Effects of Flood Hazards on Property Values: Evidence Before and After Hurricane Floyd 1

Transcription

1 Effects of Flood Hazards on Property Values: Evidence Before and After Hurricane Floyd 1 Okmyung Bin Department of Economics East Carolina University Greenville, NC Voice: Fax: bino@mail.ecu.edu Stephen Polasky Department of Applied Economics University of Minnesota St. Paul, MN Voice: Fax: spolasky@apec.umn.edu August 2003 Abstract: This study uses a hedonic property price function to estimate the effects of flood hazards on residential property values. The study utilizes data from over 8,000 single-family residential home sales in Pitt County, North Carolina between 1992 and This area experienced significant flooding from Hurricane Floyd in September Results show that a house located within a floodplain has a lower market value than an equivalent house located outside the floodplain. Furthermore, the price discount from locating within a floodplain is significantly larger after Hurricane Floyd than before. For an average valued house, the price discount for floodplain location is greater than the capitalized flood insurance premiums for sales after Floyd but less than capitalized flood insurance premiums for sales before Floyd. 1 We thank Ralph Forbes of the Pitt County Management Information Systems for helping us acquire and understand floodplain and property sales data and Afua Dei-Tutu for excellent research assistance.

2 I. Introduction Hurricane Floyd hit eastern North Carolina in September 1999 bringing torrential rains and record flooding. Hurricane Floyd directly affected over two million people and resulted in the largest peacetime evacuation in US history (Federal Emergency Management Agency (FEMA), 2002). The total amount of damage was estimated to be about $6 billion, most of it caused by flooding. The land in eastern North Carolina is relatively flat and low-lying, and most of the area is prone to flooding. Prior to Hurricane Floyd many people in eastern North Carolina did not have flood insurance and many homeowners in floodplains were not aware that they lived in a floodplain (FEMA, 2002). The large-scale damage caused by Hurricane Floyd increased awareness of flood hazards. There were also government programs to promote both awareness and the purchase of flood insurance. Sales of flood insurance policies increased by 24 percent in North Carolina after Hurricane Floyd (FEMA, 2002). In North Carolina, there were 102,876 flood insurance policies covering $14.9 billion in property as of December 2001 (FEMA, 2002). Several previous studies have documented the price reduction from location in a floodplain and compared the price reduction with the capitalized flood insurance premiums (Shilling, Benjamin, and Sirmans, 1985; MacDonald, Murdoch, and White, 1987; Donnelly, 1989; Speyrer and Ragas, 1991; Harrison, Smersh, and Schwartz, 2001). A common finding in these studies is that location within a floodplain lowers property value anywhere from four to twelve percent of average. With the exception of Harrison, Smersh, and Schwartz (2001), these studies find that the sales price reduction is more than the capitalized value of insurance premiums, indicating that there may exist non-insurable costs associated with flooding. It is interesting to note that the study area in Harrison, Smersh, and Schwartz (2001) had not experienced any major flooding in the recent past. The other studies that had found significant 2

3 price differentials for floodplain location were in areas that had experienced recent flooding. This suggests that recent experience with flooding raises perception of flood risk and the discount associated with living in flood prone areas. To the best of our knowledge, there has been no prior attempt to compare the effects of flood hazards on property values before and after a major flooding event. In this paper, we use a hedonic property price function to estimate the effects of flood hazards on property values and compare the effects both before and after Hurricane Floyd using floodplain data coupled with extensive property sales records from Pitt County in eastern North Carolina. Our data contain single-family residential home sales records between July 1992 and June 2002, which include sales both before and after Hurricane Floyd. Pitt County suffered massive flooding during Hurricane Floyd. Over 6,000 homes were flooded in Pitt County, more than 50,000 people were displaced, and about 6,000 people were housed in emergency shelters for several weeks (Colby, Mulcahy, and Wang, 2000). The results from estimating the hedonic model indicate that the market value of a house located within a floodplain is lower than a similar house located outside the floodplain. Over the entire sample from 1992 to 2002, the reduction in property values for houses in the floodplain is about $7,530 for an average valued house. On average, property values are reduced by an estimated 5.8 percent when located in a floodplain. When we distinguish between sales that occurred before Hurricane Floyd from those that occurred after Floyd, we find that the estimated discount for location in the floodplain more than doubled. The discount for floodplain location for an average valued house was about $10,770 post-floyd and about $4,930 pre-floyd. These discounts for location in the floodplain can be compared to the capitalized premium value of flood insurance. For an average valued house, assuming a five percent discount rate, the capitalized premium value of flood insurance is $6,880. For an average valued house, the 3

4 discount for post-floyd sales is larger than the capitalized insurance premiums while the discount for pre-floyd sales is smaller. When homeowners are well aware of flood hazards and are fully insured, one might expect the reduction in property values from floodplain location to be equal to or greater than the capitalized value of flood insurance premiums. Floods cause non-monetary losses that cannot be covered by insurance, including the hassle and deprivation of being displaced along with the loss of personal or family items with sentimental (if not monetary) value. Given the choice between being flooded out of house and home with full insurance coverage versus not being flooded, there is little doubt that most would prefer not to be flooded. Therefore, one might expect that housing price reductions for properties in a floodplain might exceed the capitalized cost of flood insurance. On the other hand, if property owners underestimate the costs of flooding or homeowners are relatively unaware of flood hazards, there might be little reduction in value for properties within a floodplain. Our results are consistent with the notion that recent experience with flooding awakens or reinforces the perceived risks and costs associated with flooding. Our results for flooding provide an interesting contrast with Beron, Murdoch, Thayer, Vijverberg (1997) who estimated the hedonic price of earthquake risk before and after the Loma Prieta earthquake in They found that hedonic price for earthquake risk actually fell after the earthquake. Most San Francisco Bay area residents are well aware of the risk of earthquakes. There has been extensive media coverage and educational campaigns about earthquake preparedness for decades. Prior to Hurricane Floyd, many people living in eastern North Carolina were unaware that they lived in a floodplain and did not own flood insurance. The flooding associated with Hurricane Floyd probably caused a reassessment of risks (or awakening to risks) in a way that the Loma Prieta earthquake did not. It is interesting to note that FEMA 4

5 reported that Pitt County had the highest growth rate in flood insurance policies in North Carolina following Hurricane Floyd (FEMA, 2002). In this study we utilize data from Geographic Information Systems (GIS) to measure the distance from each property to important features that could influence property values, which was not considered by previous flood-related hedonic pricing studies. We calculate the distance to the Tar River, which is the major river in Pitt County, and to the nearest stream and creek for each house in the data set. It may be desirable to live near water as long as there is not a danger from flooding. This allows us to control for the effects of being close to water resources as distinct from whether a property is in a floodplain. Although most floodplains locate along the Tar River and creeks and streams, proximity to these water resources increased property values after controlling for the effects of the floodplains. Our data also include a number of structural variables (e.g., square footage, number of bedrooms, age) as well as neighborhood attributes that affect home values, such as proximity to railroads and business centers. We find that proximity to railroads, major road and streets, and Pitt-Greenville Airport all have a negative impact on housing values. II. Study Area and Data Pitt County is located in the coastal plain of eastern North Carolina. The Tar River, which goes through the middle of the county, flows into the Pamlico River and then into Pamlico Sound. Most of the county is low and flat. The county s highest elevation is only 126 feet above sea level and about 60 percent of the area has soil that is poorly drained (Colby, Mulcahy, and Wang, 2000). Floodplains in the county are located adjacent to the Tar River and its tributary streams and creeks. Figure I presents a map showing the floodplain as well as the Tar River. Most of the higher ground in the county is located on the southern side of the Tar River. 5

6 According to the 2000 Census, the county has a population of 133,798, of which 60,476 people reside in Greenville, the County s largest city. This study combines three data sets from the Pitt County Management Information Systems (MIS): property parcel data, floodplain mapping data, and Pitt County Geographic Information Systems (GIS) data. Property parcel records contain detailed information about the structural attributes of properties such age, square feet, and heating source, as well as sale date and sale price. A total of 8,375 single-family residential homes were sold between July 1992 and June House sales prices are adjusted to June 2002 prices using a consumer price index for housing. The average selling price was $129,816 with a minimum sales price of $5,157 and a maximum of $722,018. About 70 percent of the houses have a sales price between $75,000 and $225,000. A typical home is about 22 years old and has 2,345 square feet. About 38 percent of the homes have gas heating, and about 80 percent have a fireplace. Floodplain mapping data provide the location and size of floodplains in the County. Floodplains are defined as areas that would flood in a 100-year flood event (i.e., areas with a 1.0 percent chance of flooding in a year). Each home sale in the data set is compared with the floodplain mapping data to determine whether the property is within the floodplain or not. About six percent of the homes in the data are located in a floodplain. In addition, this study calculates the distance of each property to the Tar River, Pitt-Greenville Airport, major roads and landmarks, and streams and creeks. Distances are measured as the Euclidean distance in feet from the centroid of the property to the nearest edge of a feature. This study also controls for the effects of townships. A set of dummy variables is included for the fifteen townships in the County, which equals one for a house located in a township and zero otherwise. A large fraction of house sales occurred in Winterville (41 percent), which is a newly developed residential area. Table I defines the variables used in the hedonic price function and their definitions. Summary statistics are given in table II. 6

7 In 1968, Congress passed the National Flood Insurance Program (NFIP). Prior to passage of this program, the federal government routinely paid large sums for disaster relief after floods. The purpose of NFIP was to provide insurance protection for property owners and to encourage communities to adopt floodplain management. Insurance is available only in communities with approved community floodplain management, which currently number about 20,000 communities in the US. Currently, the NFIP provides a maximum flood coverage limit of $250,000 on single-family dwellings. The coverage is written through local insurance companies that collect the premium and pay claims for the NFIP. III. Methodology This study uses a hedonic property price approach to estimate the effects of the location within a floodplain on property values. Since the pioneering work by Rosen (1974), many studies have applied the hedonic property price models to estimate the value of environmental amenities and the discount from hazards such as earthquakes or chemical plants on residential property values (e.g., Brookshire, Thayer, Tschirhart and Schulze, 1985; Thayer, Berknopf, Brookshire, and Schulze, 1985; Folland and Hough, 2000). In this section, we provide a brief discussion of key features for the theoretical model and the estimation method. Freeman (1993) provides a useful summary of the theoretical aspects of the hedonic property price models. Let X represent a vector of binary characteristics of a house such as township location, whether it is gas heated, has hardwood floors or a fireplace, and let Z represent a vector of continuous characteristics such as square footage, age, distance to river and business center. Let F represent the flood risk variable. The housing market is assumed to be in equilibrium, which requires that individuals optimize their residential choice based on the prices of alternative locations. Prices are assumed to be market clearing, given the inventory of housing choices and 7

8 their characteristics. With these assumptions, the price of any house, P, can be described as a function of structural and locational characteristics as well as flooding risk: P = P(X, Z, F). (1) Equation (1) is referred to as the hedonic price function. Assume that each individual s utility function depends upon Q, a composite commodity representing all goods other than housing and the housing characteristics. Then, each individual will choose where to live by maximizing utility, U(Q, X, Z, F) subject to a budget constraint given by M P Q = 0, where M is income. The price of Q is implicitly scaled to $1. The first-order necessary conditions that characterize a solution for the optimal choice of the housing selection yield: U / F U / Q P =. (2) F Equation (2) shows that the marginal rate of substitution between flood risk and the composite good, that is the marginal willingness to pay to reduce the flood risk, equals the implicit price of the flood risk. Thus, the estimation and partial differentiation of the hedonic price function with respect to the flood risk variable reveal the marginal willingness to pay for the flood risk reduction. We use least squares regression analysis to estimate the following hedonic price function: ln P = α + βi X i + γ j Z j + φf + ε (3) i where X i is the binary indicator of the i th house attribute, Z j is the continuous variable of the j th housing attribute, F is a binary variable equal to one for location in the floodplain and zero otherwise, and ε is a random error term. We use a natural log transformation for distance-related variables in Z j to capture the declining effect of these variables with distance. Previous studies have found that the log transformation of the distance-related variables fits the model better than j 8

9 a simple linear functional form (Iwata, Murao, and Wang, 2000; Mahan, Polasky, and Adams, 2000). We use quadratic specifications for other continuous variables such as square footage, age, and number of rooms. We experimented with several alternative specifications and found that the primary results are robust to alternative functional forms. We use White s method (White, 1980) to obtain a consistent estimator of the covariance matrix to get estimates of the standard error corrected for potential heteroskedasticity. In the following section, the estimation results are reported along with the marginal effects. The marginal effect for flood risk ( P/ F) is equal to price times the flood risk coefficient. This approach for calculating the marginal effect applies to all non-logged variables. For distance related variables, which are log-transformed, the marginal effect is price times the distance coefficient divided by the distance. All marginal effects are evaluated at the observed sample means. IV. Estimation Results The estimation results are presented in table III and table IV. In table III, we do not distinguish between sales that occurred before Hurricane Floyd and sales that occurred after Floyd. The model used to generate the results in table IV is identical to the model used to generate table III with one exception. We added an interaction term between the floodplain variable and a post-floyd dummy variable, which equals one if the sale occurs after Hurricane Floyd and zero otherwise. As shown in table III, virtually all of the variables are statistically significant at the 95 or 99 percent level and have the expected sign. The coefficient on the flood variable (FLOOD) has a negative sign and is statistically significant. The estimated marginal effect for the flood 9

10 variable implies that the location within the floodplain lowers the property value by an estimated $7,529, which represents a 5.8 percent reduction in sales price for an average valued house. Table IV shows the effect of Hurricane Floyd on the estimated discount for house prices within the floodplain. The difference in the estimated discount between pre-floyd and post- Floyd sales is statistically significant. The pre-floyd estimated discount for location in the floodplain is $4,933. The post-floyd estimated discount more than doubled to $10,774. The results indicate that the massive damage caused by Hurricane Floyd changed the perception of flood risks by property owners and the discount for properties within the floodplain. The predicted price differentials for pre-floyd and post-floyd sales due to the location within a floodplain are compared with the capitalized value of flood insurance premiums in table V. For the calculation of flood insurance premiums, three market values of houses are used: low ($75,000), average ($125,000), and high ($225,000). Reported insurance premiums are based on NFIP insurance rates. Insurance premiums largely depend on the type of flood zone, the value of contents insured, deductibles as well as the value of buildings. Each type of flood zone reflects the severity or type of flooding. We use the rates for single family home in the flood zone A without a basement, which is one of the most common types of houses with insurance coverage in Pitt County. The flood zone A is the area with a 1.0 percent chance of annual flooding and with the mandatory flood insurance purchase requirements. Deductibles are assumed to be $500 for building and contents, respectively. Annual flood insurance premiums for each type of house are given in the second column of table V. The annual insurance payments are discounted in perpetuity using discount rates of three percent, five percent, and seven 10

11 percent. If property owners of outside the floodplain purchase flood insurance, a more valid comparison would be between the price differentials and the marginal change in insurance cost between homes inside and outside the floodplain. However, most property owners in Pitt County outside of the floodplain do not buy flood insurance. The predicted price differentials for pre-floyd sales are smaller than the capitalized value of insurance costs while the price differentials for post-floyd sales are larger than the capitalized value of insurance costs. For pre-floyd sales, the only price differential that is larger than the capitalized insurance costs is a high-valued house with a seven percent discount rate. For post- Floyd sales the price differentials are greater than the capitalized insurance costs with the exception of a low-valued house with a three percent discount rate. These results are consistent with thinking that property owners may underestimate the risks from flooding if they have not had recent experience with flooding. On the other hand, the monetary valuation of risks from flooding is perceived to be greater than the capitalized value of flood insurance for property owner with recent flood experience. There are some uninsurable costs from flood risk (e.g., disruption of normal life and loss of items of sentimental value). The difference between the sales price differentials and the capitalized insurance costs may represent a risk premium reflecting these factors. Alternatively, it is possible that property owners now overestimate the risks posed by flooding. Tables III and IV also show that structural and neighborhood variables have significant influence on sales values. The coefficients of structural variables all have expected signs and are statistically significant. Evaluated at the average value of houses, the results indicate that a house price increases by $42 per an additional square foot. An additional year of age of a house lowers the estimated sales price by $1,448 evaluated at the mean value of age. Similarly, having 11

12 an additional bedroom increases estimated sales price by $2,913 and having an additional bathroom increases estimated sales price by $18,921. Having the following characteristics increases estimated house sales price: gas heat $4,414, a brick face $7,529, a fireplace $18,823, and hardwood floors $10,515. A house in good condition increases estimated sales price by $16,746. Having the property vacant at the time of sales lowers estimated sales price by $52,576. A new house sold within a year after construction lower estimated sales price by $5,192. Many new houses in Pitt County are sold without landscaping. Houses that are two or three years old may command a higher price because they are likely to have complete landscaping. The results indicate that proximity to the nearest business center, the nearest creek or stream, and the Tar River increases the property values while the proximity to the Pitt-Greenville Airport, railroads, and major roads and streets lowers property value. Note that for these distance-related variables a negative (positive) relationship to the dependent variable means that the proximity to the feature increases (decreases) the property values. Moving 300 feet closer to the nearest creek or stream increases estimated sales value by $267 evaluated at the observed mean values and similarly moving 1,000 feet closer to the Tar River from an initial distance of one mile results in a $688 increase in house value. After controlling for the floodplain effect, proximity to water resource seems to increase the property value. Without controlling the flood risk in the regression, the proximity to creek or stream may show a negative effect. Moving 1,000 feet closer to the nearest railroad lowers estimated sales value by $290 evaluated at the observed mean values, and similarly reducing the distance to the Pitt-Greenville Airport by 1,000 feet from an initial distance of one mile results in a $1,278 decrease in house value. Proximity to these features may generate greater noise, congestion, and traffic. 12

13 V. Conclusions This study estimates the effects of house location within a floodplain on residential property value using a hedonic property price approach before and after a major flood that occurred when Hurricane Floyd swept through North Carolina in September The estimation results indicate that the price of a residential housing property located within a floodplain is lower than an otherwise similar house located out of a floodplain. Location within the floodplain lowers estimated sales value for an average house by $7,529, which represents a 5.8 percent of the average house sales price. The difference in the estimated discount between pre-floyd and post-floyd sales is statistically significant. The pre-floyd estimated discount for location in the floodplain is $4,933. The post-floyd estimated discount is $10,774, which is more than double the pre-floyd discount. The estimated sales price differentials are compared to the capitalized value of flood insurance. The reduction in house price from location in a flood plain was smaller than the capitalized value of flood insurance for pre-floyd sales but larger for post-floyd sales. These results indicate that recent experience with flooding tends to increase the perceived risk associated with flooding. It remains to be seen whether the risk premium tied to recent experience will tend to erode as memories of flooding recede. 13

14 References Beron K., J. Murdoch, M. Thayer, W. Vijverberg An Analysis of the Housing Market before and after the 1989 Loma Prieta Earthquake. Land Economics 73: Brookshire, D., M. Thayer, J. Tschirhart, and W. Schulze A Test of Expected Utility Model: Evidence from Earthquake Risks. Journal of Political Economy 93: Colby, J., K. Mulcahy, and Y. Wang Modeling Flooding Extent from Hurricane Floyd in the Coastal Plains of North Carolina. Environmental Hazards 2: Cropper, M., L. Deck and K. McConnell On the Choice of Functional Form for Hedonic Price Functions. Review of Economics and Statistics 70: Donnelly, W Hedonic Price Analysis of the Effects of a Floodplain on Property Values. Water Resources Bulletin 24: Federal Emergency Management Agency After Floyd - North Carolina Progress. Folland, S. and R. Hough Externalities of Nuclear Power Plants: Further Evidence. Journal of Regional Science 40: Freeman, M The Measurement of Environmental and Resource Values: Theory and Method. Resources for the Future. Washington, D.C. Harrison, D., G. Smersh, and A. Schwartz Environmental Determinants of Housing Prices: The Impact of Flood Zone Status. Journal of Real Estate Research 21: Iwata, S., H. Murao and Q. Wang Nonparametric Assessment of the Effects of Neighborhood Land Uses on the Residential House Values. Advances in Econometrics: Applying Kernel and Nonparametric Estimation to Economic Topics. Vol. 14. JAI Press. MacDonald, D., J. Murdoch, and H. White Uncertain Hazards, Insurance and Consumer Choice: Evidence from Housing Markets. Land Economics 63: Mahan, B., S. Polasky and R. Adams Valuing Urban Wetlands: A Property Price Approach. Land Economics 76: Rosen, S Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition. Journal of Political Economy 82: Shilling, J., J. Benjamin, and C. Sirmans Adjusting Comparable Sales for Floodplain Location. The Appraisal Journal Speyrer, J. and W. Ragas Housing Prices and Flood Risk: An Examination Using Spline Regression. Journal of Real Estate Finance and Economics 4:

15 Thayer M., R. Berknopf, D. Brookshire, and W. Schulze An Economic Evaluation of Hazard Alerts: A Case Study of Mammoth, California. U.S. Geological Survey. White, H A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity. Econometrica 48:

16 TABLE I Definition and Description of the Variables Variable Description PRICE House sales price adjusted to a June 2002 level TOWN1 Dummy variable for a township (1 if Arthur, 0 otherwise) TOWN2 Dummy variable for a township (1 if Ayden, 0 otherwise) TOWN3 Dummy variable for a township (1 if Belvoir, 0 otherwise) TOWN4 Dummy variable for a township (1 if Bethel, 0 otherwise) TOWN5 Dummy variable for a township (1 if Chicod, 0 otherwise) TOWN6 Dummy variable for a township (1 if Falkland, 0 otherwise) TOWN7 Dummy variable for a township (1 if Farmville, 0 otherwise) TOWN8 Dummy variable for a township (1 if Fountain, 0 otherwise) TOWN9 Dummy variable for a township (1 if Grifton, 0 otherwise) TOWN10 Dummy variable for a township (1 if Grimesland, 0 otherwise) TOWN11 Dummy variable for a township (1 if Pactolus, 0 otherwise) TOWN12 Dummy variable for a township (1 if Carolina, 0 otherwise) TOWN13 Dummy variable for a township (1 if Swift Creek, 0 otherwise) TOWN14 Dummy variable for a township (1 if Winterville, 0 otherwise) TOWN15 Dummy variable for a township (1 if Greenville, 0 otherwise) GASHEAT Dummy variable for gas heating (1 if gas heating, 0 otherwise) FCBRICK Dummy variable for face brick (1 if face brick, 0 otherwise) FIREPLC Dummy variable for fireplace (1 if fireplace, 0 otherwise) HWFLOOR Dummy variable for hard wood floor (1 if hard wood floor, 0 otherwise) QUALITY Dummy variable for good quality (1 if good quality, 0 otherwise) VACANT Dummy variable for vacant house (1 if vacant house, 0 otherwise) NEWHOME Dummy variable for new home (1 if sold within a year after built, 0 otherwise) SQFT Total structure square footage AGE Year house was built subtracted from 2002 BEDRM Number of bedrooms BATHRM Number of bathrooms AIRPORT Distance in feet to the Pitt-Greenville Airport STREAM Distance in feet to nearest creek or stream CENTER Distance in feet to nearest business center RAILRD Distance in feet to nearest railroad RIVER Distance in feet to the Tar River TRAFFIC Distance in feet to major roads and streets FLOOD Dummy variable for house within floodplain (1 if floodplain, 0 otherwise) FLOYD Dummy variable for Hurricane Floyd (1 if sold after Floyd, 0 otherwise) 16

17 TABLE II Summary Statistics of the Variables Variable Mean Std. Dev. Minimum Maximum PRICE 129, , , , TOWN TOWN TOWN TOWN TOWN TOWN TOWN TOWN TOWN TOWN TOWN TOWN TOWN TOWN TOWN GASHEAT FCBRICK FIREPLC HWFLOOR QUALITY VACANT NEWHOME SQFT 2, , AGE BEDRM BATHRM AIRPORT 32, , , , STREAM , CENTER 4, , , RAILRD 4, , , RIVER 20, , , TRAFFIC , FLOOD FLOYD Note: Number of observations equal 8,

18 TABLE III Estimation Results of the Hedonic Price Function (No Floyd-Interaction Term) Variable Coefficient Standard Error p-value Marginal Effect CONSTANT < TOWN < $9, TOWN < $15, TOWN < $13, TOWN < $36, TOWN $4, TOWN $8, TOWN < $12, TOWN < $28, TOWN < $28, TOWN < $8, TOWN < $18, TOWN < $21, TOWN $13, TOWN $4, GASHEAT <.0001 $4, FCBRICK <.0001 $7, FIREPLC <.0001 $18, HWFLOOR <.0001 $10, QUALITY <.0001 $16, VACANT < $52, NEWHOME < $5, SQFT 4.59E E-05 <.0001 $41.56 SQFT E E-09 < AGE < $1, AGE E E BEDRM $2, BEDRM BATHRM <.0001 $18, BATHRM < ln(airport) <.0001 $0.21 ln(stream) $0.89 ln(center) $0.23 ln(railrd) $0.29 ln(river) < $0.18 ln(traffic) <.0001 $29.33 FLOOD < $7, Notes: Dependent variable is the log of sales price. Adjusted R-square value is Marginal effects are evaluated at the observed means. 18

19 TABLE IV Estimation Results of the Hedonic Price Function (Floyd-Interaction Term) Variable Coefficient Standard Error p-value Marginal Effect CONSTANT < TOWN < $9, TOWN < $15, TOWN < $13, TOWN < $36, TOWN $4, TOWN $8, TOWN < $12, TOWN < $28, TOWN < $28, TOWN < $8, TOWN < $18, TOWN < $21, TOWN $14, TOWN $4, GASHEAT <.0001 $4, FCBRICK <.0001 $7, FIREPLC <.0001 $18, HWFLOOR <.0001 $10, QUALITY <.0001 $16, VACANT < $52, NEWHOME < $5, SQFT 4.58E E-05 <.0001 $41.49 SQFT E E-09 < AGE < $1, AGE E E BEDRM $3, BEDRM BATHRM <.0001 $18, BATHRM < ln(airport) <.0001 $0.21 ln(stream) $1.04 ln(center) $0.23 ln(railrd) $0.29 ln(river) < $0.18 ln(traffic) <.0001 $30.31 FLOOD $4, FLOOD*FLOYD $5, Notes: Dependent variable is the log of sales price. Adjusted R-square value is Marginal effects are evaluated at the observed means. 19

20 TABLE V Comparison Between the Present Value of Flood Insurance Premiums and House Price Differentials Annual Flood Present Value of Insurance Premiums Sales Price Sales Price Value of Insurance Under Various Discount Rates Differential Differential Houses Premiums 3% 5% 7% Before Floyd After Floyd Low ($75K) $285 $9,513 $5,708 $4,077 $2,850 $6,225 Avg. ($125K) $344 $11,467 $6,880 $4,914 $4,750 $10,375 High ($225K) $448 $14,933 $8,960 $6,400 $8,550 $18,675 Notes: Flood insurance premium estimates are based on the post Flood Insurance Rate Maps (FIRM) for singlefamily houses in the flood zone type A without a basement and with estimated base flood elevation of 3 feet or more. The content values of $18,000, $30,000, and $50,000 are assumed for the low, average, and high valued houses, respectively. Deductibles for building and contents are assumed to be $500. Premiums include the federal policy fee of $80 and the increased cost of compliance (ICC) of $6. 20

21 FIGURE I Map of Pitt County, North Carolina Notes: Dark shading area represents floodplains. The Tar River flows through the middle of the county and is represented by the dark line. 21