Content uploaded by L.M. Martinez
Author content
All content in this area was uploaded by L.M. Martinez on Oct 09, 2023
Content may be subject to copyright.
some of the contradictory results over the years have often been
caused by differing methods of analysis, data quality, and regional
differences.
This paper examines the relationship between the availability of
transportation infrastructure and services and the house prices in an
urban area and tries to assess the impact of public investment in
transportation on residential property values. This study was devel-
oped for the Lisbon, Portugal, metropolitan area (LMA) as part of a
broader study that intends to develop a new value-capture financing
scheme for public transportation in the LMA.
The available data focus in three central municipalities in Portugal
(Amadora, Lisbon, Odivelas), where these effects could be more eas-
ily measured because of the existence of a significant variability of
public transportation services.
This study presents several hedonic pricing models to assess the
relationship between transportation accessibility and house values.
These models range from the classic model to spatial hedonic price
models (spatial lag) and include local and systemwide accessibility
indicators. The results of the different models are assessed and com-
pared with the need in mind to forecast house prices in subsequent
phases of the research project.
LITERATURE REVIEW
In the 1960s economists like Alonso and Muth developed the the-
ory for determining residential location in the urban land market (1, 2).
The theory illustrates a model in which a household chooses to
locate at a point where its bid–rent curve intersects with the actual
one, in which the bid–rent curves have a declining gradient with the
distance from the residential location to the central business district
(CBD). However, it might be necessary to consider the effect of
other variables, such as neighborhood characteristics.
The introduction of the hedonic pricing methodology by Rosen (3)
led to an easier way of attributing value to different properties’ fea-
tures. A number of studies have observed the integration of physi-
cal, neighborhood and accessibility characteristics of the property in
models trying to explain the differences in property values or house
prices (4–35). The hedonic price model is a multivariate regres-
sion model for housing values as well as a common robust indirect
approach to valuation, in that its estimates represent the implied prices
that people place on obtaining desirable features of a property and
avoiding undesirable ones (20, 36).
Most commonly, hedonic price models have used ordinary least
squares (OLS) estimation (22, 33, 37–39), but more recently these
Effects of Transportation Accessibility
on Residential Property Values
Hedonic Price Model in the Lisbon, Portugal,
Metropolitan Area
L. Miguel Martínez and José Manuel Viegas
The aim of this paper is to examine the relationship between the avail-
ability of transportation infrastructure and services and the pattern of
house prices in an urban area and to assess whether public investment in
transportation can modify residential property values. This study was
developed for the Lisbon, Portugal, metropolitan area (LMA) as part of
a broader study that intends to develop new value-capture financing
schemes for public transportation in the LMA. The paper focuses on
three central municipalities in Portugal (Amadora, Lisbon, and Odivelas),
where these effects could be more easily measured because of the exis-
tence of a significant variability of public transportation services. The
paper tries to determine, with different spatial hedonic pricing models,
the extent to which access to transportation infrastructure currently is
capitalized into house prices and isolates the influence of three different
transportation infrastructures: metro, rail, and road. The results suggest
that the proximity to one or two metro lines leads to significant property
value changes. Results further indicate that the classic hedonic price model
(ordinary least-squares estimation) leads to similar coefficient values of
the local accessibility dummy variables compared with the spatial lag
model and thus provides a steady basis to forecast the property value
changes derived from transportation investment for the study area in the
absence of a significant property value database.
For decades there has been considerable discussion about the effects
of transportation accessibility on housing prices. It is well known
that a good public transport system provides a high level of access
to work and other activities for households and to businesses for cus-
tomers and employees. The monetary value of this accessibility will
be reflected in the value of a home or a business, in addition to the
value of other features, such as the specific physical attributes of the
building and neighborhood characteristics.
The impact of public transport on property values has been studied
from many perspectives, including analyses of different types of
systems (e.g., rapid, commuter, light rail) and residential versus
commercial impacts. Studies have also attempted to isolate both
positive and negative effects. The varied approaches make it dif-
ficult to compare the results of one study with another. Furthermore,
CESUR, Department of Civil Engineering, Instituto Superior Técnico, Lisbon Technical
University, Avenido Rovisco Pais 1049-001 Lisbon, Portugal. Corresponding author:
L. M. Martínez, martinez@civil.ist.utl.pt.
Transportation Research Record: Journal of the Transportation Research Board,
No. 2115, Transportation Research Board of the National Academies, Washington,
D.C., 2009, pp. 127–137.
DOI: 10.3141/2115-16
127
128 Transportation Research Record 2115
models have been extended to incorporate spatial effects in multiple
ways: feasible generalized least-square estimation (34) and spatial
econometric models (spatial lag and spatial error models) (20, 40).
There are several points of empirical evidence relating the changes
in commercial and residential property market values and transport
investment. Parsons Brinkerhoff (41) concludes that proximity to rail
systems is valued by property owners, and there is little support that
this proximity can decrease property values.
Table 1 presents a compilation of studies from Europe, and Table 2
does the same for North America. As can be seen from the tables,
the evidence is broadly positive, with the widest difference being
found between the residential and commercial markets.
Much of the European research (Table 1) has focused on the resi-
dential market, but in the U.S. research (Table 2), the commercial
market has been the main target. Almost uniformly, the impacts are
seen as positive, with some very large percentage increases, particu-
larly in commercial property values. The enormous variability in
(positive) impact points toward either the importance of other factors,
the specificity of results, or the limitations of the methods used, or a
combination of all these factors.
DATA DESCRIPTION
The data used in this study are 2007 cross-sectional real estate data
from an online realtor’s database (Imokapa Vector) for Lisbon,
Portugal. This database presents the asking price of residential
properties on sale during February of 2007 with a total of 12,488
complete records, 70% inside Lisbon’s municipality. The real estate
data contained information on the asking sale price, structural attri-
butes, and address. The descriptive statistics of the data are pre-
sented in Table 3.
The real estate data was geocoded and imported into a geographic
information system transportation analysis network map. All the prop-
erties were connected with the road and public transport network, to
measure the several accessibility indicators used.
The dependent variable is the advertised asking selling price at
which the owner or realtor is offering the property on the market.
This can be a limitation to the model, because the dependent vari-
able is not directly linked to an equilibrium price, where supply and
demand have cleared the transaction (65). Other studies that relate
public transport accessibility to residential property or land values
also have relied on asking prices (11, 16, 24, 64–66).
The properties of accessibility to public transport and to the
road network were developed by means of two different types of
characterization: local accessibility and systemwide accessibility.
Local accessibility indicators were calculated with the network dis-
tance to public transport entry points (walking distance at a speed of
4 km/h) and to roads considered in the Lisbon Mobility Plan 2004 (67)
road hierarchy at the various distances in meters. The three presented
levels of road hierarchy represent urban motorways for Network1,
urban arterials as Network2, and collector and distributor roads as
Network3.
TABLE 1 Property Value Impacts of Public Transport Proximity in European Cities
Case and Location Impact on Impact Source
Bremen, Germany Office rents +50% in most cases (42)
Croydon, United Kingdom, Tramlink Residential property Some localized positive impacts (43)
Friburg, Switzerland Office rent +15–20% (42)
Friburg Residential rent +3% (42)
Greater Manchester, United Kingdom Not stated +10% (42)
Hannover, Germany Residential rent +5% (42)
Helsinki, Finland, Metro Property values +7.5–11% (44)
London Crossrail Residential and commercial property Positive (45)
London Docklands LRT Residential and commercial property Positive (44)
London JLE Residential and commercial property Positive (46, 47)
Manchester Metrolink House prices Unable to identify (48)
Montpellier, France Property values Positive (42)
Nantes, France, LRT Commercial property Higher values (42)
Nantes LRT Not stated Small increase (42)
Nantes LRT Number of commercial premises +13% (44)
Nantes LRT Number of offices +25% (44)
Nantes LRT Number of residential dwellings +25% (44)
Newcastle upon Tyne, United Kingdom House prices +20% (42)
Orléans, France Apartment rents None, initially negative due to noise (42)
Rouen, France Rent and houses +10% most cases (42)
Saarbrücken, Germany Not stated None (42)
Sheffield, United Kingdom, Supertram Property values Unable to identify (16, 48)
Strasbourg, France Office rent +10–15% (42)
Strasbourg Residential rent +7% (42)
Tyne and Wear Co., Metro Property values +2% (49)
Vienna, Austria, S-Bahn Housing units +18.7% (44)
NOTE: LRT = light rail transit.
Martínez and Viegas 129
These accessibility measures were built with two different
approaches. The first approach considers an all-or-nothing influence
of proximity to public transport entry points and to the road network,
resulting in a set of dummy variables for each public transport mode
or line and for each road hierarchy level.
The second approach considers a continuous decreasing function
of impedance of proximity to public transport entry points and to
the road network. To model this continuous impedance, an inverse
logistic function was used with different parameters for each pub-
lic transport line and road hierarchy. The inverse logistic function
considered is
where Xmax is a specific parameter of each public transport line or
road hierarchy, and aand bare calibrated by considering two different
points in the curve (e.g., 5-min walking distance −Y=0.90, and
15-min walking distance −Y=0.10).
An example of a comparison between both measuring approaches
can be assessed in Figure 1, where the main differences are in the
values observed for accessibility distances greater than the threshold
established for the all-or-nothing measuring approach.
These variables and their descriptive statistics are presented in the
local accessibility attributes of Table 3.
YabX X
=+− −
(
)
(
)
1
11
exp max ()
The systemwide accessibility indicators were calculated by
means of a gravitational model calibrated for Lisbon’s Mobility Plan
survey of 2004. The model equation (68) is
where
Fij =total flow between zone iand jfor each mode,
Oi=total number of trips with origin in i,
Dj=total number of trips with destination in j,
Cij =impedance between zones iand jmeasured in travel
time between zones, and
Aiand Bj=calibration variables needed in a doubly constrained
gravitational model calibration.
The calibrated βvalues for public transport (PT) and private car
(PC) are presented in Figure 2, where the difference between the
calibrated βvalues is very significant (approximately 4 times greater
FOADB C
ADB C
ij i i j j ij
i
kk l
=
()
=
ii ii i
ii i
exp
exp
β
β
1
kk
k
j
mm mj
m
BOA C
()
[]
=
()
⎡
⎣⎤
⎦
∑
∑
(
)
exp
2
1
ii i
β
TABLE 2 Property Value Impacts of Public Transport Proximity in North American Cities
Case and Location Impact On Impact Source
Atlanta, Georgia Office rents Positive (8, 50)
Baltimore, Maryland, LRT Not stated Unable to identify (44)
Boston, Massachusetts Residential property +6.7% (50, 51)
Buffalo, New York House prices +4–11% (23)
Chicago, Illinois, MTA House prices +20% (52)
Dallas, Texas, DART Commercial rents +64.8% (53)
Dallas DART Property values +25% (53, 54)
Linden, New Jersey Residential property Positive (55)
Los Angeles, California Property values Higher values (56)
Miami, Florida House prices +5% (7)
New Jersey SEPTA rail House prices +7.5–8% (57)
New Jersey PACTO rail House prices +10% (57)
New York Not stated Positive (58)
Pennsylvania SEPTA rail House prices +3.8% (57)
Portland, Oregon House prices +10% (42)
Portland Gresham Residential rent >5% (42)
Portland Metropolitan Express House prices +10.5% (17, 19)
San Diego, California, Trolley Not stated Positive (44)
San Francisco, California, Bay Area BART Property values Positive (59)
San Francisco Bay Area BART Residential rent +10–15% (60, 61)
+15–26%
Santa Clara County, California Commercial and office property +23–120% (62)
Santa Clara County House prices +45% (18)
Santa Clara County Residential rent +15% (25)
St. Louis, Missouri Property values +32% (63)
Toronto, Canada, Metro House prices +20% (29, 44)
Washington, D.C. Residential rent Positive (50, 64)
130 Transportation Research Record 2115
TABLE 3 Descriptive Statistics of Variables (Nⴝ12,488)
Variable Description Mean St. Dev.
Price Asking price (7) 223,123.11 145,408.15
Ln_Price Natural logarithm of the asking price 12.17 0.533
Structural Attributes
Bedrooms Number of bedrooms 2.393 1.068
House 1 if house 0.027 0.161
Floor Floor number 2.952 2.431
Area Area (sq. meters) 103.789 59.253
Age1 1 if property age ≤10 years 0.351 0.477
Age2 1 if 10 years < property age < 50 years 0.327 0.469
Age3 1 if property age ≥50 years 0.322 0.467
Garage 1 if garage spaces ≥1 0.470 0.499
Neighborhood Attributes
Educational index Number of undergraduate persons/population over 20 years old (500 meters radius) 0.197 0.129
Entropy index Entropy index (EI) within a walking distance of 500 meters 0.220 0.103
Local Accessibility Attributes
Metro
2MAccess10 1 if walk time to access two metro lines ≤10 minutes 0.048 0.214
2MAccess =1/(1 +exp(4.394 −0.439 *(20 – walking time))) 0.058 0.163
1MAccess10 1 if walk time to access one metro lines ≤10 minutes 0.265 0.441
1MAccess =1/(1 +exp(6.812 −0.659 *(17 – walking time))) 0.123 0.240
Road
Network1_1,000 1 if distance to Network1 ≤1,000 meters 0.425 0.494
Network1 =1/(1 +exp(8.789 −0.007 *(2,000 – access distance))) 0.266 0.351
Network2_500 1 if distance to Network2 ≤500 meters 0.438 0.496
Network2 =1/(1 +exp(8.789 −0.013 *(1,000 – access distance))) 0.273 0.353
Network3_250 1 if distance to Network3 ≤250 meters 0.558 0.500
Network3 =1/(1 +exp(8.789 −0.026 *(500 – access distance))) 0.345 0.367
Rail
Azambuja10 1 if walk time to Azambuja train station < 10 minutes and less than 20% of the distance to CBD 0.006 0.078
Azambuja =1/(1 +exp(4.394 −0.439 *(20 – walking time))) 0.005 0.057
Lisboa10 1 if walk time to Lisbon train station < 10 minutes and less than 10% of the distance to the CBD 0.014 0.119
Lisboa =1/(1 +exp(4.394 −0.439 *(20 – walking time))) 0.014 0.094
Nacional10 1 if walk time to Nacional train station < 10 minutes 0.013 0.114
Nacional =1/(1 +exp(4.394 −0.439 *(20 – walking time))) 0.010 0.082
Sintra10 1 if walk time to Sintra train station < 10 minutes and less than 20% of the distance to CBD 0.028 0.164
Sintra =1/(1 +exp(4.394 −0.439 *(20 – walking time))) 0.029 0.131
Fertagus10 1 if walk time to Fertagus train station < 10 minutes and less than 20% of the distance to CBD 0.001 0.037
Fertagus =1/(1 +exp(4.394 −0.439 *(20 – walking time))) 0.001 0.033
Cascais10 1 if walk time to Cascais train station < 10 minutes and less than 20% of the distance to CBD 0.014 0.117
Cascais =1/(1 +exp(4.394 −0.439 *(20 – walking time))) 0.011 0.078
Systemwide Accessibility Attributes
Gravitational_PT Gravitational model accessibility index with βcalibrated for public transport 0.708 0.058
Gravitational_PC Gravitational model accessibility index with βcalibrated for private car 0.493 0.084
EI500
1
=
(
)
(
)
(
)
=
∑pp
k
ii
i
kiln
ln ,69 70
Martínez and Viegas 131
for the private car, showing a much greater ease of displacement
than in public transport).
A land use database for the study area was then used for the cal-
culation of the accessibility indicator. The Djterm of the gravita-
tional model equation was replaced by the land use surface (Aj) and
standardized through the land use surface of the whole study area
∑n
j=1Aj. The accessibility indicator results then in the following
equation:
The descriptive statistics of these accessibility indicators are
presented in Table 3.
Some neighborhood attributes for each property were also cal-
culated, to improve to explanatory power of the models. These
variables are an Educational Index that calculates the percentage
of undergraduate persons in the population over 20 years old in a
500-m radius around the property, and an entropy index that mea-
sures the mixture of land use types in a radius of 500 m (69, 70).
AccGriijjj
j
j
j
n
j
nCpApA A
A
=
( ) ()()
=
=
=∑
exp ;βii
1
1
∑∑ ()3
These descriptive statistics of these neighborhood attributes are
presented in Table 3.
MODELING METHODOLOGY
Six different cross-sectional models were developed in this study.
Three specifications for the accessibility effect—local accessibility
all-or-nothing, local accessibility continuous, and systemwide
accessibility, and two modeling approaches—ordinary least-squares
regression model (OLS) and spatial lag regression model. Both pre-
sent a semi-logarithmic hedonic specification that is widely used in
the property value literature, because it usually produces robust esti-
mates and enables convenient coefficient interpretation. The general
structure of the OLS model is
where Piis the price of house i, and Xi1, . . . , Xin are the vectors of
the explanatory variables for the price of house i. The dependent
Ln PXXX
NI
iiinini
(
)
=+ + ++ +
(
)
ββ β β
σ
01122
2
0
∼
⑀
⑀,(44)
0,00
0,10
0,20
0,30
0,40
0,50
0,60
0,70
0,80
0,90
1,00
10 15 20 25 30
Accessibility indicator
Accessibility time (min.)
Continuous Approach All-or-nothing Approach
50
FIGURE 1 Comparison example of all-or-nothing and continuous approaches.
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
Exp(-beta.c (k,j))
Accessibility time Cij (minutes)
Beta PC = -0,03420 Beta PT = -0,00833
020406080
100 120 140 160 180 200
FIGURE 2 Calibrated values of gravitational model for Lisbon Mobility Plan 2004.
132 Transportation Research Record 2115
variable is given in the natural logarithmic form; thus, the values
of the coefficients represent percentage change. The specifications
used for the OLS models (for each type of accessibility index) are
given by
The spatial lag model’s general structure is presented in Equation 8.
where ρis the autoregressive coefficient and WLn(Pi)the spatial
lagged variable in order to a N×Nspatial weight matrix.
The specifications used for the spatial lag models are given by
Ln Bedrooms House
Ln BD HS
PW
iP i i
i
()
=++
′+′+
()
ραβ β ′′
+′+′+′
β
ββ β
FL
AR AG AG
Floor
Area Age Age
i
iii23
23++ ′
+′+′
β
ββ
GR
LI EI
Garage
EducationIndex Entro
i
ippyIndex
2MAccess 1MAccess
2MA 1MA
i
iiN
+′+′+′
βββ
11
23
1
23
Network
Network Network S
i
NiNi
+′+′+′
βββ
NN
CS
Sintra
Cascais
i
ii
+′+β ⑀()10
Ln Bedrooms House
Ln BD HS
PW
iP i i
i
()
=++
′+′+
()
ραβ β ′′
+′+′+′
β
ββ β
FL
AR AG AG
Floor
Area Age Age
i
iii23
23++ ′
+′+′
β
ββ
GR
LI EI
Garage
EducationIndex Entro
i
ippyIndex
2MAccess 1MAccess
2MA 1MA
i
ii
+′+′
ββ10 10
++ ′+′
+′
ββ
β
NiNi
N
12
3
1 1000 2 500Network Network__
NNetwork Sintra
Cascais
SN
CS
3 250 10
10
_ii
i
+′
+′
β
β++⑀i()9
Ln Ln
PW X X X
iP iinin
i
()
=+
′+′+′++
′+
()
ρβββ β
01122
⑀⑀i()8
Ln Bedrooms House Floor
BD HS FL
Piii
()
=+
′+′+′
αβ β β iii
ii
+′
+′+′+′
β
βββ
AR
AG AG GR
Area
Age Age Gara
23
23gge
EducationIndex EntropyIndex
LI EI
i
ii
+′+′
+
ββ
′′ +′
ββ
PT PC
Gravitational PT Gravitational PC__
iiii
+⑀()7
Ln Bedrooms House Floor
BD HS FL
Piii
()
=+
′+′+′
αβ β β iii
ii
+′
+′+′+′
β
βββ
AR
AG AG GR
Area
Age Age3 Gara
23
2gge
EducationIndex EntropyIndex
LI EI
i
ii
+′+′
+
ββ
′′ +′
+′
ββ
β
2MA 1MA
2MAccess 1MAccess
Network
ii
N11iiN iN i
i
+′+′
+′+
ββ
β
23
23Network Network
Sintra
SN ′′ +βCSCascaisii
⑀()6
Ln Bedrooms House Floor
BD HS FL
Piiii
(
)
=+ +
′+′
αβ β β ++′
+′+′+′
β
βββ
AR
AG AG GR
Area
Age Age Garag
i
ii23
23 ee
EducationIndex EntropyIndex
LI EI
i
ii
+′+′
+′
ββ
βββ
β
2MA 1MA
2MAccess 1MAccess
Netwo
10 10
1
ii
N
+′
+′rrk Network
Network
1 1000 2 500
32
2
3
__
_
iN i
N
+′
+′
β
β550 10
10 5
ii
ii
+′
+′+
β
β
SN
CS
Sintra
Cascais ⑀()
The spatial weight matrix for all spatial lag models was devel-
oped assuming constant spatial dependence between properties
until a maximum established distance. The maximum Euclidean
distance used was 1,000 m, resulting in a Moran’s I=0.144
(p-value =.000).
MODELING RESULTS AND DISCUSSION
Estimation results from the six different models are presented in
Table 4. Using the pseudo R2as goodness-of-fit measure (squared
correlation between the predicted and the observed values of the
dependent variable), one can observe a high explanation of the
dependent variable with values ranging from 0.75 and 0.80.
Langrange multiplier (LM) tests were also conducted to assess
if the omission of the spatial lag on the OLS model was erroneous
(i.e., H0: ρ=0). The LM test statistic (35) is given by
where
M=common residual maker vector (N×N) in OLS estimation,
e=spherical OLS residual vector (N×1),
σ2=e′e/N,
b=OLS coefficient vector (K×1),
tr =trace of the matrix (N×N).
The LM test is approximately χ2distributed with one degree of
freedom. As presented in Table 4, the LM test for all the spatial lag
models is significant at α=0.01, indicating a proper spatial lag
specification.
Almost all the independent variables used in the six models are
significant at α=0.05 (Sintra being the exception for both local
accessibility spatial lag models), independently of the estimation
model and of the proxy accessibility variables used. In addition, the
models consistently demonstrate the impact of each independent
variable on the natural logarithm of the asking price and a similar
magnitude of the coefficients of the structural attributes along the
different models.
The coefficient estimation for the structural attributes shows that
the Area (floor surface) is the attribute with greater impact in the
dependent variable (approximately 0.07% increase for a 1% square
meter increase), followed by the Age and Bedrooms, which present
similar values in all the models.
The neighborhood attributes are the ones that present a higher
coefficient of variation among the estimated models. As expected,
the spatial lag term “replaces” some of the explanatory power of
the neighborhood variables and the constant of the model, although
it does not affect so significantly the coefficients of the accessibil-
ity attributes, as can be seen in the local accessibility models in
Table 4.
LM tr
=′
⎧
⎨
⎩
⎫
⎬
⎭
(
)
′+′+
(
)
eWv
WXb MWXb W W W
σσ
222
112
i())
Ln Bedrooms House
Ln BD HS
PW
iP i i
i
()
=++
′+′+
()
ραβ β ′′
+′+′+′
β
ββ β
FL
AR AG AG
Floor
Area Age Age
i
iii23
23++ ′
+′+′
β
ββ
GR
LI EI
Garage
EducationIndex Entro
i
ippyIndex
Gravitational_PT Gravita
PT PC
i
i
+′+′
ββttional_PCii
+⑀(
)
11
TABLE 4 Results of OLS and Spatial Lag Models
Ordinary Least Squares Maximum Likelihood Spatial Lag
Local Accessibility Local Accessibility Systemwide Accessibility Local Accessibility Local Accessibility Systemwide Accessibility
All-or-Nothing Model Continuous Model Model All-or-Nothing Model Continuous Model Model
Coef. Std. Error Coef. Std. Error Coef. Std. Error Coef. Std. Error Coef. Std. Error Coef. Std. Error
SP_LAG_Logprice — — — — — — 0.3649* 0.0084 0.3561* 0.0085 0.3082* 0.0085
Constant 11.0898* 0.0102 11.0862* 0.0101 10.2046* 0.0345 6.8077* 0.0991 6.9089* 0.0999 6.6669* 0.1019
Structural Attributes
Bedrooms 0.0347* 0.0033 0.0340* 0.0033 0.0429* 0.0030 0.0432* 0.0030 0.0427* 0.0030 0.0492* 0.0032
House 0.1736* 0.0166 0.1740* 0.0165 0.2230* 0.0154 0.1669* 0.0154 0.1685* 0.0154 0.2124* 0.0163
Floor 0.0165* 0.0010 0.0160* 0.0010 0.0187* 0.0009 0.0158* 0.0009 0.0155* 0.0009 0.0176* 0.0010
Area 0.0069* 0.0001 0.0069* 0.0001 0.0069* 0.0001 0.0065* 0.0001 0.0064* 0.0001 0.0065* 0.0000
Age2 −0.1291* 0.0068 −0.1291* 0.0068 −0.1524* 0.0063 −0.1030* 0.0064 −0.1034* 0.0063 −0.1250* 0.0066
Age3 −0.0855* 0.0073 −0.0849* 0.0072 −0.1201* 0.0067 −0.0736* 0.0068 −0.0729* 0.0068 −0.1037* 0.0071
Garage 0.1226* 0.0063 0.1252* 0.0063 0.1449* 0.0059 0.1109* 0.0059 0.1126* 0.0059 0.1330* 0.0062
Neighborhood Attributes
Educational index 0.9348* 0.0201 0.9535* 0.0202 0.5590* 0.0186 0.3893* 0.0222 0.4160* 0.0225 0.1737* 0.0205
Entropy index 0.4249* 0.0248 0.3869* 0.0250 0.2115* 0.0222 0.2570* 0.0231 0.2312* 0.0234 0.0975* 0.0211
Local Accessibility Attributes (all-or-nothing and continuous approach)
2MAccess10 (2MAccess) 0.0518* 0.0099 0.1248* 0.0142 — — 0.0349* 0.0092 0.0916* 0.0133 — —
1Maccess10 (1MAccess) 0.0617* 0.0058 0.0875* 0.0090 — — 0.0462* 0.0054 0.0652* 0.0084 — —
Network1_1,000 (Network1) −0.0745* 0.0052 −0.1105* 0.0074 — — −0.0536* 0.0049 −0.0732* 0.0069 — —
Network2_500 (Network2) 0.0580* 0.0049 0.0863* 0.0068 — — 0.0368* 0.0046 0.0458* 0.0064 — —
Network3_250 (Network3) −0.0585* 0.0047 −0.0584* 0.0064 — — −0.0416* 0.0044 −0.0380* 0.0060 — —
Sintra10 (Sintra) −0.0916* 0.0118 −0.1001* 0.0118 — — −0.0358* 0.0110 −0.0614* 0.0134 — —
Cascais10 (Cascais) 0.1073* 0.0211 0.1099* 0.0210 — — 0.0675* 0.0196 0.1517* 0.0259 — —
Systemwide Accessibility Attributes
Gravitational_PT — — — — 0.3623* 0.0731 — — — — 0.5110* 0.0690
Gravitational_PC — — — — 1.4484* 0.0522 — — — — 1.0447* 0.0501
Pseudo R2.762 .764 .797 .795 .795 .820
LM statistic 1,272.566* 1,154.496* 921.855*
Log likelihood −409.811 −371.612 560.635 226.472 236.608 1021.560
Akaike info. criterion 853.622 777.223 −1,097.270 −416.944 −437.217 −2,017.130
NOTE: * denote coefficient significantly different from zero at the 1%, 5%, and 10% level of significance (two-tailed test), respectively. — = not applicable.
134 Transportation Research Record 2115
This fact shows the stability of the local accessibility coefficients
in the all-or-nothing approach. The metro accessibility attributes
coefficients in the two models vary between 3.49% and 5.18% for
the accessibility to two metro lines and between 4.62% and 6.17%
for the accessibility to a single metro line, showing a significant
impact of the metro proximity over the property values.
The local accessibility coefficients in the continuous approach
present stability as well. The metro accessibility attributes coeffi-
cients in the two models vary between 9.16% and 12.48% for the
accessibility to two metro lines and between 6.52% and 8.75% for
the accessibility to a single metro line, showing again a significant
impact of the metro proximity over the property values.
The rail accessibility attributes coefficients in all the local
accessibility models illustrate a positive impact for the proximity
to the Cascais Line, with coefficients ranging between 6.75% and
10.73% for the all-or-nothing accessibility measure approach and
between 10.99% and 15.17% for the continuous approach; and a
negative impact for the proximity to the Sintra Line with coeffi-
cients ranging between −9.16% and −3.58% for the all all-or-nothing
approach and between −10.01% and −6.14% for the continuous
approach.
These effects might be explained by the perception of lack of
security associated with the Sintra Line, which prevents the prop-
erties of the nearby areas from taking full advantage of the prox-
imity to this public transport system of high capacity and good
service levels and the of the proximity of the Cascais Line to a very
expensive residential area in the southeast area of Lisbon (Restelo
neighborhood).
The road accessibility attributes coefficients in the two models of
the all-or-nothing approach vary between −7.45% and −5.36% for
accessibility Road Hierarchy 1, between 3.68% and 5.80% for acces-
sibility to Road Hierarchy 2, and between −4.16% and −5.85% for
accessibility to the Road Hierarchy 3.
The road accessibility attributes present a similar impact for the
two models of continuous approach, with coefficients varying
between −11.05% and −7.32% for accessibility Road Hierar-
chy 1, between 4.58% and 8.63% for accessibility to Road Hierar-
chy 2, and between −3.80% and −5.84% for accessibility to Road
Hierarchy 3.
These estimations show that the proximity to Road Hierarchy 2
(urban ring roads and radial network) is the proximity that presents
a positive impact on the property values, while proximity to Road
Hierarchy 3 (urban distribution network) and Road Hierarchy 1
(motorways) present a negative impact.
These results can derive from the congestion and noise external-
ities perceived by the population near Road Hierarchy 1 and the
dominance of office buildings near Road Hierarchy 3 roads in Lis-
bon’s city center, reducing the residential attractiveness and supply
in these areas.
The results of the local accessibility spatial lag models show the
existence of a smaller impact of road network attributes on the depen-
dent variable that might be explained by the significant increase of
the stated preference lag coefficient.
Direct comparison of the coefficients resulting from the local
accessibility continuous models with the all-or-nothing accessibil-
ity measure models must be done with some caution, because the
continuous indicators present continuous values between 0 and 1,
and the estimated coefficient in that case already incorporates this
range of values in the independent variable. The percentage of change
on the property selling prices will result from the product between
the value of the accessibility indicator and the coefficient of the
same indicator, resulting in a property selling prices impact distri-
bution rather than in a single value. Even if this slightly increases
the complexity of the price estimates, it is thought that it improves
its accuracy.
The systemwide accessibility OLS model presents significant
differences in coefficients of the constant and neighborhood attri-
butes when compared with the local accessibility OLS model. This
might be due to the significant correlation of the systemwide acces-
sibility attributes with the neighborhood attributes (i.e., the correla-
tion between the Gravitational_TC and Entropy Index is equal to
0.465, and is equal to 0.235 with the Educational Index), which can
explain the reduction of the neighborhood attributes coefficients.
This correlation results from the fact that the systemwide acces-
sibility indicators measure accessibility to activities scattered in
the study area, which can be positively influenced by the presence
of a high land use mixture around the property measured by the
Entropy Index.
For example, the Gravitational_TC indicator simultaneously mea-
sures public transport accessibility and land use activity proximity
and their relation. This fact illustrates that with the systemwide acces-
sibility indicators, it is difficult to isolate the changes in property val-
ues derived from transportation infrastructures investment from the
neighborhood land use characteristics.
The systemwide accessibility spatial lag model also illustrates
the last statement because of the smaller value of the LM test for
the spatial lag model compared with the other spatial lag models (see
Table 4). This is because the systemwide accessibility indicators
can already explain part of the spatial dependence of the property
asking price.
With the Akaike info criterion used to rank the models, it is pos-
sible to consider the systemwide accessibility OLS model (although
with problems of correlation between variables) as the best pre-
diction model, followed by the local accessibility continuous spa-
tial lag model and the local accessibility all-or-nothing spatial lag
model.
It can thus be concluded, from the estimates of the developed
models, that the local accessibility models can better measure the
isolated effect of transportation investment on properties’ selling
prices and that the estimates from the OLS models can be suffi-
ciently accurate in the absence of a significant property values data-
base for the entire study area (needed for the calculation of the
spatial lag model).
SUMMARY AND CONCLUSIONS
This paper analyzes the effect of transportation accessibility on
property prices, as part of a broader study that intends to develop
a new value capture financing scheme for public transportation in
the LMA. Several cross-sectional hedonic price models are esti-
mated on the basis of an online realtor’s database (Imokapa Vector)
of properties’ selling asking price. The models account for struc-
tural, neighborhood, and accessibility attributes of residential prop-
erties, with accessibility structured into two types: local accessibility
attributes and systemwide accessibility attributes.
The main focus of this study is to develop a framework to fore-
cast house prices and the influence of transportation infrastructure
investment in further steps of the research project.
Martínez and Viegas 135
The estimated models revealed that
•Local accessibility hedonic price models that were developed
show the existence of spatial interactions of sale prices, presenting
a spatial autocorrelation with a significant spatial lag.
•Local accessibility models present a stability of the local acces-
sibility coefficients estimated, the significance of the Sintra attribute
being the only one affected.
•The metro accessibility attributes coefficients in the two all-or-
nothing models vary between 3.49% and 5.18% for the accessibility
to two metro lines and between 4.62% and 6.17% for the accessibil-
ity to a single metro line, showing a significant impact of the metro
proximity over the property values.
•The metro accessibility attributes coefficients in the two contin-
uous models vary between 9.16% and 12.48% for the accessibility to
two metro lines and between 6.52% and 8.75% for the accessibility
to a single metro line.
•The rail accessibility attributes coefficients illustrate a positive
impact for the proximity to the Cascais Line with coefficients rang-
ing between 6.75% and 10.73%, and a negative impact for the prox-
imity to the Sintra Line with coefficients ranging between −9.16%
and −3.58% (not significant for the usual significance levels) for the
all-or-nothing accessibility measure models.
•The rail accessibility attributes coefficients for the continuous
accessibility measure models present a positive impact for the prox-
imity to the Cascais Line with coefficients ranging between 10.99%
and 15.17%, and a negative impact for the proximity to the Sintra
Line with coefficients ranging between −10.01% and −6.14% (not
significant for the usual significance levels).
•The road accessibility attributes coefficients in the two all-
or-nothing accessibility measure models vary between −7.45% and
−5.36% for the accessibility to Road Hierarchy 1, between 3.68% and
5.80% for the accessibility to Road Hierarchy 2, and between −5.85%
and −4.16% for the accessibility to Road Hierarchy 3.
•The road accessibility attributes coefficients in the continuous
accessibility measure models vary between −11.05% and −7.32%
for accessibility to Road Hierarchy 1, between 4.58% and 8.63% for
accessibility to Road Hierarchy 2, and between −5.84% and −3.80%
for accessibility to the Road Hierarchy 3.
•The systemwide accessibility models present significant differ-
ences in coefficients of the constant and neighborhood attributes when
compared with the local accessibility models. This indicates the dif-
ficulty in isolating the accessibility effects from the neighborhood
effects on house prices with these models.
The coefficients resulting from the local accessibility continuous
models cannot be compared directly with the coefficients from the
all-or-nothing accessibility measure models because the continuous
indicators present continuous values between 0 and 1. The percent-
age of change on the property selling prices will result from the prod-
uct between the value of the accessibility indicator and the coefficient
of the same indicator.
With the Akaike info criterion used to rank the models, it is pos-
sible to consider the systemwide accessibility OLS model (although
with problems of correlation between variables) as the best pre-
diction model, followed by the local accessibility continuous spa-
tial lag model and the local accessibility all-or-nothing spatial lag
model.
The main conclusions that can be drawn from the estimates of the
developed models are that the local accessibility models can better
measure the isolated effect of transportation investment on proper-
ties’ selling prices and that the estimates from the OLS model can
be sufficiently accurate in the absence of a significant property val-
ues database for the entire study area (needed for the calculation of
the spatial lag model).
ACKNOWLEDGMENTS
This research is supported by the Portuguese National Science Foun-
dation (FCT) since 2006. Imokapa Vector made available an online
realtor’s database; TIS.pt made available the LMA Mobility Survey
from 2004; and INTERGRAPH provided the Geomedia Professional
5.2 license.
REFERENCES
1. Alonso, W. Location and Land Use: Towards a General Theory of Land
Rent. Harvard University Press, Cambridge, Mass., 1964.
2. Muth, R. F. Cities and Housing: The Spatial Pattern of Urban Residential
Land Use. University of Chicago Press, Ill., 1969.
3. Rosen, S. Hedonic Prices and Implicit Markets—Product Differentia-
tion in Pure Competition. Journal of Political Economy, Vol. 82, No. 1,
1974, pp. 34–55.
4. Cervero, R., and M. Duncan. Transit’s Value-Added Effects: Light and
Commuter Rail Services on Commercial Land Values. In Transporta-
tion Research Record: Journal of the Transportation Research Board,
No. 1805, Transportation Research Board of the National Academies,
Washington, D.C, 2002, pp. 8–15.
5. Bowes, D. R., and K. R. Ihlanfeldt. Identifying the Impacts of Rail Tran-
sit Stations on Residential Property Values. Journal of Urban Economics,
Vol. 50, No. 1, 2001, pp. 1–25.
6. Cervero, R., and M. Duncan. Neighbourhood Composition and Residen-
tial Land Prices: Does Exclusion Raise or Lower Values? Urban Studies,
Vol. 41, No. 2, 2004, pp. 299–315.
7. Gatzlaff, D. H., and M. T. Smith. The Impact of the Miami Metrorail on
the Value of Residences near Station Locations. Land Economics, Vol. 69,
No. 1, 1993, pp. 54–66.
8. Bollinger, C. R., K. R. Ihlanfeldt, and D. R. Bowes. Spatial Variation in
Office Rents Within the Atlanta Region. Urban Studies, Vol. 35 No. 7,
1998, pp. 1097–1118.
9. Armstrong, R. J., and D. A. Rodriguez. An Evaluation of the Acces-
sibility Benefits of Commuter Rail in Eastern Massachusetts Using
Spatial Hedonic Price Functions. Transportation, Vol. 33, No. 1, 2006,
pp. 21–43.
10. Weinberger, R. R. Light Rail Proximity: Benefit or Detriment in the
Case of Santa Clara County, California? In Transportation Research
Record: Journal of the Transportation Research Board, No. 1747, TRB,
National Research Council, Washington, D.C., 2001, pp. 104–113.
11. Cheshire, P., and S. Sheppard. Capitalised in the Housing Market or
How We Pay for Free Schools: The Impact of Supply Constraints and
Uncertainty. Proc., Royal Geographical Society Meeting, 2003.
12. Debrezion, G., E. Pels, and P. Rietveld. The Impact of Rail Transport
on Real Estate Prices: An Empirical Analysis of the Dutch Housing
Market. Tinbergen Institute Discussion Paper 2006-031/3. Tinbergen
Institute and Vrije Universiteit Amsterdam, Amsterdam, Netherlands,
2006.
13. Lewis-Workman, S., and D. Brod. Measuring the Neighborhood Bene-
fits of Rail Transit Accessibility. In Transportation Research Record:
Journal of the Transportation Research Board, No. 1576, TRB, National
Research Council, Washington, D.C., 1997, pp. 147–153.
14. Martins, C., and O. Bin. Estimation of Hedonic Price Functions via Addi-
tive Nonparametric Regression. Empirical Economics, Vol. 30, No. 1,
2005, pp. 93–114.
15. Epple, D. Hedonic Prices and Implicit Markets—Estimating Demand
and Supply Functions for Differentiated Products. Journal of Political
Economy, Vol. 95, No. 1, 1987, pp. 59–80.
136 Transportation Research Record 2115
16. Henneberry, J. Transport Investment and House Prices. Journal of
Property Valuation and Investment, Vol. 16, No. 2, 1998, pp. 144–158.
17. Al-Mosaind, M. A., K. J. Dueker, and J. G. Strathman. Light-Rail Tran-
sit Stations and Property Values: A Hedonic Price Approach. In Trans-
portation Research Record 1400, TRB, National Research Council,
Washington, D.C., 1993, pp. 90–94.
18. Cervero, R., and M. Duncan. Benefits of Proximity to Rail on Hous-
ing Markets: Experiences in Santa Clara County. Journal of Public
Transportation, Vol. 5, No. 1, 2002.
19. Chen, H., A. Rufolo, and K. J. Dueker. Measuring the Impact of Light
Rail Systems on Single-Family Home Values: A Hedonic Approach
with Geographic Information System Application. In Transportation
Research Record 1617, TRB, National Research Council, Washington,
D.C., 1998, pp. 38–43.
20. Kawamura, K., and S. Mahajan. Hedonic Analysis of Impacts of Traffic
Volumes on Property Values. In Transportation Research Record: Jour-
nal of the Transportation Research Board, No. 1924, Transportation
Research Board of the National Academies, Washington, D.C., 2005,
pp. 69–75.
21. Cervero, R., and M. Duncan. Transit’s Value-Added Effects. Light and
Commuter Rail Services and Commercial Land Values. In Transporta-
tion Research Record: Journal of the Transportation Research Board,
No. 1805, Transportation Research Board of the National Academies,
Washington, D.C., 2002, pp. 8–15.
22. Haider, M., and E. J. Miller. Effects of Transportation Infrastructure and
Location on Residential Real Estate Values: Application of Spatial
Autoregressive Techniques. In Transportation Research Record: Jour-
nal of the Transportation Research Board, No. 1722, TRB, National
Research Council, Washington, D.C., 2000, pp. 1–8.
23. Hess, D. B., and T. M. Almeida. Impact of Proximity to Light Rail
Rapid Transit on Station-Area Property Values in Buffalo. Urban Studies,
Vol. 44, No. 5 & 6, 2007, pp. 1041–1068.
24. Rodriguez, D. A., and F. Targa. Value of Accessibility to Bogota’s
Bus Rapid Transit System. Transport Reviews, Vol. 24, No. 5, 2004,
pp. 587–610.
25. Weinberger, R. R. Commercial Rents and Transportation Improve-
ments: The Case of Santa Clara County’s Light Rail. Lincoln Institute
of Land Policy, Cambridge, Mass., 2001.
26. Bateman, I., B. Day, I. Lake, and A. Lovett. The Effect of Road Traffic
on Residential Property Values: A Literature Review and Hedonic Pric-
ing Study. Scottish Executive Development Department, Edinburgh,
2001.
27. Bartik, T. J. Measuring the Benefits of Amenity Improvements in Hedo-
nic Price Models. Land Economics, Vol. 64, No. 2, 1988, pp. 172–183.
28. Kanemoto, Y. Hedonic Prices and the Benefits of Public Projects.
Econometrica, Vol. 56, No. 4, 1988, pp. 981–989.
29. Bajic, V. The Effects of a New Subway Line on Housing Prices in Met-
ropolitan Toronto. Urban Studies, Vol. 20, No. 2, 1983, pp. 147–158.
30. Forrest, D., J. Glen, and R. Ward. The Impact of a Light Rail System
on the Structure of House Prices: A Hedonic Longitudinal Study.
Journal of Transport Economics and Policy, Vol. 30, No. 1, 1996,
pp. 15–29.
31. Adair, A., S. McGreal, A. Smyth, J. Cooper, and T. Ryley. House Prices
and Accessibility: The Testing of Relationships Within the Belfast
Urban Area. Housing Studies, Vol. 15, No. 5, 2000, pp. 699–716.
32. Munoz-Raskin, R. C. Walking Accessibility to Bus Rapid Transit in Latin
America: Does It Affect Property Values? Case of Bogotá, Colombia.
Presented at 86th Annual Meeting of the Transportation Research Board,
Washington D.C., 2007.
33. Kockelman, K. M. Effects of Location Elements on Home Purchase
Prices and Rents in San Francisco Bay Area. In Transportation
Research Record 1606, TRB, National Research Council, Washington,
D.C., 1997, pp. 40–50.
34. Bae, C. H. C., M. J. Jun, and H. Park. The Impact of Seoul’s Subway
Line 5 on Residential Property Values. Transport Policy, Vol. 10, 2003,
pp. 85–94.
35. Shin, K., S. P. Washington, and K. Choi. Effects of Transportation
Accessibility on Residential Property Values: Application of Spatial
Hedonic Price Model in Seoul, South Korea, Metropolitan Area. In
Transportation Research Record: Journal of the Transportation
Research Board, No. 1994, Transportation Research Board of the
National Academies, Washington, D.C., 2007, pp. 66–73.
36. Forkenbrock, D. J., S. Benshoff, and G. E. Weisbrod. NCHRP Web
Document No. 31: Assessing the Social and Economic Effects of
Transportation Projects. University of Iowa, Iowa City, 2001.
37. Carey, J., and J. Semmens. Impact of Highways on Property Values: Case
Study of Superstition Freeway Corridor. In Transportation Research
Record: Journal of the Transportation Research Board, No. 1839, Trans-
portation Research Board of the National Academies, Washington, D.C.,
2003, pp. 128–135.
38. Coulson, N. E., and R. F. Engle. Transportation Costs and the Rent Gra-
dient. Journal of Urban Economics, Vol. 21, No. 3, 1987, pp. 287–297.
39. Tse, C. Y., and A. W. H. Chan. Estimating the Commuting Cost and
Commuting Time Property Price Gradients. Regional Science and
Urban Economics, Vol. 33, No. 6, 2003, pp. 745–767.
40. Frazier, C., and K. M. Kockelman. Spatial Econometric Models for
Panel Data: Incorporating Spatial and Temporal Data. In Transporta-
tion Research Record: Journal of the Transportation Research Board,
No. 1902, Transportation Research Board of the National Academies,
Washington, D.C., 2005, pp. 80–90.
41. Parsons Brinckerhoff Quade & Douglas Inc. The Effect of Rail Transit
on Property Values: A Summary of Studies. NEORail II. Cleveland,
Ohio, 2001.
42. Hass-Klau, C., G. Crampton, and R. Benjari. Economic Impact of Light
Rail: The Results of 15 Urban Areas in France, Germany, UK and North
America. Environmental and Transport Planning, Brighton, United
Kingdom, 2004.
43. ATISREAL, Geofutures, UCL, and Symonds-Group. Land Value and
Public Transport. Stage Two: Testing the Methodology on the Croy-
don Tramlink. Royal Institution of Chartered Surveyors, London,
2004.
44. Hack, J. Regeneration and Spatial Development: A Review of Research
and Current Practice. IBI Group, Toronto, Ontario, Canada, 2002.
45. Hillier Parker. Crossrail: Property Value Enhancement. Canary Wharf
Group Plc., London, 2002.
46. Chesterton. Property Market Study. Prepared for Jubilee Line Extension
Impact Study. University of Westminster, London, 2000.
47. Pharoah, T. Jubilee Line Extension Development Impact Study. Univer-
sity of Westminster, London, 2002.
48. Dabinett, G. Realising Regeneration Benefits from Urban Infrastructure
Investment: Lesson from Sheffield in the 1990s. Town Planning Review,
Vol. 62, No. 2, 1998, pp. 171–189.
49. Pickett, M. W., and K. E. Perrett. The Effect of the Tyne and Wear
Metro on Residential Property Values. Transport and Road Research
Laboratory, Crowthorne, Berkshire, United Kingdom, 1984.
50. Rail Transit and Property Values. American Public Transit Asso-
ciation, 2002, www.apta.com/research/info/briefings/briefing_1.cfm
(accessed January 8, 2006).
51. Armstrong, R. J., Jr. Impacts of Commuter Rail Service as Reflected in
Single-Family Residential Property Values. In Transportation Research
Record 1466, TRB, National Research Council, Washington, D.C.,
1994, pp. 88–98.
52. Gruen, A. The Effect of CTA and Metra Stations on Residential Prop-
erty Values. Report to the Regional Transportation Authority, Chicago,
Ill., 1997.
53. Weinstein, B. L., and T. L. Clower. The Initial Economic Impacts of the
DART LRT System. Center for Economic Development and Research,
University of North Texas, Denton, 1999.
54. Kay, J. H., and G. Haikalis. All Aboard. Planning, Vol. 66, No. 10, 2000,
pp. 14–19.
55. Diaz, R. B. Impacts of Rail Transit on Property Values. Proc., Ameri-
can Public Transit Association (APTA) 1999 Rapid Transit Conference,
1999.
56. Fejarang, R. Impact on Property Values: A Study of the Los Angeles
Metro Rail. Presented at 73rd Annual Meeting of the Transportation
Research Board, Washington D.C., 1994.
57. Voith, R. P. Transportation, Sorting and Housing Values. Journal of the
American Real Estate and Urban Economics Association (AREUEA),
Vol. 19, No. 2, 1991, pp. 117–137.
58. Anas, A., and R. J. Armstrong. Land Values and Transit Access: Model-
ing the Relationship in the New York Metropolitan Area. An Implementa-
tion Handbook. U.S. Federal Transit Administration, Office of Technical
Assistance and Safety, Springfield, Va., 1993.
Martínez and Viegas 137
59. Cervero, R., and D. Aschuer. Economic Impact of Transit Investments:
Guidebook for Practitioners. Transit Cooperative Research Program.
Federal Transit Administration, U.S. Department of Transportation,
Washington, D.C., 1998.
60. Cevero, R. Transit Based Housing in the San Francisco Bay Area: Market
Profiles and Rent Premiums. Transportation Quarterly, Vol. 50, No. 3,
1996, pp. 33–49.
61. Sedway Group. Regional Impact Study. Report commissioned by Bay
Area Rapid Transit District (BART), San Francisco, Calif., 1999.
62. Cervero, R., C. Ferrell, and S. Murphy. Transit Cooperative Research
Program (TCRP) Research Results Digest 52: Transit-Oriented Devel-
opment and Joint Development in the United States: A Literature Review.
Transportation Research Board of the National Academies, Washington,
D.C., 2002.
63. Garrett, T. A. Light-Rail Transit in America: Policy Issues and Prospects
for Economic Development. Federal Reserve Bank of St. Louis, Mo.,
2004.
64. Benjamin, J. D., and G. S. Sirmans. Mass Transportation, Apartment Rent
and Property Values. Journal of Real Estate Research, Vol. 12, No. 1,
1996, pp. 1–8.
65. Rodríguez, D. A., and C. H. Mojica. Capitalization of BRT Network
Effects into Land Prices. Lincoln Institute of Land Policy, Cambridge,
Mass., 2007.
66. Du, H. B., and C. Mulley. The Short-Term Land Value Impacts of
Urban Rail Transit: Quantitative Evidence from Sunderland, UK. Land
Use Policy, Vol. 24, No. 1, 2007, pp. 223–233.
67. Câmara Municipal de Lisboa, Lisboa: O desafio da mobilidade, Câmara
Municipal de Lisboa—Licenciamento Urbanístico e Planeamento Urbano,
2005.
68. Tischer, M. L. Special Issue on Methodological Issues in Accessibility.
Journal of Transportation and Statistics, 2001.
69. Cervero, R., and K. Kockelman. Travel Demand and the 3Ds: Density,
Diversity, and Design. Transportation Research Part D: Transport and
Environment, Vol. 2, No. 3, 1997, pp. 199–219.
70. Potoglou, D., and P. S. Kanaroglou. Modelling Car Ownership in
Urban Areas: A Case Study of Hamilton, Canada. Journal of Transport
Geography, Vol. 16, No. 1, 2008, pp. 42–54.
The Transportation and Economic Development Committee sponsored publication
of this paper.
