1

GREENBACKS FOR GREENWAYS

Introduction

In 2009, the Prisma Health Swamp Rabbit Trail opened to the public. Since opening,

the trail has become a staple of Greenville, weaving itself into the tapestry of an

up-and-coming well-planned mid-sized metro. When tourists visit Greenville, they can

access the trail along with trail-side amenities such as the Swamp Rabbit Cafe and Grocery,

try some local brews at the Swamp Rabbit Brewery, or even catch a hockey game where

they see the Swamp Rabbits take on any number of rivals in the East Coast Hockey League.

The trail itself connects downtown Greenville’s beautiful Falls Park and Main Street success

stories with its zoo and playground at Cleveland Park, and in the other direction connects

to Furman University and downtown Travelers Rest. It is used by hundreds of thousands of

people every year and is widely regarded as an asset to the community (Reed 2011).

But prior to the construction of the Swamp Rabbit trail, the project faced

considerable opposition in the community Greenville county, a conservative county in

South Carolina, had to begin construction in two of Greenville’s already well used parks,

and would connect to a dying downtown in Travelers Rest. Policy makers and elected

officials took some risk on approving the project.

The trail was successful, indicated by the half a million annual users in 2014 (Reed).

Its success warranted expansion since the original trail was laid, creating spurs to connect

to more amenities and to create a regional bicycle network.​ The Doodle Trail opened in

Easley 6 years after the Swamp Rabbit Trail; Clemson is planning its Green Crescent Trail;

and there are plans to connect the three trails together, along with a spur through Mauldin
 2
GREENBACKS FOR GREENWAYS
to Fountain Inn. With each new project comes new opposition, new questions, and

potentially reduced marginal returns.

That opposition presents a problem. Although there is a seeming advantage to

building the trail, that advantage is difficult to capture. The possibility of improved

property values, health benefits, economic development and tourism all exist with the

creation of parks and trails, as well as intangible community value. G​reenways, bike trails

and rail-to-trail conversions generally raise the values of nearby homes, although those values

tend to decline with distance. ​There is a lack of knowledge on whether the greenway is a

good public investment, or specifically if it is self-funding. The answers to these problems

would be greatly beneficial to policy-makers and elected officials in the future as they

evaluate new trails, spurs, and additions to the Swamp Rabbit Trail.

This research will determine if county government revenues attributable to the

Swamp Rabbit Trail outweigh county expenditures on the project. Specifically, the research

will analyze the trail through a fiscal impact analysis that quantifies the total cost of the

central spine of the trail along with the gains or losses in property tax revenue that resulted

from the trail’s impact on the prices of nearby homes. The effects on property values will be

estimated through several hedonic pricing models in order to answer the same question

through multiple approaches. The results of the research should address the problem of

knowing whether the trail is revenue-positive or -negative from the perspective of the

county government. More generally, it should provide policy-makers and elected officials

with knowledge and insight to make better-informed decisions in regards to greenway

expansion in the area.

 3
GREENBACKS FOR GREENWAYS
To begin answering, or rather to understand ​how​ to answer the research question,

the paper first reviews prior research on the subject. At a basic level, the research is

concerned with if and how the value of greenways is captured by county coffers. The

literature shows that a common method of analysis of government projects in terms of

costs and revenues for local projects is through fiscal impact analysis. Conventional

wisdom in the field is that bicycle facilities and parks are appropriate to include in

traditional local capital public works projects, making the fiscal impact analysis tool

appropriate for the Swamp Rabbit Trail. Next, the scope of fiscal impact analysis and some

definitional issues will be addressed, followed by a brief history and theory of the analysis

and then its defining traits. Next, a survey of literature on the intersection of urban

greenways and property assessment will be reviewed, and finally, the literature review will

discuss methods of measuring the specific impact of the trail on property values, ending on

hedonic price models as a tool to determine revenues gained through property tax. The

following section presents the specific research design and methods that will be used in

this project.

Literature Review

Valuing bicycle greenways

Public greenways are, by definition, public, and thus typically do not carry a fee. The same can

be said of the Swamp Rabbit Trail. This makes it a non-traded amenity, and determining the

value of non-traded goods in monetary terms is a difficult process.​ While ascertaining the costs is

a relatively straightforward endeavor, public policy officials, planners, engineers, bicycle

 4
GREENBACKS FOR GREENWAYS
advocates and opponents may all come to different conclusions on the overall worth of a

greenway network. Krizek suggests that the estimation of a greenway’s costs and its benefits be

evaluated no differently than any other capital expenditure or infrastructure project. Common

methods of making that determination in terms of infrastructure projects are cost-benefit

analysis, fiscal or economic impact assessments, financial risk analysis, and

return-on-investment (ROI) or social-ROI studies (2007).

Fiscal Impact Analysis

Definition and theory​.

Fiscal impact analysis was first standardized in 1978 by Burchell and Listokin in their

seminal work, ​The Fiscal Impact Handbook: Estimating Local Costs and Revenues of Land

Development​ (Raja & Verma 2010). As a matter of semantics, ​fiscal​ impact studies are not to be

confused with other similar impact studies, even though they are sometimes incorrectly used

interchangeably. The related impact analyses are ​demographic ​impact studies, or changes in

some aspect(s) of population of a given study area; ​social​ impact studies, or the impact of a

variable on groups or subgroups within a social setting; ​public service​ impact studies, or the

impact of a policy or project on the funding, level of service, or total or quality of services

provided to the public; or most commonly, ​economic​ impact studies, the measure of change in

employment, income, or business and commercial activity (Leistritz 1994).

Fiscal impacts instead measure specifically the direct costs and revenues generated by

certain decisions for a particular entity such as a city or county government. As a tool, fiscal

impact analysis is used to guide policy makers in decision-making processes which require

choosing between two or more competing land-uses. Because it is often used as a stand-alone
 5
GREENBACKS FOR GREENWAYS
analysis, it shapes discourse on issues such as the costs of sprawl or fixed costs of annexation

(Raja & Verma 2010). Fiscal impact analysis concerns itself with only ​direct​ impacts of a

specific policy, decision or land use. These include costs to the government for police and fire

protection, infrastructure, or public sector growth requisite to provide service for population

growth. Secondary impacts are not counted due to the difficulty in accurately capturing their

impacts and the likelihood of double-counting primary impacts within secondary capture

methods (Burchell & Listoken 2012). Burchell and Listoken go on to narrow the scope of fiscal

impact analyses both to within current costs and revenues and to within the public sector (2012).

That is to say, there are no private costs considered when tallying the burden or benefits of the

project—only revenues to government coffers and expenditures on the budget.

This leads to the question of how one evaluates costs and revenues to the budget. Broadly

speaking, there are two methodologies: the average cost method and the marginal cost method

(Kotval & Mullin 2006). The average cost method is typically a simpler method used when data

is hard to find at the community level. It takes a price which reflects a standard or average is

some way, and applies it to the region in question evenly over the population: this could be an

average cost of service at a per capita level, or having a standard level of service estimate and

applying it to the population as a whole (i.e. 2 officers per 1000 people means the standard level

of service for a town of 20,000 people would be 40 officers) (Bise II 2010). The second broad

method, the marginal cost, considers the existing over or under-capacity to provide services and

estimates the resulting costs accordingly. This can be done with a variety of methods, but the

most common and thorough is the local case study (Bise II 2010). This uses a particular location

and its precise data to determine marginal costs by evaluating the budget and interviewing
 6
GREENBACKS FOR GREENWAYS
officials. This is the method used in this study because the project in question is relatively

narrow (a greenway) and actual cost data were avalable. Not only costs must be measured

however, as revenues are considered as well. Revenues generally consist of three major

headings: property tax revenue from new development, miscellaneous revenues based on current

proportions, and state aid (primarily educational) (Kotval & Mullin 2006). This study will focus

primarily on the first major heading, property tax revenue.

The Assessment of Property Along Greenways

Measuring effects on property values.

Since the research question focuses on the effects of the Prisma Health Swamp Rabbit

Trail on property values (and by extension, property tax revenues), this survey of the literature

will include only research that touches on the fiscal revenues category, specifically on property

values and property tax. Lindsey et. al. (2004) have created a taxonomy of values of urban

greenways which includes the type of value and the preferred method of value-capture. For

property values, hedonic price modeling is suggested (Lindsey et. al. 2004, Krizek 2007).

Additionally, urban greenways can be thought of not as (or at least not ​only​ as) a bicycle

facility but also as park and open space, in which more literature exists on the valuation of

non-traded goods on physical property. The green space literature repeats what greenway

literature has written, namely that valuation of government revenues from non-traded goods can

be difficult especially with the multiplicity of uses for which a park can serve (McConnell &

Walls 2005). McConnell and Walls go on to divide the two primary methods of ascertaining

valuation of non-traded green space (at least within the field of economics): stated preference

and revealed preference. Stated preference relies on surveys crafted to determine their
 7
GREENBACKS FOR GREENWAYS
preferences, willingness-to-pay or willingness-to-accept certain choices or outcomes. Revealed

preference is so called because it uses actual purchases and market decisions to infer values for

proximate amenities. Revealed preference presents, at the least, a floor for what a user is willing

to pay for an amenity through market signals. Stated preference hypothetically presents a more

precise figure, but is dependent on the honesty of the person stating his or her preference.

Revealed preference is used inherently in this study because the study references home prices,

which are reflections of market signals and revealed preferences.

Hedonic Price models

Hedonic price modeling serves as the tool to measure revealed preference for a

non-traded good through price signals in the market of a related traded good. In this case, a

greenway, the Prisma Health Swamp Rabbit Trail, serves as the non-traded amenity and nearby

single-family homes will serve as the traded good through which value is inferred. McConnell

and Walls (2005) use the metaphor of a bundle of goods to describe a product, and the hedonic

model’s job is to parse out the value of each item in the bundle such as the structure itself, the

land on which it sits, the character of the surrounding community, and the access to nearby

amenities like the trail.

There are a number of forms of a hedonic model; there is the linear, the log-log,

quadratic, log-linear, and Box-Cox hedonic models (McConnell & Wall 2005). Although linear

form hedonic models carry some rsiks compared to other models, particularly with regards to

omitted variable bias, linear forms are still widely used and accepted, as are multiple hedonic

models for single study areas.Box-Cox is a statistical formula used to normalize non-normal

variables; Hiazhen (2013) asserts that the Box-Cox logarithmic form of hedonic modeling is
 8
GREENBACKS FOR GREENWAYS
superior in “fitness and forecasting” than that of the linear, semi-log and reverse semi-log forms.

However, the Box-Cox form uses more coefficients, so that when key variables are omitted, a

linear form (linear, linear-log, log-log, and a linear Box-Cox form) may return more accurate

results (Cropper et. al. 1988). Over the past decade, there has been concern that the hedonic price

model’s accuracy is highly susceptible to omitted variable bias. Building upon Cropper et. al.’s

findings, Kuminoff, et. al. (2010) have created a Monte Carlo study that allows one to move

away from linear functional forms even with omitted variables by adding fixed effects and

quasi-experimental cross-sectional designs that make up for omitted variables.

In one meta-analysis of hedonic pricing models, in which 52 studies were surveyed (most of

which were recent and within the United States), 12 studies were selected (“​the main reason for

excluding a study was that it was not possible to derive the desired effect size”)​, of which, 5

ran a linear functional form, 5 ran a semi-log, 2 ran a double log and two ran a Box-Cox

transformation (two studies ran two models) (Brander and Koetse 2011).

Findings of prior hedonic studies

The hedonic method has become a popular method of quantifying the revealed preference

for a non-traded good, as suggested by the number of studies within the literature. The following

will include an overview of fiscal and economic impact studies, hedonic pricing analyses, and

cost-benefit analyses of open spaces (with a focus on greenways). This overview will serve to

inform what to expect roughly in terms of the results of this study and will provide useful context

for a consideration of the results. As a general rule, greenways, bike trails and rail-to-trail

conversions increase the values of nearby homes, although those values tend to decline with

distance from trail or trail access point. The increase or premium tends to be in the 1-5 percent
 9
GREENBACKS FOR GREENWAYS
range but can jump as high as 15.6 percent. Crompton and Nicholls (2019) created an updated

review of the impact of greenways and trails upon proximate property values in which they

reviewed 12 studies of hedonic analyses on greenways and trails (see Table 1). There are a few

studies within Crompton & Nicholls’s review which have already been discussed, but a

comparison side by side with a number of other studies shows that most studies find that

greenways and other bike trail typologies have a small but positive affect on property value. As

with most generalizations, there are exceptions to the rule—not all trails show a positive

relationship. Two studies examined multiple trails or typologies (bike lanes compared to trails)

and found that there was a slight negative relationship between bike lanes and home value.

Author/Date Typology Proximity Result

Quayle & Hamilton 1999 3 Greenways Adjacent Premiums of: 15.6, 14.5,
11.9%

Lindsey et. al. 2003 All Greenways in Indiana ½ mile $3731 (4 percent)
- -
Greenways with trails $4348 (4.7 percent)
Conservation Corridors $5317 (5.7 percent
- -
Monon Trail $13059 (14 percent)
Other Greenway trails -$1025 (-1 percent)
Conservation Corridors $2239 (2.4 percent

Payton & Oltensmann Monon Trail ½ mile 4.1 percent (declines as

2014 Other Greenway Trails neighborhood income
increases)
1.8 percent

Munroe, parker, Campbell Catawba Trail 10 percent of SFH within .01 percent
2004 Mecklenburg and Gaston 5000 feet .03 percent
Counties

Sanders et. al. 2010 Minneapolis St. Paul Distance to nearest trail Non-significant
non-park trails

Welch et. al Portland Or: Distance to nearest path Per ft from facility:
Bike Lanes -$2.47
Local paths $0.01
Regional Paths $0.86

Asabre & Huffman 2009 San Antonio: N/A 1.7 percent

 10
GREENBACKS FOR GREENWAYS

Trails 3.9 percent

Greenbelts 4.8 percent
Greenbelts with Trails

Karadeniz 2008 Little Miami (OH) Scenic 1 mile from a trailhead $7.05 per foot closer (4.7
Trail percent avg)

Parent & vom Hofe 2013 Little Miami (OH) Scenic 10,000 feet from a $230 per foot closer
Trail trailhead (.0009 percent)

Noh 2005 Rail Trail Within ¼ mile 8.2 percent

Kashain et al. 2018 Rail Trail Within city Boundaries 8.6 percent
Table 1 – Comparison of related studies

To underscore the findings of Crompton and Nicholl’s literature review, consider a

hedonic study conducted in an area geographically and economically similar to Greenville

County, on the impact of the Catawba regional trail on properties in North Carolina. Researchers

used hedonic modeling in order to determine the increase in property values (and by extension,

property taxes). The results of the study indicated that the creation of the trail had an impact on

property values, and the impact was predominantly within the first 1000 feet of the trail, but not

completely exhausted until the 5,000 foot mark. As in most of the studies in Table 1, premiums

were found primarily within a small buffer around the trail. (Campbell & Munroe).

Brander and Koatse conducted a meta-analysis (2011) which included 12 suitable

hedonic studies out of a list of 52; while their focus was on a comparison of hedonic to

contingent valuation, the literature will be reviewed now in terms of results from the 12 hedonic

studies they reviewed. Two of the studies within the meta-analysis found similar results, in that

there was a relationship between distance to amenity and premium value, however, one study

expressed that in linear fashion while the other found most of the significance within a buffer and
 11
GREENBACKS FOR GREENWAYS
an exponential drop outside of that buffer. The later study also found that the size of the open

space was significant, and that each acre of park increases home value. This reinforces the

possibility of park space which intersects the trail could distort the premiums added by the

swamp rabbit trail itself. (Bolitzer & Netusil). A third study within a 1500 foot buffer found that

the type of parks also affect premium value: natural area parks have the most significant effect

on value, followed by golf courses, specialty parks, and finally and urban parks (Lutzenhiser and

Netusil 2001). ). The typology of the Swamp Rabbit Trail as a greenway or primarily a bicycle

facility is significant. The intersection of the Swamp Rabbit trail through several parks although

not being a park or large open space on its own, presents a challenge. Considering the size of the

open space effects value, and the trail is linear but cuts through similarly valued open space

amenities makes accurate valuation in these areas difficult as methods for measuring both green

spaces and greenways are quite similar, and the revenue gains of one could be falsely reported in

the other.

Moving away from the Brander-Koatse study, there is also a robust literature which

observes home price values for bicycle-specific amenities. Portland was once again studied via

the hedonic model, this time not for its open space but proximity to advanced bike facilities. The

study found closer proximity to advanced bike facilities, particularly a shared use bicycle trail,

tended to contribute positively to property values. Each ¼ mi closer to the nearest facility

represented a $686 premium for single family homes (Liu & Shi 2017). In a study similar to the

analysis in question of the Swamp Rabbit Trail, a trail in Muskego, Wisconsin was studied for its

effects on property values. It too was a former railroad line that was cleaned and opened as a

bike path in the 2000s, with a large number of observations along the trail corridor. Muskego
 12
GREENBACKS FOR GREENWAYS
itself is more exurban than that of Greenville city, but matches a small rural community like that

of Travelers Rest, and sits almost equidistant between Milwaukee and Madison, mirroring

Greenville’s proximity to Atlanta and Charlotte. The analysis found a statistically significant

positive impact, with home prices adjacent to the path increasing by almost 9 percent, whereas

prior to the installation of the trail, the railroad was a disamenity, where home value increased by

$0.75 for every foot farther from the railroad (Kashien et. al. 2018).

Finally, there is also some amount of literature, though not as much, on the costs and

benefits of bicycle facilities. While the analysis of total costs and total benefits is not being

conducted in this study, getting a holistic view of trail benefits may assist in further answering

the challenge posed in the introduction of giving policymakers an understanding of the value of

bicycle trails. Krizek (2007), in proposing methods to estimate the benefits of bicycle facilities,

reviews 5 cost-benefit analyses which examine a total of 10 trails in 6 different countries:

Colombia, India, the Netherlands, Norway, and the United States. All of the studies yielded a

positive benefit to cost ratio. The highest was in Delhi, India with a 20:1 ratio. The author

suggests that in countries with less developed infrastructure, gains are typically greater, due to

lower costs and higher marginal return. The smallest ratio, which supports the previous assertion,

was 1.5:1 in Amsterdam, which contains some of the most robust cycling infrastructure in the

world. The average benefit ratio of the 10 trails is 6:1.

To conclude, one common method for estimating the value of bicycle facility and

greenway amenities is through hedonic modeling. This approach uses revealed preferences,

which are thought to be more reliable than contingent valuation methods, and employs a

multivariate statistical model to isolate the specific benefit of the amenity from within a larger
 13
GREENBACKS FOR GREENWAYS
bundle of goods, in this case a single family home. Similar studies have found that greenways

typically produce a small positive but measurable effect on the values of nearby homes, in the

range of a few thousand dollars, and that this benefit drops off rather quickly as one moves away

from the amenity.

 14
GREENBACKS FOR GREENWAYS
Methodology

Overview

The methodology of this study’s fiscal impact analysis broadly compares the dollar figure

amount in 2019 dollars that was spent by the county of Greenville on the Swamp Rabbit Trail

and compares it to the revenues the county gained through the trail’s addition. The trail opened

in 2010, which is when the study will begin, and will continue for the useful life of the trail in

2040 (Wang 2005, VanBlarcom 2012), and maintenance costs will be recognized along with

revenues for each year (excluding 2010, which has capital costs in lieu of maintenance). The

revenues studied in this research are not all-encompassing, but are limited to the revenue from

property tax on single family homes. Gains on commercial land are omitted. There will be no

sales tax analysis, as the county of Greenville does not levy a local option sales tax. The hotel

occupancy tax will also not be considered, which is not a strong concern because there has not

yet been major hotel development along the trail outside of the downtown area. The

hypothesized increase in property tax revenue will be measured through multiple hedonic price

models using data from before (2005) and after (2015 and 2019) construction of the trail. The

revenue estimates produced from the models’ output will be assumed constant throughout the

life of the trail. The cost element includes determining the cost of capital investment and

maintenance across the lifespan of the trail to create a final dollar figure of costs to date.

Costs

​The Total Cost of the project will be estimated using the formula

T=C+O
 15
GREENBACKS FOR GREENWAYS
Where T is the total cost of the project in a given year, C is the capital cost in a given year, and O

is the operations and maintenance costs. Revenue from the trail is simply the property tax

revenue from the trail. Therefore the total net cost in any given year will be

N = R – (C + O)

where N is the net cost for the year and R is the revenue for the year. Sine the analysis looks over

30 years, the net revenue over the life of the trail will be
30
N = ∑ R i − (C i + O i )
i=0
Further description and sources for the variables are shown in Table 2.

Table 2 – Sources of data used in the analysis

Total Cost The total cost will be the sum of the cost of the Swamp Rabbit Trail

Capital cost Construction costs will include the total investment into capital in order

to finish the central spine of the trail from Ferris through Travelers Rest

Operations and Operation and Maintenance costs will be estimated using Greenville

Maintenance costs County Budget

This is the same total cost formula used by researchers at Appalachian State University when

performing their analysis of a greenway trail in North Carolina (Whitehead et. al.). The total cost

consists of the capital or construction costs, as well as operations and maintenance. As displayed

in figure 7, the capital costs of the portion of the Swamp Rabbit trail which opened in 2010 were

$2.7 million (Hallo 2018) and designated as total cost for 2010, when the trail officially opened.

No capital was included for any trail connections, spurs or additions built after 2010.The
 16
GREENBACKS FOR GREENWAYS
Greenville County budget does not line item the maintenance for individual parks within its

budget, so Ty Houck of Greenville County Parks and Recreation was consulted and estimated

that $60,000 dollars a year is used for the maintenance of the trail. This includes the whole trail,

even the southern trail in Fountain Inn and in Conestee Nature Preserve. The trail length used in

the study area is 15.4 miles out of the 19.9 total, or 77.4 percent. The $60,000 estimate was

multiplied by .774 to reflect the cost of maintenance for the study area. Over the 30 year life of

the trail, the total cost is estimated to be $4.5 million nominally.

Property Assessment and Revenue Measurement Techniques

Methods.​ Two hedonic models were used in order to estimate the Swamp Rabbit Trail’s

effects on property values. The two models used were difference-in-difference hedonic model

(Parmeter & Pope 2012) and a cross-sectional analysis of properties adjacent to the trail, a

multi-linear ordinary least squares (OLS) regression (Lindsey 2004, Zhang et. al. 2018, Linsey &

Knaap 2007, Campbell & Munroe 2007). The difference-in-difference model uses a pre- and

post-treatment assessment values to demonstrate what, if any, premium exists. The

difference-in-difference model detects the average difference between pre- and post-trail

assessments of houses near the trail as compared with the difference in assessed values of similar

homes in the county over the same period. In this research, that average difference was applied to

all treated properties’ to arrive at a gross premium received by the county. A similar approach

will be taken for the second model, where an average will be drawn from the results, that average

multiplied to the assessed value of all properties within the buffer, in order to have a single gross

premium. The second model’s variables will also be used for a logistical regression through a
 17
GREENBACKS FOR GREENWAYS
propensity score matching process in order to use similar control and test groups for the

quasi-experimental model.

Before discussing the models

themselves, a description of the data

preparation which the models

processed will be useful. GIS

shapefiles were obtained from

Greenville County containing every

parcel in the county, along with the

entirety of the Swamp Rabbit Trail.

In addition, the assessed value of every parcel in the county from 1999 to 2018 was collected.

Using Greenville County’s “Physical Database Design” prepared by Greenville County’s GIS

department in 2016 for reference, all parcels with a land use code beginning in “11” for

residential single family were selected and exported. The trail was then trimmed to the defined

study area as seen in figure 1 to the right. All remaining parcels within a half mile of the trail

were selected (displayed in dark purple), and a new column of data was created, designating all

selected parcels with a dummy variable of “1” to demonstrated treated value. A half mile was

selected because studies that examine property values using a hard limit as opposed to a rate of

decay, one half mile is commonly used in the literature, with studies showing most value is

captured within the first half mile (Lindsey et. Al. 2003, Payton & Oltensmann 2014). Then,

using tax parcel ID numbers, assessment value data from 2005 and 2015 were joined to the table

using inside joins, ensuring that all remaining parcels could be found in 2005, 2015, and 2019
 18
GREENBACKS FOR GREENWAYS
(the year the GIS parcel data was obtained). The years 2005 and 2015 were selected as pre- and

post-treatment years because they are each 5 years removed from the opening of the trail. Using

QGIS’s distance to nearest polygon tool, a column containing the distance in feet to the trail was

added. Finally, using street boundaries, downtown Greenville parcels were removed from the

dataset. The trail runs through downtown, which contains two large parks: Cleveland and Falls

park. Because both access to downtown Greenville’s Main Street, and access to Cleveland and

Falls park serve to increase property

values through non-traded amenity

value like the Swamp Rabbit trail,

parsing out what value was added by

the trail as compared to the parks

would be difficult to ascertain

(McConnell & Walls 2005). Thus

when running both models, parcels

adjacent to Cleveland park, Falls park, and Main Street were omitted using common arterials

surrounding the target area (see figure 2). The table was then exported from ArcGIS to excel for

cleaning variable values in the cross-sectional model (see table 4).

Of the 103,467 observations, 4021 0 bedroom values were removed, as they were

assumed null. Baths and half baths were combined (with half baths divided by two to give a

decimal value). All Bathroom values of less than 1 were removed and assumed null or invalid.

All parcels on under .02 acres of land were removed, for 552 total cases. All parcels at over 10

acres were removed, at 621 total cases. This was to ensure that the parcels measured reflected
 19
GREENBACKS FOR GREENWAYS
single family home value, and were not either merely zoned single family but uninhabitable (as

the case may be with those under .02 acres) or that the value was not primarily agricultural or

developable land with a home placed on it (as the case may be with large acre lots). All

properties with square feet of under 400 were removed for the same reason, for 17 cases. In

addition, 2 properties were removed with high square footage and low property value, suggesting

the home may not be habitable. Extreme values were removed—all 8-9 bedrooms (6 cases).

97,324 cases remain, meaning less than 6 percent of total observations were cleaned.
 20
GREENBACKS FOR GREENWAYS
Model 1.

Y = b1X1 + b2X2 + ... + A

Table 3 – Model 1 variables and definitions

Independent variables

Treatment The treatment variable is a dummy indicating

whether or not the observation is within a

half-mile of the trail

Time Time states whether or not the observation is

pre- or post treatment, expressed as a 0 for pre

and a 1 for post

Interaction The interaction variable is the treatment

variable multiplied by the time variable

Dependent Variable

Assessment The assessed value of the designated parcel

The first model, the difference-in-difference regression model (commonly DiD or DD), is

the more accurate of the two models, as it is quasi-experimental in design including a time

dimension that the cross-sectional analysis does not. It also uses the additional variables of the

cross-sectional through propensity scoring. The purpose of using this model is to proximate an
 21
GREENBACKS FOR GREENWAYS
increase in value of homes near the trail over time. In Greenville County, new property

assessments occur every 5 years or at point of sale, so one year from a five year period before the

trail’s construction and one year post-construction were used. The earliest uses of hedonic

modeling also used this basic approach. Kitchen and Hendon (1967) used distance (to a park)

and both home prices and assessed value to show correlation between home value and proximity

to parks. The weakness to the Kitchen-Hendon study is its single slice in time, making the

relationship between proximity to parks and home values weak.

The treatment is determined by feet from the trail, using Lindsey (2004) and Campbell

and Munroe (2007), with a boundary of one half-mile. The regression used is ordinary least

squares (OLS, or sometimes linear least squares. The regression draws a line through the

observations where the sum of the distance from the observations to the line is minimized.

regression will be conducted through R and R Studio statistical software. The prepared data was

exported into R and the regression was run using the following formula:

>lmDiD = lm(Assessment ~ Time + Treatment + Interaction, data = DiD)

>summary(lmDiD)

The result includes a few key outputs: the coefficient of the interaction variable, which

displays the amenity value of the trail, the significance of that value, and the multiple r squared,

or the fit of the model. This method still has its weak points; omitted variables are major issue in

most regressions and hedonic models and this is no different. In this case, omitted variables are

things that effect the price of a home such as neighborhood characteristics or Parmeter and Pope

(2012) make a strong case that the quasi-experimental Difference-in-difference methodology can

be modified to limit the problem of omitted variables.

 22
GREENBACKS FOR GREENWAYS
In addition to running a difference-in-difference model, the data set was limited to only

the treated parcels and a closely-matched set of control parcels. Those control parcels were

identified using a propensity score. This was performed in R Studio using the data from model

two. Following the process developed by Randolph et. al. (2014), the Matchit package was added

to R and the following command was run:

>m1.out = matchit(treatment ~ BEDS + BATHS + ACRES + SQFEET, data = TPData,

method = “nearest”)

>summary(m1.out)

What this does is take the independent variables used in the cross-sectional model and

compares the values of each independent variable to both the treated and control groups in the

difference-indifference model. It is a measurement of likelihood to be treated, so when the DiD

is run, the control group compared to the treated group are as similar is possible. This does mean

that not all control observations are used. Instead, the default matchit command is a 1:1 ratio of

treated to control groups, so for each treated observation (4,658), there is one control

observation, meaning the propensity score was conducted without replacement. Nearest

Neighbor method was the method in which matches were selected, using a logit as the default

distance measure.

Another threat to the validity of this method is the assessed value itself. If assessed value

is not an accurate reflection of fair market value, then the results will only demonstrate what

assessed value is meant to capture. At the very least, with method 1, it will be internally

consistent, as the purpose is to compare changes in assessed value.

 23
GREENBACKS FOR GREENWAYS
Model 2.

Y = b1X1 + b2X2 + ... + A

“Where ​Y​ is the dependent variable you are trying to predict, ​X1​, ​X2​ and so on are the

independent variables you are using to predict it, ​b1,​ ​b2​ and so on are the coefficients or

multipliers that describe the size of the effect the independent variables are having on your

dependent variable ​Y,​ and ​A​ is the value ​Y​ is predicted to have when all the independent variables

are equal to zero,” (Princeton University, 2020). A list of variables can be in Table 4.

The second model is a OLS linear regression observing parcels within a designated

distance of the Swamp Rabbit Trail. ​Along the original trail, nearby parcels are selected (see

figure 1). Crompton (2001, 2011) suggests a distance not greater than 2000-3000 feet from the

trail while examining increases in assessed property value. Correll et. al. (1978) found that there

was a decrease in value for every foot farther from a greenbelt, but only up to 3,200 feet. Lindsey

et. al. (2004) in their Hedonic GIS analysis, use a buffer of one half-mile, which this study will

replicate. The rationale behind the buffer is that beyond a half mile, most people drove to the

greenway (and the author does not speak for Lindsey, but assumes that the need to drive to the

trail diminishes the value of the trail to the property owner, or is better captured through a

travel-cost method or willingness-to-pay survey), and that the effects of parks (and thus it is

assumed in the study, greenways) diminish or cannot be sufficiently isolated beyond a

2000-3000 foot buffer.

The parcels selected for study are once again single-family homes only. Single family homes are

within a different sub-market than multi-family homes, trailers, or apartments. The variables
 24
GREENBACKS FOR GREENWAYS
provided within this sample are lot size in acres, square footage of the building, number of

buildings, assessed value, bedrooms, bathrooms, half bathrooms (which were combined with

bathrooms), and distance within the half mile buffer and also in feet.

Table 4 – Model 2 variables and definitions

DEPENDENT VARIABLE

Assessed Value Greenville County assessment value in 2019 dollars

INDEPENDENT VARIABLES

Acreage Calculated using the ​parcel polygon in GIS

Square feet Finished size of the home

Bedrooms Number of bedrooms per home

Bathrooms Number of bathrooms per home​, including half haths

Buildings Number of buildings on the lot

Distance Measured as a dummy variable with a distance of a half mile

from the trail; also feet from trail to polygon centroid

The biggest challenge faced by this model, as is the case with most hedonic modeling, is

omitted variable bias. Glaringly, this model does not have data for home age or home condition
 25
GREENBACKS FOR GREENWAYS
(often proxies for one another); other hedonic models have included such details as building

material, number of fireplaces, secondary amenities such as parks, playgrounds, or community

pools, and higher order effects that include crime rates and demographic characteristics like

number of bachelor's degrees (Zhang et. al. 2018). This study also uses a Euclidean distance

from parcel to trail, as opposed to a network analysis. Although Euclidean distance is a common

practice in hedonic modeling, network analysis can often provide a better or more accurate fit of

the data (Binbin et. al. 2013). Due to the nature of the bundle of characteristics, multicollinearity

is an issue; that is, high correlation between variables could increase standard error. However,

correction of this problem by removing the confounding variable increases the problem of

omitted variable bias.

To correct for this as much as possible, the data was tested multiple times for proper fit,

resulting in a few findings. First, in every test of the data through the R regression “lm”

command, building number was insignificant. It was dropped from the final regression and was

not used in the propensity score. Perplexingly, early regressions showed a negative coefficient

for beds, meaning more beds lowered the assessed value according to the model. A correlation

coefficient test was run between beds and baths, beds and square feet, and baths and square feet,

the suspected colinear variables. The R command used was “cor.test,” using the Pearson method.

The results showed that baths and square feet had a .80 correlation, where the other two pairs had

a .65 and .68 correlation respectively. Omitting baths and square feet in the same regression

improved the beds fit.

The final step of the methodology is to take the output from the regressions and convert it

back into a dollar figure to compare to the cost formula and to standardize that over the timespan
 26
GREENBACKS FOR GREENWAYS
of the life of the trail. Only the results from the DiD were used in this conversion. Because the

cross-sectional had so many omitted variables and a non-experimental design, the results would

include value from all non-included variables, making the DiD the more accurate of the two, and

allowing the cross sectional to serve its purpose as a means to conduct propensity score

matching. The result of the DiD regression on the interaction variable was considered the

amenity value of the trail, and was multiplied by the total number of treated parcels to get the

added assessed property value, which was then multiplied by the county millage rate. Historic

records were used for past millage rates, and the current millage rate was assumed for all future

years, making the valuation more conservative. Once multiplied by the millage rate, the resulting

number yields total added county revenue. Annual costs, taken from the county, are subtracted to

yield an annual gain or loss. That figure is then converted through the net present value formula

to factor the time value of the County’s investment over the course of the trail.

N
Rt
NPV = ∑ (1+i)t
t=0

Where R​t​ is the net return at time period t, N is the number of time periods, t is the index of time

periods (in this case, 30 years, from 2010 to 2040), and ​i​ is the discount rate, in this case 1.8%,

the seasonally unadjusted annual inflation rate in 2018 according to the federal reserve.
 27
GREENBACKS FOR GREENWAYS
Results

The results come in 5 parts: the results of the cross-sectional regression with dummy

distance; the results of the cross-sectional regression by decay with Euclidian feet; the result of

the propensity score matching; the result of the difference-in-difference; the DiD converted into

net present value.

Figure 3 to the right

shows the results of the

cross-sectional regression with

the half mile buffer. All

variables are significant and the

adjusted r​2​ is 57 percent. This

model suggests that being within half a mile of the trail adds $23,135.10 dollars in value to the

property of a single family home. An important note on the dependent variable used in the

cross-sectional model as compared to the difference-in-difference model is that the

cross-sectional uses assessed ​market​ value, as opposed to the DiD’s ​taxable​ assessed value. in

order to arrive at the taxable assessed value from market assessed value is a multiplier of 4

percent. This is what the county

applies to residential properties

before the millage rate is applied.

So while the market value of the

effect is $23,135.10, the taxable

assessed value of this difference

 28
GREENBACKS FOR GREENWAYS
shown in the model is 925.40 dollars, still significantly higher than the DiD below, and evidence

that there are omitted variables contributing to that amenity value.

The second regression uses the same variables but looks at distance as a measure of

Euclidean feet instead of a binary buffer. All coefficients are significant and the adjusted R​2 ​is a

favorable 57 percent. The Distance coefficient suggests that there is a decrease of .5530 dollars

in assessed market value for every foot further away from the trail the property is located. For

better comparison and comprehension, that means there is a loss of 2.21 dollars in taxable

assessed value for every 100 feet further away from the trail the property is located. Although

this number seems low, it measures

the decay using observations across

the whole county and assigns a linear

value, although the decay is in all

likelihood exponential as it travels

farther away; secondly, the omitted

variables are still an issue meaning

the number is likely inflated in that

regard.

The results of the propensity

score to the right use the same data, variables, and observations as the previous two results.

However, instead of value as the dependent variable, the treatment dummy variable was used, to

find an equal number of control observations with similar characteristics, or that are equally

likely to receive treatment, making any difference in the interaction variable more likely to be
 29
GREENBACKS FOR GREENWAYS
from the treatment, or addition of the

trail. The matching yielded a 1 to 1

match of treated to control

observations, without any lost treated

observations.. The mean difference of

the variables in the matched set grew

considerably closer: from a difference

of 291 to 37 square feet; .3 baths to .05 baths; .22 to .06 beds, and .28 to .05 acres.

Using the matched set of 9,316 observations, the difference-in-difference model was run.

The first thing to note is the treated variable is ​not s​ ignificant. This is an indication that the

propensity score worked well. The interaction variable, while less significant, still shows only a

1 percent chance of error. The adjusted R​2​ is on the low side. The interaction variable suggests

that the amenity value of the trail is 324.56 dollars of taxable value for each property within a

half mile of the trail.

 30
GREENBACKS FOR GREENWAYS
The final piece of the results is the conversion of the single amenity value rate of

the property to county revenue minus the cost of the trail over the life span of 30 years. This

shows that while the property values of single-family homes are increased enough to

significantly out-perform maintenance costs, they did not increase enough to recoup the total

capital cost of the trail across its expected lifetime. As seen below, the slope of the total gross

revenues line is steeper than the gross costs, but not steep enough to intersect total gross costs.

After adjusting to net present value, the County spent more than it received in property tax

revenue to the tune of 2.04 million dollars. Of course, this isn’t to say that the trail itself is a net

loss. This study examined only single-family home values and was narrow in that scope. It does

not take into consideration occupancy taxes, multi-family, commercial, or industrial lots,

economic development and growth along the trail, or more nebulous measurements of value like

health cost savings or air quality.

Discussion
 31
GREENBACKS FOR GREENWAYS
While the findings here suggest that is not the case, it is possible that an analysis that

included the revenues from multi-family residential as well as other land uses would show

revenues equal to or exceeding costs. Similar studies discussed in the literature review often

discuss the value in terms of percent of added premium value. In order to compare the value of

this study to others that do not use this specific measure of value, the value was converted into a

percentage. The average taxable assessed value of all treated parcels from the DiD regression,

which resulted in the 324.56 premium, was calculated at 5,976.38 dollars, making the premium

5.6 percent of average taxable value. This premium value sits squarely in the middle of premium

values expressed in percentages. The dollar value is considerably lower than most other studies,

but this is likely due to using taxable assessed value as opposed to fair market value as a measure

of value. Of the 8 other trails studied which used a half-mile boundary, only one was negative,

one was in double digits, and 4 were between 4-6 percent (Lindsey et. al. 2003, Payton &

Oltensmann 2014). The literature implies that if the buffer were to be made smaller, the

interaction coefficient or premium would go up, but the number of parcels would be reduced, so

the overall effect is unclear. Quayle and Hamilton studied three trails and used only adjacent

parcels and found premiums of between 12-16 percent (1999). Two trails were examined, one

within city boundaries and one with a quarter-mile buffer, and returned premiums of 8.6 and 8.2

percent respectively.

Of the studies conducted on distance-decay of the effect, only one does not use some sort

of outer limit, and compares local bike paths to regional bike paths. It found that the added value

of local paths was only .01 dollars per foot, where the regional path was .86 dollars per foot. The

value per foot of the Swamp Rabbit Trail per foot using taxable assessed value is .02 dollars
 32
GREENBACKS FOR GREENWAYS
(.553 dollars using assessed market value) which is a premium of .03 percent. The Catawba trail

studied in North Carolina used a 5,000 foot buffer for their decay study and found premiums of

.03 percent and .01 percent per foot in the two counties the trail passed through.

This assessment raises multiple questions: what should Greenville, or other local and

regional governments, do with this information? What policies should be in place given the

results found? What are the pros and cons of the results? In essence, what are the consequences

of the study?

First, consider the scope of the results. The final results show that the trail did ​not p​ ay for

itself through the channels evaluated in this study. This does not mean the trail is not

self-funding. The study is limited only to direct revenue from single-family property tax

assessment increases, and does not include any other sources of revenue. Property tax gains from

multi-family parcels or commercial parcels have not been considered. Economic development

gains have not been factored in. Tourism was not included in the results. The trail was a

renaissance to downtown Travelers Rest, which is primarily commercial in nature, featuring new

restaurants and some retail, including an outdoors retailer and a brewery which borrows the

trail’s name. Several bike shops have opened adjacent to the trail for repairs, rentals and sales. A

handful of commercial and mixed-use establishments have opened along the stretch of trail in the

county between the city of Greenville and Travelers Rest. Some apartments and residential

developments are being developed alongside the trail, using its proximity as a primary selling

point. All of these things add value to Greenville County, Greenville city, and Travelers Rest

(not to mention a small segment in Fountain Inn).

 33
GREENBACKS FOR GREENWAYS
The multi-family and commercial parcels built along the trail previous to its instillation

likely saw a positive change in value as well, but that addition of amenity value was not

considered. Considering the difference in annual maintenance and annual revenue generated is

positive, any additional gains from multi-family and commercial only work to close the gap

between initial capital costs and annual revenue gains (as depicted in figure 8). A percentage of

every dollar spent at a trail-side commercial property, job created from said business, and

property tax revenue created from new trail-oriented business can be attributed to the trail and is

not considered in the scope of this study.

Furman University’s own Dr. Julian Reed found in an economic development and public

health study (2014) of the trail over the first four years of its use that roughly 7 million dollars

can be attributed to the trail annually. While the methods used in those studies were not

particularly rigorous (a handful of interviews with nearby business owners), the author could be

off the figure by 70 percent and still find enough value in economic development and tourism

alone to make up the difference remaining in this study. This all points to the probable

conclusion that the trail, while not self-funding through single family property value increases

alone, has earned more money for the county coffers than its construction cost.

Additionally, the scope of the study only examined the gains for the county. The amenity

value found was not applied to Greenville city, Travelers Rest, or Fountain Inn millage rates,

although the value would certainly be taxed by these municipalities. The scope of the study also

removed certain parcels that would otherwise be in the study area. Single family homes in the

Cleveland Park, Falls Park, and downtown areas were removed from analysis not only during the

modeling, but also when applying the premium value back to the parcels near the trail. They
 34
GREENBACKS FOR GREENWAYS
were removed from the modeling process so as not to artificially inflate the premium value of the

trail with the premium value of the other listed amenities. It is also possible that, because of the

other amenities, that the trail is subject to the law of diminishing returns and does not add the

same amount of value to homes near all the other features as it would to a home located

elsewhere in the study area. For this reason, the parcels were not calculated into the net revenue

figure, even though it is likely they contributed some value.

There are also harder to quantify added value effects that are not measured here that

could save the local governments money in a long-term FIA approach as well. The trail is used

often for exercise (Reed 2011), and the health benefits received by additional workouts could

reduce the burden on the county and state governments in EMT and healthcare services.

Commuting, while not common on the trail, still takes place, and reduces miles driven

emissions. It is possible that the trail shifted some exercise from gyms to the trail; if the trail is

located closer than the previous workout spot, then emissions decrease. The opposite is also true.

It is also possible that the creation of the trail’s added tourism and recreation has created a net

positive number of new or longer car trips, creating a negative impact on air quality. Then there

are intangible community benefits of having access to a safe and free public park.

That isn’t to say that there are not negative consequences to the park as well, or at least

the possibility of negative effects. The trail added value to properties, making the homes and

other property near the trail more valuable. While this is generally considered a good thing,

adding value to a community, in certain areas it can bring about a rapid increase in land rents,

possibly forcing out long-time or economically disadvantaged residents. In other words, it is

important to be aware of the gentrification the trail may cause. This can take place not only in
 35
GREENBACKS FOR GREENWAYS
areas targeted by this study, that is single family homes, but also by a typically less stable and

lower income areas: multi-family, along with renters of both multi-family and single-family

homes. While property owners see an increase in their assets with the premium the trail provides,

those renting, be it in a single family or multi-family residence, do not reap the benefits of the

premium long-term, while still facing the increase in rent that accounts for both the increase in

assessed value with property tax that is passed on to the consumer, but also through an increase

in demand for units near the trail.

This leads to some future policies worth considering. First, while it is typical for public

projects to be studied and analyzed during the proposal phase in order to convince council

members or to provide cover against opposed constituents, rarely is there a post-implementation

study to determine whether the projections were accurate, and the proposal was a success. Local

governments could consider implementing a post-implementation study requirement of major

public works projects, to determine their cost-effectiveness or at the very least that they are

delivering on what they were proposed to do. A pilot program for non-traditional transportation

options is also a possibility. If the trail is popular, spurs economic development, and is revenue

positive, they should consider what the possibility of a connected network of trails and lanes

could do for commutes and businesses along the network. The region has already begun

exploring this through Swamp Rabbit Trail expansions and a regional network of connected

trails.

Finally, it is worth reflecting on the limitations of the study. Opportunities to improve the

study include additional data. This could work in concert with a policy recommendation to

improve data collection for future studies. In the cross-sectional model, omitted variables was an
 36
GREENBACKS FOR GREENWAYS
issue that made the model somewhat unreliable for over-valuing the amenity. Specifically, the

age of building in question and/or the condition of the home would have a major effect on value

and was not included. Higher order characteristics like neighborhood demographics including

median income, crime rate, school quality, racial composition, educational attainment, and share

of renters to owners was not included. As previously discussed, the study was limited in its scope

to single-family parcels and to Greenville county. The study could be expanded to include

multiple land use typologies and expanded to include municipalities such as Greenville city,

Fountain Inn, and Travelers Rest. The study omitted parcels from possibly the most wealthy part

of Greenville County—downtown Greenville. While this was done so as not to inflate or conflate

the amenity value of the trail with the premium for downtown and falls park access (as well as

Cleveland park), it is likely that these parcels had some effect on the trail premium value, and

that was not measured either in the modeling for the premium, or in the fiscal impact analysis

when the premium value was applied to treated parcels. Reasonable care was taken to measure

the amenity of the trail only, and this meant the omission of the downtown parcels, as well as the

disconnected portions of trail in Fountain Inn and Conestee were excluded. There was also no

network analysis of the treated and untreated observations. The treated parceled were measured

simply as a function of intersecting with a buffer of half a mile from the trail. In reality, some

treated observations are farther than half a mile when considering trail heads and routes to the

trail.
 37
GREENBACKS FOR GREENWAYS
References

Binbin L., Charlton, M., Harris, P., & Fotheringham A. S. (2014). Geographically weighted

regression with a non-Euclidean distance metric: a case study using hedonic house price

data,​ International Journal of Geographical Information Science, vol. 28​ (4), pp.

660-681.

Bise II, L. C. (2010). Fiscal Impact Analysis: Methodologies for Planners. ​American Planning

​ hicago, Il.
Association, C

Boarnet, M. G., Greenwald, M., McMillin, T. E. (2008). Walking, Urban Design, and Health.

Journal of Planning Education and Research, Vol. 27​ (3).

Bolitzer, B., & Netusil, N. R. (2000). The Impact of Open Spaces on Property Values in

Portland, Oregon. ​Journal of Environmental Management, Vol. 59​ pp. 185-193.

Brander, L. M. & Koetse, M. J. (2011). The Valuation of Urban Open Space: Meta-Analyses of

Contingent Valuation and Hedonic Pricing Results. ​Journal of Environmental Economics

and Management, Vol. 92​ (10) pp. 2763-2773.

Brown Jr., G. & Mendelsohn, R. (1984). The Hedonic Travel Cost Method. ​The Review of

Economics and Statistics, Vol. 66​ (3), pp. 427-433.

CAMPBELL, H., & MUNROE, D. (2007). Greenways and Greenbacks: The Impact of the

Catawba Regional Trail on Property Values in Charlotte, North Carolina. ​Southeastern

Geographer,​ ​Vol. 47 (​ 1), pp. 118-137.

Correll, M. R., Lillydahl, J. H. & Singell, L. D. (1978). The Effects of Greenbelts on Residential

Property Values: Some Findings on the Political Economy of open Space. ​Land

Economics, Vol 54 (​ 2) pp. 207-217.

 38
GREENBACKS FOR GREENWAYS
Crompton, J.L & Nicholls, S. (2019) The Impact of Greenways and Trails on Proximate Property

Values: an Updated Review. ​Journal of Parks and Recreation Administration, Vol. 37

(3).

Crompton, J. L. (2012). Estimates of the Economic Benefits Accruing From an Expansion of

Houston’s Bayou Greenway Network. ​Journal of Parks and Recreation Administration,

Vol. 30​ (4) pp. 83-93.

Cropper, M. L., Deck, L. B., McConnell, K. E. (1988). On the Choice of Functional Form for

Hedonic Price Functions. ​The Review of Economics and Statistics, Vol. 70 (​ 4) pp.

668-675.

Dallat, M. T, Soerjomataram, I., Hunter, R. F., Tully, M. A., Cairns, K. J., Kee, F. (2013). Urban

Greenways have the Potential to Increase Physical Activity Cost-Effectively. ​European

Journal of Public Health, Vol. 24​ (2) pp. 190-195.

Dannenberg, A. L., Bhatia, R., Cole, B. L., Heaton, S. K., Feldman, J. D., Rutt, K. D. (2008) Use

of Health Impact Assessments in the U.S.: 27 Case Studies, 1999-2007. ​American

Journal of Preventive Medicine, Vol. 34​ (3).

de Haan, J. and E. Diewert (2013), "Hedonic Regression Methods", in ​Handbook on Residential

Property Price Indices,​ OECD Publishing, Paris.

Everett, M. (1977). Benefit Cost Analysis for Labor Intensive Transportation Systems.

Transportation, Vol 6 ​(1) pp. 57-70.

Fishman, E., Schepers, P., Kamphius, C. B. M. (2015). ​American Journal of Public Health, Vol.

105 ​(8) pp. e13-e.15.

 39
GREENBACKS FOR GREENWAYS
Fleming, C. M., & Cook, A. (2008). The Recreational Value of Lake McKenzie, Fraser Island:

an Application of the Travel Cost Method. ​Tourism Management, Vol. 29 (​ 6) pp.

1197-1205.

Gotschi, J. (2011). Costs and Benefits of Bicycle Investments in Portland, Oregon. ​Journal of

Physical Activity and Health, Vol. 8​ (1) pp. S45-S58.

Grabow, M., Hahn. M., Whited, M. (2010). Valuing Economic and Health Impacts in Wisconsin.

Haizhen, W., Ling, Z., Xiaoqing, B. (2013). The Application of Box-Cox Transformation in

Selecting Functional Form for Hedonic Price Models. ​Applied Mechanics and Materials,

vol. 357-360 ​pp. 2869-2875.

Hallo, Lisa (2018). The Trail that Almost Wasn’t. Upstate Forever.

Kashien, R., Winden, M., Emmett Storts 92018). The Effect of a Recreational Bike Path on

Housing Values in Muskego, Wisconsin. ​Journal of Parks and Recreation

Administration. Vol 36 p​ p. 160-173.

Krizek, J. K. (2007). ​Estimating the Economic Benefits of Bicycling and Bicycle Facilities: an

Interpretive Review and Proposed Methods. ​Essays on Transport Economics (​ pp.

219-248).

Kuminoff, N. V., Parmeter, C. F., Pope, J. C. (2010). Which Hedonic Models Can We Trust to

Recover the Marginal Willingness to Pay for Environmental Amenities? ​Journal of

Environmental Economics and Management, vol. 60 (3)​ pp. 145-160.

Leistritz, F. L. (1994). ECONOMIC AND FISCAL IMPACT ASSESSMENT, ​Impact

Assessment, Vol. 12​ (3), pp. 305-317

 40
GREENBACKS FOR GREENWAYS
Lindsey, G., Man, J., Payton, S., Dickson, K. (2004). Property Values, Recreation Values, and

Urban Greenways. ​Journal of Parks and Recreation Administration, Vol. 22​ (3) pp.

69-90.

Liu, J. H. & Shi, W. 92017) Impact of Bike Facilities on Residential Property Prices.

Transportation Research Record, Vol 2662​ (1).

McConnell, V. & Walls, M. (2005). The Value of Open Space: Evidence from Studies of

Non-Market Benefits. Lincoln Institute for Land Policy.

Posner, R. A. (2000). Cost-Benefit Analysis: Definition, Justification, and Comment on

Conference Papers. ​Journal of Legal Studies, Vol. 29 (​ S2) pp. 1153-1177.

Price, A. E., Reed, J. A., Muthukrishnan, S. (2012). Trail User Demographics, Physical Activity

Behaviors, and Perceptions of a Newly Constructed Greenway Trail. ​Journal of

Community Health, Vol. 37 (​ 5).

Raja, S. & Verma, N. (2010). Got Perspective? A Theoretical View of Fiscal Impact Analysis.

Planning Theory, Vol. 9​ (2) pp. 126-136.

Randall, A. (1994). A Difficulty with the Transportation Cost Method. ​Land Economics, Vol. 70

(1) pp. 88-96.

Randolph, J. J., Falbe, K., Manuel, A. K., Balloun, J. L. (2014). ​A Step-by-Step Guide to

Proponsity Score Matching in R​, Practical Assessment, Research and Evaluation Vol. 19

(18).

Reed, J. A. (2011). ​Greenville Health System Swamp Rabbit Trail: Year 1 Findings,​ Greenville

County Parks and Recreation Department.

 41
GREENBACKS FOR GREENWAYS
Reed, J. A. (2012). ​Greenville Health System Swamp Rabbit Trail: Year 2 Findings,​ Greenville

County Parks and Recreation Department.

Reed, J. A. (2013). ​Greenville Health System Swamp Rabbit Trail: Year 3 Findings,​ Greenville

County Parks and Recreation Department.

Reed, J. A. (2014). ​Greenville Health System Swamp Rabbit Trail: Year 4 Findings,​ Greenville

County Parks and Recreation Department.

Sen, A. (2000). The Discipline of Cost-Benefit Analysis, ​Journal of Legal Studies, Vol. 29 (​ S2)

pp. 931-952.

Zerbe, R. O. (2017). A Distinction Between Benefit-Cost Analysis and Cost-Benefit Analysis:

Moral Reasoning and a Justification for Benefit-Cost Analysis. ​University of Washington,

Seattle.​