is unrealistic, then the true effect of shutting down the nuclear plant on the welfare of Plymouth residents. may differ.
VI. 3. Benefit Estimation: The Green Space As previously mentioned, Entergy C~orporation owns a wooded green space surrounding the nuclear plant that it is keeping undeveloped for tax reasons.
This green space creates a buffer bet\veen the plant a11d the local comm1..mit:i as well as acting as a source of outdoor recreation f6r Plymonth residents. Proximity to this green space generates a positive externality thaLis capitalized into house 58 OAGI0001301 00064
prices,. as evidenced by the significant effect found in the regression results, Using the same process used to (:akulatc damages above, 1 have calculated the benefits of the green space using each model's parameters, These results are presented in Table 5, The most reasonable distance cutoff of 5 km is obtained through the same method used to find the 8 km cutoff for the nuclear plant.
Estimates using roughly this cutoff are once again highlighted in yellow. The most conservative estimate is about $25,9 million in aggregate benefIts. The highest estimate is $128.4 million in aggregate benetlts, lvfore moderate parameter choices still result in substantial estimated benefits, many arol1ndS3J.*II {} million.
Using model 4 and the most reasonable distance cutoff of 5 km, my choice estimate is benents of $11,759 per house on average, and $78.8 million in aggregate.
59 OAGI0001301 00065
Table 5. Green Space Benefits Year 3km Mean Benefit per House $6,:335.3/
e ::; -0.0239 No. of obs. benefiting 3011 6875 (Model 2)
Total Benefits
~--~--------~-+--------~~~~~.
$25,908,450.60 <p.J"",JL..) ... -'- . . '-, . . . ..., $56,489,200.63 Mean Benefit per House $7,357.93 e ::; -0.0327 No. of obs. benefiting 6875 (Modell)
$35,186,546.00 $76,571,617.32 r-------~~-----+----------~----~
Mean Benefit per House $10,510.03 e =-0.0476 No. of obs. benefiting 6875 (Model 4) r----------------+--------~~~==~
efits $50,138,118.15 . . . <l>~".'!J $109,328,450.00 Mean Benefit per House $14,412.65 $12,323.29 e ::; ~0.0564 No. of obs. benefiting (Model 3) 0 G>
a0 60 0
w 0
I 0
0 0
(j)
(j)
Figure 13. Heat l\IIap of Green Space Benefits
- =:= fiighway L ] P,lglim Gree'1 I
Benefits
~~.~.~~
- ." High: 75898.1 o
61 OAGI0001301 00067
Figure 13 shows a heat map (akin to Figure 12) representing the spatial distribution of the benefits of the green space using model 4' s elasticities. In this map, greener areas correspond to higher green space benefits, with redder areas indicating areas that benefit less or are even possibly considered "negatively benefiting" by distance to the green space. While the above benefit estimates exclude these "negative benefits," the map includes them to show the degree to which the benefits fall quickly with distance near the green space, but fall slower farther away.
One may be tempted to directly compare the benefits of the green space with the damages of the nuclear plant, but this exercise would be meaningless.
The only reason to compare the two would be if their fates were inextricably linked, meaning closing the nuclear plant would necessarily imply developing the green space. There is no reason to think this is so. If the local government were to close Pilgrim Station to gain the $52.9 million in benefits, it could simply zone the green space to protect it from development, or even seize it using eminent domain.
Results Summary Under a wide range of parameter choices for my estimates, the Pilgrim nuclear plant significantly harms local house prices. Using the reasonable parameters of Dc = 8km and e = 0.0861 results in $52.9 million in aggregate damages attributable to the risks of nuclear power. Hence, consumers disvalue the negative externality of the risk of nuclear accidents, as they are willing to pay 62 OAGI0001301 00068
significant amounts to avoid it (from $2,123 to $11,937 per house on average, depending upon the parameters chosen).9 However, the green space surrounding the nuclear plant and owned by its parent company generates a significant positive externality, amounting to able $78.8 million using the same model and a 5 km distance cutoff. This suggests that the green space provides a beneficial role in providing a buffer between the plant and the community, as well as providing recreational benefits to residents. This finding can inform future policy decisions regarding the placement of potential future nuclear plants, since creating an undeveloped green space around a new plant can help mitigate the damages from the risk it poses to the community.
9 Estimates from Table 4.
63 OAGI0001301 00069
VII. Discussion In this section, I discuss issues with the methodology and data that may bias my results or cause them to misrepresent the true economic damages of the negative externality. Also discussed is the degree to which the public perception of risk embodied in house prices accurately reflects the true risks posed by the nuclear plant, and whether or not this matters.
VII. 1. Potential Problems and Biases The estimates presented above make several assumptions that are potentially incorrect and may bias my results. The issues discussed in this section are the collinearity problems posed by the inclusion of several location-based variables in the model, the need for and choice of a distance cutoff, the realistic price-distance relationships after the plant's closure, the existence of "employment effects," implicit liability and insurance, the issue of price estimates understating willingness to pay, and the costs of replacement electricity generation methods.
Several independent variables are highly correlated, most notably the variables representing the distance from the nuclear plant and from the green space or shore. This fact poses potential collinearity issues which would lead to large standard errors. Fortunately, there exist houses near the coast but not near the plant and there exist houses that are near the green space but not near the plant. This variation in the data allows reasonably precise estimates, with most specifications resulting in statistical significance at the 99% level for the variables in interest.
64 OAGI0001301 00070
The damage and benefit estimates above have assumed that the plant's closure would result in all house prices reverting to a constant level at Dc (holding all else constant). This level would be invariant to the plant, meaning that the price-distance elasticity would go to zero. This assumption could be unrealistic for several reasons.
First, the choice of a distance cutoff Dc significantly affects the estimated damages and benefits, as the reader can observe in Tables 4 and 5. The need for a distance cutoff is apparent since I must compare each house's price with its hypothetically "unaffected" price, and this "unaffected" price cannot be chosen arbitrarily. For example, choosing the price level of houses in Plymouth that are the farthest from the plant within the dataset (~20.5 km away) would lead to large damage estimates. This choice would not be wise because it would mean that damage estimates would only be limited by the size of the dataset; if my dataset included houses in the neighboring town, this method would lead to even larger damage estimates simply due to the strictly increasing logarithmic functional form assumed. In reality, we know that beyond some distance the nuclear plant should stop having an effect on house prices, but the basic logarithmic functional form does not capture this.
As a result, I need a distance cutoff to find a price level to which I can compare houses found to be affected by the nuclear plant. Basing the cutoff on the extent (in distance) to which the nuclear plant significantly affects house prices in the data provides a useful estimate of where house prices cease to be affected by 65 OAGI0001301 00071
the nuclear plant. These unaffected house prices are the ones that should be used for comparison, which is why I focus on them in my analysis above.
Second, there may be an unobserved, non-nuclear related relationship between house price and distance from the plant. If this were the case, closing the plant would cause prices to revert to the level corresponding to this relationship, whereas my methodology assumes that the elasticity would revert to zero. For example, if the plant were located in a "bad" part of town, then prices would still rise with distance even after the plant was closed. This means that the elasticity observed in my models is not due to the plant alone but also due to the plant's placement in a bad part of town. As a result, the plant's closure will not cause the price-distance elasticity to go to zero, but instead would simply decrease it. This problem does not seem likely because aspects of the local community that are most likely to cause these issues (e.g. percent Caucasian, percent of houses vacant) are observed and controlled for within the model. Further, being near the coast, the area around the plant is relatively affluent, so if anything the observed price-distance relationship is biased downward, leading the damage estimates to understate the true damages.
A related problem could also be significant. The closure of the plant may not cause the price-distance elasticity to go to zero because closing the plant may not cure all of the sources of the negative externalities. For example, after the closure of other nuclear plants in the past, nuclear waste continued to be stored on-site. This storage can continue to pose a negative externality. Indeed, one 66 OAGI0001301 00072
study found that media attention regarding the storage of nuclear waste at a non-operating plant increased the salience of this negative externality, reducing house prices (Metz et af 1997). As a result, the price-distance relationship may not go to zero after the plant's closure as some of the negative externality remains. In this circumstance, the benefits from closing the plant will in fact be lower, although the damages of its creation would be unaffected by this issue.! 0 In sum, the assumption that the price-distance elasticities go to zero with the removal of the source (the plant or green space) could be unrealistic and bias results. However, it is not clear which direction this bias would go, since the "nice neighborhood" effect could bias the observed relationship in one direction, whereas the "persisting externality" effect could bias it in the other.
Another issue is the possibility of "employment effects," that is, boons to local house prices due to employment opportunities at the nuclear plant, as workers more highly value houses near the location of their jobs at the plant. In theory, this could offset some of the damages to house prices, suggesting that people actually disvalue the risks of nuclear power more than the price-distance relationship directly implies. Despite this theoretical problem, it does not seem problematic for my estimates since there is evidence that nuclear plants do not result in significant employment effects (Rephann 1997). This is due to that fact that nuclear plants are a capital-intensive business that employs relatively few
!O On a public policy note, it may be unwise to systematically consider benefits on these grounds when deciding whether or not to allow a plant to operate. Doing so could potentially cause plants to avoid closure by irreversibly contaminating the surrounding land rather than containing their waste. This is clearly an undesirable activity.
67 OAGI0001301 00073
workers. In particular, Pilgrim Station employs about 580 permanent workers, which amounts to about 1.1 % of the population of the town of Plymouth, although not all of its workers necessarily live in the town.
There are two more related issues that must be considered: nuclear plant liability and implicit government guarantees. Interpreting the my results as measuring the value placed on the risks of nuclear power may be misleading since the nuclear power plant is liable for damages in the event of a nuclear accident, though this liability is limited by the Price-Anderson act and the corporate structure. If homebuyers take into account the fact that the nuclear plant will compensate them for damages in the event of an accident, they will have a form of insurance against nuclear risks. This will reduce the fear of living near the nuclear plant, mitigating the effect of distance on home prices. As a result, my results may not capture the full amount that Plymouth residents disvalue the risks of living near a nuclear plant, and as a result my estimates of the true valuation of these risks could be biased downward.
However, to the extent that nuclear plant is liable for damages in the event of a catastrophe, the externality of nuclear risk is internalized. If consumers are rational and markets are efficient, proximity to the nuclear plant should only harm local house prices to the extent that the externality is not fully internalized. Thus, a significant observed effect (as I find) reflects un-internalized externalities, which are the effects that matter for economic efficiency. Thus, the issue of plant liability for damages should not be a problem in interpreting results as important 68 OAGI0001301 00074
for economic efficiency or for the benefits of closing the plant.
There is also possibility that an implicit guarantee of government assistance would also act as an insurance policy, also biasing my damage estimates downward. This issue could be problematic since it does not represent internalization of the risks of nuclear power. Rather, in the event of an accident the damage is simply shifted from Plymouth residents to the government (and hence taxpayers).
Another issue that must be remembered was covered in section III: that the estimates presented here are based on house prices. Figures 2 and 3 demonstrate how prices will understate willingness to pay since there is also consumer surplus.
The representative consumer is willing to pay at least this price to avoid the risks, but he or she may also be willing to pay much more for the inframarginal units of distance and hence risk reduction. To the extent that the consumer did not have to pay this amount, he or she enjoys consumer surplus that is not being captured by the price-distance relationship. This means that the true willingness to pay to avoid the risks of nuclear power, and hence the true disvalue placed on these risks, is greater than (or equal to) damages implied by the price itself. Thus, price-based damage estimates are a lower bound for the damages attributable to the nuclear plant and for the benefits attributable to the green space.
The cost of replacing the energy currently produced by Pilgrim Station is another important issue. The electricity that the plant currently generates would have to be replaced through some other means, necessitating another source of 69 OAGI0001301 00075
energy of some sort, such as a coal plant or windmills. Environmentally friendly methods are generally more expensive than nuclear power, so their use would be costly and lead to higher energy costs for Plymouth residents. This would likely depress house prices in Plymouth as well, with homebuyers seeking other areas with lower costs of living. Other methods of generating electricity such as coal plants generate negative externalities as well, so choosing this route would simply be substituting one externality for another. Either way, Plymouth residents would suffer some costs if the nuclear plant were to close: either from higher energy costs and lower house prices, or from further negative externalities from other dirty forms of power. Thus, the decision to close the plant must be based not only on the benefits of closing the plant, but also on the costs of the alternative sources of energy.
Finally, the issue of spatial autocorrelation I I was tested for, and was found to be a very significant issue. The symptom of spatial autocorrelation would be incorrect standard errors, leading to faulty inference. Correcting for this would lead to more efficient inference regarding the statistical significance of the estimated coefficients, but fortunately it would not bias the magnitude of these estimates. An extension of this research could be done by estimating the models while correcting for spatial autocorrelation.
II Spatial autocorrelation is when a model's residuals are correlated with location.
70 OAGI0001301 00076
VII. 2. Inaccurate Public Perception of Risk Another issue is the potential lack of information and/or rationality among homebuyers in Plymouth, issues which could lead to inaccurate public perception of risk and hence inaccurate personal assessments of how much one really values the risks posed by the nuclear plant. Rational people with full information will make accurate valuations of the risks of nuclear power based on their own utility functions, resulting in an optimal willingness to pay for distance from the plant that reflects their preferences. However, if people have incorrect information that overstates (understates) the risks of nuclear power, they will in practice be willing to pay more (less) than the optimal amount. Likewise, if they are irrational and respond wildly to new information (accurate or not), there is no reason to expect their revealed willingness to pay to reflect their true preferences. As such, irrationality or imperfect information regarding the risks of nuclear power may bias my results. If people are irrationally sensitive (insensitive) to information about nuclear power, or their information sources are biased to overstate (understate) these risks, then their revealed willingness to pay as found in my results may overstate (understate) their true, rational valuations. This issue casts doubt on the usefulness of using damages to property values to accurately assess individuals' true valuations of the risks of nuclear power.
However, as Gawande and Jenkins-Smith point out, from a legal standpoint it does not matter whether these damages to property values stem from reasonable perceptions of risk (Gawande and Jenkins-Smith 2001). Indeed, residents of affected homes are still worse off when their property depreciates, 71 OAGI0001301 00077
whether it is due to rationally or irrationally perceived risks. The effects on property values may not represent true economic damages if public perceptions of risk are inaccurate, but they are damages to the local community nonetheless in the form of the depreciation of local properties.
I put this issue aside for three reasons. First, it is not clear in which direction the bias will be. While one can imagine that people may systematically overestimate the risks of nuclear power, it is also conceivable that they underestimate it. If some people vastly underestimate risk (say, because they forget to consider the plant's location in their home purchasing decision)), this will increase housing demand and hence house prices near the plant beyond what rational individuals would find optimal. The observed price-distance relationship in this case would understate damages. Second, it is methodologically and philosophically difficult (if not impossible) to un-arbitrarily correct for this if one deems it a problem. Third, it is still a matter of debate among economists and philosophers regarding which preferences should matter for welfare: those revealed by agents (whether they be considered rational or not) or the underlying (yet amorphous) enlightened, rational preferences that people supposedly have.
72 OAGI0001301 00078
VIII. Conclusion I use a hedonic model to estimate the value that individuals place on the risks of nuclear power around Pilgrim Station in Plymouth, Massachusetts. This study is motivated by the controversial issue of the renewal of the plant's license, which is set to expire in 2012, as well as the larger issue of the role of nuclear power in fighting fossil-fuel induced climate change. In making a policy decision regarding the continued operation of the plant, both costs and benefits of renewing the license must be measured and compared, and the disvalue placed on the negative externalities posed by the nuclear plant constitutes an important cost to be measured. I find that homebuyers place significant value on avoiding the risks of nuclear power and that these valuations are capitalized into property values. The price-distance elasticity for the nuclear plant is estimated at 0.0861 (significant at the 99% level), extending to about 8 km. Using the regression results and reasonable parameters, I find that the damages due to the nuclear plant (and hence the benefits from its closure) are around $7,940 per affected house, aggregating to $52.9 million, or $1,024 per Plymouth resident on average.
I also measure the positive externality generated by the green space that surrounds Pilgrim Station. The price-distance elasticity for the green space is estimated to be -0.0476 (significant at the 99% level), extending to about 5 km. I present a range of estimates, but using the most reasonable model and parameters, the benefits attributable to the green space are about $11,758 per affected house, and $78.8 million in aggregate.
73 OAGI0001301 00079
I also find anomalous results regarding the effect of media coverage, with the number of Boston Globe articles discussing accidents at or the dangers of the Pilgrim Nuclear Generating Station being significantly positively correlated with house prices, with a statistically significant elasticity of 0.0 195. The positive significant relationship holds even after controlling for year fixed effects. This anomalous media effect is explored more in the Appendix to this paper.
As previously discussed, one may be tempted to compare the damages of the nuclear plant and the benefits of the green space, but this comparison would have little meaning. Policy decisions regarding the status of the green space and the nuclear plant can be made independently. I conclude that there would be substantial external benefits to the closure of the Pilgrim nuclear plant. To make a policy decision regarding the potential closure of Pilgrim Station, these benefits must be weighed carefully with the costs, the most important of which being those associated with the production of electricity to replace that foregone by closing the nuclear plant.
74 OAGI0001301 00080
Appendix This appendix contains some explorations into the data that are not crucial to the main thrust of my thesis: the estimation of the damages attributable to Pilgrim Station. These explorations include an expanded look at the impact of the media variable on house prices, an impact that was counter-intuitively found to be positive and significant in the main models. I also test to see if the price-distance elasticity varies over time and with the median age of the community. While I do not solve the mystery of the anomalous effect of media attention, I find that the price-distance elasticity is negatively related with time and with the median age of the community.
A.i. Media-Distance Interaction The anomalous positive relationship between house price and media coverage is perplexing. One hypothesis that could theoretically explain this is that more media coverage intensifies the premium placed upon safe housing, pushing up house prices further away from the plant and driving prices down near it. If enough of these safe houses are affected strongly enough, then this positive effect due to media coverage may swamp the negative effects near the plant. If this were the case, then one should see a differing relationship between price and media for houses at different distances from the plant. Specifically, houses near the plant should be adversely affected by media coverage whereas houses far away should be positively affected.
75 OAGI0001301 00081
To test this hypothesis, I create a standardized interaction variable between Log(Pilgrim Distance) and Log(Globe Articles) (as well as standardizing the variables themselves). The standardization of the variables is meant to ease the interpretation of the interaction variable. These standardized variables are denoted Z[Log(Pi/grim Distance)], Z[Log(Globe Articles)], and Z[Log(Distance)]xZ[Log(Globe)] Interaction. Ifmy hypothesis is correct, the coefficient f3 on the interaction variable should be positive. Since my variables are standardized, houses one standard deviation above the mean distance from the plant will have a Z[Log(Pi/grim Distance)] of 1, so increasing Log(Globe articles) by one standard deviation will change the logarithm of house prices by f3. By contrast, houses relatively close to the plant (one standard deviation below the mean) will have a Z[Log(Pi/grim Distance)] of -1, so for these houses increasing Log(Globe articles) by one standard deviation will change log-house prices by (-1)f3. Since house prices far away should benefit from more media attention (and house prices nearby should be reduced) under my hypothesis, f3 is expected to be positive.
Running the regression from model 4 with these standardized variables results in the coefficients reported in Table AI. Once again, the expectation about the effect of media coverage is contradicted by the data. The coefficient on the interaction variable is negative and significant at the 99% level, suggesting that, not only is media attention positively associated with house prices, but that this effect is stronger nearby the plant. Indeed, for houses significantly far away from 76 OAGI0001301 00082
the nuclear plant (beyond about 2 standard deviations above the mean distance),
increasing Log(Globe articles) is actually associated with lower house prices, the opposite of what one would expect the effect of the interaction between the two to 12 be.
Table At. Regression Results for Model A1 Dependent Variable: Log(Real House Price)
Coefficient (I) (2)
Z[Log(Pilgrim Distance)] 0.0448*** 0.0440***
(4.82) (4.73)
Z[Log( Globe articles)] 0.0166*** 0.0167***
(3.74) (3.77)
Z[Log(Distance )]xZ[Log( Globe)] Interaction -0.0083***
(-2.76)
Constant 8.176*** 8.l70***
(21.17) (21.16)
R-squared 0.6232 0.6234 Adjusted R-squared 0.6199 0.6201 No. observations 10,428 10,428 t-statisties in parentheses
- , **, *** indicates significance at the 90%, 95%, and 99% level, respectively A.2. Lagged and Cumulative Media Effects A second hypothesis to explain the anomalous results is that using the number of Boston Globe articles in the year of the sale does not accurately 12 For example, let Z[Log(Pi/grim Distance)] = 3, indicating houses very far away from the plant
(~22 km away). Then the rate of change of Log(Real House Price) with respect to Z[Log(Globe articles)] is 0.0167-0.0083(3) = -0.0082, meaning increasing media attention hurts prices for distant, "safe" homes.
77 OAGI0001301 00083
represent the speed or mechanism through which the media transmits information.
A more realistic model may use the number of Boston Globe articles in the previous year as the meaningful measure of media attention.
Log(Lag Globe articles), representing the amount of media attention in the year preceding the sale, is introduced into model 4, and the regression results are presented in Table A2. Again, the effect of the media attention on house prices is positive and statistically significant, with a 1% increase in media attention being associated with roughly a 0.015% increase in house prices. Investigating the interaction of the lagged effect yields the same conclusion as in model A 1. The lagged media effect is positive and stronger for houses closer to the nuclear plant, counter to intuition, although this effect is less robust than before (only significant at the 90% level).
78 OAGI0001301 00084
Table A2. Regression Results for Model A2 Dependent Variable: Log(Real House Price)
Coefficient (I) (2) (3)
Log(Pilgrim Distance) 0.0868*** 0.0868***
(4.85) (4.85)
Log( Globe articles) 0.0128**
(2.24)
Log(Lag Globe articles) 0.0153*** 0.0152***
(2.99) (2.96)
Z[Log(Pilgrim Distance)] 0.0446***
(4.79)
Z[Log(Lag Globe articles)] 0.0131 ***
(2.94)
Z[Log(Distance) ]xZ[Log(Lag Globe articles)] -0.0050*
(-1.66)
Constant 10.657*** 10.700*** 9.680***
(24.16) (24.28) (21.97)
R-squared 0.6265 0.6265 0.6266 Adjusted R-squared 0.6233 0.6233 0.6233 No. observations 10,299 10,299 10,299 t-statistics in parentheses
- , **, *** indicates significance at the 90%, 95%, and 99% level, respectively A.3. Cumulative Media Effects Model A3 examines the idea that media effects may accumulate over time.
The mechanism through which media attention is hypothesized to adversely affect house prices is by gradually increasing public knowledge of the risks of nuclear power. This knowledge probably does not significantly evaporate over time; 79 OAGI0001301 00085
instead, continuing media attention as time progresses could lead to accumulating knowledge about the risks of nuclear power, increasing the effect of the nuclear plant on house prices.
To account for this, model A3 introduces a variable called Log(Cumulative Globe Articles) which is the logarithm of the total number of Boston Globe articles published since 1992 (the beginning of my sample) that seriously discuss the risks of Pilgrim Station. Table A3 presents the results, once again finding a strong significant positive relationship between cumulative media attention and house prices. Of course, since the cumulative media variable is naturally increasing with time, and the housing boom corresponds to a similar time period, this strong significant coefficient could simply be picking up time effects, even though Year-fixed effects are included in the model.
When Log(Globe articles) is included in the model, it is driven to insignificance, but this is unsurprising given the collinearity between the single-year and cumulative media variables. The distance-cumulative media interaction variable produces estimated coefficients similar to those found in models Al and A2, with a robust, counter-intuitive sign.
80 OAGI0001301 00086
Table A3. Regression Results for Model A3 Dependent Variable: Log(Real House Price)
Coefficient (1) (2) (3)
Log(Pilgrim Distance) 0.0861 *** 0.0861 ***
(4.82) (4.82)
Log( Globe articles) -0.0044
(-0.77)
Log(Cumulative Globe articles) 0.0830*** 0.0593***
(7.77) (5.97)
Z[Log(Pilgrim Distance)] 0.0430***
(4.63)
Z[Log(Cumulative Globe articles)] 0.0344***
(5.90)
Z[Log(Distance)]xZ[Log(Cum. Globe)] -0.0129***
(-4.35)
Constant 7.568*** 7.628*** 8.103***
(19.49) (19.62) (21.01)
R-squared 0.6232 0.6232 0.6238 Adjusted R-squared 0.6199 0.6199 0.6205 No. observations 10428 10428 10428 t-statisties in parentheses
- , **, *** indicates significance at the 90%, 95%, and 99% level, respectively While these models are interesting, they do not change the counter-intuitive finding that media attention is robustly related in house prices in a counter-intuitive way. Not only is media attention oddly associated with higher house prices in Plymouth in general, but this relationship also appears to be strongest near the nuclear plant, in the exact area where one would expect media attention to most harm house prices.
81 OAGI0001301 00087
A.4. Time-varying Price-Distance Relationships Another question of interest is whether or not the price-distance relationship changes over time. In the 1980s and early 1990s, Pilgrim Station was considered one of the worst run nuclear plants in the United States according to the Nuclear Regulatory Commission. Since then, it has solved many of its safety problems, and the NRC has upgraded its rating. Ifhomebuyers take this reduction in risk into account, then the price-distance relationship could have decreased over time, as Pilgrim became less risky to live near.
To investigate the changing relationship over time, model A4 uses an interaction variable for year of sale and distance from Pilgrim Station. The coefficient on the interaction variable is negative and significant at the 99% level, meaning that the strength of the price-distance relationship is decreasing with time. For example, according to this model, the estimated price-distance elasticity in 2009 would be estimated to be 0.0497. By contrast, this model would imply a price-distance elasticity of 0.1386 in 1993.
82 OAGI0001301 00088
Table A4. Regression Results for Model A4 Dependent Variable: Log(Real House Price)
Coefficient (I) (2)
Log(Pilgrim Distance) 0.0932*** 0.0900***
(5.18) (5.00)
Z[Year] 0.0790*** 0.1294***
(21.90) (10.1 0)
Z[Log(Distance)]xZ[Year] -0.0236***
(-4.10)
Constant 7.908*** 7.910***
(20.32) (20.34)
R-squared 0.6154 0.6160 Adjusted R-squared 0.6126 0.6132 No. observations 10,428 10,428 Note: Year fixed effects were removed from this modcl.
t-statistics in parentheses
- , **, *** indicates significance at the 900/0, 950/0, and 990/0 level, respectively This evidence is consistent with the hypothesis that Pilgrim Station's improving safety record was reflected in house prices, as homebuyers took into account the reduced risk in their decisions. However, it is also possible that this relationship is driven by other factors that were changing over this time. For example, if the desirability of proximity to the shore changed over the years due to changing environmental or social factors, then the correlation found in this model could be biased, with the change in the plant's price-distance elasticity reflecting changes in the shore's price-distance elasticity.
However, if the decline in the price-distance elasticity over time is truly due entirely to the plant's improved safety record, it has interesting implications for estimating damages. If one wants to estimate the benefits of closing the plant, then one should use today's smaller elasticity of 0.0497, since those are the damages that would be alleviated by the plant's closure. This could mean that the 83 OAGI0001301 00089
estimates based on the entire sample (0.0861 for the basic version of this model) could be too large, since the sample includes houses sold when the plant posed more of a risk. On the other hand, if one wants to determine legal damages to past homeowners in Plymouth, then one should use the estimated damages based on the price-distance elasticity in the year of the house's sale, since that was the time that they realized the damages of the sale in the form of lower house value. Of course, it is very difficult to tell if the decline in the price-distance elasticity over the past 17 years is due entirely to the nuclear plant's improving safety record or due to other changes in the housing market in Plymouth, making it difficult to determine whether or not using these different elasticities would be correct.
A.5. Age-Distance Interactions The disvalue an individual places on living near the nuclear plant may vary from person to person. One reason for this is differing expectations about the likelihood of a nuclear meltdown during the time that one expects to spend living at the house. While one cannot easily directly observe individuals' beliefs about this likelihood, it seems reasonable to assume that the elderly are less likely to experience a nuclear meltdown in their remaining lifetime than the young. Thus, one may expect the premium placed upon risk (and hence the magnitude of the price-distance elasticity) to be smaller for communities with more elderly residents.
To test this hypothesis, model AS includes an interaction variable between Log(Pilgrim Distance) and the median age of the Census Block. As one can see in 84 OAGI0001301 00090
Table A5, the coefficient on this variable is negative and significant at the 95%
level, implying that house prices in communities with older residents are less sensitive to distance from the nuclear plant. Specifically, increasing the Census Block's median age by I standard deviation (4.33 years) reduces the elasticity by 0.014. This is consistent with the above story about the elderly having a lower probability of experiencing a meltdown, but it is also consistent with several other hypotheses. For example, lower risk-aversion among older communities could explain this finding. Also, young people are more likely to be worried about the long-term effect of radiation on their children or on pregnant women, issues that are not as relevant to the elderly. Interestingly, it is also possible that the fall in the elasticity over time as found in model A4 above is driven at least in part by the gradual aging of the population.
85 OAGI0001301 00091
Table AS. Regression Results for Model AS Dependent Variable: Log(Real House Price)
Coefficient
- 0) (2)
Log(Pilgrim Distance) 0.1067*** 0.1102***
(5.91 ) (6.09)
Z[Log(Median Census Block Age)] 0.0400*** 0.0679***
(7.10) (5.44)
Log(Distance)xZ[Log(Med. Age)] -0.0140**
(-2.51)
Constant 7.526*** 7.537***
( 19.33) 09.36)
R-squared 0.6250 0.6252 Adjusted R-squared 0.6217 0.6219 No. observations 10,428 10,428 i-statistics in parentheses
- . **, *** indicates significance at the 90%, 95%, and 99% level, respectively 86 OAGI0001301 00092
References Allen, Scott, and Globe Staff. "NRC lowers Pilgrim rating Cites staff errors, high exposure". The Boston Globe 24 May 1996, City ed., Metro/Region sec.:
25.
Bezdek, R.H. and Wendling, R.M. (2006) "The impacts of nuclear facilities on property values and other factors in the surrounding communities", Int. J.
Nuclear Governance, Economy and Ecology, Vol. 1, No.1, pp.l22-144.
Blomquist, G. C., "The Effect of Electric Utility Power Plant Location on Area Property Value," Land Economics, 1974,50:1,97-100 Boyle, Melissa A., and Katherine A. Kie1, "A Survey of House Price Hedonic Studies of the Impact of Environmental Externalities", Journal of Real Estate Literature, 2001, (9) 2: 117-144.
Clapp R.W., Cobb S., Chan c.K., and Walker B. "Leukemia near Massachusetts nuclear power plant", Lancet 336:1324-5, 1987.
Clark, David E., and Allison, T. (1999) "Spent nuclear fuel and residential property values: the influence of proximity, visual cues and public information", Papers in Regional Science, Vol. 78, pp.403-421.
Clark, David E., Michelbrink:, L., Allison, T. and Metz, W. (1997) "Nuclear power plants and residential housing prices", Growth and Change, Vol.
28, pp.496-5l9.
Clark, David E., and Nieves, L.A. 1994. "An Interregional Hedonic Analysis of Noxious Impacts on Local Wages and Property Values", Journal of Environmental Economics and Management, 27, 235-253.
Folland, Sherman, and Robbin Hough. 1991, "Nuclear Power Plants and the Value of Agricultural Land", Land Economics, 67, 30-36.
Sherman, Folland, and Hough Robbin. "On the External Effects of Nuclear Power: Further Evidence". Journal of Regional Science 40 (2000): 735-53.
Galster, G.C. (1986) "Nuclear power plants and residential property values: a comment on short-run vs. long-run considerations", Journal of Regional Science, Vol. 26, pp.803-805.
Gamble, H.B. and Downing, R.H. (1982) "Effects of nuclear power plant on residential property values", Journal of Regional Science, Vol. 22, pp.457-478.
Gawande, K. and Jenkins-Smith, H. (2001) "Nuclear waste transport and residential property values: estimating the effects of perceived risks",
Journal of Environmental Economics and Management, Vol. 42, pp.207-233.
Gelbspan, Ross, and Globe Staff. "Pilgrim plant worker fails alcohol test on job",
The Boston Globe 11 Dec. 1991, City ed., MetrolRegion sec.: 43.
Gruber, Jonathan. (2007) Public Finance and Public Policy. New York, NY:
Worth, 2007.
Howe, Peter 1., and Globe Staff. "Pilgrim nuclear plant closed to fix faulty 87 OAGI0001301 00093
system", The Boston Globe 29 Jan. 1989, City ed., Metro/Region sec.: 27.
Johnson, Carolyn Y. "NRC looking into a nap at Pilgrim, Case spurs review at nuclear plant". The Boston Globe 14 Oct. 2004, Third ed., Metro/Region sec.: B5.
Knox, Robert. "Pilgrim owners reprimanded for worker who slept on the job",
The Boston Globe 22 Feb. 2007, First ed., South sec.: Reg4.
Nelson, J. P., "Three Mile Island and Residential Property Values: Empirical Analysis and Policy Implications", Land Economics. 1981,57:3,363-72.
"Pilgrim missing items with trace radiation", The Boston Globe 24 May 2006, Third ed., MetrolRegion sec.: B2.
"Plymouth, Massachusetts (MA) Detailed Profile - relocation, real estate, travel, jobs, hospitals, schools, crime, move, moving, houses news, sex offenders." Stats about all US cities - real estate, relocation info, house prices, home value estimator, recent sales, cost of living, crime, race, income, photos, education, maps, weather, houses, schools, neighborhoods, and more. 27 Apr. 2009 <http://www.city-data.comlcity/Plymouth-Massachusetts.html>.
Rephann, T.J. (1997) "The economic impact of LULUs" , Allegany College of Maryland.
Rosen, Sherwin. "Hedonic prices and implicit markets: Product differentiation in pure competition". Journal of Political Economy, 82(1):34-55, 1974 Sheppard, Stephen, 1999. "Hedonic analysis of housing markets," Handbook of Regional and Urban Economics, in: P. C. Cheshire & E. S. Mills (ed.),
Handbook of Regional and Urban Economics, edition 1, volume 3, chapter 41, pages 1595-1635 Elsevier.
"Small fire extinguished at nuclear plant", The Boston Globe 31 Oct. 2008, First ed., Metro sec.: B2.
Sternglass E.J. "Birthweight and Infant Mortality Changes in Massachusetts Following Releases from the Pilgrim Nuclear Power Plant." Submitted to the Massachusetts Legislature, June 10, 1986.
United States Nuclear Regulatory Commission. "NRC: Backgrounder on Emergency Preparedness For Nuclear Power Plants. United States Nuclear Regulatory Commission. 23 Apr. 2009 <http://www.nrc.gov/reading-rmldoc-collections/fact-sheets/emerg-plan-prep-nuc-power-bg.html>.
"United States and States - R190 1. Median Household Income." American FactFinder. United States Census Bureau. 27 Apr. 2009
<http://factfinder.census.gov/servlet/GRTTable?_ bm=y&-
box- head- nbr=R1901&-ds- name=ACS - 2007- 1YR- GOO-
_lang=en&-format=US-30&-CONTEXT=grt>.
88 OAGI0001301 00094