Viewscapes and Flood Hazard Coastal Housing Market Response

Viewscapes and Flood Hazard: Coastal Housing Market Response to Amenities and Risk May 21, 2007 Okmyung Bin, Tom Crawford, Jamie Kruse, and Craig Landry East Carolina University

Coastal Housing Market § Coastal amenities and risk are so highly correlated that separate identification within the hedonic framework is potentially challenging. § Bin and Kruse (2006) found that virtually all ocean front homes on the Outer Banks of North Carolina are located in the V zone.

In this study… § A three dimensional measure of ocean view, that varies independent of risk classification, is constructed to disentangle these spatially integrated attributes. § A spatial autoregressive hedonic model is developed to provide consistent estimates of the willingness to pay for coastal amenities and risk.

Theoretical Framework § Hedonic price model: H: hedonic property price schedule S: structural characteristics N: neighborhood characteristics E: environmental characteristics R: property risk factors

Theoretical Framework § Individual utility function: q: a composite commodity (numeraire) U is bounded, increasing, and strictly concave in all arguments

Theoretical Framework § Expected utility function: Individual utility is represented by q is a composite commodity (numeraire) m represents expected income converts sales price to an annual payment

Theoretical Framework § First order conditions:

New Hanover County, NC

New Hanover County, NC

New Hanover County, NC § Our study area encompasses four primary beach communities – Carolina Beach, Figure Eight Island, Kure Beach, and Wrightsville Beach. § They are among the most highly developed areas in the region and are at risk of flooding and shoreline erosion due to storms and nor’easters. § Greater vulnerability to natural disasters has resulted in a trend of rising insured disaster losses as other coastal locations.

Data § This study utilizes various spatial data including geocoded property parcel boundaries, linear ocean shoreline, and a gridded elevation surface. § Source data were obtained from the New Hanover Co Tax Office and the National Oceanic and Atmospheric Administration (NOAA). § We use a total of 1, 075 single family residential homes that were sold in the study area between 1995 and 2002.

Viewscape § A viewscape is the area visible from an observation point at a specified location while accounting for visual obstructions. § A separate elevation surface was constructed for each year using air photos and building height information from tax assessor data. § The observer point was positioned at the center of a building footprint at an elevation 10 feet below the maximum height of a building.

Illustration of Viewscape

Elevation Surfaces

Illustration of Viewscape

Illustration of Viewscape

Summary Statistics Variable Definition PRICE Adjusted house sales price AGE Age of house BATHRM Number of bathrooms SQFT Mean Std. Dev. 297, 967. 980 347, 437. 300 22. 058 20. 186 2. 491 1. 173 Total structure square footage 1784. 080 884. 541 LOTSIZE Total lot size measured in square feet 9235. 220 10431. 000 AIRCOND Central air conditioning (= 1) 0. 896 0. 306 FIREPLCE Fireplace (= 1) 0. 487 0. 500 MULTISTR Multistory house (= 1) 0. 447 0. 497 HDWDFLR Hardwood floor (= 1) 0. 084 0. 277 DETGAR Detached garage (= 1) 0. 039 0. 194 POOL Swimming pool (= 1) 0. 032 0. 175 GOODCOND Good condition (= 1) 0. 152 0. 359

Summary Statistics Variable Definition EVACRTE Distance to the evacuation route CBD Distance downtown Wilmington VIEW Degree of ocean view in one mile BEACH Distance in feet to the nearest beach SOUND Mean Std. Dev. 3188. 420 5037. 080 71, 930. 730 14, 709. 660 18. 365 41. 666 1742. 560 1366. 320 On sound front (= 1) 0. 158 0. 365 PIER Pier (= 1) 0. 095 0. 293 SFHA On Special Flood Hazard (= 1) 0. 453 0. 498

A Spatial Lag Hedonic Model § With increasingly available spatial data and analysis methods, consideration of spatial dependence has been of growing importance. § Spatial dependence in the hedonic price models refers to the interdependence among sales prices caused by their proximity. § If relevant spatial dependence is ignored, then the estimates could be inconsistent, and any inference may result in misleading conclusions.

A Spatial Lag Hedonic Model § A spatial lag hedonic model assumes that the spatially weighted sum of neighborhood housing prices enters as an explanatory variable. § The spatial weights matrix reflects the structure of potential spatial interaction. § Inclusion of the spatially lagged dependent variable induces a correlation with the error term.

MLE of Spatial Hedonic Model Variable Coefficient Standard Error P-value LAMBDA( ) 0. 114656 0. 049429 0. 020362 AGE 0. 018373 0. 003201 0. 000000 AGE 2 0. 000252 0. 000045 0. 000000 BATHRM 0. 296027 0. 065270 0. 000006 BATHRM 2 0. 028684 0. 008968 0. 001382 SQFT 0. 000029 0. 000073 0. 694468 SQFT 2 0. 000021 0. 000103 0. 837590 LOTSIZE 0. 000010 0. 000005 0. 038575 LOTSIZE 2 0. 003487 0. 005020 0. 487324 AIRCOND 0. 201767 0. 059521 0. 000700 FIREPLCE 0. 115396 0. 035515 0. 001157 MULTISTR 0. 102600 0. 041291 0. 012963 HDWDFLR 0. 003305 0. 056588 0. 953419

MLE of Spatial Hedonic Model Variable Coefficient Standard Error P-value DETGAR 0. 098230 0. 080808 0. 224139 POOL 0. 063704 0. 091142 0. 484577 GOODCOND 0. 065039 0. 048129 0. 176582 ln(EVACRTE) 0. 023042 0. 015010 0. 124761 ln(CBD) 0. 742626 0. 374364 0. 047289 VIEW 0. 002957 0. 000636 0. 000003 ln(BEACH) 0. 147332 0. 033956 0. 000014 SOUND 0. 349963 0. 065531 0. 000000 PIER 0. 143529 0. 074067 0. 052646 SFHA 0. 113402 0. 056003 0. 042876

Marginal WTP § Marginal WTP of Viewscape § Marginal WTP of Proximity to Beach § Marginal WTP of SFHA

Marginal WTP Estimates Marginal Willingness to Pay ($) Variable 95% Lower Bound Mean 95% Upper Bound VIEW 579. 93 995. 31 1432. 12 BEACH 469. 56 853. 67 1252. 78 SOUND 82627. 01 141022. 21 211971. 85 PIER 220. 72 51944. 01 114153. 81 SFHA 66665. 42 36081. 73 1896. 48 § A bootstrapping procedure is used to generate 95% confidence intervals for the marginal WTP. The reported confidence intervals are based on 1, 000 sets of random parameter vectors.

Conclusions § Failure to control for both risk and amenities will likely lead to biased results that inaccurately identify the sources of property value. § In this paper, a three dimensional GIS based viewscape is constructed to disentangle the effects of coastal amenities from risk in a coastal setting. § Our findings suggest that incorporating the GIS based view measures can be successful in isolating risk factors from spatial amenities.