Perceived Environmental Risk, Media, and Residential Sales Prices

Perceived Environmental Risk, Media, and Residential Sales Prices Authors

J u l i a F r ey b o t e a nd E ric F r ui t s

Abstract

We investigate the relationship of homebuyer risk perception and sales prices over different development stages of an environmental hazard (underground natural gas transmission pipeline) characterized by no sensory impact on homes, no accidents, and a relatively low actual risk of fatal explosions. We also investigate the moderating effect of media coverage of unrelated fatal pipeline explosions on this relationship. Using a hedonic pricing model, we find that (1) media coverage moderates the relationship of pipeline proximity (perceived risk) and sales prices in the pipeline construction phase and (2) higher perceived risk reduces sales prices once the pipeline is in operation.

Perceived risk, may it be financial, health, social or environmental, influences the purchase intentions of consumers. With regard to residential real estate, many researchers conclude that environmental risk perceived by homebuyers negatively affects residential sales prices. The closer a single-family home is to an environmental hazard (i.e., the higher the perceived risk), the lower is the sales price (e.g., Carroll, Clauretie, Jensen, and Waddoups, 1996; Boxall, Chan, and McMillan, 2005; Case, Colwell, Leishman, and Watkins, 2006; Hansen, Benson, and Hagen, 2006). The risk perception of laypersons such as homebuyers is primarily based on emotions or intuition rather than a sophisticated analysis of actual risks as conducted by experts (Slovic, 1987). Previous studies focus on homebuyer risk perception as it relates to permanent or transient environmental hazards, such as landfills, oil pipeline ruptures or nuclear plants, which tend to have a sensory impact on single-family homes either through noise, odor or unappealing views. However, we argue that even a low risk environmental hazard such as an underground gas pipeline (high-pressure transmission) without any sensory impact or accidents affects the risk perception of homebuyers. Natural gas pipelines have been subject to a great deal of controversy and uncertainty among homeowners about their safety. The natural gas pipeline system can be divided into two categories: (1) high-pressure transmission pipelines that move large volumes of liquefied natural gas in a relatively straight line and (2)

J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 1 8

u

F r e y b o t e

a n d

F r u i t s

low-pressure distribution lines that spread out like spider webs providing natural gas to homes and businesses. Historical data from the Department of Transportation (DOT) show that fatal gas transmission explosions are significantly less frequent than fatal gas distribution explosions. However, Slovic (1987) concludes that laypersons consider liquefied natural gas transport and storage to be very risky with regard to catastrophic and deadly outcomes and medium risky with regard to the immediacy of effects and unknown risks. As a consequence, we expect the risk perception of homebuyers to be affected by the presence of the underground pipeline, irrespective of the actual probability of pipeline explosions or its non-sensory character, which in turn negatively affects the sales prices of single-family homes in close proximity to the pipeline. Additionally, the environmental risk perceived by laypersons is influenced by media coverage. We argue that particularly during the construction stage of the underground pipeline, media coverage of unrelated fatal gas pipeline explosions is likely to increase the perceived risk of homebuyers and subsequently reduce the sales prices of homes in close proximity to the future pipeline location. In the construction phase, the potential environmental and health risks of the pipeline are likely to be less known to homebuyers. Media coverage of unrelated pipeline explosions increases awareness of potential pipeline risks among homebuyers and affects their risk perception through increased availability of related information and images. In the presence of media reports of deadly yet unrelated pipeline explosions, homebuyers are likely to assess the environmental and health risks of pipelines to be higher and subsequently penalize homes in close proximity to the pipeline, even before it is in operation and transporting natural gas. Our results support our expectations. During the pipeline construction period, media coverage of unrelated pipeline explosions moderates the relation of proximity to future pipeline location (perceived risk) and sales prices. In months without unrelated fatal explosions, proximity to the future pipeline location had the same effect on sales prices as in the pre-pipeline period. However, in months with fatal explosions, homes in close proximity to the future pipeline location suddenly experienced reduced sales prices. After the pipeline was in operation and transporting natural gas, homes that were closer to the pipeline also sold at a discount. Our study contributes to the following literatures. Researchers have found that amenities and disamenities affect single-family home values positively and negatively respectively (e.g., McCluskey and Rausser, 2001; Rosiers, 2002; Simons and Saginor, 2006; McKenzie and Levendis, 2010; Hoen et al., 2011; Zahirovic-Herbert and Chatterjee, 2011; Hansz and Hayunga, 2012; Wilde, Loos, and Williamson, 2012; Gordon, Winkler, Barrett, and Zumpano, 2013; Turnbull, Zahirovic-Herbert, and Mothorpe, 2013; Wyman, Hutchison, and Tiwari, 2014). Our findings complement this body of research by suggesting that the mere presence of a negative externality, even in form of a low risk and non-sensory environmental hazard, negatively affects residential sales prices.

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 1 9

A few researchers explicitly or implicitly investigate the effect of media coverage on perceived risk and home values in proximity to permanent or transient environmental hazards (Flynn, Peters, Mertz, and Slovic, 1998; McCluskey and Rausser, 2001, 2003). However, these earlier studies focus on media reports about the particular hazard, for example, following related accidents. Our study distinguishes itself from these previous studies by investigating the impact of unrelated gas pipeline explosions on sales prices of homes close to an accidentfree pipeline, particularly during different development stages of the pipeline. Our study complements Bauer, Braun, and Kvasnicka (2013), who investigate the impact of an unrelated nuclear power plant accident in Japan on sales prices of single-family homes in close proximity to existing nuclear plants in Germany. Secondly, an emerging literature shows the importance of online searches for information gathering by homebuyers (Hohenstatt, Ka¨sbauer, and Scha¨fers, 2011; Richardson and Zumpano, 2012; Seiler, Madhavan, and Liechty, 2012; Beracha and Wintoki, 2013). With the increased transition from print to online newspapers, our focus on media coverage of unrelated explosions and its effect on perceived risk provides additional evidence for the reliance of homebuyers on online (news) searches as part of their information-gathering process, which in turn affect purchasing and pricing decisions. Lastly, a number of researchers investigating factors such as loss aversion (Sun and Seiler, 2013) and strategic default (Seiler and Walden, 2014) emphasize the importance of homeowner psychology for decision-making in residential real estate markets. Purchase decisions by homeowners are not only driven by fundamentals, but also by psychological factors such as the availability heuristic or emotions. By investigating homebuyer risk perception and how it is affected by media coverage, we provide additional evidence for the impact of homebuyer psychology on residential real estate decisions. The paper is structured as follows. The next section provides a review of previous studies on environmental hazards, sales prices, and perceived risk, which is followed by a discussion of our data and methodology. We then present and discuss our results, which is followed by a conclusion.

u

Literature Review

Perceived risk has been found to negatively affect the attitude of consumers towards particular products and their purchase decisions (Laroche, Yang, McDougall, and Bergeron, 2005; Jakus and Shaw, 2003; Keh and Pang, 2010; Choi, Lee, and Ok, 2013; Henthorne, George, and Smith, 2013; Liao and Hsieh, 2013). Social responses to environmental hazards can be explained with the concept of social amplification of risk. Risk-relevant events such as a pipeline explosion or nuclear reactor accidents enter social communication processes, concerns, and awareness about a particular risk and thereby affect individual and group behaviors. This is known as ‘‘technological stigma,’’ which leads to an J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 2 0

u

F r e y b o t e

a n d

F r u i t s

avoidance behavior in the form of shunning or avoiding, for example, places with a high perceived risk, such as single-family homes close to a nuclear plant (Flynn, Peters, Mertz, and Slovic, 1998). Slovic (1987) emphasizes the difference between experts that employ sophisticated risk assessment techniques to evaluate environmental risks and laypersons that rely on intuitive risk judgments. The risk perception of laypersons and the general public is increased by media coverage of an environmental hazard (Slovic, 1987; McCluskey and Rausser, 2001; Watson, Riffe, Smithson-Stanley, and Ogilvie, 2013) and decreased by the constant exposure (frequent contact) of homeowners to a hazardous site such as nuclear reactor (Maderthaner, Guttmann, Swaton, and Otway, 1978). Accidents related to an environmental hazard thereby represent a signal whose information content about potential consequences are particularly affected by how catastrophic or deadly the outcomes are (dread risk) and how unknown or delayed risks are (unknown risk; Slovic, 1987). Media coverage of environmental hazards results in greater perceived risk as it increases awareness, perception of a danger or the number and geographical spread of individuals with knowledge of the hazard (Flynn, Peters, Mertz, and Slovic, 1998). The effect of media on perceived risk is also in line with the availability heuristic, as discussed by Tversky and Kahneman (1974). The availability heuristic allows individuals to assess the probability or frequency of an event more efficiently based on the ease with which such an event can be brought to mind. However, the availability heuristic can bias the probability judgment for an event to occur if it is based on salience, familiarity or other factors (Tversky and Kahneman, 1974). Empirically, the availability heuristic has been found to affect risk judgment and perceived risk (Pachur, Hertwig, and Steinmann, 2012). In this context, media in particular affects salience of a particular risk and distorts the frequency perception of risky events contributing to biases in risk judgments (Morton and Duck, 2001; Slater and Rasinski, 2006; Stein, Duen¸as-Osorio, and Subramanian, 2010). Many researchers have investigated the impact of negative environmental externalities such as air and water pollution, oil pipeline ruptures, leaking underground storage tanks, floodplains, landfills, and nuclear plants on residential property values (Reichert, Small, and Mohanty, 1992; Smolen, Moore, and Conway, 1992; Page and Rabinowitz, 1993; Flower and Ragas, 1994; Carroll, Clauretie, Jensen, and Waddoups, 1996; Dotzour, 1997; Simons, 1999; Boyle and Kiel, 2001; Hite, Chern, Hitzhusen, and Randall, 2001; Simons, WinsonGeideman, and Mikelbank, 2001; Rosiers, 2002; Boxall, Chan, and McMillan, 2005; Simons and Winson-Geideman, 2005; Case, Colwell, Leishman, and Watkins, 2006; Hansen, Benson, and Hagen, 2006; Simons and Saginor, 2006; McKenzie and Levendis, 2010; Hoen et al., 2011; Wilde, Loos, and Williamson, 2012; Turnbull, Zahirovic-Herbert, and Mothorpe, 2013). Contrary to the underground natural gas pipeline investigated in this study, the environmental hazards investigated in previous studies affect a single-family home through odors, noises, and/or unappealing views. Consequently, these studies find that proximity to sensory environmental hazards reduces single-family home values.

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 2 1

While the impact of proximity to an environmental hazard (perceived risk) has received considerable attention in the hedonic pricing literature, few studies, explicitly or implicitly, account for the impact of media coverage on perceived environmental risk. Flynn, Peters, Mertz, and Slovic (1998) investigate news coverage and risk perceived by the general public with regards to the highly publicized Federal Bureau of Investigation (FBI) raid of a nuclear plant. The authors find that the FBI raid increased negativity about the plant in the media and led to an emphasis on nuclear plant risks. Additionally, the raid led to an increase in perceived risk, aversion/avoidance of home purchases in close proximity to the plant (stigmatization) and reduced sales prices for these properties. Of particular relevance is the finding that individuals that knew about the raid had a more negative attitude towards the plant than those that could not recall this event. Using residential transaction data, McCluskey and Rausser (2001) investigate the effect of perceived risk of a hazardous use (waste site) on sales prices and the effect of media coverage on this perceived risk. The authors find that perceived risk reduced residential sales prices while media coverage increased perceived risk. However, perceived risk of an environmental hazard was not constant. Instead perceived risk evolved over time. In particular, perceived risk was affected by events related to the waste site, such as court rulings and news coverage, with perceptions changing accordingly. McCluskey and Rausser (2003) also show that the magnitude of housing market reactions to a hazardous waste site is affected by media coverage. Particularly in periods of new information and media reporting about the hazard, home value appreciation rates were impacted by proximity to the hazard (perceived risk). The findings of Hansen, Benson, and Hagen (2006) for the effect of a highly publicized gasoline pipeline explosion on property values suggest that media reports about the accident may have increased the risk perceived by homeowners, justifiably or not, as proximity to the pipeline prior to the accident had no significant effect on home values. Bauer, Braun, and Kvasnicka (2013) focus on how an unrelated event affects perceived environmental risk. The authors show that distant events such as the nuclear plant accident in Fukushima, Japan in 2011 influenced prices of single-family homes near active and shut down nuclear plants in Germany. The reduction of sales prices in close proximity to nuclear plants after the Fukushima accident suggests that the distant accident had an impact on the risk perception of German homebuyers. With regard to the reviewed literature, we firstly expect a negative effect of the proximity of a single-family home to the underground pipeline (perceived risk) on sales prices once the pipeline is in operation. As the evaluation of health and environmental risks by homebuyers is primarily based on emotions, the actual risks of an underground pipeline without any accidents or sensory impact are likely to be irrelevant in their assessment and pricing decisions. Thus, even a low risk environmental hazard such as an underground pipeline is hypothesized to negatively affect sales prices. J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 2 2

u

F r e y b o t e

a n d

F r u i t s

We secondly expect that media coverage of unrelated pipeline explosions moderates the relationship of perceived risk and sales prices particularly during the construction phase. While the new pipeline is still under construction, homebuyers are more likely to be uncertain or even unaware about potential pipeline risks. As a consequence, their awareness and risk judgments are likely to be affected to a larger extent by media reports about unrelated pipeline explosions. The negativity of media reports as well as images of property damage and fatalities following an explosion are expected to increase the perceived risk of homebuyers and affect their pricing decisions.

u

Data and Methodology

In our investigation, we focus on the South Mist Pipeline Extension (SMPE) in Oregon, a high-pressure intrastate transmission pipeline that was announced in September 1999 (notice of intent submitted) and went into service in September 2004. The SMPE is a 24-inch diameter natural gas pipeline, approximately 62 miles long. The SMPE is underground its entire length, with the exception of certain above ground valves and inspection points that are required by federal code. In most locations, the pipeline is buried at a depth of five feet on average. The pipeline requires a permanent easement directly over the pipeline for maintenance and safety. In general, this maintenance easement is 40 feet wide. After construction of the pipeline was completed, vegetation above the pipeline was restored to its pre-construction condition. Exceptions to the restoration of vegetation were large trees and other plants with potentially damaging root structures, which are not allowed to grow directly above the pipeline. In addition, portions of the pipeline occupy roadway rights-of-way. During the planning and construction phase of the pipeline, the operator closely worked with environmental interest groups, homeowners, and other stakeholders to ensure their concerns were addressed. We consider it reasonable to assume that homebuyers knew about the pipeline during the different development stages investigated in this study. During the construction period, potential homebuyers would have been informed of the pipeline by local media reporting on some of the controversy over placement of the pipeline. In addition, buyer’s agents required to protect their client’s interests would be inclined to disclose the prospective pipeline. After completion of pipeline construction, homebuyers adjacent to the pipeline would have been informed about the underground pipeline by an easement identified in the deed. Using geographic information system (GIS) software and residential transaction data, we collect information about single-family properties within one mile of the pipeline in Washington and Clackamas counties in Oregon. Our rationale for a one-mile radius is as follows. In cases of major pipeline accidents such as the 2010 explosion in San Bruno, California, casualties and property damage predominantly occurred close to the epicenter of the explosion. Homeowners within a mile of pipelines are also commonly evacuated after pipeline accidents.

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 2 3

With our one-mile-radius approach, we follow previous studies on pipeline accidents, such as Hansen, Benson, and Hagen (2006), that suggest a one-mile radius is sufficient to capture the proximity effect while ensuring a relative homogeneity of the sample with regard to neighborhood quality. The relatively narrow definition of our geographical area of interest also allows controlling for locational factors to a larger extent (Tu and Eppli, 2001). Information about the sales price and characteristics of these properties for the period of 1992 to 2006 is obtained from the county assessor’s offices. Our initial sample size is 11,002 transactions. We eliminate 360 outliers and vacant residential land sales from the sample, resulting in a final sample size of 10,642 transactions. Exhibit 1 shows the distribution of single-family home transactions in our sample relative to the pipeline location. Previous studies in environmental economics and real estate measure perceived risk of an environmental hazard either directly through surveys or indirectly, proxied by the distance of a home to the environmental hazard (McCluskey and Rausser, 2001; Cameron, 2006). Cameron (2006) suggests an extension of the two-dimensional price-distance to a three-dimensional perspective accounting for the directional heterogeneity in distance effects. However, this methodology is suggested for environmental hazards causing noticeable pollution such as odors, which affect surrounding properties in a multi-dimensional way. As we focus on an underground natural gas pipeline without any sensory effects on residential properties, we employ the two-dimensional price-distance proxy for perceived risk. We measure distance to the pipeline with a variable (PIPEDIS) equal to the linear distance (in feet) of a single-family home to the pipeline. We assume for this variable that the closer a single-family home is to the pipeline, the higher is the environmental risk perceived by homebuyers. We employ a quasi-experimental design in which we compare the relationship of perceived risk (pipeline proximity) and sales prices at three points in time: (1) before the pipeline was intended to be built (January 1992 to August 1999), (2) after the notice of intent to build the pipeline was submitted to the end of construction (September 1999 to August 2004), and (3) after the pipeline was brought into operation (September 2004 to December 2006). Consequently, we include three variables in our hedonic pricing model capturing (1) the period before the intent was submitted (PIPEDIS), (2) after the intent was submitted and construction began (PIPEDISCONS), and (3) after the pipeline started operating (PIPEDISOPER). We expect the relationship of perceived risk and sales prices to be linear, in line with McCluskey and Rausser (2001). To test for the moderating effect of media coverage of unrelated natural gas pipelines on the relationship of perceived risk and sales prices, we obtain information about gas pipeline explosions with fatalities for all U.S. states from the U.S. Department of Transportation (DOT). According to the DOT dataset, these pipeline accidents represent leaks, ruptures or mechanical punctures leading to explosions (ignitions of natural gas) due to a variety of causes such as J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 2 4

u

F r e y b o t e

a n d

F r u i t s

E x h i b i t 1 u Distribution of Residential Transactions around SMPE Pipeline (Washington County and Clackamas County, Oregon)

equipment or material failure, incorrect operation or external damage to the pipelines through vehicles or farming equipment. We focus on pipeline explosions with fatalities as opposed to accidents with injuries or property damage only as the media are more likely to report fatal

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 2 5

pipeline explosions. In addition, fatal pipeline explosions are likely to have a stronger impact on the ease with which homebuyers can bring gas pipeline explosions to mind and thus increase the salience of this particular environmental risk in a homebuyer’s mind (availability heuristic). Over the period covered in this study, only 20 fatal explosions (with a total of 33 fatalities) occurred in gas transmission, on which we focus with the SMPE transmission pipeline, while 169 fatal explosions (186 fatalities) occurred in connection with gas distribution. Thus, compared to fatal gas distribution explosions, the actual frequency of fatal gas transmission explosions is relatively low. Our empirical analysis of the impact of media on perceived risk is based on the assumption that fatalities from gas pipeline explosions increase the awareness and interest level of the general public. This interest manifests itself in increased web and news searches. To assess the robustness of this assumption, we conduct an additional test. For the period of January 2007 to May 2013, we collect monthly interest level data for Google web (WEB) and news (NEWS) searches for the term ‘‘pipeline explosion’’ and determine the quarterly average. In our robustness test, we use Google data instead of a news archive such as LexisNexis for the following reasons. Firstly, we are not able to review and record media reports (print and online) of all nationwide pipeline explosions over the period we cover. Secondly, Weaver and Bimber (2008) show the limitations of traditional archives such as LexisNexis compared to Google data. In particular, their findings suggest that LexisNexis insufficiently captures news stories and is inferior to Google in terms of finding news stories. With regard to our study, LexisNexis is furthermore insufficient as only major newspapers such as the Oregonian are indexed by LexisNexis, but widely-read local papers such as the Portland Tribune and Willamette Week are not. In addition to the Google data, we obtain the quarterly number of fatalities caused by gas pipeline (distribution and transmission) explosions from the DOT. We combine fatal gas distribution and transmission explosions in our analysis, as homebuyers may not be able to distinguish between these two types of gas pipeline related explosions due to their limited familiarity with the natural gas pipeline system. To account for the effect of delayed DOT accident reports on media coverage and general public interest, we also include the first and second lag of fatalities in our analysis. We subsequently regress the WEB and NEWS interest level on the number of fatalities in the same quarter and the previous two quarters. The results of our regression are shown in Exhibit 2 and support our assumption about the relationship of gas pipeline explosion fatalities, media coverage, and general public interest. While pipeline explosion fatalities significantly increase the interest level in the web searches of pipeline explosions, they have an even stronger effect on the interest of Google users in news about pipeline explosions. To assess the moderating effect of media coverage on the relationship of perceived risk and sales prices, we introduce two interaction terms between PIPEDIS and MEDIA for the construction and operation period (PIPEDIS x MEDIACONS and J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 2 6

u

F r e y b o t e

a n d

F r u i t s

E x h i b i t 2 u Gas Pipeline Explosion Fatalities, News, and General Public Interest Levels

Web Interest (WEB)

News Interest (NEWS)

Fatalitiest

3.067***

6.939***

Fatalitiest21

0.507

2.818*

Fatalitiest22

0.460

5.241***

Adj. R2

15.68%

32.86%

N

77

65

Notes: This table presents the assumption check for the relationship of gas pipeline explosion fatalities, news coverage, and general public interest for the period of January 2007 to May 2013. Web Interest represents the interest level for the term ‘‘gas explosion,’’ based on web searches, derived from Google. News Interest represents the interest level for the term ‘‘gas explosion,’’ based on news searches, derived from Google. Fatalitiest is the number of gas pipeline explosion fatalities per quarter derived from the Department of Transportation. Fatalitiest21 and Fatalitiest22 are the respective lags. The augmented Dickey-Fuller test was used to assess stationarity for each of the data series and no unit root was found. * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.

PIPEDIS x MEDIAOPER). MEDIA is a binary variable coded 1 if a natural gas pipeline explosion resulting in fatalities happened in the same month as a sales transaction. The MEDIA variable is coded based on nationwide DOT records. Our rationale for this approach is as follows: Firstly, a search of news stories about gas pipeline explosions in the leading Oregon newspaper, The Oregonian, and its online presence, oregonlive.com, revealed media coverage of explosions in California, Texas, and West Virginia. This indicates that local newspapers cover gas pipeline explosions nationwide. Additionally, Oregon experienced no fatal gas pipeline explosion during the span of our analysis. Secondly, for our quasiexperiment it is only important that homebuyers were exposed to media reports of fatal gas pipeline explosions to create salience of the associated images and information, and increase perceived risk. The location of explosions and number of fatalities are not expected to affect our results. With regard to our previous analysis of pipeline accident fatalities, general public interest, and media/web coverage, we are confident that the MEDIA variable measures media coverage sufficiently. In our hedonic pricing model, we exclude the main effect for MEDIA, as our interest is focused on the effect of MEDIA on the slope of the PIPEDIS and sales prices relationship (PIPEDIS x MEDIACONS and PIPEDIS x MEDIAOPER). Additionally, fatal pipeline explosions covered by the media are only expected to have an impact on the sales prices of single-family homes in proximity to a pipeline, but not on residential sales prices in general. Our approach is in line with previous studies that include interaction terms, but

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 2 7

exclude one of the main effects in their analysis (Hartzell, Sun, and Titman, 2006; Xu, Han, and Yang, 2012). In line with the existing hedonic pricing model literature, we use the natural log of sales price as a dependent variable (Carroll, Clauretie, Jensen, and Waddoups, 1996; Aydin and Smith, 2008; Aroul and Hansz, 2012; Turnbull, ZahirovicHerbert, and Mothorpe, 2013). In our analysis, we control for property-specific attributes, such as age, property size, living room size, bedrooms, and bathrooms, slope above 25%, as well as for neighborhood-specific attributes, such as distance to urban growth boundary border, river, and floodplain. We additionally control for the higher order effects of variables used in the model, such as quadratic terms and interactions. We also include interactions of physical features, as they are likely to have an impact on home values (Meese and Wallace, 1997; Hansen, 2006). Portions of the pipeline occupy roadway rights-of-way. In addition to the pipeline, the proximity to arterial roads may negatively affect residential real estate values. To control for the effect of arterial roads on home prices and isolate the effect of pipeline distance on sales prices, we include a variable measuring the distance from arterial roads, as well as an interaction effect of the distance to the pipeline and arterial roads. Furthermore, we include the natural log of employed individuals in the county as a measure of general economic conditions that affect house prices (Clapp and Giaccotto, 1994). We control for the locational characteristics of each of the 17 neighborhoods covered by including binary variables for individual neighborhoods (Sirmans, Macpherson, and Zietz, 2005). To capture changes in property market and economic conditions, we include a trend variable in line with previous hedonic pricing studies (Do and Grudnitski, 1997; Sirmans, MacDonald, Macpherson, and Zietz, 2006; Clauretie and Daneshvary, 2009). An overview of our variables is presented in Exhibit 3. We do not have detailed property market data at the county level for the period covered, which limits our ability to control for housing market conditions. However, both counties (Clackamas and Washington) are part of the Portland, Oregon MSA, and the Case-Shiller index for Portland shows a continuous increase in home prices over the period of 1992 to 2006. A second limitation of our study is that we only focus on the characteristics of single-family houses in our hedonic pricing model and do not account for the effects of bargaining power between buyer and seller (Harding, Rosenthal, and Sirmans, 2003) due to the unavailability of data about buyer and seller characteristics in our sample. Exhibit 4 presents the descriptive statistics. As shown in Panel A, single-family homes in our sample sold on average for $203,693, have an average age of nine years, an average property size of 15,496 square feet, and living area of 1,203 square feet. The average distance to the natural gas pipeline is 3,484 feet. In Panel B, we provide the mean sales prices for quartiles of pipeline distance for the three time periods of interest. For all three periods, single-family homes closest to the pipeline (first quartile) have the highest sales prices, followed by the second and J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 2 8

u

F r e y b o t e

a n d

F r u i t s

E x h i b i t 3 u Variable Descriptions

Variable

Description

Ln(PRICE)

Natural log of sales price; dependent variable.

TREND

Measures the passage of time, in years, with 1992 5 0.

AGE

Age of the house at the time of sale; equal to the year of sale minus the year the house was built.

Property Size

Land area of the property, in square feet.

Living Area

Living area of the house on the property, in square feet.

Property Size 3 Living Area

Interaction between land area and living area.

Bedrooms

Number of bedrooms.

Full Baths

Number of full bathrooms.

Bedroom 3 Living Area

Interaction between living area and the number of bedrooms.

Full Baths 3 Living Area

Interaction between living area and the number of full bathrooms.

Bedrooms 3 Full Baths

Interaction between the number of bedrooms and the number of full bathrooms.

SLOPE

Indicator variable equal to 1 if a substantial portion of the property has a slope greater than 25% and zero otherwise.

UGBDIST

Censored variable equal to the linear distance in feet to the urban growth boundary border; equal to zero if the property is within the urban growth boundary.

UGBDIST 2

Squared value of UGBDIST.

ARTDIST

Censored variable equal to the linear distance in feet to nearest arterial road border; equal to zero if the property is adjacent to arterial road.

ARTDIST 3 PIPEDIS

Interaction effect of distance to arterial road and pipeline.

FLOOD

Indicator variable equal to 1 if a substantial portion of the property is in a floodplain.

RIVERDIST

Linear distance from center of property to nearest river, in feet.

FLOOD 3 RIVERDIST

Interaction of proximity to river and whether a substantial portion of the property is in a floodplain.

PIPEDIS

Linear distance in feet to the future pipeline location before pipeline construction intent was submitted.

PIPEDISCONS

Linear distance in feet to the pipeline / future pipeline location for the period after the pipeline construction intent was submitted and before the pipeline went into service (pipeline construction period).

PIPEDISOPER

Linear distance in feet to the pipeline for the period after the pipeline started operations (operation period).

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 2 9

E x h i b i t 3 u (continued) Variable Descriptions

Variable

Description

PIPEDIS 3 MEDIACONS

Interaction effect of linear distance to pipeline and MEDIA, a binary variable coded 1 if a natural gas pipeline explosion with fatalities occurred in the same month as a transaction, 0 otherwise; in the pipeline construction period.

PIPEDIS 3 MEDIAOPER

Interaction effect of linear distance to pipeline and MEDIA, a binary variable coded 1 if a natural gas pipeline explosion with fatalities occurred in the same month as a transaction, 0 otherwise; in the pipeline operation period.

NHOODxx

Indicator variables for 16 of the 17 neighborhoods; neighborhood 15 is the reference group.

ln(EMPL)

Natural log of county employment.

third quartiles. One explanation is that, in our dataset, properties in the first quartile of pipeline distance have larger living areas and property sizes than properties in the other quartiles. We employ a hedonic pricing model following Rosen (1974) and subsequent studies (e.g., Carroll, Clauretie, Jensen, and Waddoups, 1996; Aydin and Smith, 2008; Aroul and Hansz, 2012; Turnbull, Zahirovic-Herbert, and Mothorpe, 2013). This methodology assumes that the value of a property can be decomposed into the values of its constituent characteristics (e.g., physical, locational) and allows us to estimate the marginal effect of the perceived risk on sales prices. Our models for perceived risk and the impact of media coverage are shown in equations (1) and (2), respectively. In line with previous studies (e.g., Aroul and Hansz, 2012), we employ ordinary least squares (OLS) regression. While the log transformation of sales price reduces heteroscedasticity to some extent, we also employ White heteroscedasticity-robust standard errors and covariance to address heteroscedasticity.

ln(SalesPrice) 5 b0 1 o bRisk Xi 1 o bYi 1 «,

(1)

where Xi represents perceived risk (PIPEDIS, PIPEDISCONS, and PIPEDISOPER). Yi represents a set of trend, property- and neighborhood-specific control variables.

J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 3 0

u

F r e y b o t e

a n d

F r u i t s

E x h i b i t 4 u Descriptive Statistics

Variable

Mean

Std. Dev.

Max.

Min.

203,692.60

122,302.40

2,080,000

20,000

9.26

18.46

111

0

15,495.89

85,984.21

3,479,278

1,203.20

527.86

7,232

448

Bedrooms

3.28

0.70

8

1

Bathrooms

2.61

0.75

11

0

Panel A: Full sample Sales price Age Property size Living area size

300.61

Distance from arterial road

742.21

589.89

7,795.00

Distance from UGB

840.47

4,488.47

51,849.82

Distance from river

8,251.60

4,882.52

21,945.71

35.23

Distance from pipeline

3,483.60

1,123.93

5,277.63

3.66

Slope of greater than 25%

0.02

0.12

1

0

Floodplain

0.02

0.14

1

0

54,775.12

1,109,206

815,065

Employment

1,001,928

52.19 0

Panel B: Mean sales prices separated by quartiles of pipeline distance Pre-Pipeline Period

Construction Period

Operation Period

First Quartile

161,099.80

231,959.80

372,747.20

Second Quartile

134,465.40

221,989.30

293,634.70

Third Quartile

123,461.60

207,568.80

300,700.80

Fourth Quartile

113,325.00

185,485.90

273,638.10

Notes: Panel A presents the descriptive statistics for sales price (in dollars), age at time of sale (in years), property size (in square feet), living area (in square feet), number of bedrooms and full bathrooms, employment (in thousand), slope of greater than 25% (binary variable), floodplain (binary variable) as well as distance from arterial road, UGB, river, and pipeline in feet for residential houses sold in Clackamas County and Washington County, Oregon from 1992 to 2006. Mean sales prices per quartiles of pipeline distance are presented in Panel B.

ln(SalesPrice) 5 b0 1 o bRisk Xi 1 o bMediaYi 1 o bZi «,

(2)

where Xi represents perceived risk (PIPEDIS, PIPEDISCONS, and PIPEDISOPER). Yi represents the interaction terms of perceived risk and media coverage (PIPEDIS 3 MEDIACONS and PIPEDIS 3 MEDIAOPER) and Zi represents a set of trend, property- and neighborhood-specific control variables.

P e r c e i v e d

u

E n v i r o n m e n t a l

R i s k

u

2 3 1

Results Impact of Perceived Risk

The results of our model based on equation (1) are presented in Exhibit 5. The coefficients of our PIPEDIS variables measuring pipeline distance in different pipeline development stages are significant at the 1% level and differ in size and direction. The coefficient on PIPEDIS is negative. Prior to the announced intention to construct the pipeline (PIPEDIS), the closer a single-family home was to the future pipeline location, the higher was the sales price. This effect may be due to pre-existing characteristics of the future pipeline location that are not further investigated in this study. The relationship of distance to pipeline and sales price holds in the construction period, as indicated by the negative coefficient on PIPEDISCONS. This finding suggests that homebuyers did not perceive the future pipeline as a potential environmental hazard that diminished home values. However, the negative coefficients on PIPEDIS and PIPEDISCONS differ in size, with the latter being smaller. This difference may be due to factors such as negative externalities from the pipeline construction. However, in the operation period, the relationship of distance to pipeline and sales prices changes. The coefficient on PIPEDISOPER is positive and significant at the 1% level. The larger the distance between a single-family home and the environmental hazard (i.e., the lower the perceived risk), the higher was the sales price. This finding is in line with previous studies, which find a negative effect of the proximity to an existing environmental hazard on single-family house prices (e.g., Carroll, Clauretie, Jensen, and Waddoups, 1996; McCluskey and Rausser, 2001, 2003; Boxall, Chan, and McMillan, 2005; Simons and Saginor, 2006; Hoen et al., 2011). While these previous studies investigate environmental hazards affecting homes through odor, noise or unappealing visuals, our findings suggest that the mere presence of an accident-free underground hazard, which solely affects houses through an easement and is relatively safe based on past DOT data, reduces sales prices. Our findings contradict Hansen, Benson, and Hagen (2006), who find no impact of (1) a gasoline pipeline prior to its rupture and (2) an accident-free crude oil pipeline on home values. Impact of Perceived Risk and Media Coverage

The results of our model based on equation (2) are presented in Exhibit 6. The coefficients of PIPEDIS, PIPEDISCONS, and PIPEDISOPER are significant at the 1% level and in line with our findings in Exhibit 5. In the pre-intent (PIPEDIS) and construction period (PIPEDISCONS), proximity to the (future) pipeline location increased sales prices, albeit to a lower extent in the construction period, J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 3 2

u

F r e y b o t e

a n d

F r u i t s

E x h i b i t 5 u Hedonic Pricing Estimation Results for Perceived Risk

Coeff.

T-Value

Perceived Risk PIPEDIS PIPEDISCONS PIPEDISOPER

20.055 20.015 0.008

27.56*** 23.10*** 2.86***

Physical Characteristics Age Property size Living area Property size 3 Living area Bedrooms Full baths Bedrooms 3 Living area Full baths 3 Living area Bedrooms 3 Full baths

0.002 0.005 0.344 20.001 0.191 0.119 20.058 0.004 20.015

0.54 1.95** 7.36*** 20.76 8.53*** 5.08*** 25.95*** 0.23 22.45***

Other Control Variables TREND Ln(EMPL) ARTDIST ARTDIST 3 PIPEDIS FLOOD RIVERDIST FLOOD 3 RIVERDIST SLOPE UGBDIST UGBDIST 2 NHOOD01 NHOOD02 NHOOD03 NHOOD04 NHOOD05 NHOOD06 NHOOD07 NHOOD08 NHOOD09 NHOOD10 NHOOD11 NHOOD12 NHOOD13 NHOOD14 NHOOD16 NHOOD17 Constant

0.086 1.193 20.029 0.015 0.112 20.007 20.005 20.201 20.005 0.000 0.014 0.034 0.532 0.302 0.152 1.036 0.025 0.653 0.003 20.336 20.328 20.730 20.140 20.162 20.241 20.245 26.648

31.22*** 8.59*** 21.47 2.57*** 2.88*** 24.49*** 20.39 24.08*** 20.49 1.71* 0.05 0.16 7.13*** 4.94*** 6.10*** 4.41*** 0.77 8.19*** 0.02 22.41** 21.40 22.23** 28.52*** 22.83*** 21.97** 22.45*** 23.51***

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 3 3

E x h i b i t 5 u (continued) Hedonic Pricing Estimation Results for Perceived Risk

Notes: The table presents the results of our hedonic pricing model using the natural log of sales price as the dependent variable and OLS regression (White heteroscedasticity-robust standard errors) for our overall sample for Washington County and Clackamas County, Oregon. PIPEDIS, PIPEDISCONS and PIPEDISOPER are variables equal to the linear distance in feet to the pipeline for the pre-intent period, after pipeline construction began and after the pipeline went into service. PIPEDIS 3 MEDIACONS and PIPEDIS 3 MEDIAOPER are interaction effects between MEDIA, a binary variable coded 1 if an unrelated gas pipeline explosion happened in the same month as a residential home transaction, and PIPEDIS, for the respective periods. Age is the age of a singlefamily home at the time of sale. Property size is the land area of the property, in square feet. Living area is the livable area of the house on the property, in square feet. Bedrooms and Full baths are given in numbers. Property size 3 Living area, Bedrooms 3 Living area, Full baths 3 Living area, and Bedrooms 3 Full baths represent interaction effects of the respective variables. TREND is the trend component. Ln(EMPL) is the natural log of county employment. ARTDIST is the linear distance in feet to nearest arterial road border; equal to zero if the property is adjacent to arterial road. ARTDIST 3 PIPEDIS is the interaction effect of the two variables. FLOOD is a binary variable coded 1 if a substantial portion of the property is located on a flood plain. RIVERDIST is the linear distance from center of property to nearest river, in feet. FLOOD 3 RIVERDIST is the interaction effect of the two variables. SLOPE is equal to 1 if a substantial portion of the property has a slope greater than 25% and zero otherwise. UGBDIST is equal to the linear distance in feet to the urban growth boundary border; equal to zero if the property is within the urban growth boundary. UGBDIST2 is the respective quadratic term. NHOODx is a binary variable coded 1 for each of 17 neighborhoods covered by the study; neighborhood 15 is the reference group. Adj. R2 5 55.52%; the Sum of squared residuals 5 1,906.29; log-likelihood 5 25,950.09; the number of observations 5 10,642. * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.

while proximity to the pipeline (higher perceived risk) reduced prices once the pipeline was in service (PIPEDISOPER). As hypothesized, introducing media coverage of unrelated gas pipeline explosions with fatalities moderates the relationship of pipeline proximity and sales prices in the construction period. The coefficient on PIPEDIS 3 MEDIACONS is significant at the 5% level and positive. In months with unrelated fatal gas pipeline explosions, homebuyers penalized proximity to the pipeline with a discount in sales prices due to their higher perceived risk. The positive effect of PIPEDIS 3 MEDIACONS reduces the negative effect of PIPEDISCONS to some extent, leading to a reduced premium for properties in close proximity to the future pipeline location in months of fatal pipeline explosions compared to months without any explosion. The fact that homebuyers were willing to pay a higher price for homes farther away from the pipeline in months of pipeline explosions J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 3 4

u

F r e y b o t e

a n d

F r u i t s

E x h i b i t 6 u Hedonic Pricing Estimation Results for Perceived Risk and Media Coverage

Perceived Risk PIPEDIS PIPEDISCONS PIPEDISOPER Media Coverage PIPEDIS 3 MEDIACONS PIPEDIS 3 MEDIAOPER

Coeff.

T-Value

20.055 20.015 0.011

27.61*** 23.21*** 3.25***

0.008 0.000

2.43** 0.03

Physical Characteristics Age Property size Living area Property size 3 Living area Bedrooms Full baths Bedrooms 3 Living area Full baths 3 Living area Bedrooms 3 Full baths

0.002 0.004 0.343 20.001 0.191 0.119 20.058 0.004 20.015

0.56 1.93* 7.34*** 20.76 8.52*** 5.07*** 25.93*** 0.24 22.45**

Other Control Variables TREND Ln(EMPL) ARTDIST ARTDIST 3 PIPEDIS FLOOD RIVERDIST FLOOD 3 RIVERDIST SLOPE UGBDIST UGBDIST 2 NHOOD01 NHOOD02 NHOOD03 NHOOD04 NHOOD05 NHOOD06 NHOOD07 NHOOD08 NHOOD09 NHOOD10 NHOOD11 NHOOD12 NHOOD13 NHOOD14 NHOOD16 NHOOD17 Constant

0.085 1.239 20.029 0.015 0.112 20.007 20.005 20.201 20.005 0.000 0.014 0.036 0.533 0.302 0.151 1.039 0.022 0.655 0.008 20.331 20.318 20.720 20.141 20.161 20.238 20.244 27.256

30.10*** 8.74*** 21.46 2.56*** 2.87*** 24.55*** 20.39 24.07*** 20.51 1.71* 0.05 0.17 7.16*** 4.95*** 6.06*** 4.43*** 0.64 8.22*** 0.05 22.37** 21.36 22.20** 28.55*** 22.83*** 21.95* 22.44** 23.76***

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 3 5

E x h i b i t 6 u (continued) Hedonic Pricing Estimation Results for Perceived Risk and Media Coverage

Notes: The table presents the results of our hedonic pricing model using the natural log of sales price as the dependent variable and OLS regression (White heteroscedasticity-robust standard errors) for our overall sample for Washington County and Clackamas County, Oregon. PIPEDIS, PIPEDISCONS and PIPEDISOPER are variables equal to the linear distance in feet to the pipeline for the pre-intent period, after pipeline construction began and after the pipeline went into service. PIPEDIS 3 MEDIACONS and PIPEDIS 3 MEDIAOPER are interaction effects between MEDIA, a binary variable coded 1 if an unrelated gas pipeline explosion happened in the same month as a residential home transaction, and PIPEDIS, for the respective periods. Age is the age of a singlefamily home at the time of sale. Property size is the land area of the property, in square feet. Living area is the livable area of the house on the property, in square feet. Bedrooms and Full baths are given in numbers. Property size 3 Living area, Bedrooms 3 Living area, Full baths 3 Living area, and Bedrooms 3 Full baths represent interaction effects of the respective variables. TREND is the trend component. Ln(EMPL) is the natural log of county employment. ARTDIST is the linear distance in feet to nearest arterial road border; equal to zero if the property is adjacent to arterial road. ARTDIST 3 PIPEDIS is the interaction effect of the two variables. FLOOD is a binary variable coded 1 if a substantial portion of the property is located on a flood plain. RIVERDIST is the linear distance from center of property to nearest river, in feet. FLOOD 3 RIVERDIST is the interaction effect of the two variables. SLOPE is equal to 1 if a substantial portion of the property has a slope greater than 25% and zero otherwise. UGBDIST is equal to the linear distance in feet to the urban growth boundary border; equal to zero if the property is within the urban growth boundary. UGBDIST2 is the respective quadratic term. NHOODx is a binary variable coded 1 for each of 17 neighborhoods covered by the study; neighborhood 15 is the reference group. Adj. R2 5 55.52%; the sum of squared residuals 5 1,905.78; log-likelihood 5 25,948.66; the number of observations 5 10,642. * Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.

reflects the increase in perceived risk. The moderating effect of media on the relationship of perceived risk and sales prices in the construction period is visualized in Exhibit 7. Media coverage has no moderating effect on the relationship of perceived risk and sales prices in the pipeline operation period, as evidenced by the insignificant coefficient on PIPEDIS 3 MEDIAOPER. The respective main effect (PIPEDISOPER) is likely to entirely capture the effect of perceived risk (proximity to pipeline) on sales prices, which suggests that homebuyers already perceived the pipeline as an environmental risk and were not additionally influenced by media coverage in their risk judgment. As hypothesized, we find evidence that media coverage affects the risk judgment of homebuyers and consequently asset pricing. In months with fatal gas pipeline explosions covered in the media, homebuyers are likely to have a biased judgment J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 3 6

u

F r e y b o t e

a n d

F r u i t s

E x h i b i t 7 u Moderating Effect of Media on the Relationship of Perceived Risk and Sales Price

200000

195000

190000 Without media With media

185000

180000

175000 0

100

250

500

1000

2500

5280

Note: This figure shows the moderating effect of media coverage, measured as binary variable coded 1 for a month with an unrelated fatal gas pipeline explosion, on the relationship of perceived risk and sales price, during the pipeline construction period. The graph is based on a hypothetical single-family house with a sales price of $200,000. The x-axis represents distance from the pipeline in feet and the y-axis represents sales price.

of the probability of a gas pipeline explosion due to the ease with which gas pipeline explosions come to mind, in line with the availability heuristic (Tversky and Kahneman, 1974). Thus, our findings provide additional support that the availability heuristic affects perceived risk and is affected by media (Morton and Duck, 2001; Slater and Rasinski, 2006; Stein, Duen¸as-Osorio and Subramanian, 2010; Pachur, Hertwig, and Steinmann, 2012). To assess the robustness of our findings with regard to seasonal influences, we include binary variables for each month of the year, except January, in our full model, as shown in equation (2). The results of this regression (not reported) are consistent with our results in Exhibit 6, in terms of directionality, significance, and effect sizes, and suggest that our results are robust to seasonality. Compared to the majority of previous hedonic pricing studies reporting an adjusted R2 of 70% to 80% (e.g., Carroll, Clauretie, Jensen, and Waddoups, 1996; Hansen, Benson, and Hagen, 2006; McKenzie and Levendis, 2010; Hoen et al., 2011; Noonan and Krupka, 2011), our adjusted R2 is smaller. However, it is relatively similar to the adjusted R2 of Boxall, Chan, and McMillan (2005), who investigate the impact of oil and natural gas facilities on residential real estate prices.

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 3 7

One explanation for our relatively low adjusted R2 may be the omission of other important variables. Sirmans, Macpherson, and Zietz (2005) present an overview of the variables most frequently included in hedonic pricing models. A review of this list indicates that our model as shown in equations (1) and (2) includes the most important physical and locational control variables, based on previous studies, except time on market (TOM). The omission of this variable from our model results from the lack of TOM data in our dataset derived from assessors’ offices. As discussed by Sirmans, Macpherson, and Zietz (2005), previous studies find that listing price, brokers, listings, market conditions, and the atypical physical characteristics of homes are correlated with TOM. While TOM is expected to be correlated with our dependent variable, it is not expected to be significantly correlated with the independent variables in our model. Consequently, we are confident that an omitted variable bias does not threaten our results. Other excluded variables that could be used to explain additional variation in sales prices are financing and other marketing issues. Analogously to TOM, these variables were excluded due to lack of data. They are not expected to bias results, as correlations with included predictors are likely to be small. Pricing of Environmental Hazard

While the hedonic pricing model allows us to estimate the marginal effect of perceived risk and media coverage on sales prices, it fails to provide a more detailed analysis of the environmental hazard pricing by homebuyers. As the price differential between homes closer to the pipeline and farther away indicates the pricing of environmental hazard by homebuyers (McCluskey and Rausser, 2001), we conduct a simulation of the impact of perceived risk and media coverage on sales price for the construction and operation period. The starting point of our simulation is a hypothetical property located on or adjacent to the pipeline. Based on our regression results as shown in Exhibit 6, we then calculate the effect of perceived risk based on PIPEDISCONS and PIPEDISOPER, as well as the effect of media coverage based on PIPEDIS 3 MEDIACONS and PIPEDIS 3 MEDIAOPER on a randomly chosen home value ($350,000) for varying distances to the pipeline. To get the sales price discount or premium, we determine the price difference between a house of a particular distance to the pipeline (e.g., half a mile) and the house located on or adjacent to the pipeline ($350,000), and divide it by $350,000. The results of our simulation are shown in Exhibit 8. During the construction period, homebuyers priced a distance of 500 feet, 1,000 feet, and a mile from the pipeline with a discount of 0.7%, 1.4%, and 7.3%, respectively, for months without pipeline explosions. However, for months with unrelated fatal pipeline explosions, single-family homes within a distance of 500 feet, 1,000 feet, and 1 mile to the pipeline sold at a discount of 0.3%, 0.6%, and 3.2%. These reduced discounts for months with pipeline explosions reflect the moderating effect of media coverage. J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 3 8

u

F r e y b o t e

a n d

F r u i t s

E x h i b i t 8 u Simulation of Hazard Pricing

Distance from Pipeline

Pipeline Construction

Pipeline Operation

Sales Price

Percent Difference

Sales Price

Percent Difference

Panel A: Without media coverage 100

349,500

20.1%

350,379

0.1%

250

348,751

20.4%

350,946

0.3%

500

347,506

20.7%

351,887

0.5%

1,000

345,031

21.4%

353,757

1.1%

2,500

337,709

23.5%

359,268

2.6%

5,280

324,547

27.3%

369,099

5.5%

Panel B: With media coverage 100

349,785

20.1%

350,385

0.1%

250

349,462

20.2%

350,963

0.3%

500

348,924

20.3%

351,927

0.6%

1,000

347,851

20.6%

353,852

1.1%

2,500

344,653

21.5%

359,621

2.7%

5,280

338,803

23.2%

370,285

5.8%

Note: This table presents the results for a simulation of the impact of perceived risk and media coverage on residential sales prices, based on the respective regression coefficients in Exhibit 6. Starting with a sales price of $350,000 of a hypothetical property on or adjacent to the pipeline, the sales price and percentage change in sales prices with increasing distance from the pipeline are reported for months with and without media coverage of fatal pipeline explosions in the construction and operation period. 1 mile 5 5,280 feet.

The increase in perceived risk from media coverage increased the sales prices of homes farther away from the pipeline. For example, during the construction period, a home in the 1,000 feet distance to the pipeline sold for $345,031 in months without pipeline explosions and for $347,851 in months with fatal pipeline explosions. The difference of $2,820 reflects the impact of perceived risk increased by media coverage on sales prices. Once the pipeline is in operation, homebuyers were willing to pay higher sales prices as the distance to the environmental hazard increased (i.e., perceived risk decreased). In months without fatal pipeline explosions, single-family houses with a distance of 500 feet, 1,000, feet and 1 mile to the environmental hazard sold at a premium of 0.5%, 1.1%, and 5.5%, respectively. In months with fatal pipeline explosions, single-family homes with these distances sold at a premium of 0.6%, 1.1%, and 5.8%, respectively.

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 3 9

While media coverage of fatal pipeline explosions had no significant effect on sales prices in the operation period (Exhibit 6), the simulation of hazard pricing suggests that in months of fatal pipeline explosions, distance to the hazard was priced slightly higher. For example, a home in a 500 feet distance from the pipeline sold for $351,887 in months without media coverage and for $351,927 in months with media coverage. The slight difference likely results from media coverage and its effect on perceived risk.

u

Conclusion

We investigate whether risk perceived from a non-sensory environmental hazard, in the form of an accident-free underground gas transmission pipeline with a relatively low actual risk of explosion, affects the sales prices of single-family homes in close proximity to it. In our analysis, we account for the different development stages of the pipeline and the moderating effect of media coverage of unrelated pipeline explosions on this relationship. Our empirical findings for the operation period are in line with previous findings for sensory permanent or transient environmental hazards (e.g., Carroll, Clauretie, Jensen, and Waddoups, 1996; McCluskey and Rausser, 2001, 2003; Boxall, Chan, and McMillan, 2005; Simons and Saginor, 2006; Hoen et al., 2011) and suggest that the actual risk or sensory impact of a hazard do not reduce the environmental risk perception of homebuyers or eliminate its negative impact on pricing decisions. Laypersons, such as homebuyers, who evaluate environmental and health risks primarily based on emotions rather than sophisticated risk analysis, even penalize single-family homes in proximity to a low risk, accident-free and non-sensory disamenity. Our results for the construction period are particularly interesting. In months without unrelated fatal pipeline explosions, the relationship of proximity to future pipeline location and sales prices is the same as in the pre-pipeline period. However, in months of fatal accidents, homebuyers suddenly required a discount for properties in close proximity to the pipeline (i.e., homes with higher perceived risk). This finding indicates that media coverage of unrelated fatal explosions increased homebuyer awareness and risk perception, which is in line with previous studies (Flynn, Peters, Mertz, and Slovic, 1998; McCluskey and Rausser, 2001, 2003; Bauer, Braun and Kvasnicka, 2013). Our findings show that technological stigma and risk perception are not only affected by accidents or news coverage of a particular hazard, but also by media coverage of unrelated events, which complements the findings of Bauer, Braun, and Kvasnicka (2013). The impact of media coverage on perceived risk appears to be particularly high during periods of increased uncertainty and limited understanding of potential risks by homebuyers, such as during the construction period. Our findings have implications for homebuyers, agents, operators of environmental hazards, and advocacy groups. Pipeline operators, for example, may have to J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 4 0

u

F r e y b o t e

a n d

F r u i t s

provide additional data and information about the safety of transmission pipelines, potentially as early and frequently as possible, in order to reduce perceived risk and the potential effects of negative media reports. Homeowners and agents may have to be sensitive to media coverage of events directly or indirectly affecting a particular property or neighborhood when selling a single-family home. Homeowners also have to be aware that even a low-risk non-sensory disamenity may permanently reduce their property values. We understand our study as a starting point for further investigations into risk perception, media and residential real estate pricing. Future studies may investigate the persistence and duration of media effects on perceived risk and asset pricing. Other studies may build upon Maderthaner, Guttmann, Swaton, and Otway (1978) and investigate whether the perceived risk decreases with constant exposure to the pipeline.

u

References Aroul, R.R. and J.A. Hansz. The Value of ‘‘Green’’: Evidence from the First Mandatory Residential Green Building Program. Journal of Real Estate Research, 2012, 34:1, 27–49. Aydin, R. and B.A Smith. Evidence of the Dual Nature of Property Value Recovery Following Environmental Remediation. Real Estate Economics, 2006, 36:4, 777–812. Bauer, T.K., S. Braun, and M. Kvasnicka. Distant Event, Local Effects? Fukushima and the German Housing Market. Ruhr Economic Papers, 2013, Number 433. Beracha, E. and M.B. Wintoki. Forecasting Residential Real Estate Price Changes from Online Search Activity. Journal of Real Estate Research, 2013, 35:3, 283–312. Boxall, P.C., W.H. Chan, and M.L. McMillan. The Impact of Oil and Natural Gas Facilities on Rural Residential Property Values: A Spatial Hedonic Analysis. Resource and Energy Economics, 2005, 27, 248–69. Boyle, M.A. and K.A. Kiel. A Survey of House Price Hedonic Studies of the Impact of Environmental Externalities. Journal of Real Estate Literature, 2001, 9:2, 117–44. Cameron, T.A. Directional Heterogeneity in Distance Profiles in Hedonic Property Value Models. Journal of Environmental Economics and Management, 2006, 51, 26–45. Carroll, T.M., T.M. Clauretie, J. Jensen, and M. Waddoups. The Economic Impact of a Transient Hazard on Property Values: The 1988 PEPCON Explosion in Henderson, Nevada. Journal of Real Estate Finance and Economics, 1996, 13, 143–67. Case, B., P.F. Colwell, C. Leishman, and C. Watkins. The Impact of Environmental Contamination on Condo Prices: A Hybrid Repeat-Sale/Hedonic Approach. Real Estate Economics, 2006, 34:1, 77–107. Choi, J., A. Lee, and C. Ok. The Effects of Consumers’ Perceived Risk and Benefit on Attitude and Behavioral Intention: A Study of Street Food. Journal of Travel & Tourism Marketing, 2013, 30:3, 222–37. Clapp, J.M. and C. Giaccotto. The Influence of Economic Variables on Local House Price Dynamics. Journal of Urban Economics, 1994, 36:2, 161–83. Clauretie, T.M. and N. Daneshvary. Estimating the House Foreclosure Discount Corrected for Spatial Price Interdependence and Endogeneity of Marketing Time. Real Estate Economics, 2009, 37:1, 43–67.

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 4 1

Do, A.Q. and G. Grudnitski. The Impact on Housing Values of Restrictions on Rights of Ownership: The Case of an Occupant’s Age. Real Estate Economics, 1997, 25:4, 683–93. Dotzour, M. Groundwater Contamination and Residential Property Values. Appraisal Journal, 1997, 65:3, 279–86. Flower, P.C. and W.R. Ragas. The Effects of Refineries on Neighborhood Property Values. Journal of Real Estate Research, 1994, 9:3, 319–38. Flynn, J., E. Peters, C.K. Mertz, and P. Slovic. Risk, Media, and Stigma at Rocky Flats. Risk Analysis, 1998, 18:6, 715–27. Gordon, B.L., D. Winkler, J.D. Barrett, and L. Zumpano. The Effect of Elevation and Corner Location on Oceanfront Condominium Value. Journal of Real Estate Research, 2013, 35:3, 345–63. Hansen, J.L., E.D. Benson, and D.A. Hagen. Environmental Hazards and Residential Property Values: Evidence from a Major Pipeline Event. Land Economics, 2006, 82:4, 529– 41. Hansen, J. Australian House Prices: A Comparison of Hedonic and Repeat-Sales Measures. Reserve Bank of Australia Economic Research Department, Research Discussion Paper, 2006-03. Hansz, J.A. and D.K. Hayunga. Club Good Influence on Residential Transaction Prices. Journal of Real Estate Research, 2012, 34:4, 549–75. Harding, J.P., S.S. Rosenthal, and C.F. Sirmans. Estimating Bargaining Power in the Market for Existing Homes. The Review of Economics and Statistics, 2003, 85:1, 178–88. Hartzell, J.C., L. Sun, and S. Titman. The Effect of Corporate Governance on Investment: Evidence from Real Estate Investment Trusts. Real Estate Economics, 2006, 34:3, 343–76. Henthorne, T.L., B.P. George, and W.C. Smith. Risk Perception and Buying Behavior: An Examination of Some Relationships in the Context of Cruise Tourism in Jamaica. International Journal of Hospitality & Tourism Administration, 2013, 14:1, 66–86. Hite, D., W. Chern, F. Hitzhusen, and A. Randall. Property-Value Impacts of an Environmental Disamenity: The Case of Landfills. Journal of Real Estate Finance and Economics, 2001, 22:2/3, 185–202. Hoen, B., R. Wiser, P. Cappers, M. Thayer, and G. Sethi. Wind Energy Facilities and Residential Properties: The Effect of Proximity and View on Sales Prices. Journal of Real Estate Research, 2011, 33:3, 279–316. Hohenstatt, R., M. Ka¨sbauer, and W. Scha¨fers. ‘‘Geco’’ and its Potential for Real Estate Research: Evidence from the U.S. Housing Market. Journal of Real Estate Research, 2011, 33:4, 471–506. Jakus, P.M. and W. D. Shaw. Perceived Hazard and Product Choice: An Application to Recreational Site Choice. Journal of Risk and Uncertainty, 2003, 26:1, 77–92. Keh, H.T. and J. Pang. Customer Reactions to Service Separation. Journal of Marketing, 2010, 74, 55–70. Laroche, M., Z. Yang, G.H.G. McDougall, and J. Bergeron. Internet versus Bricks-andMortar Retailers: An Investigation into Intangibility and Its Consequences. Journal of Retailing, 2005, 81:4, 251–67. Liao, C.-H. and I.-Y. Hsieh. Determinants of Consumer’s Willingness to Purchase GrayMarket Smartphones. Journal of Business Ethics, 2013, 114, 409–24. Maderthaner, R., G. Guttmann, E. Swaton, and H.J. Otway. Effect of Distance upon Risk Perception. Journal of Applied Psychology, 1978, 63:3, 380–82. J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

2 4 2

u

F r e y b o t e

a n d

F r u i t s

McCluskey, J.J. and G.C. Rausser. Estimation of Perceived Risk and Its Effect on Property Values. Land Economics, 2001, 77:1, 42–55. ——. Hazardous Waste Sites and Housing Appreciation Rates. Journal of Environmental Economics and Management, 2003, 45, 166–76. McKenzie, R. and J. Levendis. Flood Hazards and Urban Housing Markets: The Effects of Katrina on New Orleans. Journal of Real Estate Finance and Economics, 2010, 40, 62– 76. Meese, R.A. and N.E. Wallace. The Construction of Residential Housing Price Indices: A Comparison of Repeat-Sales, Hedonic-Regression, and Hybrid Approaches. Journal of Real Estate Finance and Economics, 1997, 14:1/2, 51–73. Morton, T.A. and J.M. Duck. Communication and Health Beliefs—Mass and Interpersonal Influences on Perception of Risk to Self and Others. Communication Research, 2001, 28: 5, 602–26. Noonan, D.S. and D.J. Krupka. Making- or Picking- Winners: Evidence of Internal and External Price Effects in Historic Preservation Policies. Real Estate Economics, 2011, 39: 2, 379–407. Pachur, T., R. Hertwig, and F. Steinmann. How Do People Judge Risks: Availability Heuristic, Affect Heuristic, or Both? Journal of Experimental Psychology: Applied, 2012, 18:3, 314–30. Page, G.W. and H. Rabinowitz. Groundwater Contamination: Its Effects on Property Values and Cities. Journal of the American Planning Association, 1993, 59:4, 473–81. Reichert, A.K., M. Small, and S. Mohanty. The Impact of Landfills on Residential Property Values. Journal of Real Estate Research, 1992, 7:3, 297–314. Richardson, H. and L. Zumpano. Further Assessment of the Efficiency Effects of Internet Use in Home Search. Journal of Real Estate Research, 2012, 34:4, 515–48. Rosen, S. Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition. Journal of Political Economy, 1974, 82:1, 34–55. Rosiers, F.D. Power Lines, Visual Encumbrance and House Values: A Microspatial Approach to Impact Measurement. Journal of Real Estate Research, 2002, 23:3, 275–301. Seiler, M.J. and E. Walden. Lender Characteristics and the Neurological Reasons for Strategic Mortgage Default. Journal of Real Estate Research, 2014, 36:3, 341–62. Seiler, M.J., P. Madhavan, and M. Liechty. Toward an Understanding of Real Estate Homebuyer Internet Search Behavior: An Application of Ocular Tracking Technology. Journal of Real Estate Research, 2012, 34:2, 211–41. Simons, R.A. The Effect of Pipeline Ruptures on Noncontaminated Residential EasementHolding Property in Fairfax County. Appraisal Journal, 1999, 67:3, 255–63. Simons, R.A. and J.D. Saginor. A Meta-Analysis of the Effect of Environmental Contamination and Positive Amenities on Residential Real Estate Values. Journal of Real Estate Research, 2006, 28:1, 71–104. Simons, R.A. and K. Winson-Geideman. Determining Market Perception on Contamination of Residential Property Buyers Using Contingent Valuation Surveys. Journal of Real Estate Research, 2005, 27:2, 194–220. Simons, R.A., K. Winson-Geideman, and B.A. Mikelbank. The Effects of an Oil Pipeline Rupture on Single Family House Prices. Appraisal Journal, 2001, 69:4, 410–18.

P e r c e i v e d

E n v i r o n m e n t a l

R i s k

u

2 4 3

Sirmans, G.S., L. MacDonald, D.A. Macpherson, and E.N. Zietz. The Value of Housing Characteristics: A Meta-Analysis. Journal of Real Estate Finance and Economics, 2006, 33, 215–40. Sirmans, G.S., D.A. Macpherson, and E.N. Zietz. The Composition of Hedonic Pricing Models. Journal of Real Estate Literature, 2005, 13:1, 3–43. Slater, M.D. and K.A. Rasinski. Media Exposure and Attention as Mediating Variables Influencing Social Risk Judgments. Journal of Communication, 2006, 55:4, 810–27. Slovic, P. Perception of Risk. Science (New Series), 1987, 236:4799, 280–85. Smolen, G.E., G. Moore, and L.V. Conway. Economic Effects of Hazardous Chemical and Proposed Radioactive Waste Landfills on Surrounding Real Estate Values. Journal of Real Estate Research, 1992, 7:3, 283–95. Stein, R.M., L. Duen¸as-Osorio, and D. Subramanian. Who Evacuates When Hurricanes Approach? The Role of Risk, Information, and Location. Social Science Quarterly, 2010, 91:3, 816–34. Sun, H. and M.J. Seiler. Hyperbolic Discounting, Reference Dependence, and its Implications for the Housing Market. Journal of Real Estate Research, 2013, 35:1, 1–23. Tu, C.C. and M.J. Eppli. An Empirical Examination of Traditional Neighborhood Development. Real Estate Economics, 2001, 29:3, 485–501. Turnbull, G.K., V. Zahirovic-Herbert, and C. Mothorpe. Flooding and Liquidity on the Bayou: The Capitalization of Flood Risk into House Value and Ease-of-Sale. Real Estate Economics, 2013, 41:1, 103–29. Tversky, A. and D. Kahneman. Judgment under Uncertainty: Heuristics and Biases. Science, 1974, 185:4157, 1124–31. Watson, B.R., D. Riffe, L. Smithson-Stanley, and E. Ogilvie. Mass Media and Perceived and Objective Environmental Risk: Race and Place of Residence. Howard Journal of Communications, 2012, 24:2, 134–53. Weaver, D.A. and B. Bimber. Finding News Stories: A Comparison of Searches Using LexisNexis and Google New. Journalism & Mass Communication Quarterly, 2008, 85:3, 515–30. Wilde, L., C. Loos, and J. Williamson. Pipelines and Property Values: An Eclectic Review of the Literature. Journal of Real Estate Literature, 2012, 20:2, 245–59. Wyman, D., M. Hutchison, and P. Tiwari. Testing the Waters: A Spatial Econometric Pricing Model of Different Waterfront Views. Journal of Real Estate Research, 2014, 36: 3, 363–82. Xu, P., Y. Han, and J. Yang. U.S. Monetary Policy Surprises and Mortgage Rates. Real Estate Economics, 2012, 40:3, 461–507. Zahirovic-Herbert, V. and W. Chatterjee. What is the Value of a Name? Conspicuous Consumption and House Prices. Journal of Real Estate Research, 2011, 33:1, 105–25.

Julia Freybote, Florida International University, Miami, FL 33199 or jfreybote@ gmail.com. Eric Fruits, Nathan Associates Inc., Economics International Corp., and Portland State University, Portland, OR, 97207 or [email protected].

J R E R

u

Vo l .

3 7

u

N o .

2 – 2 0 1 5

This page intentionally left blank