Stigmatized Asset Value: Is It Temporary or Long-Term?

STIGMATIZED ASSET VALUE IS IT TEMPORARY OR LONG-TERM

Jill J McCluskey and Gordon C Rausser

AbstractmdashStigma is a negative attribute of real estate acquired by envi-ronmental contamination and re ected in its value (Elliot-Jones 1996)Using a model of neighborhood turnover with external economies weshow that both temporary stigma and long-term stigma are possibleequilibrium outcomes after the discovery and cleanup of a hazardouswaste site The existence and duration of stigma are examined usinghedonic price techniques with data from housing sales prices in DallasCounty Texas We nd that results depend critically on distance from thehazardous waste site Neighborhood turnover due to changes in the levelof poverty also appears likely

I Introduction

IN the past three decades the public has become increas-ingly aware of environmental risks This awareness is

re ected in the negative impact of environmental contami-nation on property values Stigma is a loss in property valuebeyond the cleanup cost of the contamination (Patchin1991) There are two externality effects that cause stigmaThe rst is an environmental externality on the propertiesadjacent to a hazardous waste site the contamination causesneighboring property owners to be concerned about healthissues The second is a neighborhood externality the asso-ciation with a hazardous waste site can affect the composi-tion of residents in the neighborhood and other attributesthat determine neighborhood quality and property values Ifthe neighborhood externality is the source of the stigmathen remediating the hazardous waste site may not result inincreased property values Although some analysts haveargued that uncertainty or concern over whether the prop-erty is still contaminated is a cause of stigma (Mundy1992) the neighborhood externality effect has not beenconsidered

When a triggering event results in direct damage theremay be a spillover or multiplier effect The resulting addi-tional damage is called consequential damage In the case ofa hazardous waste site the environmental externality causeschanges in the composition of a neighborhood which inturn results in property value diminution above and beyondthe immediate damage Once environmental contaminationbecomes associated with a particular neighborhood itsproperty values may be stigmatized Even if a potentialhomebuyer believes that the formerly contaminated area hasbeen cleaned up he or she will probably demand a discountThe resale value will most likely be lower than that of a

comparable property without a history of contamination ifthere is a market for the property at all This reluctance tobuy can be re ected in lower residential property values andmay be based on perceived risk that may or may not have ascienti c foundation

A neighborhood may be distinguished as undesirable if itis identi ed as contaminated Real estate has an intangiblecomponent that is determined by the publicrsquos perception ofthe location similar to the intangible asset of goodwill on acorporationrsquos balance sheet When the public perceives thata neighborhood is no longer fashionable the value of theintangible component falls By making the neighborhoodless desirable a hazardous waste site decreases the value ofthe neighborhoodrsquos property making it more affordable tolower-income families and less attractive to higher-incomefamilies Over time higher-income residents may relocateAs a result the by-products of high-income residents suchas social status good schools low crime rates quick policeresponse and well-maintained owner-occupied homesmay disappear Therefore although the environmental prob-lems are temporary they affect the character of a neighbor-hood creating long-term stigma1 In the worst-case sce-nario outside businesses may ldquoredlinerdquo the area causingneighborhood businesses to relocate In such a scenario it isunlikely property values will rebound2

Previous studies have attempted to measure bene ts fromthe cleanup of hazardous waste by showing that residentialproperty values decline as the distance to a hazardous wastesite decreases (for example Ketkar 1992 Thayer Albers ampRahmatian 1992) Extending this argument if the hazard-ous waste site is remediated then the discount for a locationthat is close to a former hazardous waste site should berecouped A signi cant bene t of cleanup is then the differ-ence between what property values were before the hazard-ous waste site and what they are with the hazardous wastesite However if the neighborhood externality effect is thesource of the stigma then this reasoning is faulty Conse-quently if stigma effects from a site exist then past studiesthat have made this value-recoupment argument may haveoverestimated the property-value bene ts of cleanup ofhazardous waste sites

Received for publication June 28 1999 Revision accepted for publica-tion January 7 2002

Washington State University and University of California Berkeleyrespectively

This paper has bene ted from discussions with Michael HanemannZhihua Shen Greg Adams Larry Dale Ken Rosen Larry Karp IlyaSegal Robert Edelstein David Zilberman John Harris Ron Mittelham-mer and Phil Wandschneider Any remaining errors are the responsibilityof the authors This work was supported by an EPAndashNSF grant JillMcCluskey gratefully acknowledges nancial support she received fromthe Fisher Center for Real Estate and Urban Economics at the UCBerkeleyrsquos Haas School of Business

1 Hall (1994) provides a case study of the effect of contamination on thevalue of an apartment property An underground oil spill created unhealthyconditions which led to a number of tenants leaving Rents then declinedand the property became ldquoseedyrdquo Even after the spill was cleaned up theproperty had substantially declined in value

2 Most changes to a neighborhood are reversible It is usually possiblefor gentri cation to take place There is usually some reason for gentri- cation to occur such as an amenity an attractive location or a desirableemployer or business moving into the area Bond (2001) found thatformerly contaminated water-view lots in Australia fared better withrespect to property diminution than lots without water views

The Review of Economics and Statistics May 2003 85(2) 276ndash285copy 2003 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology

There are two major sources of uncertainty that arisefrom environmental contamination (1) whether the prop-erty is still a health risk even after the property has beenremediated and (2) what future potential liabilities exist andwho is responsible for them Using an expected-utilityapproach it can be shown that the uncertainty surroundinghazardous waste sites can result in lower property values(Boyd Harrington and Macauley 1996) More generally amonetary value can be placed on irreversible events such asa permanent change in health status and loss of life basedon the choices an individual makes about income consump-tion and risk A potential buyer must be compensated forthe expected value of future damage to his health and anamount equal to a certainty equivalent to compensate for therisk associated with the contaminated site

Uncertainty can also make it dif cult for prospectivebuyers to obtain nancing Due to uncertainty about therisks of mortgaging contaminated properties lendersrsquo will-ingness to provide nancing on contaminated properties fellfrom the late 1960s to a low point in the early 1980s whereit stayed until the early 1990s when it began to improve3

During the low-willingness-to- nance period the vast ma-jority of lenders would not consider providing nancinguntil the property has been remediated and satis ed certaintolerance limits The result of the loss of mortgagability isoften that the property is held off the market4 However arecent increase in the understanding of the management ofthe risk surrounding contaminated properties has led to agreater willingness to provide nancing for these proper-ties5

In this paper we use a theoretical model with externalityeffects to show that both temporary stigma and long-termstigma are possible equilibrium outcomes after the discov-ery and cleanup of a hazardous waste site The former isdriven only by the environmental externality whereas thelatter is driven by the neighborhood externality The exis-tence and duration of stigma are then tested for by estimat-ing a hedonic price model with data from housing salesprices in Dallas County Texas In the second stage of thehedonic price analysis the bid functions are estimated toevaluate the externality characteristics of the model Thefocus of the empirical analysis is the RSR lead smelter inWest Dallas which operated from 1934 to 1984 and causedsoil contamination from air emissions and slag material Thepooled data set used in this analysis covers the period 1979to 1995 and includes over 200000 observations

II Previous Empirical Evidence

Many authors have used property transaction data tovalue environmental attributes and more speci cally studythe impact of hazardous waste sites Researchers such asKetkar (1992) Kiel (1995) Kiel and McClain (1996)Kohlhase (1991) Smith and Desvousges (1986) andThayer et al (1992) have consistently found that proximityto hazardous waste sites and other locally undesirable landuses has a negative impact on property values6

The current body of literature on the empirical effects oflocally undesirable land uses does not address whether thediminution of property values caused by these land uses istemporary or long-term or whether externality effects existAlthough there have been many previous studies that at-tempt to measure the effect of environmental contaminationand cleanup on property values they generally focus on theshort run Most importantly almost no studies have ana-lyzed postcleanup property values Typically impacts ofcontamination on property values are examined with across-sectional data set at a single point in time7 Withoutincluding postcleanup property values studies cannot struc-ture the event analysis correctly to analyze the effects ofcleanup In contrast this analysis examines the impact ofenvironmental contamination on residential property valuesby analyzing data from before identi cation of the hazard-ous waste site before during and after cleanup has beencompleted and after new concerns were raised in thepostcleanup years Consequently it is possible to considerthe longer-run effects of contamination and remediationprocesses

The exceptions to the lack of postcleanup analysis in theliterature are Kohlhase (1991) and Dale et al (1999)Kohlhase (1991) includes a toxic site in her study that wasldquoalmost cleaned uprdquo at the time and nds statisticallyinsigni cant coef cient on price and distance Dale et al(1999) analyze the effect of the RSR smelter but take adifferent approach than is taken in the current analysis Theydo not provide a theoretical foundation or a correspondingempirical analysis for distinguishing between long-term andtemporary stigma Moreover they quantify the discontinuityof the distance price gradient surrounding the smelter byincluding indicator variables for two speci c neighbor-hoods As a result they conclude that there was no long-term stigma associated with the RSR smelter We approachthis problem with a linear spline function which speci callyallows for the in uence of the smelter to diminish withincreasing distance and arrive at different conclusions3 This lack of mortgagability has been discussed in the appraisal and

legal literatures Patchin (1991 p 169) writes ldquoThis reluctance to lendapplies not only to Superfund sites but to sites with comparatively lowlevels of contamination rdquo Lewandrowski (1994) discusses that in caseswhere nancing is still available interest rates will be higher

4 Patchin (1991 p 169)5 A specialty of appraisal for contaminated properties was developed

during the 1990s During that time the credit market went through aperiod of learning

6 For additional cites and a comprehensive survey of empirical resultssee Farber (1998)

7 Exceptions include Dale et al (1999) Kohlhase (1991) and Kiel(1995) who examine property values at more than one point in time Kieland McClain (1996) examine housing prices before and after a failedincinerator siting

STIGMATIZED ASSET VALUE IS IT TEMPORARY OR LONG-TERM 277

III Multiple-Equilibrium Hedonic Model

Sometimes property values recover following environ-mental contamination and sometimes they do not In theformer instance the environmental externality results inonly a temporary stigma Accordingly our theoretical modelallows for environmental contamination to lead to either arecovery equilibrium or long-term stigma on property val-ues in formerly contaminated neighborhoods 8

Much of the current literature focuses only on uncertaintyof health risk and liability as a potential cause of stigma(Mundy 1992) If uncertainty is the only cause of stigmathen once the uncertainty is removed values should recoverHowever in the residential succession literature externalitymodels have been developed that emphasize the role ofincome levels or racial composition in explaining neighbor-hood turnover (Miyao 1978 Coulson and Bond 1990) Iflow-income residents enter a high-income neighborhoodthen the composition of the neighborhood may tip from highincome to low income Such a neighborhood transformationcould well result in long-term rather than temporarystigma

The price of housing and land re ects consumersrsquo valu-ations of all the attributes that are associated with housingincluding environmental quality The level of environmentalcontamination is a quality attribute of a differentiated prod-uct Consumers can choose the level of environmentalquality through their choice of a particular residential loca-tion Housing prices may include premiums for locations inareas with high environmental quality If so the pricedifferentials may be viewed as implicit prices for differentlevels of environmental quality Given these issues measur-ing these differentials requires a neighborhood turnovermodel with externalities (Coulson and Bond 1990)

In the model speci cation there are two types of house-holds high-income (Yh) and low-income (Y l) with Yh Y l Households consume one unit of housing and bid forhousing in competitive rental markets9 Household prefer-ences are represented by a strictly quasi-concave utilityfunction that is increasing in the quality of housing Qprovided by the unit rented and the quantity of all othergoods X The quality of housing is a function of theperceived environmental damage (E) and of the averageincome of the neighborhood in which the house is located(Y) which introduces the neighborhood externality Thelevel of environmental damage plays the role of lteringthat the age of housing stock represents in Coulson andBondrsquos (1990) model Both types of households have thesame preferences the only difference is income Both typeswould like to live in a neighborhood with a large proportionof high-income people

In the proposed model utility is generated by

U U~Q~E Y X (1)

where QY 0 and QE 0 Income goes toward rent andall other goods Following Rosen (1974) preferences forresidential quality can be expressed by bids for variousqualities of housing u that result in the same amount ofutility Assuming that there is no borrowing or lending ahouseholdrsquos budget constraint is then X 1 u 5 Y Substi-tuting the budget constraint in equation (1) and invertingallows the bid function to be derived

u 5 u ~E Y U Y (2)

It is assumed that housing quality is a normal good inequation (2) thus uEY 0

Assuming an open city each household has a utility levelUi (i 5 l h) that is the default utility that is obtainable ifthat household locates outside the city The neighborhoodrental price for housing in equilibrium is

P~E Y Ul Uh Yl Yh 5 maxi

ui~E Y (3)

Housing will turn over from one income group to the otherat the environmental damage level D at which

u l~D Y Ul 5 uh~D Y Uh (4)

From uEY 0 it follows that uEl uE

h at D This means thathigh-income households place a higher value on housingwith less environmental damage than D for a given Y

Mean neighborhood income will depend on the exoge-nous distribution of environmental contamination in thevicinity of the housing stock F(E) 5 0

E f(u) du Assumingthat the bid functions of low-income households are suf -ciently high so that all housing is occupied the meanneighborhood income is

Y 5 F~EY l 1 1 2 F~EYh (5)

The neighborhood equilibrium occurs at the values of Y andE that solve equations (4) and (5) The possibility ofmultiple equilibria in a similar model with ltering by ageof the house is discussed in Coulson and Bond (1990) Therequirement for uniqueness of the equilibrium condition isthat

uEh 2 uE

l 1 ~uYh

2 uYl~Yh 2 Y l f~E 0 (6)

The interpretation is that for the equilibrium to be uniquethe external effects of neighborhood average income mustbe suf ciently small relative to the effects of the environ-mental damage If the initial income level of the neighbor-hood is equivalent to the post-environmental-damage in-come level the stigma will only be temporary However ifthe initial income was high and the post-environmental-damage income is low then a long-term stigma equilibriumresults

8 There is also a third unstable equilibrium9 In the case of owner-occupied housing the rental can be thought of as

implicit

THE REVIEW OF ECONOMICS AND STATISTICS278

The model can be easily generalized to include a contin-uum of income types Following the standard hedonic pricemodel the price of housing P in Dallas County Texas isassumed to be described by a hedonic price function P 5P( z) where z is a vector of structural neighborhood andenvironmental attributes The hedonic price of an additionalunit of a particular attribute is determined as the partialderivative of the hedonic price function with respect to thatparticular attribute Each consumer chooses an optimalbundle of housing attributes and all other goods in order tomaximize utility subject to a budget constraint The chosenbundle will place the consumer so that his indifferencecurve is tangent to the price gradient Pz Therefore themarginal willingness to pay for a change in a housingattribute is then equal to the derivative of the hedonic pricefunction with respect to that attribute

IV Empirical Study of a Stigma Equilibrium

The model indicates that if the externality effect ofaverage neighborhood income is strong enough then it ispossible for long-run stigma to result when environmentaldamage has occurred The empirical study includes ananalysis of which equilibrium emerged for the residentialhousing market near the RSR hazardous waste site inDallas Texas Further household bid functions are esti-mated in order to determine whether the neighborhoodexternality model is consistent with market behavior

A The Data Set

The data set includes 205397 observations10 with vari-ables describing price and attributes of all single-familydetached homes sold over the period 1979 to 1995 in DallasCounty Texas (Dallas County Appraisal District) Eachobservation includes information (table 1) about the saleprice11 of the homes and different variables that affect thesale price including house neighborhood and environmen-tal quality attributes As usual housing quality is describedby the square footage of living space number of bathroomslot size and indicator variables indicating the presence of apool central air conditioning house condition and similarvariables Neighborhood quality is based upon variablessuch as the percentage of households below the povertylevel school district ethnic composition and accessibilityto the DallasndashFt Worth airport the Dallas central businessdistrict (CBD) and the Galleria Mall Environmental qual-ity is described by proximity to the RSR lead smelter Using

a Geographic Information Systems (GIS) database DallasCounty is set up as a grid of X and Y coordinates Coordi-nates are assigned to each house the airport the CBD theGalleria Mall and the RSR hazardous waste site Distancecan then be calculated between any two points The GISdatabase is also used to link each house to its census tract(and the corresponding demographic information12) and itsschool district

10 As part of our data protocols we exclude observations that seemunreasonable The unreasonable observations are those with any of thefollowing characteristics price less than $4000 lot size greater than43560 square feet or less than 400 square feet and living area less than400 square feet This differs from Dale et alrsquos (1999) data protocolswhich delete observations with a selling price of less than $10000 and aprice per square foot of less than $40

11 Prices are de ated using the shelter housing price index (1982ndash1984 5 100) from the Economic Report of the President

12 For the non-Census years (1979 1981ndash1989 1991ndash1995) weightedaverages are used where the weights are based on the trend from 1980 to1990

TABLE 1mdashVARIABLE DEFINITIONS AND DESCRIPTIVE STATISTICS

Variable Description Mean Std Dev

Price Sales price of the home 104921 98168Dprice De ated sales price of the

home86010 78940

Landarea Lot size in square feet 930187 396960Livarea Living area in square feet 179743 755Dalcbd Miles to the Dallas central

business district1090 392

Dfwair Miles to DallasndashFort WorthAirport

1797 615

Galleria Miles to the Galleriashopping center

1089 576

Distrsr Miles to the RSR facility 1173 422Age Age of the house in years 1997 1618Pool 1 if pool 0 otherwise 014 034Garg 1 if attached garage 0

otherwise087 033

Baths Number of bathrooms 203 074Pblack of the census tract that are

African-American1105 1698

Phisp of the census tract that areHispanic

1155 1305

Pbpov of the census tract belowthe poverty line

768 720

Heatc 1 if central heating 0otherwise

088 032

Aircon 1 if central air conditioning0 otherwise

087 033

Good 1 if good condition 0otherwise

030 046

Average 1 if average condition 0otherwise

068 047

School Districts

CF 1 if CarrolltonFarmersBranch 0 otherwise

007 026

Dallas 1 if Dallas school district 0otherwise

032 047

Cedar Hill 1 if Cedar Hill 0 otherwise 001 011Garland 1 if Garland 0 otherwise 014 035HP 1 if Highland Park 0

otherwise003 015

Irving 1 if Irving 0 otherwise 006 023LWH 1 if LancasterWilmer

Hutchins 0 otherwise001 011

No district 1 if no district 0 otherwise 007 026MS 1 if MesquiteSunnyvale 0

otherwise005 002

Coppell 1 if Coppell 0 otherwise 002 015GP 1 if Grand Prairie 0

otherwise004 019

Richardson 1 if Richardson 0 otherwise 013 034Desoto 1 if Desoto 0 otherwise 002 015Duncan 1 if Duncanville 0 otherwise 003 018


The RSR lead smelter is located in the middle of DallasCounty approximately six miles west of the CBD Thesmelter operated from 1934 to 1984 and was purchased in1971 by the RSR Corporation The smelter emitted airbornelead which contaminated the soil in the surrounding areasLead debris created by the smelter was used in the yards anddriveways of some West Dallas residences In 1981 theUS Environmental Protection Agency (EPA) found healthrisks and RSR agreed to remove any contaminated soil inthe neighborhoods surrounding the RSR site using standardsthat were considered protective of human health at the timeIn 1983 and 1984 additional controls were imposed by theCity of Dallas and the State of Texas In 1984 the smelterwas sold to the Murmur Corporation which closed it downpermanently In 1986 a court ruled that the cleanup wascomplete

In 1991 the Center for Disease Control (CDC) loweredthe blood level of concern for children from 30 to 10micrograms of lead per deciliter of blood Low-level leadexposure during childhood may cause reductions in intel-lectual capacity and attention span reading and learningdisabilities hyperactivity impaired growth or hearing loss(Kraft and Scheberle 1995) Also in 1991 the State ofTexas found hazardous waste violations at the smelter In1993 the RSR smelter was placed on the Superfund Na-tional Priorities List (NPL)

Although initially high the percentage of minority resi-dents in close proximity to the smelter grew during theperiod of study Within one mile of the RSR site during1979ndash1980 the mean percentage of Hispanic residents inthe Census tracts where houses were sold was 548During 1991ndash1994 within one mile of the RSR site thismean percentage of Hispanic residents in the Census tractswhere houses were sold increases to 692 The compara-ble percentages for Dallas County as a whole are 88 and138 respectively

B Event Analysis

The analysis covers the impact of the smelter on propertyvalues over four event-driven time periods (1) pre-1981when the smelter operated but health risks were not of -cially identi ed or publicized (2) 1981ndash1986 when healthrisks from soil contamination were of cially identi edcleanup was initiated and a court ruled cleanup was com-pleted (3) 1987ndash1990 after cleanup was again ruled com-plete and (4) 1991ndash1995 when new concerns arose andadditional cleanup occurred The event analysis allows us toanalyze which of the equilibria occurred Slovic et al(1991) provide support for the use of event-driven timeperiods They write ldquoSocial ampli cation [of risk] is trig-gered by the occurrence of an adverse eventrdquo13 Kiel andMcClain (1995) also divide their data into event-driven timeperiods in order to analyze the effect of changes in infor-

mation over time about an incinerator siting on propertyvalues

In addition to considering division by event-driven timeperiods Chow tests are performed to evaluate whetherstructural changes occurred The results indicate that everyyear with one exception is signi cantly different from theprevious one The exception is that the data from sales in1993 are not signi cantly different from those from sales in1994 In addition Wald tests for structural change which donot assume that the disturbance variance is the same acrossregressions are performed to test if the event-driven periodsare the same The results indicate that each period issigni cantly different from the others In order to partiallycontrol for the differences across years within the event-driven time periods indicator variables are included to theindicate year of sale

In addition to the statistical measures there are obviousdifferences across the event-driven time periods Before theidenti cation of the smelter as a hazardous waste sitehouses were sold as close as 017 miles to it In the rstperiod after cleanup (1987ndash1990) no houses within a mileof the RSR site were sold14 These results can be interpretedto mean that home sellers and buyers have different expec-tations during each of our time periods During the rstpostcleanup period (1987ndash1990) it is possible to look atpostcleanup stigma After 1991 when the new concernsarise there should be different expectations

C Estimation Techniques and Functional Form

Rosen (1974) who proposes a two-step process developsthe hedonic model The rst step is to estimate the hedonicprice function In the second step the hedonic prices of theattributes are used as dependent variables in the estimationof (inverse) bid functions In order to solve the identi cationproblem and achieve consistent estimation of the two-stepmodel it is assumed that supplies of housing attributes areexogenously given and vary across submarkets and house-hold demand is a function of observable household charac-teristics and a single unobserved taste variable (Coulson andBond 1990) This taste variable has the same conditionaldistribution across submarkets These assumptions are stan-dard in hedonic demand studies and can be defended to theextent that most of the housing stock is accumulated capitalin place and adjustment costs are high Consequently thecurrent state of an arearsquos housing market is largely deter-mined by the accumulated effects of historical accidents inthat market This means that the marginal implicit prices ofattributes will vary independently of the other demand-shiftvariables (Coulson and Bond 1990) Under this speci ca-tion ordinary least squares can be used to estimate thehedonic price equation but not the bid functions because

13 Slovic et al (1991 p 685)

14 The usual explanation for a lack of sales around a locally undesirableland use is that there are no buyers However it may also be the case thatpotential sellers are holding on to their properties with the hope thatproperty values will rise


their error term is in uenced by the unobserved taste com-ponent An instrumental variables approach can be used inthe second step if the conditions for identi cation are met

Our study follows the previously cited literature on theempirical effects of locally undesirable land and considersonly linear and semilog (natural logarithm of the dependentvariable) functional forms A linear speci cation has theobvious interpretation that a unit increase in an attributecauses the price to rise by an amount equal to the coef -cient with a semi-log speci cation the coef cients can beinterpreted as percentages of the average house price Giventhe presence of independent indicator variables a Box-Coxtransformation of the dependent variable is used to choosebetween the linear or natural logarithmic forms for thedependent variable The hypothesis that the linear form ispreferred could be rejected for every year Although thehypothesis that the semi-log speci cation is best could berejected for most years the estimates of l are always closeto zero15 Given this limited analysis of functional form thefollowing semi-log speci cation below is reported

ln P~x 5 b0 1 Obixi 1 e (7)

where P is the sale price of the home the xirsquos are the variousattributes of the house and e is a white-noise error term

D Quantitative Measures of Distance Stigma

We rst estimate the standard distance model given byequation (7) for the entire Dallas County housing marketThe estimation results are presented in table 2 The esti-mated coef cients have the expected signs and are statisti-cally signi cant in each period with only a few exceptionsThese exceptions are that a few of the school-district indi-cator variables are not signi cant at conventional levels inthe rst and third time periods In addition a few of theschool-district indicator coef cients vary signi cantly overtime This can be explained by the fact that there aredifferent information sets for each time period for thetradeoffs between the value of location in a speci c schooldistrict and health risks Additional baths living area andlot size increase the sale price of a home Also the presenceof a pool garage central air conditioning and central heateach have a positive effect on home values Houses in goodor average condition command a price premium Housesthat are located in census tracts with higher percentages ofpoverty African-American residents and Hispanic resi-dents sell at lower prices Houses that are close to the

Galleria Mall sell at higher prices than those that are fartheraway in all four periods The signs on the coef cients of thedistance variables (distance to the airport the RSR smelterand the central business district) change over time in thismodel

Our rst hypothesis is that people pay a premium fordistance from the RSR smelter The price gradient fordistance starts out signi cantly positive before the EPAidenti cation of the RSR site and during cleanup of the siteindicating that buyers are willing to pay a premium for alocation that is farther away from the RSR site The positivesign on distance before EPA identi cation can be interpretedto mean that some effect of the RSR site was alreadycapitalized in property values in 1979ndash1980 However aftercleanup this coef cient turns signi cantly negative Thisdiffers from the expected sign of the distance coef cientwhich is either positive or zero

15 The Box-Cox transformation is

p~l 5 H Pl 2 1

l l THORN 0

ln l l 5 0

Using Box-Cox maximum likelihood analysis l was estimated for eachyear The yearly estimates of l range from 2009 to 021 A value of l 50 (l 5 1) implies that a semi-log (a linear) speci cation is best

TABLE 2mdashDISTANCE MODEL HEDONIC ESTIMATION RESULTS

Variable

Value

1979ndash1980 1981ndash1986 1987ndash1990 1991ndash1995

Livarea 439E24 431E24 421E24 408E24(9631) (17954) (15764) (15619)

Baths 0090 0064 0057 0064(2158) (2778) (2190) (2388)

Pool 0050 0079 0098 0072(853) (2677) (3017) (2071)

Landarea 950E27 290E26 445E26 664E26(188) (1127) (1501) (2275)

Garage 0087 0090 0093 0138(2037) (3392) (2566) (3951)

Aircon 0107 0125 0135 0200(1542) (2830) (2181) (3263)

Heatc 0110 0085 0101 0164(1541) (1850) (1570) (2708)

Good 0218 0293 0452 0642(1597) (3538) (4388) (7978)

Average 0150 0164 0284 0426(1175) (2046) (2807) (5432)

Galleria 20019 20014 20014 20021(21316) (22173) (21954) (23045)

Dalcbd 20016 20032 0004 0016(2559) (22144) (205) (876)

Dfwair 20007 20014 0002 0007(2342) (21430) (178) (616)

Pbpov 20008 20003 20002 20003(21526) (21055) (2561) (21349)

Pblack 20002 20003 20005 20007(21481) (23640) (24404) (27345)

Phisp 20003 20004 20007 20007(21123) (22915) (24206) (25062)

Distrsr 0011 0018 20033 20048(350) (1078) (21587) (22393)

The coef cients for years and school districts are suppressed due to space limitations Please contactthe corresponding author to obtain these results t-Statistics are in parentheses


There are a number of explanations for the negative signson distance in the postcleanup periods The most compellingexplanation is that sphere of in uence of the smelter islimited This issue is explored with an examination of thecontinuity price gradient for distance Another possibleexplanation is that very few houses in close proximity to thesmelter were sold after cleanup These discounted houses nolonger affect the coef cient on the distance variable aftercleanup16

Our second hypothesis is that the coef cient on distancedid not change over the different event-driven time periodsThis hypothesis is a crude test of the duration of stigma17

For example if the coef cient on distance starts out posi-tive and then after remediation it is no longer positive thenstigma is only temporary Although we already know thatthe coef cients change from signi cantly positive to signif-icantly negative F-tests have been conducted to testwhether the distance coef cient in each period is equal tothe coef cient on distance in every other period All of thesehypotheses can be rejected which indicates that the coef -cients on distance did signi cantly change across periodsThe F-statistics range from 866 to 61992

Continuity of the Price Gradient Previous studiessuch as McClelland et al (1983) nd that the impact of thewaste site on property values dissipates rapidly with dis-tance Following Thayer et al (1992) we use linear splinefunctions to allow for discontinuities of the price gradientLinear splines allow for there to be one premium fordistance up to a critical point and then an adjustment to thepremium after that point We rst allow the price gradient tobe discontinuous at one point and then attempt to choosethe critical point from the set 1 11 12 13 10using a grid search with the entire data set18 Our criterionfor choosing the critical point is minimizing the sum ofsquared errors There are two values that result in localminima and a third value that results in a global minimumFrom these ndings we allow the price gradient to bediscontinuous at three points Thayer et al (1992) concludefrom their ndings that there is more than one shift in thehedonic function for their data but they do not allow foradditional shifts in the price gradient for distance Using asecond grid search we choose the following three criticalpoints 12 26 and 46 miles

Formally the linear spline is structured as follows Let x1

be the distance to the site let x2 x3 and x4 be the distancesat which the price gradient is allowed to be discontinuousand let x5 to xn be the other attributes of the house Thelinear spline can be represented as

ln P~x 5 b0 1 b1x1 1 b2d2~x1 2 x2

1 b3d3~x1 2 x3 1 b4d4~x1 2 x4

1 Obixi 1 e

(8)

where

d2 5 H 1 if x1 x20 otherwise



The coef cient associated with each critical point becomesan adjustment term in the distance coef cient The adjust-ments to the price gradient are cumulative

d ln P~x

dx15 b1 1 b2d2 1 b3d3 1 b4d4 (9)

Hedonic Empirical Results We estimate the hedonicmodel with linear splines given by equation (8) The esti-mation results are presented for the distance coef cient b1

and the adjustment coef cients b2 b3 and b4 in table 3 andare depicted graphically in gure 1 The hypothesis that theeffect of the smelter is constant with distance can be rejectedif the coef cients for the adjustment variables are differentfrom zero The coef cients on the adjustment variables aresigni cantly different from zero for each critical point in

16 Dale et al (1999) offer two possible explanations for a negativedistance coef cient in a previously stigmatized area First the economet-ric model may be misspeci ed Second the cleanup could have providedamenities not found elsewhere in the urban area We agree that the simpledistance model is misspeci ed because of the discontinuity in the pricegradient We did not nd any evidence that the cleanup provided uniqueamenities

17 The reason that this is a crude test for the duration of stigma is that theprice gradient for distance from the smelter will be discontinuous if thesphere of in uence of the smelter dissipates rapidly with distance

18 One mile is the lowest possible starting point because the number ofsales transactions is very small in close proximity to the smelter

TABLE 3mdashHEDONIC PRICE REGRESSIONS WITH LINEAR SPLINE FUNCTION

Period Quantity Value t-StatisticPrice

Gradient

1979ndash1980 Distance from RSR 0880 488 0880Adjustment 1 (12 miles) 21100 2568 20220Adjustment 2 (26 miles) 0279 809 0059Adjustment 3 (46 miles) 20048 2627 0011





each period As expected in each period the distancecoef cient is positive and the adjustment at 12 miles isnegative The interpretation is that the rst adjustmentcoef cient is negative because the in uence of the smelterdiminishes with increasing distance The second adjustmentto the price gradient is positive in each case This is causedby some higher-price houses that are close but outside the12 mile sphere of in uence of the smelter The thirdadjustment is sometimes positive and sometimes negativebut always small in relative magnitude

The duration of stigma in close proximity to the smeltercan be tested while allowing for discontinuities in the pricegradient We conduct F-tests to discern whether the distancecoef cient in each period is equal to that in every otherperiod The distance coef cient for period 3 (1987ndash1990) isnot signi cantly different from that for period 1 (1979ndash1980) and the distance coef cient for period 2 (1981ndash1986)is not signi cantly different from that for period 4 (1991ndash1994) The rest of these hypotheses can be rejected whichindicates that the effect of proximity to the smelter doessigni cantly change for these periods The F-statistics rangefrom 481 to 2323

The positive coef cient on the distance coef cient in1979ndash1980 should be interpreted to mean that the marketwas aware of the smelter as a disamenity However afterEPA identi cation of the RSR site as a hazardous waste sitethis coef cient signi cantly increases The fact that the

coef cient greatly increases in magnitude provides furthersupport that a neighborhood externality occurred

The nding that the price gradient for a distance of lessthan 12 miles is positive after remediation of the site andbefore the period of new concern is important It means thatwithin the 12 mile radius of the RSR site home buyers paya premium for a home that is located farther away from theRSR site even after a court ruled that the site was cleanedup In other words with the linear spline model we nd thatthere was a postcleanup stigma (1987ndash1990) within a verylimited (no greater than 12 miles) sphere of in uenceThere is also evidence of stigma during the period 1991ndash1995 but the additional concerns and new informationcomplicate this result

E Bid-Function Estimation

In order to estimate a two-step hedonic model with bidfunctions we segment the Dallas County housing marketinto four submarkets by combining similar school districtsin the same geographic area The groupings with pricestatistics are presented in table 4 Segmenting the marketsolves the identi cation problem because similar individu-als must choose in markets with different hedonic pricefunctions (Freeman 1993) To implement this approach we rst estimate separate hedonic price functions for eachhousing submarket using the same speci cation Next the

FIGURE 1mdashLOG OF PRICE PREMIUM FOR DISTANCE FROM THE RSR SMELTER

TABLE 4mdashDALLAS COUNTY HOUSING SUBMARKETS

Submarket School DistrictsMean (Std Dev)Real Sales Price

Mean (Std Dev) HedonicPrice for Distance

Mean (Std Dev) HedonicPrice for Poverty

Northeast Garland Richardson MesquiteSunnyvale no district

84142 (58463) 20012 (00004) 20001 (745E26)

Northwest CarrolltonFarmers BranchCoppell

85720 (36410) 2148 (0590) 0038 (0004)

South Duncanville Cedar Hill DesotoLancasterWilmer Hutchins

67069 (34644) 1570 (0567) 20031 (0004)

Central Dallas School District Highland ParkIrving Grand Prairie

91046 (106515) 20183 (0217) 20016 (0003)


hedonic prices of the attributes are used as dependentvariables in the estimation of (inverse) bid functions Pre-suming that other goods are separable from housing at-tributes the parameters of the bid functions can be esti-mated using housing characteristics and demographicvariables that shift preferences as explanatory variables

In this stage we compute the prices of the attributes asthe derivative of the hedonic price function with respect tothe appropriate attribute The attributes we focus on are thedistance from the hazardous waste site and the percentage ofpoverty in the census tract in which the house is locatedThe independent variables in these attribute demand equa-tions include attribute quantities for distance percentage ofpoverty living area and the following census tract infor-mation to represent demographics percentage of AfricanAmericans and percentage of Hispanics19 Since the at-tribute quantities are endogenous variables two-stage leastsquares estimation is used Variables that are uncorrelatedwith the error term but highly correlated with the endoge-nous variable are appropriate instruments This means thatvariables that vary across sub-markets can be included asinstruments Therefore following Bartik (1987) and Coul-

son and Bond (1990) we use cross products betweendemand characteristics and submarket indicator variables asinstruments

The regression for the price of distance from theRSR smelter site (see table 5) indicates that householdsfrom higher-income census tracts are willing to bid sig-ni cantly more for houses that are located farther awayfrom the smelter than are households from lower-income census tracts This supports the ltering aspectof the model with respect to distance from the hazardouswaste site

The neighborhood turnover hypothesis from the theoret-ical model is re ected in the coef cients on the income-related variable (poverty) in the inverse demand functionsThe positive and highly signi cant coef cients on povertyin the price (discount) for poverty in the census tract (seetable 6) in all but the rst postcleanup period (1987ndash1990)can be interpreted to mean that people from higher-incomeareas require a larger discount to live in a lower-incomeneighborhood The estimation results for periods 1 2 and 4are consistent with the neighborhood externality aspect ofour theoretical model and with Coulson and Bondrsquos (1990) ndings The one sign that contradicts the model is in period3 the period in which few sales occurred in close proximityto the RSR site Therefore we conclude that our empiricalresults support the hypothesis that the neighborhood exter-nality effect is strong and the stigma may be long-termwithin 12 miles of the RSR site

19 Since individual demographic data for the homebuyer associated witheach transaction are unavailable census tract data are used as a proxy Theloss of detail from using census tract data is a drawback HoweverFreeman (1993) points out that census tract boundaries are chosen in aneffort to construct relatively homogeneous units in terms of housing andsocioeconomic characteristics If within-tract variation is small comparedwith the variation among tracts then little is lost with the use of censustract data

TABLE 5mdashPARAMETER ESTIMATES OF INVERSE DEMAND FUNCTION FOR

DISTANCE FROM RSR DEPENDENT VARIABLE

Period 1(1979ndash1980)




Intercept 3850 64031 57310 888089(2735) (217) (1519) (29269)

Distance fromRSR

23847 8424 248077 2750377(23114) (302) (21506) (229156)

Adjustment 1(starts at12 miles)

4345 264659 48371 755669(3274) (2289) (1491) (28796)


20558 45449 20176 22909(22353) (171) (2128) (1998)


0108 29591 20065 22369(2207) (398) (2202) (26967)

Poverty 20004 29657 20031 20006(21253) (22757) (22214) (2473)

Black 86E204 1229 20002 817E204(1109) (1134) (2465) (184)

Hispanic 0004 3427 0009 0010(2119) (1600) (1073) (1435)

Living area 228E205 159E203 2402E205 267E206(1240) (082) (2511) (031)

t-Statistics in parentheses


PERCENTAGE OF POVERTY IN CENSUS TRACT DEPENDENT VARIABLE





Intercept 20010 0002 20118 0058(2289) (040) (2269) (348)

Distance fromRSR

20006 20009 0093 20059(2195) (2197) (252) (2419)


0007(227)

0010(213)

20100(2267)

0064(447)


20004(2668)

20002(2802)

0015(939)

20012(21462)


0003(2710)

0001(1974)

20007(21925)

0009(4890)

Poverty 206E204 213E204 20002 628E204(2106) (3611) (29046) (5365)

Black 743E206 2959E206 545E204 2188E204(377) (2673) (8025) (26133)

Hispanic 194E205 2854E205 806E204 2216E204(391) (23234) (6490) (24111)

Living area 743E207 2195E208 268E207 996E207(1639) (105) (2745) (2094)



VIII Conclusion

In this paper we present a simple theoretical model withexternalities to show that it is possible to arrive at either along-term stigma equilibrium or a recovery equilibrium(temporary stigma) after the detection and cleanup of ahazardous waste site If the recovery equilibrium is the onethat emerges there will only be a temporary drop in prop-erty values In our empirical analysis we use hedonic pricefunctions to analyze whether a stigma equilibrium or arecovery equilibrium emerges for the residential propertiesand estimate bid functions for two attributes in order to ndempirical support for the theoretical model

Our empirical evidence shows that long-term stigmaexists in a very limited area The sphere of in uence of thesmelter is no larger than a circle around the smelter with a12-mile radius In the years directly following cleanupproperties within 12 miles of the RSR sell at signi cantlylower prices than properties located farther away From ourestimation of the bid functions in the second stage of ourhedonic price analysis we nd evidence that supports theneighborhood externality model Households from higher-income neighborhoods require a larger discount to live inclose proximity to the remediated hazardous waste site anda larger discount to live in a low-income neighborhood

REFERENCES

Bartik Timothy J ldquoThe Estimation of Demand Parameter in HedonicPrice Modelsrdquo Journal of Political Economy 95 (February 1987)81ndash88

Bond Sandra G ldquoStigma Assessment The Case of a Remediated Con-taminated Siterdquo Journal of Property Investment and Finance 19(2001) 188ndash210

Boyd James Winston Harrington and Molly K Macauley ldquoThe Effectsof Environmental Liability on Industrial Real Estate Develop-mentrdquo Journal of Real Estate Finance and Economics 12 (January1996) 37ndash58

Coulson N Edward and Eric W Bond ldquoA Hedonic Approach toResidential Successionrdquo this REVIEW 72 (August 1990) 433ndash443

Dale Larry James C Murdoch Mark A Thayer and Paul A WaddellldquoDo Property Values Rebound from Environmental Stigmas Ev-idence from Dallas County Texasrdquo Land Economics 75 (1999)311ndash326

Elliot-Jones Michael ldquoStigma in Light of Recent Casesrdquo Natural Re-sources amp Environment (Spring 1996) 56ndash59

Farber Stephen ldquoUndesirable Facilities and Property Values A Summaryof Empirical Studiesrdquo Ecological Economics 24 (January 1998)1ndash14

Freeman A Myrick The Measurement of Environmental and ResourceValues Theory and Methods (Washington Resources for the Fu-ture 1993)

Hall Robert W ldquoThe Causes of Loss in Value A Case Study of aContaminated Propertyrdquo Real Estate Issues 19 (April 1994) 32ndash38

Ketkar Kusum ldquoHazardous Waste Sites and Property Values in the Stateof New Jerseyrdquo Applied Economics 24 (June 1992) 647ndash659

Kiel Katherine A ldquoMeasuring the Impact of the Discovery and Cleaningof Identi ed Hazardous Waste Sites on House Valuesrdquo LandEconomics 71 (November 1995) 428ndash435

Kiel Katherine A and Katherine T McClain ldquoHouse Prices during SitingDecision States The Case of an Incinerator from Rumor throughOperationrdquo Journal of Environmental Economics and Manage-ment 28 (March 1995) 241ndash255ldquoHouse Price Recovery and Stigma after a Failed Sitingrdquo Applied

Economics 28 (November 1996) 1351ndash1358Kohlhase Janet E ldquoThe Impact of Hazardous Waste Sites on Housing

Valuesrdquo Journal of Urban Economics 30 (July 1991) 1ndash26Kraft Michael E and Denise Scheberle ldquoEnvironmental Justice and the

Allocation of Risk The Case of Lead and Public Healthrdquo PolicyStudies Journal 23 (Spring 1995) 113ndash123

Lewandrowski Lorraine ldquoToxic Back re Appraisal Techniques andCurrent Trends in Valuationrdquo Albany Law Journal of Science andTechnology (1994) 1ndash32

McClelland Gary H William D Schulze and Brian Hurd ldquoThe Effectsof Risk Beliefs on Property Values A Case Study of a HazardousWaste Siterdquo Risk Analysis 10 (December 1990) 485ndash497

Miyao Takahito ldquoDynamic Instability in a Mixed City in the Presence ofNeighborhood Externalities rdquo American Economic Review 68 (June1978) 454ndash463

Mundy Bill ldquoStigma and Valuerdquo The Appraisal Journal (January 1992)7ndash13

Patchin Peter J ldquoContaminated PropertiesmdashStigma Revisitedrdquo TheAppraisal Journal (April 1991) 167ndash172

Rosen Sherwin ldquoHedonic Prices and Implicit Marketsrdquo Journal ofPolitical Economy 90 (JanuaryndashFebruary 1974) 34ndash55

Slovic P M Layman N Kraus J Flynn J Chalmers and G GesellldquoPerceived Risk Stigma and Potential Economic Impacts of aHigh-level Nuclear Waste Repository in Nevadardquo Risk Analysis 11(December 1991) 683ndash696

Smith V Kerry and William H Desvousges ldquoThe value of avoiding aLULU Hazardous Waste Disposal Sitesrdquo The Review of Econom-ics and Statistics 68 (May 1986) 293ndash299

Thayer Mark Heidi Albers and Morteza Rahmatian 1992 ldquoThe Bene tsof Reducing Exposure to Waste Disposal Sites A Hedonic HousingValue Approachrdquo Journal of Real Estate Research 7 (Summer1992) 265ndash282


There are two major sources of uncertainty that arisefrom environmental contamination (1) whether the prop-erty is still a health risk even after the property has beenremediated and (2) what future potential liabilities exist andwho is responsible for them Using an expected-utilityapproach it can be shown that the uncertainty surroundinghazardous waste sites can result in lower property values(Boyd Harrington and Macauley 1996) More generally amonetary value can be placed on irreversible events such asa permanent change in health status and loss of life basedon the choices an individual makes about income consump-tion and risk A potential buyer must be compensated forthe expected value of future damage to his health and anamount equal to a certainty equivalent to compensate for therisk associated with the contaminated site

Uncertainty can also make it dif cult for prospectivebuyers to obtain nancing Due to uncertainty about therisks of mortgaging contaminated properties lendersrsquo will-ingness to provide nancing on contaminated properties fellfrom the late 1960s to a low point in the early 1980s whereit stayed until the early 1990s when it began to improve3

During the low-willingness-to- nance period the vast ma-jority of lenders would not consider providing nancinguntil the property has been remediated and satis ed certaintolerance limits The result of the loss of mortgagability isoften that the property is held off the market4 However arecent increase in the understanding of the management ofthe risk surrounding contaminated properties has led to agreater willingness to provide nancing for these proper-ties5

In this paper we use a theoretical model with externalityeffects to show that both temporary stigma and long-termstigma are possible equilibrium outcomes after the discov-ery and cleanup of a hazardous waste site The former isdriven only by the environmental externality whereas thelatter is driven by the neighborhood externality The exis-tence and duration of stigma are then tested for by estimat-ing a hedonic price model with data from housing salesprices in Dallas County Texas In the second stage of thehedonic price analysis the bid functions are estimated toevaluate the externality characteristics of the model Thefocus of the empirical analysis is the RSR lead smelter inWest Dallas which operated from 1934 to 1984 and causedsoil contamination from air emissions and slag material Thepooled data set used in this analysis covers the period 1979to 1995 and includes over 200000 observations

II Previous Empirical Evidence

Many authors have used property transaction data tovalue environmental attributes and more speci cally studythe impact of hazardous waste sites Researchers such asKetkar (1992) Kiel (1995) Kiel and McClain (1996)Kohlhase (1991) Smith and Desvousges (1986) andThayer et al (1992) have consistently found that proximityto hazardous waste sites and other locally undesirable landuses has a negative impact on property values6

The current body of literature on the empirical effects oflocally undesirable land uses does not address whether thediminution of property values caused by these land uses istemporary or long-term or whether externality effects existAlthough there have been many previous studies that at-tempt to measure the effect of environmental contaminationand cleanup on property values they generally focus on theshort run Most importantly almost no studies have ana-lyzed postcleanup property values Typically impacts ofcontamination on property values are examined with across-sectional data set at a single point in time7 Withoutincluding postcleanup property values studies cannot struc-ture the event analysis correctly to analyze the effects ofcleanup In contrast this analysis examines the impact ofenvironmental contamination on residential property valuesby analyzing data from before identi cation of the hazard-ous waste site before during and after cleanup has beencompleted and after new concerns were raised in thepostcleanup years Consequently it is possible to considerthe longer-run effects of contamination and remediationprocesses

The exceptions to the lack of postcleanup analysis in theliterature are Kohlhase (1991) and Dale et al (1999)Kohlhase (1991) includes a toxic site in her study that wasldquoalmost cleaned uprdquo at the time and nds statisticallyinsigni cant coef cient on price and distance Dale et al(1999) analyze the effect of the RSR smelter but take adifferent approach than is taken in the current analysis Theydo not provide a theoretical foundation or a correspondingempirical analysis for distinguishing between long-term andtemporary stigma Moreover they quantify the discontinuityof the distance price gradient surrounding the smelter byincluding indicator variables for two speci c neighbor-hoods As a result they conclude that there was no long-term stigma associated with the RSR smelter We approachthis problem with a linear spline function which speci callyallows for the in uence of the smelter to diminish withincreasing distance and arrive at different conclusions3 This lack of mortgagability has been discussed in the appraisal and

legal literatures Patchin (1991 p 169) writes ldquoThis reluctance to lendapplies not only to Superfund sites but to sites with comparatively lowlevels of contamination rdquo Lewandrowski (1994) discusses that in caseswhere nancing is still available interest rates will be higher

4 Patchin (1991 p 169)5 A specialty of appraisal for contaminated properties was developed

during the 1990s During that time the credit market went through aperiod of learning

6 For additional cites and a comprehensive survey of empirical resultssee Farber (1998)

7 Exceptions include Dale et al (1999) Kohlhase (1991) and Kiel(1995) who examine property values at more than one point in time Kieland McClain (1996) examine housing prices before and after a failedincinerator siting








U U~Q~E Y X (1)


u 5 u ~E Y U Y (2)




ui~E Y (3)







Y 5 F~EY l 1 1 2 F~EYh (5)


uEh 2 uE

l 1 ~uYh

2 uYl~Yh 2 Y l f~E 0 (6)



implicit





A The Data Set









home86010 78940




1797 615


1089 576


otherwise087 033




1155 1305


768 720


088 032


087 033


030 046


068 047

School Districts


007 026


032 047


otherwise003 015




otherwise005 002


otherwise004 019






B Event Analysis







13 Slovic et al (1991 p 685)












p~l 5 H Pl 2 1

l l THORN 0

ln l l 5 0



Variable

Value


Livarea 439E24 431E24 421E24 408E24(9631) (17954) (15764) (15619)

Baths 0090 0064 0057 0064(2158) (2778) (2190) (2388)

Pool 0050 0079 0098 0072(853) (2677) (3017) (2071)

Landarea 950E27 290E26 445E26 664E26(188) (1127) (1501) (2275)

Garage 0087 0090 0093 0138(2037) (3392) (2566) (3951)

Aircon 0107 0125 0135 0200(1542) (2830) (2181) (3263)

Heatc 0110 0085 0101 0164(1541) (1850) (1570) (2708)

Good 0218 0293 0452 0642(1597) (3538) (4388) (7978)

Average 0150 0164 0284 0426(1175) (2046) (2807) (5432)

Galleria 20019 20014 20014 20021(21316) (22173) (21954) (23045)

Dalcbd 20016 20032 0004 0016(2559) (22144) (205) (876)

Dfwair 20007 20014 0002 0007(2342) (21430) (178) (616)

Pbpov 20008 20003 20002 20003(21526) (21055) (2561) (21349)

Pblack 20002 20003 20005 20007(21481) (23640) (24404) (27345)

Phisp 20003 20004 20007 20007(21123) (22915) (24206) (25062)

Distrsr 0011 0018 20033 20048(350) (1078) (21587) (22393)









ln P~x 5 b0 1 b1x1 1 b2d2~x1 2 x2

1 b3d3~x1 2 x3 1 b4d4~x1 2 x4

1 Obixi 1 e

(8)

where





d ln P~x

dx15 b1 1 b2d2 1 b3d3 1 b4d4 (9)








Gradient



















84142 (58463) 20012 (00004) 20001 (745E26)


85720 (36410) 2148 (0590) 0038 (0004)


67069 (34644) 1570 (0567) 20031 (0004)


91046 (106515) 20183 (0217) 20016 (0003)














Intercept 3850 64031 57310 888089(2735) (217) (1519) (29269)

Distance fromRSR

23847 8424 248077 2750377(23114) (302) (21506) (229156)


4345 264659 48371 755669(3274) (2289) (1491) (28796)


20558 45449 20176 22909(22353) (171) (2128) (1998)


0108 29591 20065 22369(2207) (398) (2202) (26967)

Poverty 20004 29657 20031 20006(21253) (22757) (22214) (2473)

Black 86E204 1229 20002 817E204(1109) (1134) (2465) (184)

Hispanic 0004 3427 0009 0010(2119) (1600) (1073) (1435)

Living area 228E205 159E203 2402E205 267E206(1240) (082) (2511) (031)








Intercept 20010 0002 20118 0058(2289) (040) (2269) (348)

Distance fromRSR

20006 20009 0093 20059(2195) (2197) (252) (2419)


0007(227)

0010(213)

20100(2267)

0064(447)


20004(2668)

20002(2802)

0015(939)

20012(21462)


0003(2710)

0001(1974)

20007(21925)

0009(4890)

Poverty 206E204 213E204 20002 628E204(2106) (3611) (29046) (5365)

Black 743E206 2959E206 545E204 2188E204(377) (2673) (8025) (26133)

Hispanic 194E205 2854E205 806E204 2216E204(391) (23234) (6490) (24111)

Living area 743E207 2195E208 268E207 996E207(1639) (105) (2745) (2094)



VIII Conclusion



REFERENCES
































U U~Q~E Y X (1)


u 5 u ~E Y U Y (2)




ui~E Y (3)







Y 5 F~EY l 1 1 2 F~EYh (5)


uEh 2 uE

l 1 ~uYh

2 uYl~Yh 2 Y l f~E 0 (6)



implicit





A The Data Set









home86010 78940




1797 615


1089 576


otherwise087 033




1155 1305


768 720


088 032


087 033


030 046


068 047

School Districts


007 026


032 047


otherwise003 015




otherwise005 002


otherwise004 019






B Event Analysis







13 Slovic et al (1991 p 685)












p~l 5 H Pl 2 1

l l THORN 0

ln l l 5 0



Variable

Value


Livarea 439E24 431E24 421E24 408E24(9631) (17954) (15764) (15619)

Baths 0090 0064 0057 0064(2158) (2778) (2190) (2388)

Pool 0050 0079 0098 0072(853) (2677) (3017) (2071)

Landarea 950E27 290E26 445E26 664E26(188) (1127) (1501) (2275)

Garage 0087 0090 0093 0138(2037) (3392) (2566) (3951)

Aircon 0107 0125 0135 0200(1542) (2830) (2181) (3263)

Heatc 0110 0085 0101 0164(1541) (1850) (1570) (2708)

Good 0218 0293 0452 0642(1597) (3538) (4388) (7978)

Average 0150 0164 0284 0426(1175) (2046) (2807) (5432)

Galleria 20019 20014 20014 20021(21316) (22173) (21954) (23045)

Dalcbd 20016 20032 0004 0016(2559) (22144) (205) (876)

Dfwair 20007 20014 0002 0007(2342) (21430) (178) (616)

Pbpov 20008 20003 20002 20003(21526) (21055) (2561) (21349)

Pblack 20002 20003 20005 20007(21481) (23640) (24404) (27345)

Phisp 20003 20004 20007 20007(21123) (22915) (24206) (25062)

Distrsr 0011 0018 20033 20048(350) (1078) (21587) (22393)









ln P~x 5 b0 1 b1x1 1 b2d2~x1 2 x2

1 b3d3~x1 2 x3 1 b4d4~x1 2 x4

1 Obixi 1 e

(8)

where





d ln P~x

dx15 b1 1 b2d2 1 b3d3 1 b4d4 (9)








Gradient



















84142 (58463) 20012 (00004) 20001 (745E26)


85720 (36410) 2148 (0590) 0038 (0004)


67069 (34644) 1570 (0567) 20031 (0004)


91046 (106515) 20183 (0217) 20016 (0003)














Intercept 3850 64031 57310 888089(2735) (217) (1519) (29269)

Distance fromRSR

23847 8424 248077 2750377(23114) (302) (21506) (229156)


4345 264659 48371 755669(3274) (2289) (1491) (28796)


20558 45449 20176 22909(22353) (171) (2128) (1998)


0108 29591 20065 22369(2207) (398) (2202) (26967)

Poverty 20004 29657 20031 20006(21253) (22757) (22214) (2473)

Black 86E204 1229 20002 817E204(1109) (1134) (2465) (184)

Hispanic 0004 3427 0009 0010(2119) (1600) (1073) (1435)

Living area 228E205 159E203 2402E205 267E206(1240) (082) (2511) (031)








Intercept 20010 0002 20118 0058(2289) (040) (2269) (348)

Distance fromRSR

20006 20009 0093 20059(2195) (2197) (252) (2419)


0007(227)

0010(213)

20100(2267)

0064(447)


20004(2668)

20002(2802)

0015(939)

20012(21462)


0003(2710)

0001(1974)

20007(21925)

0009(4890)

Poverty 206E204 213E204 20002 628E204(2106) (3611) (29046) (5365)

Black 743E206 2959E206 545E204 2188E204(377) (2673) (8025) (26133)

Hispanic 194E205 2854E205 806E204 2216E204(391) (23234) (6490) (24111)

Living area 743E207 2195E208 268E207 996E207(1639) (105) (2745) (2094)



VIII Conclusion



REFERENCES





























A The Data Set









home86010 78940




1797 615


1089 576


otherwise087 033




1155 1305


768 720


088 032


087 033


030 046


068 047

School Districts


007 026


032 047


otherwise003 015




otherwise005 002


otherwise004 019






B Event Analysis







13 Slovic et al (1991 p 685)












p~l 5 H Pl 2 1

l l THORN 0

ln l l 5 0



Variable

Value


Livarea 439E24 431E24 421E24 408E24(9631) (17954) (15764) (15619)

Baths 0090 0064 0057 0064(2158) (2778) (2190) (2388)

Pool 0050 0079 0098 0072(853) (2677) (3017) (2071)

Landarea 950E27 290E26 445E26 664E26(188) (1127) (1501) (2275)

Garage 0087 0090 0093 0138(2037) (3392) (2566) (3951)

Aircon 0107 0125 0135 0200(1542) (2830) (2181) (3263)

Heatc 0110 0085 0101 0164(1541) (1850) (1570) (2708)

Good 0218 0293 0452 0642(1597) (3538) (4388) (7978)

Average 0150 0164 0284 0426(1175) (2046) (2807) (5432)

Galleria 20019 20014 20014 20021(21316) (22173) (21954) (23045)

Dalcbd 20016 20032 0004 0016(2559) (22144) (205) (876)

Dfwair 20007 20014 0002 0007(2342) (21430) (178) (616)

Pbpov 20008 20003 20002 20003(21526) (21055) (2561) (21349)

Pblack 20002 20003 20005 20007(21481) (23640) (24404) (27345)

Phisp 20003 20004 20007 20007(21123) (22915) (24206) (25062)

Distrsr 0011 0018 20033 20048(350) (1078) (21587) (22393)









ln P~x 5 b0 1 b1x1 1 b2d2~x1 2 x2

1 b3d3~x1 2 x3 1 b4d4~x1 2 x4

1 Obixi 1 e

(8)

where





d ln P~x

dx15 b1 1 b2d2 1 b3d3 1 b4d4 (9)








Gradient



















84142 (58463) 20012 (00004) 20001 (745E26)


85720 (36410) 2148 (0590) 0038 (0004)


67069 (34644) 1570 (0567) 20031 (0004)


91046 (106515) 20183 (0217) 20016 (0003)














Intercept 3850 64031 57310 888089(2735) (217) (1519) (29269)

Distance fromRSR

23847 8424 248077 2750377(23114) (302) (21506) (229156)


4345 264659 48371 755669(3274) (2289) (1491) (28796)


20558 45449 20176 22909(22353) (171) (2128) (1998)


0108 29591 20065 22369(2207) (398) (2202) (26967)

Poverty 20004 29657 20031 20006(21253) (22757) (22214) (2473)

Black 86E204 1229 20002 817E204(1109) (1134) (2465) (184)

Hispanic 0004 3427 0009 0010(2119) (1600) (1073) (1435)

Living area 228E205 159E203 2402E205 267E206(1240) (082) (2511) (031)








Intercept 20010 0002 20118 0058(2289) (040) (2269) (348)

Distance fromRSR

20006 20009 0093 20059(2195) (2197) (252) (2419)


0007(227)

0010(213)

20100(2267)

0064(447)


20004(2668)

20002(2802)

0015(939)

20012(21462)


0003(2710)

0001(1974)

20007(21925)

0009(4890)

Poverty 206E204 213E204 20002 628E204(2106) (3611) (29046) (5365)

Black 743E206 2959E206 545E204 2188E204(377) (2673) (8025) (26133)

Hispanic 194E205 2854E205 806E204 2216E204(391) (23234) (6490) (24111)

Living area 743E207 2195E208 268E207 996E207(1639) (105) (2745) (2094)



VIII Conclusion



REFERENCES





























B Event Analysis







13 Slovic et al (1991 p 685)












p~l 5 H Pl 2 1

l l THORN 0

ln l l 5 0



Variable

Value


Livarea 439E24 431E24 421E24 408E24(9631) (17954) (15764) (15619)

Baths 0090 0064 0057 0064(2158) (2778) (2190) (2388)

Pool 0050 0079 0098 0072(853) (2677) (3017) (2071)

Landarea 950E27 290E26 445E26 664E26(188) (1127) (1501) (2275)

Garage 0087 0090 0093 0138(2037) (3392) (2566) (3951)

Aircon 0107 0125 0135 0200(1542) (2830) (2181) (3263)

Heatc 0110 0085 0101 0164(1541) (1850) (1570) (2708)

Good 0218 0293 0452 0642(1597) (3538) (4388) (7978)

Average 0150 0164 0284 0426(1175) (2046) (2807) (5432)

Galleria 20019 20014 20014 20021(21316) (22173) (21954) (23045)

Dalcbd 20016 20032 0004 0016(2559) (22144) (205) (876)

Dfwair 20007 20014 0002 0007(2342) (21430) (178) (616)

Pbpov 20008 20003 20002 20003(21526) (21055) (2561) (21349)

Pblack 20002 20003 20005 20007(21481) (23640) (24404) (27345)

Phisp 20003 20004 20007 20007(21123) (22915) (24206) (25062)

Distrsr 0011 0018 20033 20048(350) (1078) (21587) (22393)









ln P~x 5 b0 1 b1x1 1 b2d2~x1 2 x2

1 b3d3~x1 2 x3 1 b4d4~x1 2 x4

1 Obixi 1 e

(8)

where





d ln P~x

dx15 b1 1 b2d2 1 b3d3 1 b4d4 (9)








Gradient



















84142 (58463) 20012 (00004) 20001 (745E26)


85720 (36410) 2148 (0590) 0038 (0004)


67069 (34644) 1570 (0567) 20031 (0004)


91046 (106515) 20183 (0217) 20016 (0003)














Intercept 3850 64031 57310 888089(2735) (217) (1519) (29269)

Distance fromRSR

23847 8424 248077 2750377(23114) (302) (21506) (229156)


4345 264659 48371 755669(3274) (2289) (1491) (28796)


20558 45449 20176 22909(22353) (171) (2128) (1998)


0108 29591 20065 22369(2207) (398) (2202) (26967)

Poverty 20004 29657 20031 20006(21253) (22757) (22214) (2473)

Black 86E204 1229 20002 817E204(1109) (1134) (2465) (184)

Hispanic 0004 3427 0009 0010(2119) (1600) (1073) (1435)

Living area 228E205 159E203 2402E205 267E206(1240) (082) (2511) (031)








Intercept 20010 0002 20118 0058(2289) (040) (2269) (348)

Distance fromRSR

20006 20009 0093 20059(2195) (2197) (252) (2419)


0007(227)

0010(213)

20100(2267)

0064(447)


20004(2668)

20002(2802)

0015(939)

20012(21462)


0003(2710)

0001(1974)

20007(21925)

0009(4890)

Poverty 206E204 213E204 20002 628E204(2106) (3611) (29046) (5365)

Black 743E206 2959E206 545E204 2188E204(377) (2673) (8025) (26133)

Hispanic 194E205 2854E205 806E204 2216E204(391) (23234) (6490) (24111)

Living area 743E207 2195E208 268E207 996E207(1639) (105) (2745) (2094)



VIII Conclusion



REFERENCES



































p~l 5 H Pl 2 1

l l THORN 0

ln l l 5 0



Variable

Value


Livarea 439E24 431E24 421E24 408E24(9631) (17954) (15764) (15619)

Baths 0090 0064 0057 0064(2158) (2778) (2190) (2388)

Pool 0050 0079 0098 0072(853) (2677) (3017) (2071)

Landarea 950E27 290E26 445E26 664E26(188) (1127) (1501) (2275)

Garage 0087 0090 0093 0138(2037) (3392) (2566) (3951)

Aircon 0107 0125 0135 0200(1542) (2830) (2181) (3263)

Heatc 0110 0085 0101 0164(1541) (1850) (1570) (2708)

Good 0218 0293 0452 0642(1597) (3538) (4388) (7978)

Average 0150 0164 0284 0426(1175) (2046) (2807) (5432)

Galleria 20019 20014 20014 20021(21316) (22173) (21954) (23045)

Dalcbd 20016 20032 0004 0016(2559) (22144) (205) (876)

Dfwair 20007 20014 0002 0007(2342) (21430) (178) (616)

Pbpov 20008 20003 20002 20003(21526) (21055) (2561) (21349)

Pblack 20002 20003 20005 20007(21481) (23640) (24404) (27345)

Phisp 20003 20004 20007 20007(21123) (22915) (24206) (25062)

Distrsr 0011 0018 20033 20048(350) (1078) (21587) (22393)









ln P~x 5 b0 1 b1x1 1 b2d2~x1 2 x2

1 b3d3~x1 2 x3 1 b4d4~x1 2 x4

1 Obixi 1 e

(8)

where





d ln P~x

dx15 b1 1 b2d2 1 b3d3 1 b4d4 (9)








Gradient



















84142 (58463) 20012 (00004) 20001 (745E26)


85720 (36410) 2148 (0590) 0038 (0004)


67069 (34644) 1570 (0567) 20031 (0004)


91046 (106515) 20183 (0217) 20016 (0003)














Intercept 3850 64031 57310 888089(2735) (217) (1519) (29269)

Distance fromRSR

23847 8424 248077 2750377(23114) (302) (21506) (229156)


4345 264659 48371 755669(3274) (2289) (1491) (28796)


20558 45449 20176 22909(22353) (171) (2128) (1998)


0108 29591 20065 22369(2207) (398) (2202) (26967)

Poverty 20004 29657 20031 20006(21253) (22757) (22214) (2473)

Black 86E204 1229 20002 817E204(1109) (1134) (2465) (184)

Hispanic 0004 3427 0009 0010(2119) (1600) (1073) (1435)

Living area 228E205 159E203 2402E205 267E206(1240) (082) (2511) (031)








Intercept 20010 0002 20118 0058(2289) (040) (2269) (348)

Distance fromRSR

20006 20009 0093 20059(2195) (2197) (252) (2419)


0007(227)

0010(213)

20100(2267)

0064(447)


20004(2668)

20002(2802)

0015(939)

20012(21462)


0003(2710)

0001(1974)

20007(21925)

0009(4890)

Poverty 206E204 213E204 20002 628E204(2106) (3611) (29046) (5365)

Black 743E206 2959E206 545E204 2188E204(377) (2673) (8025) (26133)

Hispanic 194E205 2854E205 806E204 2216E204(391) (23234) (6490) (24111)

Living area 743E207 2195E208 268E207 996E207(1639) (105) (2745) (2094)



VIII Conclusion



REFERENCES
































ln P~x 5 b0 1 b1x1 1 b2d2~x1 2 x2

1 b3d3~x1 2 x3 1 b4d4~x1 2 x4

1 Obixi 1 e

(8)

where





d ln P~x

dx15 b1 1 b2d2 1 b3d3 1 b4d4 (9)








Gradient



















84142 (58463) 20012 (00004) 20001 (745E26)


85720 (36410) 2148 (0590) 0038 (0004)


67069 (34644) 1570 (0567) 20031 (0004)


91046 (106515) 20183 (0217) 20016 (0003)














Intercept 3850 64031 57310 888089(2735) (217) (1519) (29269)

Distance fromRSR

23847 8424 248077 2750377(23114) (302) (21506) (229156)


4345 264659 48371 755669(3274) (2289) (1491) (28796)


20558 45449 20176 22909(22353) (171) (2128) (1998)


0108 29591 20065 22369(2207) (398) (2202) (26967)

Poverty 20004 29657 20031 20006(21253) (22757) (22214) (2473)

Black 86E204 1229 20002 817E204(1109) (1134) (2465) (184)

Hispanic 0004 3427 0009 0010(2119) (1600) (1073) (1435)

Living area 228E205 159E203 2402E205 267E206(1240) (082) (2511) (031)








Intercept 20010 0002 20118 0058(2289) (040) (2269) (348)

Distance fromRSR

20006 20009 0093 20059(2195) (2197) (252) (2419)


0007(227)

0010(213)

20100(2267)

0064(447)


20004(2668)

20002(2802)

0015(939)

20012(21462)


0003(2710)

0001(1974)

20007(21925)

0009(4890)

Poverty 206E204 213E204 20002 628E204(2106) (3611) (29046) (5365)

Black 743E206 2959E206 545E204 2188E204(377) (2673) (8025) (26133)

Hispanic 194E205 2854E205 806E204 2216E204(391) (23234) (6490) (24111)

Living area 743E207 2195E208 268E207 996E207(1639) (105) (2745) (2094)



VIII Conclusion



REFERENCES







































84142 (58463) 20012 (00004) 20001 (745E26)


85720 (36410) 2148 (0590) 0038 (0004)


67069 (34644) 1570 (0567) 20031 (0004)


91046 (106515) 20183 (0217) 20016 (0003)














Intercept 3850 64031 57310 888089(2735) (217) (1519) (29269)

Distance fromRSR

23847 8424 248077 2750377(23114) (302) (21506) (229156)


4345 264659 48371 755669(3274) (2289) (1491) (28796)


20558 45449 20176 22909(22353) (171) (2128) (1998)


0108 29591 20065 22369(2207) (398) (2202) (26967)

Poverty 20004 29657 20031 20006(21253) (22757) (22214) (2473)

Black 86E204 1229 20002 817E204(1109) (1134) (2465) (184)

Hispanic 0004 3427 0009 0010(2119) (1600) (1073) (1435)

Living area 228E205 159E203 2402E205 267E206(1240) (082) (2511) (031)








Intercept 20010 0002 20118 0058(2289) (040) (2269) (348)

Distance fromRSR

20006 20009 0093 20059(2195) (2197) (252) (2419)


0007(227)

0010(213)

20100(2267)

0064(447)


20004(2668)

20002(2802)

0015(939)

20012(21462)


0003(2710)

0001(1974)

20007(21925)

0009(4890)

Poverty 206E204 213E204 20002 628E204(2106) (3611) (29046) (5365)

Black 743E206 2959E206 545E204 2188E204(377) (2673) (8025) (26133)

Hispanic 194E205 2854E205 806E204 2216E204(391) (23234) (6490) (24111)

Living area 743E207 2195E208 268E207 996E207(1639) (105) (2745) (2094)



VIII Conclusion



REFERENCES






































Intercept 3850 64031 57310 888089(2735) (217) (1519) (29269)

Distance fromRSR

23847 8424 248077 2750377(23114) (302) (21506) (229156)


4345 264659 48371 755669(3274) (2289) (1491) (28796)


20558 45449 20176 22909(22353) (171) (2128) (1998)


0108 29591 20065 22369(2207) (398) (2202) (26967)

Poverty 20004 29657 20031 20006(21253) (22757) (22214) (2473)

Black 86E204 1229 20002 817E204(1109) (1134) (2465) (184)

Hispanic 0004 3427 0009 0010(2119) (1600) (1073) (1435)

Living area 228E205 159E203 2402E205 267E206(1240) (082) (2511) (031)








Intercept 20010 0002 20118 0058(2289) (040) (2269) (348)

Distance fromRSR

20006 20009 0093 20059(2195) (2197) (252) (2419)


0007(227)

0010(213)

20100(2267)

0064(447)


20004(2668)

20002(2802)

0015(939)

20012(21462)


0003(2710)

0001(1974)

20007(21925)

0009(4890)

Poverty 206E204 213E204 20002 628E204(2106) (3611) (29046) (5365)

Black 743E206 2959E206 545E204 2188E204(377) (2673) (8025) (26133)

Hispanic 194E205 2854E205 806E204 2216E204(391) (23234) (6490) (24111)

Living area 743E207 2195E208 268E207 996E207(1639) (105) (2745) (2094)



VIII Conclusion



REFERENCES


























VIII Conclusion



REFERENCES


























Stigmatized Asset Value: Is It Temporary or Long-Term?

Documents

Transcript of Stigmatized Asset Value: Is It Temporary or Long-Term?