R
V
Na
b
c
h
•••
a
ARR2AA
KGHLSP
1
aNrtvaoi(
foBP
(
0h
Landscape and Urban Planning 124 (2014) 14–21
Contents lists available at ScienceDirect
Landscape and Urban Planning
j o ur na l ho me pag e: www.elsev ier .com/ locate / landurbplan
esearch Paper
aluing green infrastructure in Portland, Oregon
oelwah R. Netusil a,∗, Zachary Levina, Vivek Shandasb, Ted Hartc
Reed College, Department of Economics, 3203 SE Woodstock Boulevard, Portland, OR 97202, USAPortland State University, Nohad A. Toulan School of Urban Studies and Planning, Portland, OR 97201, USAPortland State University, Environmental Science and Management, Portland, OR 97207, USA
i g h l i g h t s
A property’s sale price increases as distance from a green street increases.A facility’s age, size, and amount of tree canopy affects a property’s sale price.Census block or tract is the appropriate scale for measuring green street abundance.
r t i c l e i n f o
rticle history:eceived 3 June 2013eceived in revised form9 December 2013ccepted 1 January 2014vailable online 14 February 2014
a b s t r a c t
This study uses the hedonic price method to examine if proximity, abundance, and characteristics ofgreen street facilities affect the sale price of single-family residential properties in Portland, Oregon.Different methods for measuring proximity and abundance are explored with distance based on streetnetwork, and abundance of green streets at the census tract and census block level, producing statisticallysignificant results. A property’s sale price is estimated to increase as distance from the nearest green streetfacility increases although the magnitude of this effect is small. Facility type does not have a statistically
eywords:reen streetsedonic price methodow impact development
significant effect on a property’s sale price, but characteristics such as facility size, proportion of thefacility covered by tree canopy, and design complexity are estimated to influence sale price.
© 2014 Elsevier B.V. All rights reserved.
tormwaterortland, Oregon
. Introduction
In 2000, approximately 80% of the U.S. population lived in urbanreas—a number that is expected to reach 90% by 2050 (Unitedations, 2008). Urbanization puts pressure on urban ecosystems
esulting in changes in the amount, timing, and quality of stormwa-er runoff, loss and fragmentation of native habitat, and increasedulnerability to invasive species. The amount and placement of just
few key landscape features — such as trees, shrubs, and impervi-us surfaces — can influence the type of wildlife that can surviven urban areas and the quantity and quality of ecosystem servicesHennings & Soll, 2010).
While research is ongoing to examine the impact of landscapeeatures on urban ecosystems (Nelson et al., 2009), the value
f land cover (Kadish & Netusil, 2011), water quality (Leggett &ockstael, 2000; Poor, Pessagno, & Paul, 2007), wetlands (Mahan,olasky, & Adams, 2000), and trees (Donovan & Butry, 2010; Netusil,∗ Corresponding author. Tel.: +1 503 517 7306.E-mail addresses: [email protected] (N.R. Netusil), [email protected]
Z. Levin), [email protected] (V. Shandas), [email protected] (T. Hart).
169-2046/$ – see front matter © 2014 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.landurbplan.2014.01.002
Chattopadhyay, & Kovacs, 2010) have been estimated for urbanareas. One area that has not been well studied, however, is theimpact of green infrastructure — green streets, ecoroofs (or greenroofs), green walls, rain gardens and pervious surfaces — on the saleprice of single family residential properties.
Green infrastructure projects are being embraced by manyU.S. and European cities as a cost-effective way to control urbanstormwater (U.S. Environmental Protection Agency, 2010, 2011)among other challenges. The potential impact of green stormwatersystems on biodiversity, traffic accidents, and other city attributesis currently unknown, so valuing the effect of these facilities isgermane to policy discussions that attempt to link city greeningefforts with quantifiable measures that can be incorporated intothe decision making process (Nelson et al., 2009).
In this study we examine one city — Portland, Oregon — andthe effect of proximity, density and characteristics of green infra-structure, in the form of ‘green street stormwater facilities’ on thesale price of single family residential properties. Our study area
provides an ideal case study for several reasons. First, Portland isranked as the most sustainable city of the fifty largest cities in theUnited States (Revkin, 2008), so its policies may offer insights intoimproving the process of urban growth and how growth impacts
d Urb
tuomit
otCsSWbcuRQ
iflramfwrf“gapn2
adaswaApDsh“pimCWPnCaD
mttrNsp&
N.R. Netusil et al. / Landscape an
he social and biophysical conditions of other cities. Statewide landse planning goals, combined with policies implemented by the cityf Portland and Metro — the only directly elected regional govern-ent in the United States with regulatory power — have produced
nnovative programs that may serve as models for cities throughouthe country and the world.
Second, Portland also struggles, as many cities do, with seri-us environmental challenges: the main water body in Portland,he Willamette River, has a Superfund site and Steelhead Trout,oho, and Chinook Salmon, which use the Willamette River to reachpawning grounds, are listed as threatened under the Endangeredpecies Act (NOAA Fisheries Office of Protected Resources, 2010).hile combined systems were state of the art when constructed,
y 1993 even a moderate rainfall exceeded the Portland system’sapacity, triggering combined sewer overflows (CSOs) of dilute,ntreated sewage through 55 outfall points into the Willametteiver and Columbia Slough (Oregon Department of Environmentaluality, 2006).
Finally, over the past 20 years Portland has invested $1.4 billionn physical infrastructure projects to reduce combined sewer over-ows. These projects, which were completed in December 2011,educed the number of CSOs to the Willamette River from fifty ton average of four times each winter and once every third sum-er (Environmental Services City of Portland, 2011). Projects are
unded, in large part, by Portland’s combined sewer/water bills,hich are amongst the highest in the country (Frank, 2011). Further
ate increases to fund large capital projects may not be politicallyeasible, so in 2008 the city launched a new strategy, the $55 millionGrey to Green” program to control stormwater runoff. Programoals include planting 33,000 yard trees and 50,000 street trees,dding 43 acres of ecoroofs, controlling invasive plant species,urchasing over 400 acres of natural areas, and constructing 920ew green street facilities (Environmental Services City of Portland,010a).
Green streets, which is a term used by the City of Portlandnd the Environmental Protection Agency, refer to low-impactevelopment techniques that use “vegetated facilities to man-ge stormwater runoff at its source” and include curb extensions,treet planters, and rain gardens as well as “simple” green streets,hich involve changes to existing planting areas between curbs
nd sidewalks (Environmental Services City of Portland, 2008).dditional benefits attributed to these facilities include increasedroperty values (Wise et al., 2010), traffic calming (Ewing &umbaugh, 2009), better bike access and enhanced pedestrian
afety (Maas et al., 2009), and added green space and wildlifeabitat (Kazemi, Beecham, Gibbs, & Clay, 2009). These facilitiesare more cost-effective than piping stormwater to a treatmentlant” (Environmental Services City of Portland, 2010b) and are
ncreasingly being promoted by city managers as an effectiveeans for controlling stormwater runoff. Cities such as San Jose,
alifornia, Chicago, Illinois, Philadelphia, Pennsylvania, and Seattle,ashington are expanding green street programs (Environmental
rotection Agency, 2010a) and a low impact development ordi-ance, which includes green street facilities, passed in Los Angeles,alifornia (City of Los Angeles Stormwater Program, 2011) and as
statewide ordinance in Washington State (State of Washingtonepartment of Ecology, 2012).
Although cities are moving forward with green street facilities,any questions remain, including whether these facilities affect
he sale price of nearby houses. While green space and wildlife habi-at have been estimated to increase the sale price of single-familyesidential properties (Donovan & Butry, 2010; Mahan et al., 2000;
etusil, 2006), literature examining the relationship between greentreet facilities and the sale price of single-family residentialroperties is extremely limited. Ward et al. (Ward, MacMullan,
Reich, 2008) estimate that properties located in low-impact
an Planning 124 (2014) 14–21 15
development project areas in Seattle, Washington sold for 3.5–5%more than properties in the same zip code located outside projectareas. Williams and Wise (2009) reach the opposite conclusionfinding that lots in Gainesville, Florida with low-impact devel-opment stormwater systems are valued less than lots that useconventional approaches.
Our research contributes to this nascent literature by combin-ing a data set of single-family residential properties sold withinthe city of Portland, Oregon from January 1, 2005 to December31, 2007 with detailed information about green street facilitiescollected by project researchers. We explore the following ques-tions: First, is a property’s sale price influenced by its distance tothe nearest green street facility and is Euclidean (“straight line”)or street network the preferred distance measure? Second, doesthe abundance of facilities near a property affect its sale price andwhat is the appropriate scale for measuring abundance? Third,do green street characteristics such as facility type, size, the pro-portion of the facility covered by tree canopy, and other designfeatures affect a property’s sale price? When addressing thesequestions we examine how our results might contribute to the cur-rent debate about green infrastructure and the extent to whichimplementation can help to mitigate current challenges facingcities.
2. Methods
2.1. Study area
The study area is the part of the city of Portland, Oregon locatedwithin Multnomah County. Portland is divided into five quadrants:North (N), Northeast (NE), Southeast (SE), Southwest (SW), andNorthwest (Fig. 1). Northwest is excluded from our data set becauseonly one green street facility existed in that quadrant during thestudy period. The remaining quadrants had 4297 (N), 9232 (NE),12,594 (SE), and 3589 (SW) single-family residential home salesbetween January 1, 2005 and December 31, 2007.
2.2. Data
The data set includes sale price, property characteristics, loca-tion, land cover, and green street information. Sale price andproperty characteristics were obtained from the MultnomahCounty Assessor. Observations were screened to make sure thattransactions occurred at arms length and were free of recordingerrors; duplicate transactions were dropped with the most recenttransaction retained for the analysis. Sale prices were convertedto 2007 dollars using the Consumer Price Index-Urban (Bureauof Labor Statistics, 2011). Table 1 contains summary statistics forprice, structural, property, and neighborhood variables for the29,712 single-family residential property transactions used in theanalysis.
Green street facilities were identified using a map from the Cityof Portland Bureau of Environmental Services (2010) that containsgeoreferenced coordinates for all publicly built facilities. Greenstreet variables were created using information about publicly builtfacilities that existed in the year a property sold, for example, homessold in 2006 were associated with green street facilities built in2006 or earlier. Additional data sets from Metro were used to deter-mine the number of facilities for each census block, census tract,and neighborhood in the study area (Metro Data Resource Center,2009). In 2007 there were 614 green streets in the study area. These
facilities are sometimes built close together in “clusters,” so only asubset — 318 facilities using Euclidean distance and 262 facilitiesusing street network — are the closest facilities to the propertiesin our data set. Street network calculations use the street grid
16 N.R. Netusil et al. / Landscape and Urban Planning 124 (2014) 14–21
ea: Po
tcd(cesnes1
mlaostiO
TS
Fig. 1. Study ar
o determine distance, so the increase in average distance to thelosest green street facility from 4408 ft (1344 m) using Euclideanistance to 5,894 feet (1797 m) using street network is expectedTable 2). Network Analyst Extension (ArcGIS 10) was used to cal-ulate the abundance of green streets within a ¼-mile distance ofach property using the street grid—the average number of greentreets using this approach is 0.35 which is less than the averageumber of facilities (0.90) within a ¼-mile buffer around each prop-rty. The average age of the nearest green street, using Euclidean ortreet network distance, is 1.10 years although some facilities were2 years old when properties transacted.
Land cover information on a property, and within 200-foot, 1/4-ile, and 1/2-mile buffers, was derived using a high resolution
and cover layer that classifies each 3 ft × 3 ft (0.91 m × 0.91 m) cells high structure vegetation, low structure vegetation, impervi-us surface, or water (Metro Data Resource Center, 2007). High
tructure vegetation includes trees, while low structure vegeta-ion includes grass, shrubs, and small trees. Impervious surfacencludes hard surfaces such as roads, rooftops, and driveways.n average, 44.36% of the lots in our database are covered byable 1tructural, property, and neighborhood variables.
Variable name Definition Units
Price Real sale price $ (2007)
Lotsqft Lot square footage Square FeBldsqft Building square footage Square FeFullBath Total full bathrooms Count
HalfBath Total half bathrooms Count
Fireplaces Total fireplaces Count
Age Age of building Count
Elevation Elevation of property Feet
Income Median income at census tract $ (2000)
%White Percentage white at census tract ProportioNDist Distance from properties in N Portland to CBD Feet
NEDist Distance from properties in NE Portland to CBD Feet
SWDist Distance from properties in SW Portland to CBD Feet
SEDist Distance from properties in SE Portland to CBD Feet
rtland, Oregon.
impervious surfaces such as roofs, driveways, and patios; the per-centage of impervious surface increases to over 46% for surroundingbuffers.
Project researchers visited 254 of the 262 green street facili-ties that were closest to the properties in our data set based onstreet network distance and recorded detailed characteristics foreach facility. Most plants were identified to species; otherwise, afamily or genus was recorded. The percentage of the facility coveredby plants and trees was recorded (Anderson, 1986) as this metrichas been suggested as a more meaningful measure of vegetationthan individual counts in an urban setting (Wise et al., 2010). Facil-ities were categorized into one of four types: sidewalk bioswale,grassy bioswale, curb extension, or corner curb extension. Side-walk bioswales are located between sidewalks and streets. Grassybioswales are similar to sidewalk bioswales but have 70% or moreof the area between the sidewalk and street covered by grass. Curb
extensions extend from the curb into the street, are usually largerthan sidewalk bioswales, and are designed to calm traffic. Cornercurb extensions are similar to curb extensions, but they extend tothe corner of the street. Average facility size is approximately 465Mean Standard deviation Minimum Maximum
307,553 162,766 58,921 3,224,974et 6624 5219 808 365,750et 1904 809 396 12,061
1.58 0.66 0 60.29 0.49 0 60.82 0.67 0 658.66 30.63 0 137487.28 171.45 25 1,04045,389 16,546 10,000 136,102
n 0.78 0.14 0.26 0.9724,330 6316 7987 37,82124,394 9310 7591 49,15019,321 7070 3415 34,87328,360 10,592 6093 52,089
N.R. Netusil et al. / Landscape and Urban Planning 124 (2014) 14–21 17
Table 2Summary statistics: land cover and green street distance and abundance.
Variable name Description Units Mean Standarddeviation
Min Max
EucDist Euclidean distance from a property to nearest green street Feet 4408 2848 14 16,234NetDist Street network distance from a property to nearest green street Feet 5894 3638 0 22,257EucAge Age of nearest green street facility using Euclidean distance Count 1.10 1.96 0 12NetAge Age of nearest green street facility using street network Count 1.10 2.03 0 12GSBlock Number of green street facilities in a property’s census block Count 1.23 7.02 0 137GSTract Number of green street facilities in a property’s census tract Count 3.01 11.92 0 137EucGSQtrMile Number of green street facilities within a ¼ mile of property (Euclidean) Count 0.90 7.05 0 172NetGSQtrMile Number of green street facilities within a 1/4 mile of property (Street Network) Count 0.35 2.61 0 59GSNeigh Number of green streets in a property’s neighborhood Count 6.79 19.70 0 179High High structure vegetation on property Proportion 0.2615 0.2181 0 1Low Low structure vegetation on property Proportion 0.2948 0.1902 0 0.8926Imperv Impervious area on property Proportion 0.4436 0.1941 0 1High200 High structure within 200-foot buffer Proportion 0.2545 0.1352 0 0.9921Low200 Low structure within 200-foot buffer Proportion 0.2845 0.0980 0 0.7801Imperv200 Impervious within 200-foot buffer Proportion 0.4607 0.1251 0 0.9664High1–4 High structure within 1/4-mile buffer Proportion 0.2543 0.1191 0.0243 0.8902Low1–4 Low structure within 1/4-mile buffer Proportion 0.2780 0.0745 0.0233 0.6749Water1–4 Water within 1/4-mile buffer Proportion 0.0022 0.0181 0 0.5310Imperv1–4 Impervious within 1/4-mile buffer Proportion 0.4655 0.1150 0.0434 0.9278High1–2 High structure within 1/4-mile buffer Proportion 0.2537 0.1131 0.0562 0.8355
s3oeptmaasi(mc
2
captiPaftTesitm
TG
Low1–2 Low structure within 1/4-mile buffer
Water1–2 Water within 1/4-mile buffer
Imperv1–2 Impervious within 1/4-mile buffer
quare feet (43.2 square meters) with the largest facility measuring000 square feet (279 square meters) (Table 3). The vast majorityf property transactions, over 95%, have tree canopy in the near-st green street facility; tree canopy coverage averaged across allroperties is 52%. Dams are a raised barrier inside a green streethat slows the movement of water within the facility. They are
ade from a variety of materials (soil, rocks, or concrete), provide pathway for people to cross from the road to the sidewalk, andre typically associated with more complex planting designs. Thesetructures are present in 11.30% of the green street facilities lead-ng to an overall average number of dams of 0.31. The vast majority77.64%) of facilities are sidewalk bioswales. The next most com-
on facility type is grassy bioswale (11.16%) followed by cornerurb extensions (5.80%) and curb extensions (5.40%).
.3. Methodological approach
We use the hedonic price method, a well-established statisti-al technique, to estimate if green street proximity, abundance,nd characteristics affect the sale price of single-family residentialroperties in our study area. Because the functional form is uncer-ain (Freeman, 2003), we use a semi-log functional form, whichs quite common in the literature (Donovan & Butry, 2010; Kim,hipps, & Anselin, 2003; Mahan et al., 2000; Ward et al., 2008). Inddition to green street variables, all model specifications controlor home characteristics, location, and other environmental fea-ures on each property and in surrounding buffers (Tables 1 and 2).he relationship between several variables (Lotsqft, Bldsqft, Age,tc.) and sale price likely changes with increases in those variables,
o we use quadratic terms to capture those effects. We correct fornflation by using the real sale price as our dependent variable, butime trends may be an important factor, so we include thirty-fiveonth dummy variables in all specifications.
able 3reen street characteristics, N = 29,644.
Variable name Description Units
Vegetation taxa Number of taxa in a facility Count
Tree canopy Area of green street facility covered by trees ProportionDams Number of dams in facility Count
Area Area of nearest green street Square fee
Proportion 0.2715 0.0662 0.0351 0.6315Proportion 0.0085 0.353 0 0.5962Proportion 0.4663 0.1131 0.0661 0.7556
Many researchers (Anderson & West, 2006; Lutzenhiser &Netusil, 2001; Mahan et al., 2000) use Euclidean distance to mea-sure proximity to environmental amenities, but recent hedonicresearch on open spaces emphasizes the superiority of mea-suring distance based on street network (Sander, Ghosh, vanRiper, & Manson, 2010). Because green street facilities arelocated next to roads and sidewalks we believe that street net-work distance will provide a more accurate representation than“straight-line” Euclidean distance but we report results using bothapproaches.
Green street facilities may change the “greenness” of a neigh-borhood by increasing the amount of grass, shrubs, and trees near aproperty. Facility abundance may be an important factor in explain-ing a property’s sale price, but the appropriate scale for measuringabundance in unknown. We explore this question by including thenumber of green street facilities measured at the census block, cen-sus tract, 1/4 mile Euclidean buffer, 1/4 mile street network, andneighborhood scale. Because we expect property owners to observethese facilities while walking or driving, we prefer, a priori, the 1/4mile street network measurement based on research by Alshalalfahand Shalaby (2007) and Iacono, Krizek, and El-Geneidy (2010) thatempirically demonstrates preferred walking distance measures forurban areas of the United States.
Spatial relationships (spatial lag or spatial error) are testedand corrected for in all models. Failing to correct for spatial rela-tionships may produce biased coefficients if spatial lag is presentor biased standard errors and t-statistics if spatial error exists.We follow the process described in Anselin (2005) and first testfor spatial relationships by computing LM error and LM lag val-
ues using a 4-nearest neighbor and 8-nearest neighbor weightsmatrix. The LM error and LM lag test statistics were significantin all models, so robust LM error and robust LM lag tests wereperformed.Mean Standard deviation Minimum Maximum
4.13 2.35 0 15 0.52 0.28 0 1
0.31 1.07 0 8t 464.58 392.74 52 3000
1 nd Urb
3
gutnofd
tifittsigoasiiiaeobCttdt1tdss
asms4yapim2
ifg4spotipst
8 N.R. Netusil et al. / Landscape a
. Results
Our models incorporate distance from a property to the nearestreen street, and the abundance of green streets within 1/4 mile,sing Euclidean distance (Models 1 and 2) or street network dis-ance (Models 3 and 4). Model 4 includes characteristics of theearest green street facility such as facility type, the proportionf the facility covered by tree canopy, facility size, and landscapeeatures such as the number of taxa and the total number ofams.
Home characteristics are of the expected sign and magni-ude across specifications—a property’s sale price is estimated toncrease at a diminishing rate as lot size and building squareootage increase. Additional full and half bathrooms, increasesn elevation (a proxy for views), and neighborhood characteris-ics such as percentage white and median income at the censusract level, are also found to have a significantly positive effect onale price. Land cover variables on a property and in surround-ng buffers are included to avoid omitted variable bias becausereen streets are often located in areas with a high percentagef impervious surface area and land cover has been found to ben important factor influencing the sale price of properties in thetudy area (Kadish & Netusil, 2011). Tree canopy on a property, andn surrounding buffers, is found to have a positive but diminish-ng effect on a property’s sale price; water, which is only presentn the 200-foot to 1/4 mile and 1/4 mile to 1/2 mile buffers, has
large and significantly positive effect on sale price in all mod-ls. Robust LM error and Robust LM lag tests provide evidencef spatial lag with the 8 nearest neighbor (NN) model having theest fit, so results from that specification are presented in Table 4.orrecting for spatial lag changes the sign and significance ofhe green street proximity and abundance variables compared tohe OLS results (Model 1). Modeling proximity using street networkistance (Models 3 and 4) is preferred, a priori, because these facili-ies are located on streets; the proximity terms are significant at the% level in Models 3 and 4, but the abundance terms are not statis-ically significant. Model 4 was also estimated using the approachescribed in Kelejian and Prucha (1998). The estimated magnitude,ign and significance were similar to the spatial lag (8NN model),o only results from the spatial lag model are presented in Table 4.
Table 5The magnitude and significance of the variables capturing
ge of the nearest green street are fairly consistent acrosspecifications—the age of the nearest green street facility isodeled as a quadratic, so increases in the age of a nearest green
treet facility are estimated to decrease a property’s sale price until.4 (Model 1), 4.69 (Model 2), 4.85 (Model 3), and 3.54 (Model 4)ears after which additional increases in age are estimated to have
positive effect on a property’s sale price. Age may be serving aroxy for the maturity and density of vegetation since fast grow-
ng trees in the study area reach full maturity in ten years withoderate growing trees maturing in 10–20 years (McPherson et al.,
002).Model 4 includes characteristics of the nearest green street facil-
ty. Detailed information is available for 254 of the 262 green streetacilities in our data set. The matching of property sales to nearestreen street facility resulted in the loss of 68 observations, so Model
uses information on 29,644 transactions. Facility types were nottatistically significant, but facility characteristics such as the pro-ortion of the facility covered by trees, facility size, taxa and numberf dams are significant. The estimated coefficient on the quadraticree canopy variable (Tree Canopy Squared) is not significant, so
ncreasing tree canopy is estimated to increase a property’s salerice. Each additional dam is estimated to increase a property’sale price by 0.60% and increasing the size of a facility is estimatedo decrease a property’s sale price up to a facility size of 1,344an Planning 124 (2014) 14–21
square feet (125 square meters) past which increasing facility sizeis estimated to increase a property’s sale price.
Statistically significant estimated coefficients from Model 4were used to calculate the implicit prices of the full, indirect anddirect effects of proximity and green street characteristics usingthe methodology described in Small and Steimetz (2012). Using themean values reported in Table 2, an increase in distance of 1 footaway from the nearest green street facility is estimated to increasea property’s sale price by $0.30 of which $0.20 is a direct effect and$0.10 is an indirect effect. A 10 percentage point increase in treecanopy at the closest green street facility is estimated to increase aproperty’s sale price by $18,707 of which $12,590 is a direct effectand $6117 is an indirect effect. The mechanism by which the saleprice of a property affects the sale price of nearby properties deter-mines if the full or direct effect is the most appropriate welfaremeasurement. If property owners use the sale price of nearby prop-erties to inform their offer on a property then only the direct effectshould be used (Small & Steimetz, 2012).
The appropriate spatial scale for measuring the abundance ofgreen streets near a property is unknown, so we explore fivedifferent abundance measures—census block, census tract, 1/4mile Euclidean distance, 1/4 mile street network distance, and aproperty’s neighborhood using the Model 3 specification (streetnetwork, spatial lag, 8NN). The 1/4 mile Euclidean distance is thesmallest area covering 0.21 square miles (54.39 ha), followed bycensus block at 0.51 square miles (132.09 ha), neighborhood at1.57 square miles (406.63 ha), and census tract at 1.73 square miles(448.07 ha).
The linear distance coefficient (NetDist) is positive in all speci-fications although not significant for census tract (p-value = 0.13);the quadratic distance coefficient is significant and negative in allspecifications. The economic magnitude of proximity, however, issmall—increasing a property’s distance from a green street by 1000feet (305 meters) is estimated to increase its sale price by $430for the 1/4 mile street network model and $851 for the 1/4 mileEuclidean model, which equals 0.14% (1/4 mile street distance) and0.28% (1/4 mile Euclidean) of the mean sale price of properties inour data set.
Signs on the linear and quadratic abundance terms are con-sistent across specifications although only the census block andcensus tract coefficients are statistically significant. We estimate apositive effect on sale price at the census tract level when the num-ber of facilities exceeds 143 (133 for census block). The averagefacility size is approximately 465 square feet (43.2 square meters),which translates into building, in aggregate, 66,495 square feet(6,278 square meters) (1.53 acres or 0.0024 square miles (0.619 ha))of green street facilities in a census tract. The average size of cen-sus tracts in our study area is 1.73 square miles (448.07 ha), so thisamounts to covering around 0.14% of the census tract with greenstreet facilities.
4. Discussion
Our study uses the hedonic price method to investigate ifproximity, abundance and characteristics of green street facilitieshave an effect on the sale price of nearby properties. The resultssuggest that our study goes beyond the existing literature by pro-viding three insights. First, from a research methods perspective,the best way to measure proximity to a green street facility wasuncertain, so we explored two alternatives: street network andEuclidean. We found a statistically significant but small effect on
a property’s sale price as distance from the nearest green streetfacility increases (Models 3 and 4, Table 4; Table 6). Estimatedcoefficients for street network were significant in all models (Mod-els 3 and 4, Table 4; Table 6) while the Euclidean measures were not
N.R. Netusil et al. / Landscape and Urban Planning 124 (2014) 14–21 19
Table 4Regression results, t-statistics in parentheses.
Variable Model 1: Euclidean OLS Model 2: EuclideanSpatial Lag (8 NN)
Model 3: Street NetworkSpatial Lag (8 NN)
Model 4: Street NetworkTaxonomy Spatial Lag (8 NN)
GSDist −1.25 E−6(−0.86) 9.80 E−8(0.07) 3.16 E−6***(3.0) 2.84 E−6***(2.63)GSDist2 2.00 E−10*(1.71) 5.74 E−11(0.49) −1.94 E−10***(−3.0) −1.85 E−10***(−2.81)GSQtrMile −1.59 E−3***(−3.18) −7.90 E−4*(−1.66) 1.50 E−4(0.1) −7.56 E−4(−1.48)GSQtrMile2 1.11 E−5***(2.73) 2.95 E−6(0.76) −1.53 E−5(−0.6) 2.60 E−6(0.61)GSAge −5.96 E−3***(−3.39) −5.89 E−3***(−3.52) −4.87 E−3***(−2.90) −3.92 E−3**(−2.23)GSAge2 6.74 E−4***(3.54) 6.28 E−4***(3.47) 5.02 E−4***(2.80) 5.53 E−4***(3.04)Corner curb extension −2.93 E−3 (−0.49)Curb extension −4.35 E−3(−0.95)Grassy bioswale −1.44 E−3(−0.24)Tree canopy 4.30 E−2**(2.02)Tree canopy squared −2.65 E−2(−1.30)Facility size −3.97 E−2***(−4.37)Facility size squared 1.48 E−8***(3.13)Taxa 4.39 E−3**(2.36)Taxa squared −2.28 E−4(−1.55)Dam 5.31 E−3***(4.00)Lotsqft 8.19 E−6***(20.58) 8.13 E−6***(21.46) 8.12 E−6***(21.4) 8.17 E−6***(21.56)Lotsqft2 −2.04 E−11***(−10.38) −1.88 E−11***(−10.05) −1.88 E−11***(−10.00) −1.89 E−11***(−10.10)Bldgsqft 2.46 E−4***(54.58) 2.24 E−4***(51.94) 2.24 E−4***(52.00) 2.23 E−4***(51.78)Bldgsqft2 −1.15 E−8***(−15.17) −1.10 E−8***(−15.34) −1.10 E−8***(−15.30) −1.09 E−8***(−15.14)FullBath 0.096***(39.57) 0.090***(39.05) 0.090***(39.10) 0.090***(39.08)HalfBath 0.059***(20.12) 0.052***(18.82) 0.052***(18.80) 0.052***(18.73)Fireplaces 0.027***(12.41) 0.017***(8.43) 0.017***(8.40) 0.017***(8.19)Age −1.29 E−3***(−7.65) −1.31 E−3***(−8.16) −1.31 E−3***(−8.10) −1.40 E−3***(−8.66)Age2 8.89 E−6***(5.91) 8.94 E−6***(6.25) 8.90 E−6***(6.20) 9.68 E−5***(6.75)Elevation 9.20 E−5***(5.33) 6.62 E−5***(4.04) 6.18 E−5***(3.80) 6.50 E−5***(3.94)Income 2.61 E−6***(21.81) 1.30 E−6***(11.31) 1.30 E−6***(11.40) 1.29 E−6***(11.27)White 0.3005***(24.11) 0.198***(16.63) 0.2016***(16.90) 0.1817***(14.59)NE 0.2963***(6.42) 0.1223***(3.52) 0.1212***(3.50) 0.1680***(4.73)SW 0.1222**(2.56) 0.836**(2.51) 0.824**(2.50) 0.1132***(3.30)SE 0.0883*(1.92) 0.0217(0.67) 0.0174(0.50) 0.0284(0.85)NDist −4.35 E−5***(−11.77) −2.71 E−5***(−12.22) −2.75 E−5***(−12.20) −2.27 E−5***(−9.78)NDist2 6.56 E−10***(8.74) 4.36 E−10***(9.08) 4.40 E−10***(9.10) 3.45 E−10***(6.77)NEDist −5.81 E−5***(−42.46) −3.07 E−5***(−23.04) −3.10 E−5***(−23.30) −3.04 E−5***(−22.05)NEDist2 7.53 E−10***(31.57) 3.67 E−10***(15.92) 3.71 E−10***(16.10) 3.62 E−10***(15.23)SWDist −6.47 E−5***(−21.825) −3.99 E−5***(−13.98) −4.04 E−5***(−14.20) −3.70 E−5***(−12.15)SWDist2 11.89 E−10***(16.27) 7.22 E−10***(10.30) 7.33 E−10***(10.50) 6.27 E−10***(8.20)SEDist −4.70 E−4***(−46.79) −2.60 E−5***(−26.20) −2.59 E−5***(−26.10) −2.34 E−5***(−21.46)SEDist2 5.71 E−10***(35.25) 2.98 E−10***(18.88) 2.95 E−10***(19.70) 2.61 E−10***(15.30)High 0.0605***(3.50) 0.0572***(3.48) 0.0581***(3.50) 0.0567***(3.44)High2 −0.1239***(−5.52) −0.1148***(−5.37) −0.1156***(−5.40) −0.1131***(−5.29)Low −0.0038(−0.17) −0.005(−0.24) −0.0044(−0.20) −0.0012(−0.06)Low2 −0.0742**(−2.45) −0.0706**(−2.44) −0.0709**(−2.50) −0.0749***(−2.59)High200 0.2000***(5.59) 0.1872***(5.50) 0.1881***(5.50) 0.1900***(5.58)High200Sqr −0.1392***(−2.67) −0.1503***(−3.03) −0.1492***(−3.00) −0.1567***(−3.15)Low200 0.3918***(6.00) 0.3762***(6.06) 0.3804***(6.10) 0.3704***(5.94)Low200Sqr −0.3951***(−3.85) −0.3759***(−3.85) −0.3811***(−3.90) −0.3704***(−3.80)High1–4 0.0019(0.03) −0.1227**(−2.07) −0.1213**(−2.00) −0.1085*(−1.82)High1–4Sqr 0.3874***(4.40) 0.3204***(3.83) 0.3283***(3.90) 0.3186***(3.80)Low1–4 0.2856**(2.22) 0.187(1.53) 0.1851(1.50) 0.1798(1.47)Low1–4Sqr −0.1919(−0.93) −0.1864(−0.95) −0.1803(−0.90) −0.1671(−0.85)Water1–4 0.3928***(4.90) 0.2357***(3.08) 0.2424***(3.20) 0.2490***(3.25)High1–2 0.5381***(8.07) 0.3381***(5.33) 0.3483***(5.50) 0.3521***(5.51)High1–2Sqr −0.6359***(−6.78) −0.4842***(−5.44) −0.4923***(5.50) −0.5080***(−5.67)Low1–2 1.0637***(7.01) 0.6775***(4.689) 0.6901***(4.80) 0.6404***(4.39)Low1–2Sqr −1.6037***(−6.29) −1.06***(−4.37) −1.0594***(−4.40) −1.0025***(−4.09)Water1–2 0.3967***(8.83) 0.2396***(5.59) 0.2508***(5.80) 0.2439***(5.55)R2 0.7675 0.7721 0.772 0.7735Rho 0.326 0.325 0.37Likelihood 1.71 E 4 1.71 E 4 1.71 E 4Observations 29,712 29,712 29,712 29,644
Table 5Implicit prices for the full, indirect, and direct effects of green street characteristics.
Variable Full effect ($) Full effect as a % ofsale price
Indirect effect Indirect effect as a% of sale price
Direct effect Direct effect as a %of sale price
Green street distance (1 additional footincrease in distance from facility)
$0.30 9.79E−5% $0.10 3.87E−5% $0.20 6.59E−5%
Green street age (1 additional year) −$978 −0.32% −$320 −0.13% −$659 −0.21%Tree canopy (increase of .10) $18,707 6.08% $6,117 1.99% $12,590 4.09%Facility size (1 square foot increase) −$11.85 −3.85E−3% −$3.88 −1.26E−3% −$7.98 −2.59E−3Taxa (additional 1) $2,011 0.65% $657 0.26% $1,353 0.44%Dam (additional 1) $2,433 0.79% $796 0.31% $1,637 0.53%
20 N.R. Netusil et al. / Landscape and Urban Planning 124 (2014) 14–21
Table 6Abundance measures—spatial lag 8 nearest neighbors.
Census block Census tract 1/4-mile(Euclidean)
1/4-mile(StreetNetwork)
Neighbor-hood
NetDist 2.12 E−6**(2.0) 1.53 E−6(1.5) 2.60 E−6**(2.4) 3.16 E−6***(3.0) 2.78 E−6***(2.7)NetDist2 −1.52 E−10**(−2.4) −1.29 E−10**(−2.0) −1.67 E−10***(−2.6) −1.94 E−10***(−3.0) −1.78 E−10***(−2.8)Abundance measure −2.00 E−3***(−5.3) −1.57 E−3***(−6.1) −5.71 E−4(−1.2) 1.50 E−4(0.1) −1.56 E−4(−0.10)Abundance measure Squared 1.51 E−5***(4.6) 1.10 E−5***(5.2) 1.54 E−6(0.4) −1.53 E−5(−0.6) 9.00 E−8(0.01)NetAge −4.12 E−3**(−2.4) −3.89 E−3**(−2.3) −4.80 E−3***(−2.8) −4.87 E−3***(−2.9) −4.67 E−3***(−2.8)NetAge2 4.27 E−4**(2.4) 4.07 E−4**(2.3) 4.94 E−4***(2.8) 5.02 E−4***(2.8) 4.85 E−4***(2.7)
smmmoti
cDiltes$eatttasui
dpgatpacmcfrf2c
istfpfAicT
R2 0.7725 0.7727
Rho 0.324 0.324
Observations 29,712 29,712
ignificant in either model (Models 1 and 2, Table 4). We prefer theodels that use street network based on the significance of esti-ated coefficients and our understanding that these facilities areost likely to be encountered while driving or walking. This rec-
mmendation is consistent with Sander et al. (2010) who concludehat street network is a better for modeling open space proximityn urban areas.
Second, we find that property sale prices increase when treeanopy coverage at the nearest green street facility increases.onovan and Butry (2010) estimate that existing street trees
ncrease the sale price of properties located on the east side of Port-and, Oregon by about 3% based on the median sale price. Usinghe median sale price ($263,624) for properties in our data set westimate the full effect of existing tree canopy coverage in greentreet facilities is $11,583 (4.39% of median sale price) of which5955 (2.26%) is a direct effect and $5628 (2.13%) is an indirectffect—estimated effects that are similar in magnitude to Donovannd Butry (2010). Our results go further, however, by suggestinghat planting trees within green street facilities can help overcomehe market failure identified in Donovan and Butry (2010) — thathe benefits of street trees outweigh their maintenance cost — via
public investment in green infrastructure. As municipalities con-ider green infrastructure, gaining public support may be easiersing an argument that vegetation in green street facilities may
ncrease a property’s sale price.Related to location of vegetation is our finding that the abun-
ance of facilities is an important factor with positive effects on salerice once 0.14% (land area) or more of a census tract is covered byreen street facilities. While 0.14% may represent a small percent-ge of the overall landscape in a census tract, finding suitable areashat are not privately owned, do not impact existing water/sewer orower facilities, and yet are acceptable by census tract residents inn already highly urbanized setting may prove challenging. Whilereating large scale plans for green infrastructure, albeit expensive,ay be an important part of program design, one formidable obsta-
le for urban planners is the need to create a case-by-case scenarioor each neighborhood for effective implementation. Extensive out-each at a neighborhood level has, nevertheless, proven effectiveor implementing large scale stormwater facilities (Shandas et al.,012), yet many municipalities may need extensive resources toreate similar outreach programs.
Finally, facility characteristics such as the number of taxa, facil-ty area, age and number of dams were found to have statisticallyignificant effects suggesting that it is important to create facili-ies with a diversity of taxa and structural complexity. In addition,acilities that were older than around 4 years were found to have aositive effect on sale price, which may be due to the fact that olderacilities may contain mature vegetation and structural complexity.
t the same time, older facilities may also become less functionaln terms of draining water, so facility maintenance is an importantonsideration (Le Coustumer, Fletcher, Deletic, & Barraud, 2007).he effect on sale price from increasing taxa could enhance other
0.7721 0.7720 0.77210.326 0.325 0.32529,712 29,712 29,712
ecosystem services such as providing habitat for macroinverte-brates that require diverse natural spaces (Kazemi et al., 2009).
The estimated effects reflect use values only, are specific to thetime period of our analysis and study area, and may change overtime as the housing market evolves and as property owners learnmore about these facilities. Caution must be used when interpretingcoefficients in hedonic studies because of omitted variable bias andendogeneity. The time period of our analysis (2005–2007) includesa period of strong housing demand in the study area, so it is impor-tant to consider market conditions when interpreting our results(Coulson and Zabel, 2013). Nevertheless, the implications of thesefindings provide some of the first evidence that the quantity andcharacteristics of green street facilities in urban neighborhoods canhave profound implications for the process and outcomes of land-scape planning programs. One important and summative findingfrom this study is that developing large scale green infrastruc-ture programs can prove beneficial for residents in terms of realgains in property sale prices, yet such programs also depend onthe abundance and characteristics of facilities, their location, andthe concomitant outreach that ensures that these facilities will bemaintained throughout the life of each facility.
5. Conclusion
The EPA estimates that between $331 and $450 billion of invest-ment is needed over a 20-year period (2000–2019) to replaceor update the existing sewer infrastructure in the United States(Environmental Protection Agency, 2002). Low-impact develop-ment, which includes bioswales, rain gardens, pervious pavers,and green streets, is being promoted by the EPA (EnvironmentalProtection Agency, 2010b), and adopted by cities throughoutthe United States, as a potentially cost-effective way to reducestormwater runoff. Since many low-impact development facilitiesare highly visible — unlike the traditional underground pipe net-work — they have the potential to improve neighborhood aestheticsin addition to addressing the stormwater challenges facing manycities.
Portland, Oregon has a long history of developing low-impactdevelopment facilities through its ‘green streets’ program andcontinues to promote these facilities as improving air and waterquality, enhancing pedestrian and bicycle access and safety,expanding the amount of urban green space and wildlife habi-tat, and increasing property values (Environmental Services Cityof Portland, 2008). Findings from the present study provide sup-port for some of these claims but challenge others. For example,we find a small but statistically significant increase in a property’ssale price as distance from the nearest facility increases. Facilitytype (sidewalk bioswale, corner curb extension, curb extension, or
grassy bioswale) is not found to have a statistically significant effecton a property’s sale price, but a facility’s size, age, design complex-ity, tree canopy coverage, and abundance of facilities at the censustract or census block level do have a statistically significant effect.
d Urb
rcoieabH
A
CSJW
R
A
AA
A
B
B
BC
C
D
E
E
EE
E
EEE
E
F
F
H
I
K
K
to storm-water management and land development. Journal of Water Resources
N.R. Netusil et al. / Landscape an
If predictions that cities will face increasingly severe storms andelated impacts are correct, then municipal managers need to findost-effective and publicly acceptable approaches to reduce the riskf flooding and other stormwater impacts. Our research providesnsight into how these facilities affect the sale price of nearby prop-rties, yet we still have a limited understanding about the socialcceptance of these facilities (Shandas et al., 2010, 2012) and theiriophysical impacts (Booth & Jackson, 1997; Booth et al., 2004; May,orner, Karr, Mar, & Welch, 1997).
cknowledgements
Funding provided by an ULTRA-Ex NSF award #0948983, a Reedollege Opportunity Grant, and a grant from the Reed Collegetillman-Drake Fund. Sadie Carney, Alex Desroches, Donnych Diaz,onathan Kadish, and Bethany Pratt assisted with data generation.
e thank Jon Rork for providing helpful comments.
eferences
lshalalfah, B. W., & Shalaby, A. S. (2007). Case study: Relationship of walk accessdistance to transit with service, travel, and personal characteristics. Journal ofUrban Planning and Development, 133, 114.
nderson, E. W. (1986). A guide for estimating cover. Rangelands, 236–238.nderson, S. T., & West, S. E. (2006). Open space, residential property values, and
spatial context. Regonal Science and Urban Economics, 36(6), 773–789.nselin, L. (2005). Exploring spatial data with GeoDa: A workbook. Urbana, IL: Univer-
sity of Illinois Urbana-Champaign.ooth, D. B., & Jackson, C. R. (1997). Urbanization of aquatic systems: Degradation
thresholds, stormwater detection, and the limits of mitigation. Journal of theAmerican Water Resources Association, 33(5), 1077–1090.
ooth, D. B., Karr, J. R., Schauman, S., Konrad, C. P., Morley, S. A., Larson, M. G.,et al. (2004). Reviving urban streams: Land use, hydrology, biology, and humanbehavior. Journal of the American Water Resources Association, 40(5), 1351–1364.
ureau of Labor Statistics. (2011). Consumer price indexes overview.ity of Los Angeles Stormwater Program. (2011). Draft Stormwater low impact devel-
opment ordinance. Los Angeles: City of Los Angeles Stormwater Program.oulson, N. E., & Zabel, J. E. (2013). What Can We Learn from Hedonic Models When
Housing Markets Are Dominated by Foreclosure? Annual Review of ResourceEconomics, 5(1), 261–279.
onovan, G. H., & Butry, D. T. (2010). Trees in the city: Valuing street trees in Portland,Oregon. Landscape Urban Planning, 94, 77–83.
nvironmental Protection Agency. (2002). The clean water and drinking water infra-structure gap analysis.
nvironmental Protection Agency. (2010a). Green infrastructure case studies: Munic-ipal policies for managing stormwater with green infrastructure.
nvironmental Protection Agency. (2010b). Low impact development (LID).nvironmental Services City of Portland. (2008). Stormwater management for clean
rivers: Green Streets.nvironmental Services City of Portland. (2010). Green Street facilities in Portland,
Oregon.nvironmental Services City of Portland. (2010a). What is grey to Green?nvironmental Services City of Portland. (2010b). Grey to Green elements.nvironmental Services City of Portland. (2011). Combined sewer overflow program
progress report: January 2011 (Environmental Services, ed.). City of Portland.wing, R., & Dumbaugh, E. (2009). The built environment and traffic safety: A review
of empirical evidence. Journal of Planning Literature, 23(4), 347–367.rank, R. (2011). Why do Portland water customers pay so much? Because of big, needed
projects—And small ones that some question. Portland: The Oregonian.reeman, A. M. I. (2003). The measurement of environmental and resource values:
Theory and methods. Washington, DC: Resources for the Future.ennings, L., & Soll, J. (2010). Wildlife corridors and permeability: A literature review.
Metro, Portland: Metro Sustainability Center.acono, M., Krizek, K. J., & El-Geneidy, A. (2010). Measuring non-motorized accessi-
bility: Issues, alternatives, and execution. Journal of Transport Geography, 18(1),133–140.
adish, J., & Netusil, N. R. (2011). Valuing vegetation in an urban watershed. Land-scape Urban Planning, 104(1), 59–65.
azemi, F., Beecham, S., Gibbs, J., & Clay, R. (2009). Factors affecting terrestrial inver-tebrate diversity in bioretention basins in an Australian urban environment.Landscape Urban Planning, 92(3), 304–313.
an Planning 124 (2014) 14–21 21
Kelejian, H. H., & Prucha, I. R. (1998). Generalized spatial two-stage least squaresprocedure for estimating a spatial autoregressive model with autoregressivedisturbances. Journal Real Estate Finance and Economics, 17(1), 99–121.
Kim, C. W., Phipps, T. T., & Anselin, L. (2003). Measuring the benefits of air qualityimprovement: A spatial hedonic approach. Journal of Environmental Economicsand Management, 45, 24–39.
Le Coustumer, S., Fletcher, T., Deletic, A., & Barraud, S. (2007). Hydraulic performanceof biofilters for stormwater management: First lessons from both laboratory andfield studies. Water Science Technology, 56(10), 93–100.
Leggett, C. G., & Bockstael, N. E. (2000). Evidence of the effects of water qualityon residential land prices. Journal of Environmental Economics and Management,39(2), 121–144.
Lutzenhiser, M., & Netusil, N. R. (2001). The effect of open spaces on a home’s saleprice. Contemporary Economic Policy, 19(3), 291–298.
Maas, J., Verheij, R. A., de Vries, S., Spreeuwenberg, P., Schellevis, F. G., & Groenewe-gen, P. P. (2009). Morbidity is related to a green living environment. Journal ofEpidemiology Community Health, 63(12), 967–973.
Mahan, B. L., Polasky, S., & Adams, R. M. (2000). Valuing urban wetlands: A propertyprice approach. Land Economics, 76(1), 100–113.
May, C. W., Horner, R. R., Karr, J. R., Mar, B. W., & Welch, E. B. (1997). Effects of urban-ization on small streams in the Puget Sound ecoregion. Watershed ProtectionTechniques, 2(4), 483–494.
McPherson, E. G., Maco, S. E., Simpson, J. R., Peper, P. J., Xiao, Q., & VanDerZanden, A.M. (2002). Western Washington and Oregon community tree guide: Benefits, costs,and strategic planting. Center for Urban Forest Research, USDA Forest Service,Pacific Southwest Research Station.
Metro Data Resource Center. (2007). 2007 high resolution landcover—Shapefile ver-sion.
Metro Data Resource Center. (2009). RLIS lite. METRO Data Resource Center.Nelson, E., Mendoza, G., Regetz, J., Polasky, S., Tallis, H., Cameron, D. R., et al. (2009).
Modeling multiple ecosystem services, biodiversity conservation, commodityproduction, and tradeoffs at landscape scales. Frontiers in Ecology and the Envi-ronment, 7(1), 4–11.
Netusil, N. R. (2006). Economic valuation of riparian corridors and upland wildlifehabitat in an urban watershed. Journal of Contemporary Water Research & Edu-cation, 134, 39–45.
Netusil, N. R., Chattopadhyay, S., & Kovacs, K. (2010). Estimating the demand fortree canopy: A second-stage hedonic price analysis in Portland, Oregon. LandEconomics, 86(2), 281–293.
NOAA Fisheries Office of Protected Resources. (2010). Marine/anadromous fishspecies.
Oregon Department of Environmental Quality. (2006). Portland harbor superfund site.Poor, P. J., Pessagno, K. L., & Paul, R. W. (2007). Exploring the hedonic value of ambi-
ent water quality: A local watershed-based study. Ecological Economics, 60(4),797–806.
Revkin, A. (2008). Portland again tops a sustainable cities list. The New York Times.Sander, H. A., Ghosh, D., van Riper, D., & Manson, S. M. (2010). How do you measure
distance in spatial models? An example using open-space valuation. Environ-ment and Planning B: Planning and Design, 37(5), 874–894.
Shandas, V., Nelson, A., Arendes, C., Cibor, C. (2010). Tabor to river: An evaluation ofoutreach efforts and opportunities for engaging residents in stormwater manage-ment.
Shandas, V., Steele, R., & Nelson, A. (2012). Tabor to the River program. City of Portland:Bureau of Environmental Services.
Small, K. A., & Steimetz, S. S. (2012). Spatial hedonics and the willingness to pay forresidential amenities. Journal of Regional Sciences, 52(4), 635–647.
State of Washington Department of Ecology. (2012). Municipal stormwater generalpermit reissuance.
U.S. Environmental Protection Agency. (2010). Green infrastructure case studies:Municipal policies for managing stormwater control with green infrastructure.Washington, DC: U.S. Environmental Protection Agency Office of Wetlands,Oceans and Watersheds.
U.S. Environmental Protection Agency. (2011). A strategic agenda to protect watersand build more livable communities through infrastructure. Washington, DC: U.S.Environmental Protection Agency Office of Water.
United Nations. (2008). World urbanization prospects: The 2007 revision. New York,NY: United Nations.
Ward, B., MacMullan, E., & Reich, S. (2008). The effect of low-impact-developmenton property values. Proceedings of the Water Environment Federation, 2008(6),318–323.
Williams, E. S., & Wise, W. R. (2009). Economic impacts of alternative approaches
Planning and Management, 135(6), 537–546.Wise, S., Braden, J., Ghalayini, D., Grant, J., Kloss, C., & MacMullan, E. (2010). Integrat-
ing valuation methods to recognize green infrastructure’s multiple benefits. Centerfor Neighborhood Technology. April.
