DEMAND FOR URBAN FORESTS: A NATIONAL AND REGIONAL STUDY 
 
 
 
 
Except where reference is made to the work of others, the work described in this thesis is 
my own or was done in collaboration with my advisory committee. This thesis does not 
include proprietary or classified information. 
 
 
 
 
______________________________ 
Pengyu Zhu 
 
 
 
 
Certificate of Approval: 
 
 
 
 
________________________                              __________________________  
David Laband                                                       Yaoqi Zhang, Chair 
Professor                                                               Assistant Professor 
Forestry and Wildlife Sciences                   Forestry and Wildlife Sciences  
 
 
 
 
________________________                              __________________________ 
Brenda Allen                                                         Stephen L. McFarland 
Assistant Professor                                               Acting Dean 
Forestry and Wildlife Sciences                            Graduate School
DEMAND FOR URBAN FORESTS: A NATIONAL AND REGIONAL STUDY 
 
 
 
Pengyu Zhu 
 
 
 
 
 
A Thesis  
Submitted to  
the Graduate Faculty of  
Auburn University 
 in Partial Fulfillment of the  
Requirements for the  
Degree of  
Masters of Science 
 
 
 
 
 
 
Auburn, Alabama 
August 7, 2006 
 
 
 iii
DEMAND FOR URBAN FORESTS: A NATIONAL AND REGIONAL STUDY 
 
 
 
Pengyu Zhu 
 
 
 
 
Permission is granted to Auburn University to make copies of this thesis at its discretion, 
upon request of individuals or institutions and at their expense. The author reserves all 
publication rights. 
 
 
 
 
_______________________ 
Signature of Author 
 
 
 
 
_______________________ 
  Date of Graduation 
 
 
 iv
VITA 
 
Pengyu Zhu, son of Xueqin Zhu and Yuefen Yang, was born on January 20, 1979, 
in Danyang, China. He graduated from Danyang Advanced High School, Danyang, 
China in 1997. He entered the Department of Urban and Resources Sciences, Nanjing 
University, China in September, 1997, graduating with a Bachelor of Design degree in 
Urban Planning in June, 2002. After getting his bachelor?s degree, he entered the 
graduate school in Nanjing University in 2002. After enrolled in Master of Science in 
Urban and Regional Planning at Nanjing University for two years, he entered Graduate 
School, Auburn University, in August, 2004.  
 
 v
THESIS ABSTRACT 
DEMAND FOR URBAN FORESTS: A NATIONAL AND REGIONAL STUDY 
 
Pengyu Zhu 
 
Master of Science, August 7, 2006 
(B.S. Nanjing University, 2002) 
 
62 Typed Pages 
 
Directed by Yaoqi Zhang 
 
Various public policies have been designed to protect our green spaces and 
stimulate the demand for urban forests within communities. However, very few studies 
have been conducted to uncover the factors influencing the demand for urban forests and 
the sensitivity of demand to price and income. This thesis will shed some light by 
empirically estimating the demand for urban forests. It comprises two interrelated parts, 
in which two different models and two different data sets are used to test the relationship 
between various socioeconomic factors and demand for urban forests. 
In Part I, a theoretical economic model is specified and per capita urban forest 
data as well as other socioeconomic data from all cities with population over 100,000 
across the nation are collected and used for the estimation. The empirical findings suggest 
that the demand for urban forests is elastic with respect to price and highly responsive to 
 
 vi
changes in income. The results also show that population growth might induce a negative 
influence, while urban sprawl has a positive impact on the demand for urban forests. 
In Part II, I examine the relationship between percentage of urban forests and 
household income and population density, focusing on cities with populations over 
40,000 within the southeast region. Our empirical results show that urban forest 
percentage across the cities has characteristics of the Environmental Kuznets Curve. We 
find that household income around $39,000 is a turning point that changes the 
relationship between income and urban forest coverage from negative to positive; 
whereas the impact of population density on urban forests is just the opposite, from 
positive to negative when population density is around 180 persons per square kilometer. 
 
 vii
ACKNOWLEDGEMENTS 
 
I would like to express my sincere gratitude to my advisor, Dr. Yaoqi Zhang. 
Without his academic guidance, this thesis would have been impossible. I would also like 
to express my appreciation to the rest of my advisory committee members, Dr. David 
Laband and Dr. Brenda Allen, for their immense help to me which undoubtedly improved 
the quality and significance of this study. I also acknowledge the help I received from Mr. 
Paul Holm. Moreover, there is a special person I want to appreciate. That is my wife, 
Mingxia Sheng, who has accompanied me through the 2 years in Auburn University. 
Without her care in these 2 years, it?s impossible for me to make these achievements. 
Finally, I appreciate the financial and other support provided by the Center for Forest 
Sustainability of Auburn University, Forest Policy Center, the Challenge Cost-Share 
Grant Program of The National Urban & Community Forestry Advisory Council of 
USDA Forest Service, and Alabama Urban & Community Financial Assistance Program.  
 
 viii
Style manual or journal used:   Landscape and Urban Planning 
 
Computer software used:     ArcGIS 
    LIMDEP 
Microsoft Word 
    Microsoft Excel  
Photoshop
 
 ix
TABLE OF CONTENTS 
LIST OF TABLES ??????????????????????????.? x 
LIST OF FIGURES ??????????????????????????.. xi 
INTRODUCTION ???????????????????????????. 1 
PART I ? DEMAND FOR URBAN FORESTS IN UNITED STATES CITIES ???. 4 
1. Introduction ?????????????????????????? 4 
2.  Economic Model of the Demand for Urban Forests ??...???????.. 9 
3. Data ????????????????????????????.. 15 
4. Results ???????????????????????????.. 20 
5. Conclusions and discussion ???????????????????. 24 
PART II ? DEMAND FOR URBAN FORESTS AND ECONOMIC WELFARE: 
EVIDENCE FROM THE SOUTHEASTERN US CITIES ??????????.? 27 
1. Introduction .????????????????????????..... 27 
2. Econometric Model ??????????????????????.. 30 
3. Data .????????????????????????????. 33 
4. Results .???????????????????????????. 35 
5. Conclusions ?????????????????????????.. 38 
LITERATURE CITED ????????????????????????? 39 
 
 x
LIST OF TABLES 
Table 1-1: Results for the Regression of Residential Land Value ................................... 45 
Table 1-2: Data Description of Variables ?????????????????.... 46 
Table 1-3: Regression Results for the Demand for Urban Forests ????????.. 47 
Table 2-1: Data Description of Variables ?????????????????? 48 
Table 2-2: Regression Results ??????????????????????. 49 
 
 xi
LIST OF FIGURES 
Figure 1-1: Ecoregions of selected cities in Continental US ??????????.. 50 
Figure 2-1: Income Elasticity in Sample Cities ???????????????.. 51 
 1
INTRODUCTION 
 
Trees have been recognized as an important component of urban landscapes 
throughout the history of urbanization. As early as in 1902, Ebenezer Howard brought 
forward the ?Garden City? concept in his book entitled Garden Cities of Tomorrow. This 
pervasive theme has run through all the various dimensions of urbanization in twentieth-
century America. ?Garden City?, conceived as an oasis of civilization not too far from a 
major urban center, connected to the center by some sort of congestion-free 
transportation system but separated from the center by a greenbelt, has placed a great 
emphasis on urban trees.  
Now, one hundred years have passed. In this most important century of human 
development, various public policies have been designed to protect our green spaces and 
stimulate the demand for urban forests within communities. However, very few studies 
have been conducted to uncover the factors influencing the demand for urban forests and 
the sensitivity of demand to price and income. This thesis will shed some light by 
empirically estimating the demand for urban forest. It comprises two interrelated parts, in 
which two different models and two different data sets are used to test the relationship 
between various socioeconomic factors and demand for urban forests.  
In Part I, a theoretical economic model is specified and per capita urban forest 
data as well as other socioeconomic data from all cities with population over 100,000 
 2
across the nation are collected and used for the estimation. The empirical findings suggest 
that the demand for urban forests is elastic with respect to price and highly responsive to 
changes in income. The results also show that population growth might induce a negative 
influence, while urban sprawl has a positive impact on the demand for urban forests. 
In Part II, we examine the relationship between percentage of urban forests and 
household income and population density, focusing on cities with populations over 
40,000 within the southeast region. Our empirical results show that urban forest 
percentage across the cities has characteristics of the Environmental Kuznets Curve. We 
find that household income around $39,000 is a turning point that changes the 
relationship between income and urban forest coverage from negative to positive; 
whereas the impact of population density on urban forests is just the opposite, from 
positive to negative when population density is around 180 persons per square kilometer. 
The sample cities selected in the above two parts are different. In the first part, 
cities across the whole nation are studied. Due to the different temperature, precipitation 
and geographic factors, the possibility is greatly increased that natural environmental 
factors will have an impact on the accuracy of our estimation. Therefore, an ecoregion 
dummy controlling these natural environmental factors is used in our model. In the 
second part, we try to reduce the natural environmental impact to the minimum level. 
Cities within the southeast region are selected, due to their relatively homogenous climate 
and environment.  
This thesis will contribute to our knowledge of how socioeconomic factors 
influence the demand for urban forests. This knowledge is essential to the formulation of 
effective urban forest policies in the future. It will also help to expand the scope of 
 3
research within forestry. Previous work has largely focused on the value of urban forest 
and demand for other environmental goods. Now, it is time to elevate the discussion by 
beginning to focus on the demand for urban forests.  
 4
PART I ? DEMAND FOR URBAN FORESTS IN UNITED STATES CITIES 
 
1. Introduction 
Trees have been recognized as an important component of urban landscapes 
throughout the history of urbanization. As early as in 1902, Ebenezer Howard brought 
forward the ?Garden City? concept in his book entitled Garden Cities of Tomorrow. This 
pervasive theme has run through all the various dimensions of urbanization in twentieth-
century America. ?Garden City?, conceived as an oasis of civilization not too far from a 
major urban center, connected to the center by some sort of congestion-free 
transportation system but separated from the center by a greenbelt, has placed a great 
emphasis on urban trees.  
As urbanization continues, scientific understanding of how urban trees, forests 
and green spaces benefit people has expanded substantially to include environmental, 
social and economic domains. Sociologists and economists found that urban trees, in 
addition to providing environmental and aesthetic benefits, also brought a broad range of 
economic, social, and even psychological benefits. In recent years, urban forest valuation 
studies have addressed many facets of urban forest benefits. These economic valuations 
translate urban forest functions and benefits into terms that enhance public values. 
 5
Environmental Functions and Benefits 
Trees in urban landscapes moderate temperature and microclimates, thereby 
reducing needs for air conditioning and saving energy (Heisler, 1986; McPherson, 1990; 
Meier, 1991; Oke, 1989). Urban trees help improve air quality and sequester carbon 
(Nowak, 1993; Nowak and McPherson, 1993; Rowntree and Nowak, 1991; Smith, 1981), 
help stabilize soils, reduce erosion, improve groundwater recharge, control rainfall runoff 
and flooding (Sanders, 1986), reduce urban noise levels (Cook, 1978), and provide 
habitat that increases biodiversity (Johnson, 1988). Based on modeling of air pollution, 
stormwater mitigation and energy impacts, the annual values of urban forest services are 
estimated. For instance, the Urban Ecosystem Analysis of the Washington D.C. 
Metropolitan Area concluded that tree cover had reduced storm water storage costs by 
US$4.7 billion and generated annual air quality savings of $49.8 million (American 
Forests, 2002).  
Micro-scale studies which focus on street tree costs and benefits have also been 
conducted. Costs include tree planting, irrigation, pruning and other maintenance. 
Calculated benefits include energy savings, reduced atmospheric carbon dioxide, 
improved air quality, and reduced stormwater runoff. Economists have shown that these 
types of benefits may outweigh the costs of urban forestry programs by considerable 
margins (McPherson, 1992, 1994; McPherson and Biedenbender, 1991; McPherson et al., 
1997; McPherson et al., 1998). For instance, a 2002 analysis for Seattle, WA, indicated 
that per tree average annual net benefits were $1 to $8 for a small tree, $19 to $25 for a 
medium-sized tree, and $48 to $53 for a large tree (CUFR 2002). 
 
 6
Economic Functions and Benefits 
Urban forests can be planned to directly affect the economic development of a 
municipality or region. The most direct valuation is to estimate marketable goods, or the 
value of purchase substitutes. For example, urban forests can produce human and animal 
foods, building materials, fuels, and medicinal materials, thus contributing to the 
reduction of the costs of distribution systems needed if they are transported from rural 
areas.  
Urban trees also make neighborhoods aesthetically more appealing and add to the 
value of property (Schroeder, 1989). Hedonic or amenity pricing is often used to measure 
a price increment that correlates with a desirable condition or situation. Previous hedonic 
price analyses showed clearly that trees increase the value of residential properties and 
that people are willing to pay more for housing with trees (Anderson and Cordell, 1985, 
1988; Morales, 1980; Payne and Strom, 1975). More recently, Crompton (2001) 
concluded that a quality forest or green space has a positive economic ripple effect on 
nearby properties. Appraised property values of homes that are adjacent to parks and 
open spaces are typically about 8 to 20 percent higher than those of comparable 
properties elsewhere. Rental rates of commercial office properties were about 7 percent 
higher on sites having a quality landscape, which included trees (Crompton, 2001). 
Studies on how trees affect shoppers? behavior in retail business districts employ 
the contingent valuation method. Consumers claim they are willing to pay about 9 to 12 
percent more for products in downtown shopping areas with trees, versus in comparable 
districts without trees. Customer service, merchant helpfulness, and product quality are 
all judged to be better by shoppers in places with trees (Crompton, 2001). 
 7
Social Functions and Benefits 
Evidences also have been shown that urban forests may reduce human stress 
levels (Ulrich, 1984), promote social integration of older adults with their neighbors 
(Kweon et al., 1998), and provide local residents with opportunities for emotional and 
spiritual fulfillment that help them cultivate a greater attachment to their residential areas 
(Chenoweth and Gobster, 1990). Furthermore, the presence of trees and ?nearby nature? 
in human communities generates numerous psychosocial benefits. Kuo (2003) have 
found that having trees within high density neighborhoods lowers levels of fear, 
contributes to less violent and aggressive behavior, encourages better neighbor 
relationships and better coping skills. Hospital patients recover more quickly and require 
fewer painkilling medications when having a view of nature. Office workers with a view 
of nature are more productive, report fewer illnesses, and have higher job satisfaction. 
These are important, but often unnoticed, effects for urban people who have views of 
trees and nature in the course of their normal, everyday activities and experiences.  
As these findings have emerged, urban forests exhibit many characteristics which 
contribute to the public good. Studies about its public value have been conducted broadly, 
as listed above. However, as a commodity, its supply and demand haven?t been well 
studied from an economic view. What factors contribute to the variation in the demand 
for urban forest is a topic with critical implications to the enactment of effective urban 
forest policies. Although researchers have noticed that urban forest canopy cover 
correlates with ecological and geographic factors as well as with urban form, they have 
not shown how canopy cover varies with socioeconomic conditions that are known to 
 8
vary across all regions. The absence of this critical point makes it impossible to fully 
understand the various supply-demand balances reached in different cities. In this paper, 
a theoretical economic model will be introduced. Then the urban forest canopy cover data 
for all big cities, as well as other socioeconomic data, will be used to test and illustrate 
factors contributing to variations in the demand for urban forests.  
 9
2.  Economic Model of the Demand for Urban Forest 
Many public choice theorists, including Borcherding and Deacon (1972), 
Bergstrom and Goodman (1973), and Perkins (1977), have assumed a simple majority 
rule voting system and have estimated public demands by observing the behavior of the 
median voter. A median voter is assumed to pay the median tax price and has the median 
income level in a jurisdiction.  Another means for obtaining estimates of the demand for 
public goods proposes a two-stage procedure which uses the marginal prices from the 
first-stage hedonic regression to estimate the second-stage demand functions for 
characteristics. Studies conducted using this method include Nelson (1978), Harrison and 
Rubinfeld (1978), Palmquist (1982, 1983), Chattopadhyay (1999). 
The demands for urban forests reflect the behaviors of public as well as individual 
choice and decision. Urban forests are supplied by both public as well as private entities, 
it is very difficult to observe the demand for urban forests from the market and at 
individual household level. We have to aggregate the demand at city level, and estimate 
the per capita demand at the average level. Aggregate demand for urban forests is the 
total amount of money that each city spends to have varying quantities of urban trees at 
different socio-economic and price levels. To investigate at the city level is a good 
alternative. So each city can be viewed as an approximate decision maker.  
Our theoretical economic model, closely following Bates and Santerre (2001), 
begins by supposing that a typical person receives utility from units of urban forest 
enjoyment, q, and all other public and private goods as a composite good, y. Now 
suppose that local officials allocate the public budget and land, while individuals choose 
 10
their communities and homes to maximize the utility, U, of the average person in Eq. 1 
subject to his or her income constraint in Equation (2) , 
U = U (q, y)                                         (1) 
I = payment for enjoyed urban forest + P
y
 y                         (2) 
Where, I is the typical person?s income; P
f
 is per unit price of urban forest; P
y
 is per unit 
price of the composite good, y. 
The amount of enjoyable units of urban forest for a typical person depends on the 
total amount of urban forest, Q, population in the city, N, and the degree of publicness 
associated with the urban forest, p. Mathematically, the enjoyable units of urban forest 
can be expressed as 
                                    q = Q / N
p
                                             
When p=0, urban forest is considered a pure public good and everyone in the city can 
enjoy Q units of urban forest. When p=1, urban forest is considered a private good and 
everyone in the city can on the average enjoy Q/N units of urban forest. Intermediate 
values of p represent some congestability or rivalry in consumption. 
The price share for a unit of enjoyable urban forest can be expressed as  
P
f
 *N
p-1
                                     
When p=0, urban forest is a public good and the price share of a typical citizen for a unit 
of enjoyable urban forest is P
f
 /N, because this person only needs to pay his/her share. 
When p=1, urban forest is a private good and the price share of a typical citizen for a unit 
of enjoyable urban forest is P
f
 , because he/she is the only person who enjoyed this urban 
forest and needs to pay all the price.  
 11
The payment of a typical citizen for his/her urban forest enjoyment is the price 
share for a unit of enjoyable urban forest times the amount of urban forest he/she can 
enjoy. It can be expressed as  
(P
f
 *N
p-1
 )* (Q / N
p
) =P
f
 Q / N 
It means that, in a specific city where P
f
, Q and N are determinate, every citizen 
should on the average pay P
f
 Q / N for their urban forest enjoyment, regardless of the 
publicness of the urban forest. This is reasonable. In a city, there are at least three kinds 
of urban forest: 1) private urban forest, e.g., trees in individual backyards; 2) public urban 
forest, e.g., trees in city parks; and 3) half-public urban forests. e.g., trees in community 
parks or residential communities. The urban forest enjoyed by every person always 
comprises these three categories. The enjoyment amount of every category can vary from 
person to person. For example, rich people may consume more private urban forest than 
poor people. However, the summation of every citizen?s payment for their consumed 
urban forest should equal the aggregate urban forest price in the city. On average, every 
citizen pays P
f
 Q / N for the urban forest.  
Based on the above analysis, the income constraint in equation (2) can be 
expressed as 
I = (P
f
 *N
p-1
) q + P
y
 y                       (3) 
P
f
 *N
p-1
 is the shared price for one unit of  urban forest enjoyment and q is the units of 
urban forest enjoyment captured by this typical person. The typical person?s demand for 
units of urban forest enjoyment can be derived from the utility maximization process and 
given in a general form as 
q = q [P
f
 *N
p-1
, P
y
 , I ]                (4) 
 12
Assuming that the demand function in equation 4 can be written in constant 
elasticity form and that P
y
 =$1, the demand function can be specified as  
q = k (P
f
 *N
p-1
)
a
 I
b
                      (5) 
In equation (5), a and b denote the price elasticity and income elasticity of demand for 
urban forests. Because q is not measurable, equation (5) is multiplied by N
p
 on both sides 
to determine Q, the total amount of urban forest, and then divided by N to express urban 
forest in per capita terms, yielding 
Q= k(P
f
 *N
p-1
)
a
I
b 
N
p
Q/N= k(P
f
 *N
p-1
)
a
I
b 
N
p-1
Q/N= kP
f
a
I
b
N
(p-1)(a+1)
                       (6) 
The dependent variable becomes the per capita urban forest, which is available to us.  
Taking the natural logarithmic transformation gives the final estimation equation for 
econometric analysis,  
ln(Q/N) = b
0
+b
1
ln(P
f 
)+b
2
ln(I)+b
3
ln(N)       (7)           
where b
0
=ln(k), b
1
=a, b
2
=b, b
3
=(p-1)(a+1).   
Demand for urban forests should negatively respond to its price (b
1
<0), while a 
positive relation between per capita income and the demand for urban forest can be 
expected (b
2
>0). In fact, the positive effect of income on the demand for urban forest 
might be specified from two aspects. Firstly, with higher per capita income, the city has 
more budget for urban tree programs. Secondly, wealthy individuals are affordable to 
have bigger lot for their house and are able to spend more money in their budgets for 
landscaping in the construction of their houses, more importantly thereby causing more 
trees to be planted or maintained.  
 13
Besides, an interesting question concerns whether urban forest represents a luxury 
good with the income elasticity greater than 1. As discussed later, many researchers have 
found empirically that parks and recreation services, complements to urban forests, 
resemble a luxury good. The estimated coefficient on population gives us an indicator for 
the effect of population growth on urban forest demand. Population growth tends to 
indicate higher rates of landscape fragmentation as more competing demands are placed 
on limited resources. Because concentrated populations place many pressures on 
vegetation growth, a negative relationship between population growth and per capita 
demand for urban forest is hypothesized (b
3
<0). 
Another necessary control variable that must be considered in our model is the 
natural environmental factor. It is well known that natural vegetation in undisturbed 
environments is primarily a function of temperature and precipitation or geographic 
factors such as ecoregion or altitude that correlate with them. A large area that includes 
generally similar ecosystems and that has similar types, qualities, and quantities of 
environmental resources is known as an ecoregion. Nowak et al. (1996) and Dwyer et al. 
(2000) show that urban tree canopy cover also is highest in forested ecoregions, followed 
by grasslands and deserts, thus confirming ecoregion as a indispensable contributor to 
urban canopy variation at a national scale. These findings are not surprising, but they are 
important, for they make it clear that human modification of vegetation in urban places 
through activities like irrigation have not fully superseded the general effects of 
ecoregion characteristics.  
Moreover, in a dynamic context, it becomes more understandable that the 
ecoregion condition might influence the amount of urban forest under urbanization or 
 14
urban sprawl. In forested ecoregions, the cities are surrounded by forestland. Once the 
city grows, more forestland will be delimited within city limits.  Although part of the 
forestland will be converted into other uses such as residential or even commercial use, 
the newly added area will greatly contribute to the increase of urban forest. However, in 
grassland or desert ecoregions, the situation will be different. Most regions outside the 
city limit will have a lower forest coverage than urban area, leading to a mathematical 
decrease of its average urban forest coverage under urban growth. Of course, once the 
area has been converted into urban use, tree canopy coverage is expected to increase in 
the future, due to the impact of human behavior. In conclusion, the ecoregion factor will 
have a significant contribution to our model. After adding a dummy of ecoregion, D
eco
, 
equation (7) changes into:  
ln(Q/N) = b
0
+D
eco
+b
1
ln(P
f 
)+b
2
ln(I)+b
3
ln(N)     (8)           
 15
                                                
3. Data 
Our research will address all the big cities with population over 100,000 in the 
United States. After deleting some cities with missing data or incorrect data
1
, we finally 
get 242 cities. The locations of these sample cities are exhibited in Figure 1-1. 
Urban Forest Canopy Cover 
The United States Department of Agriculture (USDA) Forest Service collected 
and published canopy cover data (Dwyer et al., 2000) in accordance with the Forest and 
Rangeland Renewable Resources Planning Act (1974), which requires the Forest Service 
to assess ?the current and expected future conditions of all renewable resources in the 
Nation?(USDA Forest Service, 1989). The Forest Service has summarized results at state, 
county, metropolitan statistical area (MSA), urban area, and census-designated place 
(CDP) levels for the entire contiguous United States. These estimates of canopy cover are 
based on the USDA?s national resources inventory (NRI) and advanced very high-
resolution radiometer (AVHRR) data. Urban forest canopy cover, on a 0-100 percentage 
scale, was calculated for every 1 km
2
 in the United States using statistical models for 
particular physiographic regions and 1991 AVHRR data.  
These statistical models predict forest density per square kilometer based on the 
proportion of individual AVHRR pixels or cells within it with particular land cover. 
Selected jurisdictional boundaries (e.g., state, county, urban area) were added to the data 
set after the complete coverage for the United States was generated. The accuracy of the 
 
1
 The urban tree coverage in some cities is less than 0.05%. In these cases, the coverage 
percentage is regarded as 0 in the National Urban Forest Report (Dwyer et al., 2000). 
 
 16
estimates of canopy cover was determined through comparisons with canopy inventories 
of selected urban areas around the United States, based on aerial photography (Nowak et 
al., 1996). However, the urban forest canopy cover data are statistical estimates and are 
most suitable for large areas (Dwyer et al., 2000). Despite this limitation, the data are 
well suited for our analysis since the minimum land area of our sample cities is 27.1 km
2
.  
Based on the urban forest canopy cover data, land area data, and population data, we can 
calculate our dependent variable, per capita urban forest amount, for each sample city. 
Ecoregion Classification Data 
In the mid-1990s, the National Interagency Technical Team (NITT) was formed 
to develop a common framework of ecological regions for the nation. The intention is 
that the framework will foster an ecological understanding of the landscape, rather than 
an understanding based on a single resource, single discipline, or single agency 
perspective. Till now, there are two broadly recognized ecoregion division systems: 
Omernik?s ecoregion system and Bailey's ecoregion system. After comparing their 
different classification criteria, we find Omernik?s ecoregions are more suitable to our 
analysis.   
The Omernik ecoregion system is hierarchical and considers the spatial patterns 
of both the living and non-living components of the region, such as geology, 
physiography, vegetation, climate, soils, land use, wildlife, water quality, and hydrology. 
There are four levels in the Omernik ecosystem hierarchy. Level I ecoregions were 
mapped and described by the North American Commission for Environmental 
Cooperation (CEC) in 1997. A combined data set in Arc/INFO Export format, with Level 
 17
                                                
I, Level II, and Level III ecoregions for all of North America, is available from the EPA 
Ecoregions of North America download page
2
.  
In this study, a mixed use of Level I and Level II ecoregions was proposed. In 
southern Florida, the Level I ecoregion system classifies this region as ?Tropical Wet 
Forests?. But in Level II, this region is defined as ?Everglades?, which is not apt for tree 
growth. The tree canopy coverage data collected from Dwyer et al. (2000) also attests the 
low canopy percentage in this region. All the sample cities in this region have their tree 
canopy coverage below 5%, with some of them even below 1%. Moreover, in the central 
US, Level I generally classifies this region as ?Great Plains?. But as stated in the Level II 
ecoregion system, ?Great Plains? includes temperate prairies, west-central semi-arid 
prairies, south central semi-arid prairies, Texas-Louisiana coastal plain, Tamaulipas-
Texas semi-arid plain. Urban forest coverage varies greatly among these regions, with 
normally over 10% in temperate prairies or Texas-Louisiana coastal plain and less than 
5% in others. In these cases, the Level I classification of ecoregion is neither sufficient 
nor accurate for our study. Based on Omernik?s Level I and Level II ecoregion divisions, 
a revised ecoregion classification for our specific study is presented in Figure 1-1.  
 
Figure 1-1 Ecoregions of selected cities in Continental US 
 
As soon as we ascertain the ecoregion division, we can use the ArcMap to match 
each sample city with the ecoregion map and extract the information of which ecoregion 
 
2
 http://www.epa.gov/wed/pages/ecoregions/na_eco.htm#Downloads 
 18
each city belongs to. This information is then used to build an ecoregion index with 
values shown in Table 1-2. 
Economic and Demographic Data 
Demographic and socio-economic data, such as population, land area, per capita 
income, etc., come from the U.S. Census Bureau. Since the price of urban forest is 
unavailable, we use the opportunity cost of urban forest as its price. Urban forest, as one 
kind of land use within city limits, competes with other land use types such as 
commercial land and residential land. After purchasing one lot of residential land, the 
owner can decide what percent of this lot will be used to build house and what percent 
will be used to plant trees or lawns. In this case, the price or opportunity cost of urban 
forest is best exhibited by the residential land price. 
Unfortunately, the residential land price for these sample cities is also unavailable. 
At the national level, researchers have concluded that the logarithms of the nominal price 
index for residential land, disposable income, and interest rates are cointegrated (Davis 
and Heathcote, 2004). However, this research addresses the aggregate residential land 
price across the whole nation. At the city level, very few studies have been conducted. 
Davis (2005) conducted a research on land value of an average owner-occupied single-
family lot in 44 large cities by Metropolitan Statistical Area. This is the only available 
data of the residential land price in specific cities. We will use this available residential 
land price in 44 cities, and single-family owner-occupied house value which is available 
in the US Census, to estimate the residential land price for each sample city in our study.  
Previous studies have shown that residential land price is mainly correlated to 
house value, population, and city land area. Based on the existent residential land price of 
 19
44 big cities, we regress the residential land price on house value, population and land 
area to get the coefficients of every independent variable. Data are logged prior to 
estimation of the model to correct for nonnormality of the distributions. 
The results of this regression including the values of each coefficients and t-ratio 
are listed in Table 1-1. The R
2
 of 0.87 indicates the strong explanation power of our 
model and the high reliability of our forthcoming estimation for residential land price in 
other cities which is based on this model.  
 
Table 1-1: Results for the Regression of Residential Land Value 
 
Based on the coefficients of every independent variables: population (Pop), land 
area (LA), and single house value(HV), we estimate the residential land price (LV
resi
) for 
each sample city in our study using the following equation:  
Log (LV
resi
)= -10.848+0.31*Log (Pop) -0.395*Log(LA) +1.160* Log (HV).  
The estimated residential land value is described in Table 1-2. 
 
Table 1-2: Data Description of Variables 
 
 20
4. Results 
Table 1-2 presents the data description of all the variables in our empirical 
analysis. The ecoregion index, as a control variable capturing the availability of natural 
environmental effect, is inappropriate to be expressed in log form. The values of other 
variables are logged prior to estimation, according to the analysis of our theoretical model. 
Standard ordinary least square estimates are obtained for the demand equation and 
presented in Table 1-3.  
 
Table 1-3: Regression Results for the Demand for Urban Forests 
 
The regression results show that all of the estimated coefficients have their 
expected signs and are statistically significant at the 1% level. As expected, the 
coefficient on population provides a negative relationship between population growth and 
the per capita demand for urban forest. Ecoregion index in our model exhibits a very 
significant influence on the demand for urban forest. It supports that natural 
environmental factors still haven?t been completely replaced by the increasingly intense 
impact of human behavior. The positive sign before ecoregion index attests to the 
conclusion that Nowak et al. (1996) and Dwyer et al. (2000) suggested: urban tree canopy 
cover is also highest in forested ecoregions, followed by other ecoregions such as 
grasslands and deserts, 
As hypothesized, the demand for urban forest varies positively with income. The 
income elasticity of the demand for urban forest is 2.4, indicating urban forest is highly 
responsive to changes in income and may exhibit some characteristics of a luxury good. 
 21
This income elasticity estimate means that a 1% increase in per capita income would 
cause a 2.4% increase in the demand for urban forest. For a typical city with 100,000 
people and an average of 100 m
2
 urban forest canopy coverage per person, that 1% 
increase in per capita income will cause an increase of 2.4 m
2
 in per capita urban forest 
demand and an aggregate of 0.24 km
2
 additional demand for urban forest.   
Similarly, the demand for urban forest varies inversely with its price, as we 
expected. According to the regression results, the price elasticity of the demand for urban 
forest is approximately -1.37, indicating that demand for urban forests is relatively 
sensitive to changes in its price. This price elasticity estimate means that with a 1% 
increase in the price of urban forest, the demand for urban forest will decrease 1.37%. For 
the same typical city, a 1% decrease in the price of urban forest will cause an increase of 
1.37 m
2
 in per capita urban forest demand and a total amount of 0.137 km
2
 additional 
demand for urban forest. 
Another important objective of this paper is to find the effect of urban sprawl on 
the demand for urban forest. Although urban sprawl may convert some forestland into 
urban use and thus decrease the amount of aggregate forest area, it also extends the city 
limits and more forest becomes defined as urban forest. For example, developers today 
often try to incorporate rural tree stands into new urban developments. However, opposite 
conclusions also exist. In studying the fate of natural rural vegetation during urbanization 
from 1937-1975 in southeastern Wisconsin, Sharpe et al. (1986) concluded that rural 
landscapes undergo complex transformation during urbanization with most rural forests 
being destroyed rather than being incorporated into urban areas. Whether the effect of 
urban sprawl on the demand for urban forest is positive or negative remains unknown and 
 22
the research on this issue will have critical implications to the enactment of future urban 
forest policy.  
Urban sprawl, as a very popular topic, has been studied for a long time. Many 
researches have been conducted covering every aspect of this topic since the 1970s. A 
more recent comprehensive research report on urban sprawl index has been published by 
Smart Growth America in 2002. The title of the report is Measuring Sprawl and its 
Impact (Ewing et al. 2002). This comprehensive and academically rigorous study truly 
breaks new ground by going a step beyond the index to demonstrate how sprawling 
development patterns affect the way people live. It uses 22 variables to rate metro areas 
on four different aspects of their development. The ?scores? for each factor indicate how 
badly those regions have sprawled in terms of housing and population spread; segregating 
homes from the activities of daily life; lacking the focus of strong economic and social 
centers; and building poorly connected street networks. The final report ranks 83 
metropolitan areas according to their overall sprawl index scores. Based on this sprawl 
index score, we will test the effect of urban sprawl on urban forest.  
From the regression results in the second column of Table 1-3, we find the 
coefficients for income, price, population and ecoregion index still keep close to the 
results without considering the urban sprawl index (the first column of results in Table   
1-3). The adjusted R
2
 increases very little, indicating that this additional variable doesn?t 
contribute strongly to the explanatory power of our model.  However, the t-ratio shows 
this new variable is statistically significant. Since a higher index score means less 
sprawling, the negative sign before the sprawl index score indicates a positive 
 23
relationship between urban sprawl and demand for urban forest. The more sprawl a city 
has, the more demand for urban forest it has.  
While the results of using this sprawl index were satisfying, this method also 
greatly reduced the number of our sample, since the final report of Measuring Sprawl and 
its Impact only ranks 83 metropolitan areas. After reviewing the regression process, we 
find our observations in the regression model decreased from 242 to 133. Moreover, 
some of our sample cities must have the same sprawl index score, if they are located in 
the same Metropolitan Statistic Area.  
Although these limitations of the sprawl index used in our model exist, the 133 
samples can still give us a solid conclusion. Frankly, it is possible that the coefficient of 
sprawl index might change if we use different sprawl indicators such as population 
density change as well as a different number of samples. However, what we are 
concerned with in our analysis is whether urban sprawl has a positive or negative 
influence on the demand for urban forest. How seriously urban sprawl will affect the 
demand for urban forest is too difficult to identify and is well beyond the focus of this 
paper.  
 
 24
5.  Conclusions and discussion 
One of the most important empirical conclusions we make from this study is that 
higher income will cause more demand for urban forests. Our empirical findings in this 
study suggest that the demand for urban forest is elastic with respect to price and highly 
responsive to changes in income. The status of urban forest is a good indicator of urban 
environmental quality. Higher income will lead to higher environmental quality at the 
expense of alternative land use and the planting and maintaining of urban trees. Therefore, 
although economic development may convert more land, including open and green space, 
for construction purpose, societal wealth is significant in affording a higher quality 
environment. This is another evidence in support of the Environmental Kuznet?s Curve, 
which claims that in the beginning of economic development pollution increases along 
with industrialization; after a threshold, when basic physical needs are met, interest in a 
clean environment rises, reversing the trend. 
Our conclusion is consistent with other empirical evidence concerning the 
demand for public parks, recreation services, and environmental quality. Borcherding and 
Deacon (1972) found the own price elasticity for Park-Recreation is -.50 and -.41. 
Bergstrom and Goodman (1973) reported an average price elasticity estimate of -.19 for 
parks and recreation services. Perkins (1977) found a price-elastic demand for park and 
recreation with an average elasticity estimate of -2.12, while Santerre (1985) uncovered 
price elasticity estimates of -.35 on average. Other research concerning environmental 
quality also concluded similar own price elasticity. Nelson (1978) found that air quality 
price elasticity ranges from -1.2 to -1.4, while Bender et al. (1980) reported a range from 
-0.262 to -0.503. Zabel and Kiel (2000) found a price elasticity of -0.479 for ozone and -
 25
0.128 for particulates. More recently, Brasington and Hite (2005) concluded their price 
elasticity of demand for environmental quality to be -0.12. Our price elasticity estimate of 
-1.37 for urban forest is comparable to the results of these studies. 
As far as income elasticity is concerned, Borcherding and Deacon (1972) reported 
estimates ranging from 1.29 to 2.74 for parks and recreation services whereas Bergstrom 
and Goodman (1973) estimated an income elasticity of 1.32. Other findings about income 
elasticity estimates for parks and recreation services were relatively lower, with an 
average of 0.65 for Perkins (1977), and 0.71 for Santerre (1985). Our income elasticity 
estimate of 2.4 for urban forest compares a little bit higher to most of the estimates for 
parks and recreation services. This is reasonable because that urban forest is more private 
compared to other public goods such as parks and recreation services. Privately owned 
urban forest, such as trees in the backyard, can be seen everywhere and will greatly 
contribute to the whole urban forest system. However, this is not the case for parks or 
other recreation services. Private parks are probably only possessed by wealthy people 
and average income families can barely afford them. In this case, personal income 
contributes more to the demand for urban forest, due to its more private characteristics, 
thereby inducing higher income elasticity relative to parks and other recreation services.  
Another interesting and reasonable finding is that urban sprawl may impose a 
growing demand for urban forest. The more sprawl a city is experiencing, the more 
demand for urban forest it has. This result proves that urban sprawl is not as bad as many 
researchers have stated, at least from the perspective of urban forest. One driving force of 
urban sprawl is that people look for bigger green space and house lots. People who prefer 
to live in suburban areas are often wealthier and place more emphasis on environmental 
 26
quality. But we need to keep in mind that the urban sprawl index used in this paper 
reflects many other social factors in addition to urban expansion, since we were not able 
to find another index that measures urban expansion more precisely. It could be an 
important aspect that we should examine in future studies. 
We must point out some weaknesses of this study. Even though we used the 
ecoregion dummy, this only provides a rough estimate due to a great variety of natural 
conditions such as landscape, soil, climate, etc. Moreover, different model specifications 
or different sample cities will possibly change the size of coefficients. So we need to be 
cautious when we interpret income as well as price elasticity.  However, the results will 
not vary essentially and will still have applicable significance.
 27
PART II ? DEMAND FOR URBAN FORESTS AND ECONOMIC WELFARE: 
EVIDENCE FROM THE SOUTHEASTERN US CITIES 
 
1. Introduction 
Economics is the study of how individuals, as well as societies, allocate scarce 
resources to satisfy their various needs. Economic decisions are reflected not only in 
individual choices, but also in public decisions such as public budgets, policies and 
regulation. An important aspect of economic choices is associated with enjoyment of 
environmental amenities versus traditional economic goods. The Environmental Kuznets 
Curve (EKC) was coined after the relationship between environmental quality, such as air 
quality, and income was found following the analogous relationship between income 
inequality and national income first observed by Simon Kuznets (1955). EKC has been 
tested in many studies (e.g., Stern et al., 1996; De Groot et al., 2004; Lindmark, 2004; 
Rupasingha et al., 2004).  
Studies of forests in this empirical framework have focused on the relationship 
between forest coverage and income at the national level and regional level. The results 
were mixed. Shafik and Bandyopadhyay (1992) found that net change in forest cover did
 28
not significantly relate to income in 149 countries between 1961 and 1986. Panayotou 
(1993) used strictly cross-sectional international data and found a turning point in 
deforestation at $1275 (in 1985 prices) of household income. Cropper and Griffiths (1994) 
created pooled time series cross-section data for three separate regions of the world and 
found that per capita national income was a significant factor in both Africa and Latin 
America, but not in Asia.  
So far we have not found any similar studies on urban forests. In fact, one of the 
best indicators of the urban environment and amenities is the status of trees present in a 
city. Trees have been recognized as an important component of urban landscapes 
throughout the history of urbanization. Sociologists and economists found that urban 
trees, in addition to providing environmental and aesthetic benefits, also brought a broad 
range of economic, social, and even psychological benefits. Trees in urban landscapes 
moderate temperature and microclimates, thereby reducing the needs for air conditioning 
and thus saving energy (Heisler, 1986; McPherson, 1990; Meier, 1991; Oke, 1989).  
Urban trees help improve air quality and sequester carbon (Nowak, 1993; Nowak 
and McPherson, 1993; Rowntree and Nowak, 1991; Smith, 1981), help stabilize soils, 
reduce erosion, improve groundwater recharge, control rainfall runoff and flooding 
(Sanders, 1986), reduce urban noise levels (Cook, 1978), and provide habitat that 
increases biodiversity (Johnson, 1988). Urban trees also make neighborhoods 
aesthetically more appealing and add to the value of property (Schroeder, 1989). 
Evidence has also been shown that urban forests may reduce human stress levels (Ulrich, 
1984), promote social integration of older adults with their neighbors (Kweon et al., 
1998), and provide local residents with opportunities for emotional and spiritual 
 29
fulfillment that help them cultivate a greater attachment to their residential areas 
(Chenoweth and Gobster, 1990). 
Trees in cities are good but are not free. They require space that is usually very 
costly in a city as well as planting and maintenance. Any community has to face the 
difficulties in allocation of its limited budget for planting trees and other purposes, in 
allocation of the urban land for planting trees and other alternative uses. Individuals have 
to make the decision of what size of lot to purchase for their homes and in which kinds of 
urban settings. So lot size and tree presence reflect, to some extent, the market forces 
determined by the welfare of the city citizens and their preferences.  This study tests the 
relationship between the economic welfare and the tree presence in urban areas.  
At the city level, which factors contribute to the variation in status of urban 
forests is interesting and may have some policy implications. Although researchers have 
noticed that urban forest canopy cover correlates with ecological and geographic factors 
as well as with urban form, they have not shown how canopy cover varies with 
socioeconomic conditions across all regions. Is there an EKC for urban forests?  In the 
following sections, we first introduce econometric models and data, and then the results 
are presented and conclusions are made.  
 
 30
2. Econometric Model  
Urban forests are either public goods, private goods or a combination of both. 
They are determined by demand and supply. Unfortunately, it is impossible to get the 
prices and costs. Neither shadow prices nor instrumental prices or indicators, such as the 
residential land values, are available for each city.  Only two variables, population 
density and income, that should be strongly related to the presence of urban forests, are 
obtainable for all cities. Since other variables (such as residential land value) might be 
fundamentally determined by these two variables, we simply use the reduced form of 
urban forests (FOR) as a function of population density and income: 
FOR= F (POD, INC) + e 
(1) lnFOR=a
0
+a
1
lnINC+a
2
lnPOD+a
11
(lnINC)
2
+a
22
(lnPOD)
2
+a
12
(lnINC*lnPOD)+e
i
      
Where FOR represents the percentage of urban forest canopy coverage; INC is the 
median household income in 2000; POD represents the population density in the city; 
a
1
,a
2
,a
11
,a
22
,a
12
 are the coefficients of the variables respectively; e
i
 is the error term. It 
should be noted that a
11
 and a
22
 measure the second-order effect of income and 
population density on the urban forest canopy cover percentage, respectively; and a
12
 
measures the cross effect.  
There are no studies on urban forests, but some studies on other issues may be 
relevant. For example, most studies have concluded that public parks or recreation 
services, a substitute for urban forests, is a normal good with a positive income elasticity, 
either less than 1 or greater than 1 (Borcherding and Deacon 1972, Bergstrom and 
Goodman 1973, Perkins 1977, Santerre 1985). Basically, all these conclusions agree that 
higher income will result in more demand for environmental amenity. The difference 
among them is only whether environmental amenity represents a luxury good with the 
income elasticity greater than 1.  
EKC suggests that urban forest would decrease firstly with economic 
development since people choose to sacrifice environment in order to get other uses, but 
later it would increase with economic development because wealthy people can afford 
more environmental amenities. In fact, this subsequent positive effect of income on the 
demand for environmental amenity might be specified from two aspects. Firstly, with 
higher income, the city gets richer and has more money in the budget for urban 
environmental programs. Secondly, rich people will also have more money in their 
budgets for landscaping in the construction of their houses, thereby causing more trees to 
be planted or maintained.  
Our economic model can test whether there is a threshold that can change the 
impacts of income on urban forests. To get the turning point, we simply derive the 
function by income:  
0lnln2
ln
ln
12111
=++=
?
?
PODaINCaa
INC
FOR
 
(2)  
11
121
2
ln
*
a
PODaa
eINC
??
=                                                
We suppose that a similar relationship may exist between urban forests and 
population density. The biggest difference between urban areas and rural areas is that 
there exist various urban management programs in cities. With people firstly clustered in 
cities, urban services and programs, including urban forest programs, start to provide 
citizens abundant urban civilizations. At the beginning of urbanization, the clustering of 
people doesn?t actually reduce the urban forest volume. Inversely, various urban forest 
 31
programs will have an overwhelming influence on the volume and health of our urban 
forests.  
To get the turning point, we derive the function by population density:  
0lnln2
ln
ln
12222
=++=
?
?
INCaPODaa
POD
FOR
 
(3)  
22
122
2
ln
*
a
INCaa
ePOD
??
=                                                
 
 32
 33
3. Data 
Considering that the natural environment will have a great impact on the urban 
tree situation, we try to find a region with relatively more homogenous climate and 
environment. So we decide to select nine southeastern US states (Alabama, Florida, 
Georgia, Kentucky, Mississippi, North Carolina, South Carolina, Tennessee and Virginia). 
Cities with populations below 40,000 are considered rural communities and their 
surrounding areas become strong substitutes to urban forests, therefore these cities are 
excluded from our analysis. We selected a total of 149 cities for this study. Demographic 
and economic data, such as population, land area and median household income, are 
obtained from the U.S. Census Bureau.  
The United States Department of Agriculture (USDA) Forest Service collected 
and published forest canopy cover data (Dwyer et al., 2000) in accordance with the Forest 
and Rangeland Renewable Resources Planning Act (1974), which requires the Forest 
Service to assess ?the current and expected future conditions of all renewable resources 
in the Nation?(USDA Forest Service, 1989). The Forest Service has summarized results 
at state, county, metropolitan statistical area (MSA), urban area, and census-designated 
place (CDP) levels for the entire contiguous United States. These estimates of canopy 
cover are based on the USDA?s national resources inventory (NRI) and advanced very 
high-resolution radiometer (AVHRR) data. Urban forest canopy cover, on a 0-100 
percentage scale, was calculated for every 1 km
2
 in the United States using statistical 
models for particular physiographic regions.  
These statistical models predict forest canopy per square kilometer based on the 
proportion of individual AVHRR pixels or cells within particular land cover. After the 
 34
complete coverage for the United States was generated, selected jurisdictional boundaries 
(e.g., state, county, urban area) were added to the data set to extract the urban forest 
canopy cover percentage within these boundaries.  Table 2-1 presents the data description 
of all the variables in our empirical analysis. 
 
Table 2-1: Data Description of Variables 
 
 35
4. Results  
Standard ordinary least square estimates are used for the regressions. The results 
are presented in Table 2-2. Moreover, we compare two models in our estimation to infer 
the significance of the cross effect of income and population density on the demand for 
urban forest, one with the interaction part (Model A), another one without the interaction 
part (Model B).  
 
Table 2-2:  Regression Results  
 
We find that including the cross effect term in the empirical model decreases the 
adjusted R
2
 value (from 0.408 to 0.405), indicating that the cross effect term doesn?t 
contribute to the explanation power of the model. Moreover, the t ratio of the cross effect 
term is as low as 0.58, suggesting that its value is not statistically significant at all. 
Therefore, we will use Model B to interpret our results.  
 
The regression results in Model B show that all of the estimated coefficients are 
statistically significant at the 1% level. As expected, the positive coefficient on the 
second-order effect of income suggests a first negative and then positive impact of 
income on the demand for urban forests. Inversely, the negative coefficient on the 
second-order effect of population density suggests a first positive and then negative 
influence of population density on the demand for urban forest. Based on Model B, the 
equations used to calculate the threshold income value and population density influence 
in equations 2 and 3 will be transformed as below: 
(4)   
11
1
2
*
a
a
eINC
?
=                                                
(5)   
22
2
2
*
a
a
ePOD
?
=                                             
Substituting the coefficients estimated for Model B into the above equations, we get that 
the income threshold value is $38,739 per household and the population density threshold 
value is 179 persons per square kilometer.  
The existence of income threshold value provides more powerful evidence in 
support of the EKC. When the household income is less than $38,739, the percentage of 
urban forest cover decreases as income increases, indicating a negative income elasticity. 
As the income approaches the critical point, the income elasticity also approaches 0. 
After the income surpasses the threshold value, the income elasticity becomes positive 
and the demand for urban forest increases with the increasing income. 
Similarly, there also exists a population density threshold value --179 persons per 
square kilometers. When population density is less than 179 persons per square kilometer, 
the percentage of urban forest increases as population density increases. This is because 
the urbanized areas use land more efficiently than rural areas and save more land for 
urban forest development. After the population density surpasses the critical value, the 
demand for urban forest decreases with the increasing population density, due to the 
increasing stress on providing sufficient accommodation. 
The income elasticities of the demand for urban forests for all the sample cities 
are calculated using the following equation: 
(6)  
INCaa
INC
FOR
Income
ln2
ln
ln
111
+=
?
?
=?
 
 
 36
 37
Results are presented in Figure 2-1. The income elasticities vary from -2.86 to +4.92. The 
critical value of the income influence locates on the point where income elasticity equals 
0. As the income gets farther away from this critical value on both sides, the absolute 
value of income elasticity also increases. The highest (+4.92) and lowest (-2.86) income 
elasticities are reached where the highest ($93,561) and lowest ($23,483) income stand. 
The income elasticity for the mean household income ($39,787) in our sample cities is 
0.11, indicating urban forest coverage is inelastic to income. However, we must point out 
that this mean income elasticity doesn?t have much applicable significance, compared to 
the income threshold value found in our analysis. 
 
Figure 2-1: Income Elasticity in Sample Cities 
 38
5. Conclusions 
This paper analyzes the relationships between urban forest presence and income 
and population density. Our results indicate a similar trend of EKC in urban forests.  
With continuous economic development and urbanization, its impacts on urban forests 
are mixed. In general, population growth will cause urban forests to be replaced by other 
land uses. As a result, although urban forest programs still endeavor to protect urban 
forests, many urban forests and green spaces are inevitably converted for construction 
purposes to accommodate the increasing population. In this period, the demand for urban 
forests will continuously decrease due to the increasing population density which places a 
higher and higher pressure on the urban land use. 
Economic welfare will finally play a positive role in urban forest after reaching a 
certain level. Better economic welfare will make people more affordable to have more 
urban forests and other green spaces.  Higher income will lead to higher environmental 
quality at the expense of alternative land use and the planting and maintaining of urban 
trees. Therefore, although economic development may convert more land, including open 
and green spaces, for construction purposes, societal wealth is significant in affording a 
higher quality environment. 
We must point out some weaknesses of this study. Even though we limit our 
sample cities to the southeastern US, the climate and natural conditions, such as 
landscape and soil, still vary significantly from city to city. However from the relative 
good R square, we can say that income and population density are good indicators of the 
variation in urban forests.  
 
 39
LITERATURE CITED 
American Forests, 2002. Urban Ecosystem Analysis for the Washington DC Metropolitan 
Area. American Forests. 
Anderson, L. M., Cordell, H. K., 1985. Residential property values improve by 
landscaping with trees. Southern Journal of Applied Forestry 9, 162-166. 
Anderson, L. M., Cordell, H. K., 1988. Influence of trees on residential property values in 
Athens, Georgia (U.S.A.): A survey based on actual sales prices. Landscape and 
Urban Planning 15, 153-164. 
Bates, L. J., Santerre, R. E., 2001. The public demand for open space: the case of 
Connecticut communities. Journal of Urban Economics 50, 97-111. 
Bender, B., Gronberg, T.J., Hwang, H.-S., 1980. Choice of functional form and the 
demand for air quality. Review of Economics and Statistics 62 (4), 638?643. 
Bergstrom, T. C., Goodman, R. P., 1973. Private demands for public goods. American 
Economic Review 63, 280-296. 
Borcherding, T. E., Deacon, R. T., 1972. The demands for the services of non-federal 
governments. American Economic Review 62, 891-901. 
Brasington, D.M., Hite, D., 2005. Demand for environmental quality: a spatial hedonic 
analysis. Regional Science and Urban Economics 35 (1), 57-82.
 40
Center for Urban Forest Research (CUFR), 2002. USDA Forest Service. 
http://wcufre.ucdavis.edu/ 
Chattopadhyay, S. 1999. Estimating the Demand for Air Quality: New Evidence Based 
on the Chicago Housing Market. Land Economics 75 (Feb.), 22-38. 
Chenoweth, R. E., Gobster, P.H., 1990. The nature and ecology of aesthetic experiences 
in the landscape. Journal of Landscape 9, 1-18. 
Cook, D. I., 1978. Trees, solid barriers, and combinations: Alternatives for noise control. 
Proceedings of the National Urban Forestry Conference (pp. 330-339),  G. 
Hopkins, Ed., New York: USDA Forest Service. 
Crompton, J. L., 2001. Parks and Economic Development. American Planning 
Association. Chicago, IL. 
Cropper, M., Griffiths, C., 1994.  The Interaction of Population Growth and 
Environmental Quality. The American Economic Review 84: 1358-1393. 
Davis, M.A.,  Palumbo, M., 2005. The Price of Land in 46 U.S. Cities. Federal Reserve 
Board (unpublished). 
Davis, M.A., Heathcote, J., 2004.  The Price and Quantity of Residential Land in the 
United States. Federal Reserve Board of Governors and Georgetown University. 
De Groot, H.L.F., Withagen, C. A., Minliang, Z., 2004. Dynamics of China's Regional 
Development and Pollution: An Investigation into the Environmental Kuznets 
Curve. Environment and Development Economics 9: 507-37. 
Dwyer, J.F., Nowak, D.J., Noble, M.H., Sisinni, S.M., 2000. Connecting people with 
ecosystems in the 21st century: and assessment of our nation?s urban forests. Gen. 
 41
Tech. Rep. PNW-GTR-490. Portland, OR. U.S. Department of Agriculture, Forest 
Service, Pacific Northwest Research Station.  
Ewing R., Pendall, R., and Chen, D., 2002. Measuring Sprawl and its Impact. Smart 
Growth America. 
http://www.smartgrowthamerica.com/sprawlindex/sprawlreport.html. 
Forest and Rangeland Renewable Resources Planning Act. Act of Aug. 17, 1974. 88 Stat. 
476, as amended; 16 U.S.C. 1600-1614. 
Harrison, D., Rubinfeld, D. L., 1978. Hedonic Housing Prices and the Demand for Clean 
Air. Journal of Environmental Economics and Management 5(2), 81-102. 
Heisler, G. M., 1986. Effects of individual trees on the solar radiation climate of small 
buildings. Urban Ecology 9, 337-359. 
Johnson, C. W., 1988. Planning for avian wildlife in urbanizing areas in America 
desert/mountain valley environments. Landscape and Urban Planning 16, 245-252. 
Kuo, F. E., 2003. The role of arboriculture in a healthy social ecology. Journal of 
Arboriculture 29(3), 148-155. 
Kuznets, S., 1955. Economic Growth and Income Inequality. The American Economic 
Review 45: 1-28. 
Kweon, B. S., Sullivan, W. C., Wiley, A. R., 1998. Green common spaces and the social 
integration of inner-city older adults. Environment and Behavior 30(6), 832-858. 
Lindmark, M., 2004. Patterns of Historical CO
2
 Intensity Transitions among High and 
Low-Income Countries.? Explorations in Economic History 41(4): 426-47. 
McPherson, E. G., 1990. Modeling residential landscapewater and energy use to 
evaluatewater conservation policies. Journal of Landscape 9, 122-134. 
 42
McPherson, E. G., 1992. Accounting for benefits and costs of urban greenspace. 
Landscape and Urban Planning 22, 41-51. 
McPherson, E. G., 1994. Using urban forests for energy efficiency and carbon storage. 
Journal of Forestry 92(10), 36-38, 40-41. 
McPherson, E. G., Biedenbender, S., 1991. The cost of shade: Cost-effectiveness of trees 
versus bus shelters. Journal of Arboriculture 17(9), 233-241. 
McPherson, E. G., Nowak, D. J., Heisler, G. M., Grimmond, C. S., Souch, C., Grant, R., 
and Rowntree, R. A., 1997. Quantifying urban forest structure, function and value: 
The Chicago Urban Forest Climate Project. Urban Ecosystems 1, 49-61. 
McPherson, E. G., Scott, K. I., Simpson, J. R., 1998. Estimating cost effectiveness of 
residential yard trees for improving air quality in Sacramento, California, using 
existing models. Atmospheric Environment 32(1), 75-84. 
Meier, A. K., 1991. Measured cooling savings from vegetative landscapes. In E. Vine, D. 
Crawley, & P. Centolella(Eds.), Energy efficiency and the environment: Forging 
the link (pp. 321-334).Washington, DC: American Council for an Energy-
Efficient Economy. 
Morales, D. J., 1980. The contribution of trees to residential property value. Journal of 
Arboriculture 1, 109-112. 
Nelson, J.P., 1978. Residential choice, hedonic prices, and the demand for urban air 
quality. Journal of Urban Economics 5 (3), 357?369. 
Nowak, D. J., 1993. Atmospheric carbon-reduction by urban trees. Journal of 
Environmental Management, 37(3), 207-217. 
 43
Nowak, D. J., McPherson, E. G., 1993. Quantifying the impact of trees: The Chicago 
Urban Forest Climate Project. Unasylva, 173(44), 39-44. 
Nowak, D. J., Rowntree, R.A., McPherson, E.G. [and others], 1996. Measuring and 
analyzing urban tree cover. Landscape and Urban Planning 36, 49-57. 
Oke, T. R., 1989. The micrometeorology of the urban forest. Philosophical Transactions 
of the Royal Society of London, 324(Series B), 335-349. 
Palmquist, R. B. 1982. Estimating the Demand for Air Quality from Property Value 
Studies. Department of Economics, North Carolina State University. Mimeo.  
Palmquist, R. B. 1983. Estimating the Demand for Air Quality from Property Value 
Studies: Further Results. Department of Economics, North Carolina State 
University. Mimeo. 
Panayotou, T., 1993. Empirical Tests and Policy Analysis of Environmental Degradation 
at Different Stages of Economic Development.  Geneva: International Labor 
Office,  Technology and Employment Programme Working Paper WP238, 1993. 
Payne, B. R., Strom, S., 1975. The contribution of trees to the appraised value of 
unimproved residential land. Valuation 22(2), 36-45. 
Perkins, G. M., 1977. The demand for local public goods: Elasticities of demand for own 
price, cross prices, and income. National Tax Journal 30, 411-422. 
Rowntree, R. A., Nowak, D. J., 1991. Quantifying the role of urban forests in removing 
atmospheric carbon dioxide. Journal of Arboriculture 17(10), 269-275. 
Rupasingha, A., Goetz, S.J., Debertin, D.L., Pagoulatos, A., 2004. The Environmental 
Kuznets Curve for US Counties: A Spatial Econometric Analysis with Extensions. 
Papers in Regional Science 83(2): 407-24. 
 44
Sanders, R. A., 1986. Urban vegetation impacts on the hydrology of Dayton, Ohio. Urban 
Ecology, 9, 361-376. 
Santerre, R. E., 1985. Spatial differences in the demands for local public goods. Land 
Economics 61, 119-128. 
Schroeder, H. W., 1989. Environment, behavior and design research on urban forests. In 
E. H. Zube & G. T. Moore (Eds.), Advances in environment, behavior and design 
(pp. 87-117). New York: Plenum. 
Shafik, N., Bandyopadhyay, S., 1992.  Economic Growth and Environmental Quality: 
Time Series and Cross-country Evidence. Washington D.C.: The World Bank, 
Background Paper for the World Development Report, 1992. 
Sharpe, D.M., F. Stearns, L.A. Leitner and J.R. Dorney, 1986. Fate of natural vegetation 
during urban development of rural landscapes in southeastern Wisconsin. Urban 
Ecology 9, 267-287. 
Smith, W. H., 1981. Air pollution and forests. New York: Springer. 
Stern, D. I., Common, M. S., Barbier, E. B., 1996. Economic Growth and Environmental 
Degradation: The Environmental Kuznets Curve and Sustainable Development. 
World Development 24(7): 1151-1160. 
Ulrich, R. S., 1984. View through a window may influence recovery from surgery. 
Science 224, 420-421. 
USDA Forest Service, 1989. Draft 1990 RPA program. Washington, DC. 186p. 
Zabel, J.A., Kiel, K.A., 2000. Estimating the demand for air quality in four US cities. 
Land Economics 76 (2), 174?194.
 45
Table 1-1: Results for the Regression of Residential Land Value 
 Coeff. t-ratio 
ONE -10.8479 -9.89513 
Log of Population 0.309519 3.95187 
Log of Land Area -0.39496 -5.50185 
Log of House Value 1.16041 12.6048 
Adjusted R
2
 =0.867   
 
Table 1-2: Data Description of Variables 
Mean S.D. Min. Max. Sample 
Number
Urban Forest Canopy Cover Percentage (%)
b
17.6475 14.9355 0.1 69 242 
Urban Forest Area per capita (m
2
/person) 193.211 305.548 0.21 2126.44 242 
Population 2000
a
303565 620720 82026 8.01E+06 242 
Land Area (km
2
)
a
214.506 263.089 19.5 1965 242 
Population Density 2000 (persons/km
2
) 1716.33 1244.51 225.73 10007.8 242 
per capita Income ($)
a
21009.8 6055.96 9762 68365 242 
Residential Land Value (an average owner-
occupied single-family lot in 44 big cities 
(thousands of current dollars)
c
119.636 121.592 19 602 44 
Single-family owner-occupied house 
value($)
a
138766 76388.3 40900 495200 242 
Estimated Residential Land price  
a. U.S. Census Bureau (2000). 
(thousands of current dollars) 
125.17 103.52 24.81 615.44 242 
1=forest, temperate prairie, coastal plain  Ecoregion index
d
0=desert, semi-arid plain, everglade, and others 
b. Dwyer et al. (2000). 
c. Davis and Palumbo (2005) 
d. Omernik?s ecoregion system  
 46
 47
Table 1-3: Regression Results for the Demand for Urban Forests 
Independent Variables Coefficient 
(t value) 
 Model 1 
(without sprawl index) 
(Sample =242) 
Model 2 
(with sprawl index) 
(Sample =133) 
Constant -11.345 
(5.09) 
-9.522 
(3.36) 
Log of Income 2.405 
(10.48) 
2.386 
(8.43) 
Estimated Log of Urban 
          Forest Price 
-1.368 
(15.85) 
-1.258 
(11.63) 
Log of Population -0.294 
(4.42) 
-0.295 
(4.00) 
Ecoregion Index 2.187 
(18.90) 
2.062 
(14.69) 
Log of Sprawl Index Score  
 
 -0.4488 
(2.33) 
Adjusted R
2
  0.795 0.809 
Notes: 
1) The dependent variable is the log of urban forest area per capita. 
2) Urban forest price is substituted by its opportunity cost -- residential land price. 
3) Ecoregion index is inappropriate to be expressed in log from. 
4) The higher sprawl index score indicates less sprawl of the city  (see Ewing et al. 2002). 
5) The sprawl index score was logged prior to estimation of the model to correct for 
nonnormality of the distribution 
 
Table 2-1: Data Description of Variables 
 Mean S.D. Min. Max. Sample 
Number 
Urban Forest Canopy Cover Percentage 
 (%)
b
27.4 19.7 0.2 74.4 149 
Urban Forest Area per capita  
(m
2
/person) 
422.56 776.002 0.85 8559.47 149 
Population 2000
a 
 
112118 112989 40214 735617 149 
Land Area  
(km
2
)
a
146.24 235.655 12.9 1965 149 
Population Density in 2000  
(persons/km
2
) 
1208.56 796.47 61.45 4831.48 149 
Median Household Income  
a. U.S. Census Bureau (2000). 
($)
a
39786.5 12924.4 17206 93561 149 
b. Dwyer et al. (2000). 
 48
 49
Table 2-2: Regression Results  
Independent Variables Coefficient 
(t value) 
 Model A 
(with cross effect) 
(Sample =149) 
Model B 
(without cross effect) 
(Sample =149) 
Constant 322.093  
(3.53) 
328.696  
(3.64) 
Log of Income -66.1883  
(3.85) 
-64.5134 
(3.82) 
Log of Population Density 8.86332 
(1.12) 
4.3839  
(2.69) 
Square of Log of Income  3.26349  
(3.74) 
3.05329  
(3.85) 
Square of Log of Population Density -0.444164  
(3.53) 
-0.422496  
(3.53) 
Log (Income) *Log (Population Density) -0.398351  
(0.58) 
 
 
Adjusted R
2
  
 
0.405 
 
0.408 
Notes: 
1) The dependent variable is the log of urban forest canopy cover percentage. 
2)   Cross effect is represented by the last term in Model A: a
12
(lnINC*lnPD). 
Figure 1-1: Ecoregions of selected cities in Continental US 
 
 50
Figure 2-1: Income Elasticity in Sample Cities 
 
 51