Sei sulla pagina 1di 29

Urban Studies

Impact of Proximity to Light Rail Rapid Transit on Station-area Property Values in

Buffalo, New York
Daniel Baldwin Hess and Tangerine Maria Almeida
Urban Stud 2007 44: 1041
DOI: 10.1080/00420980701256005

The online version of this article can be found at:

Published by:

On behalf of:
Urban Studies Journal Foundation

Additional services and information for Urban Studies can be found at:

Email Alerts:





>> Version of Record - May 1, 2007

What is This?

Downloaded from at Biblioteca Universitaria de Granada on October 10, 2013

Urban Studies, Vol. 44, Nos. 5/6, 1041–1068, May 2007

Impact of Proximity to Light Rail Rapid Transit on

Station-area Property Values in Buffalo, New York

Daniel Baldwin Hess and Tangerine Maria Almeida

[Paper first received, May 2006; in final form, September 2006]

Summary. This study assesses the impact of proximity to light rail transit stations on residential
property values in Buffalo, New York, where light rail has been in service for 20 years, but
population is declining and ridership is decreasing. Hedonic models are constructed of assessed
value for residential properties within half a mile of 14 light rail stations and independent
variables are included that describe property characteristics, neighbourhood characteristics and
locational amenities. The model suggests that, for homes located in the study area, every foot
closer to a light rail station increases average property values by $2.31 (using geographical
straight-line distance) and $0.99 (using network distance). Consequently, a home located within
one-quarter of a mile radius of a light rail station can earn a premium of $1300–3000, or 2– 5
per cent of the city’s median home value. Model results further suggest that three independent
variables—the number of bathrooms, size of the parcel and location on the East side or West
side of Buffalo—are more influential than rail proximity in predicting property values.
Individual regression models for each of the light rail system’s 14 stations suggest that effects
are not felt evenly throughout the system. Proximity effects are positive in high-income station
areas and negative in low-income station areas. An analysis of the actual walking distance to
stations (along the street network) versus the perceived proximity to stations (measured by
straight-line distance) reveals that the results are statistically more significant in the network
distance than the straight-line distance model, but the effects are greater in the straight-line
distance model, which suggests that apparent proximity to rail stations is an added locational
advantage compared with physical walking distance to the station.

The relationship between a transit system, the to an entire region through which a transit
location of transit stations and property values system traverses. Transit stations are the
are, in theory, fundamental to land markets points of access to a transit system and the
and urban structure. This is because public ability to access them conveniently and
transit provides access to central business dis- quickly influences property values. Conven-
tricts, employment locations, healthcare, tional urban economic theory—derived from
schools and colleges and entertainment and utility maximisation—suggests a relationship
recreation, and thereby increases accessibility between improvements to transport systems

Daniel Baldwin Hess is in the Department of Urban and Regional Planning, University at Buffalo, State University of New York, 3435
Main Street, 116 Hayes Hall, Buffalo, New York 14217-3087, USA. Fax: 716 829 3256. E-mail: Tangerine
Maria Almeida is with LSC Transportation Consultants, Inc., 516 North Tejon Street, Colarado Springs, Colorado 80903, USA. Fax:
719 633 5430. E-mail: An earlier version of this research was presented at the 85th Annual Meeting of the
Transport Research Board in January 2006 in Washington, DC. The authors thank Kelly Dixon and Jody Pollot for research assist-
ance and Alex Bitterman, Kate Foster, Peter Lombardi and Beverly McLean for helpful comments and suggestions. Samina Raja
provided assistance with the model.
0042-0980 Print/1360-063X Online/07/5-61041 –28 # 2007 The Editors of Urban Studies
DOI: 10.1080/00420980701256005

and the value of property; in other words, resi- Knowledge about the influence that transit
dential property values rise with decreasing dis- proximity has on property values is important,
tance to a transit station and with decreasing because directing growth and development is
distance to the central business district of great interest to planners. Rail transit
(Alonso, 1964). Herbert Mohring (1961, substantially increases property values where
p. 244) argues that “the basic value of an invest- transit investments can be co-ordinated with
ment—be it in highway or anything else—is the new development in nearby neighbourhoods;
value of the resources it releases for other uses”, on the other hand, where rail transit margin-
suggesting that travel time savings which result ally affects or does not affect property
from proximity to a transit station should be values, transit investments serve primarily to
capitalised in the value of property and relieve congestion and improve accessibility
perhaps in the arrangement of land uses sur- and mobility but do not contribute to econ-
rounding a station (Boyce et al., 1972). omic growth and urban revitalisation.
Assuming that access to destinations reach- However, co-ordination between land use,
able by public transit is an important determi- zoning, development and public transport is
nant in the value of property, higher property difficult to achieve, even under the best
values are expected in places with superior circumstances.
access. Residential property immediately This study is structured as follows. We
adjacent to a transit station is expected to review relevant literature and use lessons
decrease in value due to congestion effects, learned from previous research as a frame-
while properties radiating out from the work for developing a conceptual model of
station and within walking distance should the relationship between property value and
increase in value.1 proximity to light rail transit that includes
Previous research investigating the impact explanatory variables. We then operationalise
of rail transit on property values reveal the theoretical model by undertaking hedonic
results of varying magnitude that are some- regression models for all Metro Rail stations
times contradictory and most studies are grouped together as well as each of the 14
unique to a given locale and do not provide stations individually. Finally, we provide con-
a firm basis to judge future impacts. Further- cluding observations and recommendations.
more, within the US, much of the research is
performed in thriving cities in the Sunbelt
Location: Theory and Practice
area, but such research is just as important in
Rustbelt cities experiencing population and Proximity to public transit has been proved to
economic decline. offer economic advantages—because of pre-
This study estimates the impact of proxi- miums realised by superior access—for most
mity to Metro Rail stations in Buffalo’s 20- types of real estate (Alonso, 1964; Mills,
year old light rail system on residential prop- 1972). A vast body of previous research esti-
erty values. We seek to understand how mates the impact of proximity to transit
effects in Buffalo—where population loss stations on land values.
has been extreme—differ from effects in Accessibility is usually measured as the dis-
cities experiencing population and economic tance from property to transit stations (or
growth. We also explore how effects differ sometimes transit corridors) that pedestrians
near various station areas, as Metro Rail in would encounter along the network of streets
Buffalo travels through diverse and uneven and sidewalks; a popular research method is
neighbourhoods. We test two methods for to study buffer areas around transit stations
measuring the distance between a residential using a radial distance of up to one mile
property and a light rail station—straight- around stations, and travel distance—defined
line distance and network distance—to as a continuous variable—is commonly used
explore the difference between actual distance as a proxy for travel time. In some studies, a
to a station and perceived distance to a station. cruder approximation is used to determine

transit proximity; studies that use city-wide anticipation of a rail line before construction
property samples often assign a binary vari- or service begins (Damm et al., 1980;
able to indicate whether a house is or is not Dunphy, 1982; Ferguson et al., 1988; Grass,
within one-quarter or half a mile of a transit 1992; Grether and Miezkowski, 1974; Knaap
stop or station (Cervero and Duncan, 2002a). et al., 1996; McDonald and Osuji, 1995).
There are a number of published literature However, our current inquiry focuses on
reviews about the effect of proximity to studies of light rail and heavy rail, as transit
transit on property value (Anas, 1982, 1983; patrons may not distinguish between the two
Cervero et al., 2002; Diaz, 1999; Huang, when stations are below ground, as they are
1994; Knaap, 1998; Parsons Brincker- in Buffalo’s residential neighbourhoods
hoff, 2001; Rice Centre for Urban Mobility where Metro Rail operates in a tunnel.
Research, 1987; Ryan, 1999; Smith and We focus on research that explores the link
Gihring, 2004; TCRP, 1998; Vessali, 1996). between rail transit and residential property
Observers have noted that inconsistent values, ignoring for the moment research on
results and weak evidence of a transport– the connection with commercial-office
land-use connection may be due to the com- property values (Bollinger and Ihlandfeldt,
plexity of metropolitan development and 1997; Cervero, 1994; Cervero and Duncan,
unpredictable travel patterns (Ryan, 1999). 2002b; Cervero and Landis, 1995; Landis
In addition, many studies are context-specific et al., 1995; Hickling Lewis Brod Inc., 2002;
and often fail to produce generalisable results Nelson, 1999; Ryan, 2005; TCRP, 1998;
because of unique research methods, unique Weinberger, 2000, 2001) and retail (Cervero,
local transport systems and land use 1994; Damm et al., 1980; TCRP, 1998)
environments. which have generally demonstrated advan-
Researchers have also quantified land use tages for commercial property to rail proxi-
premiums achieved for various modes of mity. The effects on nearby property values
public transit, including bus, light rail, heavy have been well-documented for BART, a
rail and commuter rail. There is general agree- heavy-rail system serving San Francisco and
ment amongst most researchers (including Oakland, California, where the effect on prop-
Cervero and Duncan, 2002; Lewis-Workman erty values increased as the rail system’s
and Brod, 1997) that property near heavy maturation strongly influenced commuting
rail accrues greater benefits than property patterns (Cervero and Landis, 1995; Landis
near light rail, owing to the faster speeds, fre- et al. 1995) and affected, to a lesser degree,
quent trains and greater geographical cover- land use (Cervero and Landis, 1995).
age of heavy rail. Studies of the effect of Keeping in mind that we have carefully nar-
proximity to commuter rail have shown that rowed the ample literature to match closely
property near commuter rail stations can our focus of inquiry, Table 1 summarises pre-
have higher premiums than light rail or vious empirical research on the effect of
heavy rail, especially when a commuter rail proximity to light rail and heavy rail transit
station is at the centre of, or within easy on residential property values. Numerous
walking distance of, a commercial core or studies appearing regularly in scholarly publi-
main street (Armstrong, 1994; Cervero, cations since the 1980s provide transport
1984; Cervero and Duncan, 2002a, 2002b; researchers with ample evidence for observing
Parsons Brinckerhoff, 2001). Several studies how both light and heavy rail systems affect
(Barker, 1998; Goodwin and Lewis, 1997; real estate markets. Some of these studies
Koutsopoulos, 1977) have shown that prop- examine the value of property near transit
erty values near bus routes have only modest stations and then make comparisons with
gains, if any, from transit proximity, because similar properties not near transit stations.
most bus routes lack the permanence of Other studies develop hedonic estimation
fixed infrastructure. Researchers have also models, using residential property values or
investigated the effect on property values in apartment rent prices as the dependent
Table 1. Effect of proximity to rail transit on residential property values

(transit system) Property value data Findings: effect of proximity to rail transit
Authors source Access measurement on property value or average home price


Atlanta (MARTA) Sales prices (DeKalb Straight-line distance to station Property value increased in low-income
Nelson, 1992 County tax Measured in 100-ft units from station neighbourhoods but decreased in high-
assessor) income neighbourhoods
Atlanta (MARTA) Sales prices (TRW- Distance rings of one-quarter mile, one- Property value between one to three miles of
Bowes and REDI data quarter to one-half mile, one-half mile to stations increased relative to comparable
Ihlanfeldt, 2001 services) one mile, one to two miles, and two to properties located more than three miles.
three miles from station Properties within one-quarter mile
All property transactions in the city decreased by 19 per cent compared with
properties beyond three miles
Dallas, Texas Assessed values Straight-line distance to station Property value increased 32 per cent near
(DART) Weinstein (Dallas County One-quarter mile from station, compared DART stations compared with 20 per cent
and Clower, 2002 Central Appraisal with properties located in the control in control group areas not served by rail
District) group
Miami, Florida Sales prices (Dade Distance to the nearest station Property value in high-income
(Miami Metro Rail) County property One square mile from station neighbourhoods slightly higher but
Gatzlaff and Smith, tax records) unaffected in low-income
1993 neighbourhoods
Study conducted for eight stations
Queens, New York Sales prices (TRW Network distance to station Property value decreased $2300 for every
(New York City real estate One mile radius 100 feet further from station
MTA) Lewis- database) Study conducted for three stations in Queens
Workman and
Brod, 1997
Philadelphia, Sales prices Proximity to rail service measured for Property value of single-family homes with
Pennsylvania (Montgomery census tracts. access to rail stations is approximately
(SEPTA) Voith, County tax 8 per cent higher than other homes
1993 assessment file)
Portland, Oregon Sales prices (RLIS Properties in rail corridor compared with Median house value increases at a faster rate
(Eastside MAX) database) properties along a parallel bus corridor, closer to stations.
Dueker and which serves as a control group A house located at a station will decrease by
Bianco, 1999 5 per cent, 2 per cent and 1 per cent, if
located at 400 feet, 600 feet and 800 feet
further from station, respectively
Portland, Oregon Sales prices (RLIS Straight-line distance to station Property value decreased $32.20 per metre
(Eastside MAX) Lite and One km radius further from station beginning at a
Chen et al., 1998 Metroscan) distance of 100 metres from station


Portland, Assessed values (city Network distance to station Property value increased $76 for every 100
Oregon(Eastside property tax rolls) One mile radius feet closer (within a one-half to one mile
MAX)Lewis- radius) to three stations that were studied
Workman and
Brod, 1997
Portland, Oregon Sales prices Distance rings based on walking Property values were $4.32 higher within
(Eastside MAX) One km radius (Model 1)Network distance the 500 metres than outside the 500 metre
Al-Mosaind et al., to station radius (Model 1).
1993 One-half km radius (Model 2) Property value decreased $2175 for every
100 metres further from station (Model 2)
Study conducted at seven suburban stations.
Sacramento, Sales prices (TRW- Network distance to station No statistically significant effect on home
California REDI data All property transactions in the city prices
(Sacramento Light services)
Rail) Landis et al.,
San Diego, California Sales prices (TRW- Network distance to station Property value increased $272 for every 100
(San Diego REDI data All property transactions in the city metres closer to station
Trolley) Landis services)
et al., 1995
San Diego, California Sales prices Distance rings of one-quarter and one-half Property value increased 10 per cent (East
(San Diego (Metroscan) mile from station Line) to 17 per cent (South Line) for
Trolley) Cervero multi-family homes
and Duncan, 2002a

(Table continued)
Table 1. Continued
(transit system) Property value data Findings: effect of proximity to rail transit
Authors source Access measurement on property value or average home price


San Francisco Sales prices Network distance to station Property value decreased $1578 for every
(BART) Lewis- One mile radius 100 feet further from station
Workman and Study conducted for one station
Brod, 1997 (Pleasant Hill)
San Francisco, Sales prices (TRW- Network distance to station Property value (1990) for single-family
California (BART) REDI data All property transactions in the two counties homes decline by $1.00 to $2.00 for every
Landis et al., 1995 services) metre further from BART station
San Jose, California Sales prices (TRW- Network distance to station Property value decreased $197 for every 100
(San Jose Light REDI data All property transactions in the city metres closer to station
Rail) Landis et al., services) Effect may be due to commercial and
1995 industrial uses
San Jose (Santa Clara Sales prices Straight-line distance to station Property value of apartments and homes
VTA) Cervero and (Metroscan) One-quarter to one-half mile from station increased 1 –4 per cent, but decreased
Duncan, 2002a 6 per cent for condominiums
St Louis, Missouri Sales prices (First Straight-line distance to station Property value increased 32 per cent or $140
(MetroLink) American Real One mile buffer around the light rail for every 10 feet closer to station,
Garrett, 2004 Estate Solutions) route and stations beginning at 1460 feet
Washington, DC Apartment rents Distance to station is measured in tenths of Apartment rent decreased 2.5 per cent for
(Metro) Benjamin (various property miles from station every one-tenth mile further from station
and Sirmins, 1996 management and All properties in the city There were a total of 250 observations from
apartment locator 81 apartment complexes

variable, to estimate premiums on property in Research Plan

rail station areas. Most of the studies use stat-
istical controls when investigating property
values and the explanatory variables usually Much of the research about this topic has been
include characteristics of properties and the conducted in growing metropolitan regions in
neighbourhood quality in which properties high-growth states such as California and
are located. Some studies consider suburban Texas (see Table 1). Our current investigation,
stations, but most consider stations in urban however, fills a gap in this research by exam-
neighbourhoods, as all rail networks serve a ining Buffalo, a slow-growth city that has
central business district. The studies—which experienced steep economic decline and stag-
vary in their degrees of sophistication—inves- gering population loss in the past several
tigate property values in a wide range of decades.
metropolitan settings, including downtowns, Research about the impact of rail transit on
urban neighbourhoods, mature suburbs, new adjoining property values is important in a
suburbs and exurbs. Rustbelt metropolitan area with decreasing
In general, the studies do indeed find a stat- and sprawling population (Fairbanks, 2004)
istically significant and positive correlation as it helps to shed light on the promise of
between access to rail stations and residential public transit for revitalising cities and the
property values—although magnitudes vary— potential for planning future rail transit invest-
and the greatest effects are found within areas ments. We intend the findings from this
nearest the transit stations, although being too research to be useful to urban planners and
close to a station can sometimes result in a transit officials in Buffalo at a time when
price decrease. Findings of these studies indi- they are considering various methods to
cate that property in transit station areas experi- increase transit ridership, including station
enced up to a 32 per cent premium (Garrett, area development (NFTA, 2004) and to plan-
2004) (see also Debrezion et al., 2006). ners and policy-makers anywhere who seek
However, a few studies found no significant knowledge about the complexities of the inter-
effect on property values (Gatzlaff and Smith, relationship between housing markets and
1993) and studies in San Diego and San Jose public transit accessibility.
found a decrease in property values, which Perhaps no metropolitan region with a rela-
may be explained by the industrial character tively ‘new’ light rail system has experienced
of certain neighbourhoods (Ihlanfeldt and urban decline to the degree that Buffalo has.
Boehm, 1987) served by light rail aligned We consider the 20-year population growth
along freight corridors. Property values or decline—for the period between 1980 and
increased in low-income neighbourhoods in 2000—for the cities and metropolitan
Atlanta and increased in high-income neigh- regions in Table 1 with new light rail
bourhoods in Miami, reinforcing the obser- systems. Most of these light rail systems,
vation that the studies are highly context- and subsequent extensions, were constructed
specific, with effects on land values sometimes in the 1980s and 1990s. Figure 1 shows that
realised unevenly throughout transit systems the population loss of 18 per cent in the City
and unevenly across various neighbourhood of Buffalo and 6 per cent in its metropolitan
types. In some cases, negative premiums on area over the 20-year period is unparallelled.
property values are observed and attributed to Only one other central city—St Louis—lost
nuisance effects, including noise, safety, aes- population, but that was offset to some
thetics and traffic (Al-Mosaind et al., 1993; degree by population growth in the metropoli-
Landis et al., 1995). While nuisance effects tan area. All of the central cities increased
may reduce property prices in immediate population by 30 per cent or more and all of
station areas, the net positive effect on land the metropolitan areas (except for St Louis)
values from a regional perspective is often con- increased in population by 30 per cent or
sidered to outweigh localised devaluations. more. Population loss in the Buffalo

Figure 1. Population change, 1980 –2000. Source: US Census, 1980, 2000.

metropolitan area has certainly been tremen- transformation to land use near rail stations
dous, especially in the central city, and this are stabilised. While effects from transport
provides a unique opportunity for studying improvements are generally expected to
the influence of light rail proximity on prop- develop over a long period of time (Giuliano,
erty values. 2004), we assume that the residential property
value response to light rail proximity has by
now reached equilibrium.
Previous research shows that accessibility
benefits accrue for properties—in a variety of Methodology
contexts and settings—by proximity to a rail
Study Area
station and that these benefits capitalise in
land value. We therefore assume that, as the A rapid transit system was first proposed for
distance to a light rail station decreases, acces- Buffalo in the late 1960s to connect downtown
sibility benefits accrued by homeowners or Buffalo with various suburban locations like
occupants will be greater, resulting in a the Greater Buffalo International Airport, the
higher price of property value. Our null new State University of New York suburban
hypothesis is that as the distance to the rail campus in Amherst, downtown Niagara Falls
station increases, there is no impact on property and southern suburbs. One of the key objec-
values, implying that proximity to Buffalo’s tives of the new transit plan was to anchor
light rail accrues insignificant accessibility the revitalisation of the central business dis-
benefits for nearby properties. Taking into trict and arrest population decline (Berechman
account economic decline and population loss and Paaswell, 1983; Gargan, 1984; Wallace
in Buffalo, however, we expect the light rail McHarg Roberts & Todd, 1971). In 1969,
transit effect on nearby property values to be the state’s Office of Planning Co-ordination
less than it might be in other places. Metro concluded that the Buffalo –Amherst Transit
Rail has served Buffalo for more than 20 Corridor between downtown Buffalo and the
years, so we assume that adjustments in newly proposed State University at Buffalo
people’s travel behaviour to rail service and in Amherst had the greatest potential returns

among other alternative routes and alignments suggest that planners hoped to revitalise
(New York State Office of Planning downtown and urban neighbourhoods with
Co-ordination, 1969). Urban density along Metro Rail rather than locating Metro Rail in
the corridor was believed to be sufficient to the path of new development, most of which
support heavy rail service, although planners was occurring in the suburbs.2 Figure 2
subsequently downgraded the system to light shows that the light rail system was conceptu-
rail in the 1970s because of high cost projec- alised, planned, constructed and began service
tions and lower ridership expectations. when the City of Buffalo and its suburbs were
Buffalo’s light rail system began service in in a period of serious decline—characterised
1985 and cost $1.3 billion (2005 dollars). It by out-migration of population and intraregio-
follows the alignment of Main Street, a com- nal suburbanisation (Banister and Berechman,
mercial spine, from the Inner Harbor in down- 2000)—that unarguably dampened housing
town Buffalo to the State University at Buffalo demand.3
south campus. Metro Rail enhances accessi- Like many Northeastern and Rustbelt cities,
bility by providing additional commuter Buffalo has experienced abandonment and
capacity, but it did not provide transit access economic decline—especially a loss of blue-
in new places when it was built, as the route collar employment—since the 1950s as its
serves downtown and established neighbour- manufacturing and industrial base became
hoods that already had bus service (Banister obsolete. Population declined by 50 per cent
and Berechman, 2000). between 1950 and 2000, and 2000 marked
The light rail line departs from its original the first time since 1890 that the city’s popu-
conception (New York State Office of Plan- lation fell below 300 000 people. Demo-
ning Co-ordination, 1969) by offering no graphic data reveal that the median annual
service in the suburbs; as a consequence, the household income is $24 500 and that 21 per
6.4-mile route operates at-grade downtown cent of working-age adults are in poverty
and below-grade en route to the city’s north- (Hess, 2005; US Census 2000, Summary
eastern boundary. The location and route Files 1 and 3). Like other Rustbelt cities,

Figure 2. Population of City of Buffalo and Erie County, 1810–2010. Sources: see note 3.

Buffalo experiences racial segregation, revenue. Beginning in 2001– 02, residential

poverty, low incomes and low automobile property values jumped 22 per cent in 4
ownership, and limited access and mobility years, while commercial assessments
for many segments of the population increased only 2 per cent (Cervantes and
(Buckham, 2004; Hess, 2005). The region Dolan, 2005).
also has seen sluggish job growth, loss of Despite a presumed high degree of transit
high-income jobs, population decline in the dependency—31 per cent of households do
important 20- to 34-year-old age-group and not have a vehicle available (Census
high out-migration (Fiscal Policy Institute, 2000)—ridership on Metro Rail has declined
2003; Pendall, 2003). The Buffalo metropoli- steadily (Lakamp, 1999) and has consistently
tan area lags behind the US in both earnings underperformed expectations (Kain, 1988;
and investment (Pendall and Christopherson, Pickrell, 1992; Richmond, 2005; UMTA,
2004) and high local and county taxes and per- 1989).5 Figure 3 shows that average
sistent tax increases have hurt the competi- weekday boardings for 2002 and 2003, the
tiveness of the region in attracting most recent years for which data are available,
businesses and residents (Duncombe, 2002). are below 20 000, which is lower than the
The patterns have become entrenched and number of boardings in 1986, the first full
now negatively affect property values city- year of operation.
wide (Potepan, 1996), which are further Figure 4 shows the 6.4-mile Metro Rail
depressed by an expectation of little growth route and 14 stations. The six southerly
(Capozza and Helsley, 1989).4 In recent stations are located downtown in a free-fare
years, homeowners, not commercial zone along an automobile-free pedestrian
businesses, have shouldered most of the mall, through which Metro Rail operates at
burden of the increase in property tax grade. Although hundreds of residential units

Figure 3. Average weekday unlinked trips on Metro Rail, 1985– 2003. Source: NFTA (2005).

Figure 4. Study area.

are planned or under construction, Buffalo’s Pittsburgh. Metro Rail descends into a tunnel
downtown residential population is small to serve the eight northerly stations, which
and has not expanded in recent years as are located underground. Pedestrian access
downtown neighbourhoods have in other around stations is generally good, with
rustbelt cities such as Cleveland and mature street networks and sidewalks,

although station amenities—especially retail central business district. GIS was also used
and commuter services—are lacking (Hess to determine whether a parcel is located on
and Lombardi, 2004). Metro Rail riders can the east side or west side of the light rail
make direct transfers to bus routes at all corridor.
eight of the below-ground stations and can Planners typically assume that people will
easily transfer to bus routes from the six at- comfortably walk approximately one-quarter
grade stations; from Metro Rail stations, of a mile to reach transit stops or stations
riders can transfer to 46 of the system’s 55 (Untermann, 1984). We expand the pedestrian
bus routes. Free parking is provided at park- access distance to a half mile radius around
and-ride lots at the two stations furthest from stations, in order to capture the variation in
the central business district, LaSalle and Uni- property values not necessarily observed
versity; no other Metro Rail stations possess within the one-quarter mile radius. However,
adjacent parking lots. the nearness of station spacing in Buffalo
means that half-mile buffer areas around
stations sometimes overlap, as shown in
Data Sources
Figure 5. The average distance between adja-
While many previous studies have used real cent stations is approximately half a mile.6
estate transaction databases, we instead use GIS is used to draw continuous straight-line
the 2002 assessed value of properties from distances from each parcel to the nearest
the assessor’s database for the City of station, thereby assigning each parcel to the
Buffalo, the most comprehensive and reliable nearest station. That is, each parcel is assigned
government data source for residential prop- to the nearest station; if a parcel is within half
erty values. Our sample thus includes the uni- a mile of two stations, it is used in the model
verse of residential properties in our study only for its assigned station.
area, while real estate transactions would
only have included properties that changed
Property Value Estimation
hands. In Buffalo, residential properties are
assessed based upon 100 per cent of market The price of property is influenced by a
value. This assessor’s database contains number of factors, including population
descriptive variables such as the number of growth, economic conditions, quality of life
bedrooms, the number of bathrooms and the and nearby amenities (Brigham, 1965;
year the structure was built. The benefit of Cervero et al., 2002; Landis et al., 1995;
this database, compared with other relevant Quigley, 1985). In Buffalo, property prices
real estate databases such as sales trans- are affected by factors relevant to the declin-
actions, is the absence of significant bias. ing economy such as population loss, crime
We consider further only residential parcels in neighbourhoods, low incomes and high
for which all attributes are available. vacancy rates. To specify an appropriate
To include the effect of other variables such model, we first plot on a scattergram the
as incidence of crime and types of crime, 2002 values of the dependent variable (assessed
data were collected from the Erie County property value) as a function of the indepen-
Central Police Service and the Buffalo dent variable (distance from property to
Police Department. Occupancy rate change nearest station). As the distance from a rail
and population growth rate change from station increases, the assessed property value
1990 to 2000 were computed using demo- changes proportionally by either increasing
graphic data collected from the 1990 and or decreasing. The linear relationship revealed
2000 US Census. in the scattergram is not surprising, consider-
GIS was used to measure continuous geo- ing Buffalo’s relatively flat housing market,
graphical distances such as distances from where within the Metro Rail corridor there
the parcels to the nearest station, to the are no pockets of high housing demand associ-
nearest park, to Delaware Park and to the ated with economic activity that one might

Figure 5. Half-mile radius buffer areas surrounding metro rail stations.

expect to find in a high-growth region. We value attributed to proximity to light rail

choose hedonic regression analysis7 for stations (Rosen, 1974; Can, 1990).8 The con-
hypothesis testing, since this method isolates ceptual hedonic regression model is
the factors that might affect property values
and estimates the influence on property P ¼ f(Pr, H, L, N) (1)

where, the dependent variable P is the nearest rail station are shown as straight-
assessed property value in dollars, which is a line distances on the left-hand side and
function of four vectors of independent as street network distances on the right-
variables. The four integrated vectors are: hand side.
Pr, a vector of variables that measure the (2) Property characteristics (H) are attributes
proximity of properties to light rail stations; unique to a parcel and the residential
H, a vector of variables that describe structure on the parcel (Kain and
housing characteristics; L, a vector of Quigley, 1970b; Follain and Jimenez,
variables that describe locational amenities; 1985). They include lot area (AREA),
and N, a vector of variables that describe age of house (AGE), number of bedrooms
neighbourhood characteristics. (BEDRMS), number of bathrooms
In this way, we estimate the importance of (BATHRMS) and number of fireplaces
housing characteristics, locational amenities (FIRE). Dichotomous variables indicat-
and neighbourhood characteristics in defining ing single- as opposed to multi-family
the relationship between residential property housing (SINGLE) and the presence of a
value and light rail station proximity. basement (BASEMT) are included.12
Table 2 provides a list of the 18 independent Parcels that do not contain all of these
variables in the 4 vectors, along with the defi- attributes (approximately 2 per cent) are
nition, unit of measurement and data source eliminated from the study.
for each variable. The property value, P, is (3) Accessibility and location (AL) includes
reported for each of the 7357 residential variables that describe city-wide access,
parcels in 2002 dollars. Properties are proximity to amenities (Fausold and
assessed at 2002 market values.9 Lilieholm, 1999; Schroeder, 1982) and
The four vectors of independent variables proximity to the central business district
that influence property values are described based on a standard model of urban land
in greater detail below. use (Alonso, 1964; Brigham, 1965;
Muth, 1969). In this study, Niagara
(1) Rail Proximity (Pr) measures distances Square was assumed as the focal point
(in feet) from each residential parcel to of the central business district (CBD)
the nearest rail station.10 The variable is which controls property values. This
measured using two methodologies variable was calculated by measuring
(Clifton and Handy, 2001). In the the straight-line distance from Niagara
straight-line distance model, the distance Square in downtown Buffalo to the cen-
(STR_DIST) in feet is measured along troid of each residential parcel. Other
the shortest possible straight line—some- variables include proximity (measured
times called crow-flight distance—drawn by straight-line distance) to the nearest
using GIS and connecting each residential park (NPRK), proximity to Delaware
parcel to the nearest rail station. In the Park (DELPRK )—the largest park in
network distance model, the distance the city and surrounded by cultural
(WALK_DIST) in feet is measured along attractions—and a dummy variable
the street network, simulating the actual (EAST) to indicate whether parcels are
pedestrian route of travel that a commuter located on the east side or the west
would walk between home and the rail side of Buffalo. Main Street acts as an
station.11 The network distance between historical segregating divider between
a property and the nearest station may the city’s dilapidated East Side, where
be the same as the straight-line distance, there are high shares of Blacks/African
but it can never be shorter. These two Americans and low-income families,
methods are shown side-by-side in from the West Side, where there are
Figure 6 for four station areas. The dis- more Whites and middle- and high-
tances drawn between all parcels to the income families.13
Table 2. Definition of variables and data sources
Vector Variable Operational definition Units Data source
Dependent variable
Assessed property PRICE Assessed value (per square foot) of land $ (2002) City assessor
value (PRICE) plus improvements


Independent variables
Rail station proximity STR_DIST Straight-line distance Feet Calculated using GIS
(Pr) WALK_DIST Walking distance along street network Feet Calculated using GIS network analyst
Property AREA Area of parcel Square feet City assessor
characteristics(H) AGE Age of structure Years City assessor
BEDS Number of bedrooms Number City assessor
BATHRMS Number of bathrooms Number City assessor
SINGLE Single-family housing Dichotomous variable City assessor and NY State ORPS
(0 or 1)
FIRE Number of fireplaces Number City assessor
BASEMT Presence of basement Dichotomous variable City assessor
(0 or 1)
Locational amenities CBD Straight-line distance to CBD Feet Calculated using GIS
(L) NPARK Straight-line distance to the nearest park Feet Calculated using GIS
DELPRK Straight-line distance to Delaware Park Feet Calculated using GIS
EAST East side dummy variable Dichotomous variable Calculated using GIS
(0 or 1)
Neighbourhood PROPCMR Property crime rate Crimes per capita (for Buffalo Police Dept (ECCPS) 2002 and
characteristics (N) block group) US Census 2000
VCMR Violent crime rate Crimes per capita (for Buffalo Police Dept (ECCPS) 2002 and
block group) US Census 2000
MEDINC1999 Median income $ (1999) (for block US Census 1990, 2000
OCCRTE Occupancy rate change Fraction per decade US Census 1990, 2000
POPRTE Population growth rate Fraction per decade US Census 1990, 2000


Figure 6. Straight-line distance and network distance.

(4) Neighbourhood characteristics (N) con- variables are designed to capture

tains attributes unique to a group of changes in housing demand.14
contiguous properties that may affect
property values (Bowes and Ihlandfeldt,
2001; Ding et al., 2000; Malpezzi,
1996; Muth and Goodman, 1989). The The cross-sectional assessment database
vector includes median household includes 7357 residential parcels—both
income (MEDINC1999) (Yinger, 1979) single-family and multi-family houses—
and two measures of neighbourhood located within a half mile radius of each of
crime (Manning, 1988)—property crime the rail stations.15 We enrich the parcel
rate (PROPCMR) and violent crime dataset by combining it with GIS to allow
rate (VCMR). The crime rates are calcu- measurement of distances between properties
lated by dividing the number of crimes and rail stations.
that occurred in a census tract by the Previous research (see Table 1) suggests
population of that census tract. Other that there are two methods for measuring the
neighbourhood characteristics include distance from a property to a transit station.
occupancy rate change (OCCRTE)—the The first method uses the straight-line distance
change in housing occupancy rate from a property to the nearest rail station and
between 1990 and 2000—and population thus measures perceived distance. That is, a
growth rate change (POPRTE)—the property owner may perceive a transit station
population growth rate between 1990 to be within a certain distance of the property,
and 2000 (US Census 1990 and 2000, even though natural or man-made barriers may
Summary Files 2 and 3). The latter two prevent a pedestrian from travelling along the

shortest straight-line route to reach the transit Findings

station. The second method uses a route
All Stations Model
along the street network from a property to
the nearest rail station and measures actual We first report the results of the all stations
walking distance. We assume that pedestrians model—a summation of the relationships
only walk where there are streets, sidewalks, observed at all of the stations along the light
pathways or pedestrian bridges. We choose rail—in Table 4. The model fits the data
to measure proximity to rail using two well and explains 71 per cent of the variation
methods to investigate whether perceived dis- in residential property price. The various attri-
tance or actual walking distance to transit is butes that influence property values are sig-
more important in influencing price variations. nificant at the 0.10 level and have the
An individual regression equation is devel- expected positive or negative signs.
oped for each of the 14 stations and one The model results suggest that, system-
regression model is developed that encompasses wide, property located within a half-mile
all of the stations.16 We calculate Pearson’s cor- radius of rail stations is valued $2.31 higher
relation coefficient ‘r’ and use the results to (using geographical straight-line distance)
eliminate multicollinearity among independent and $0.99 higher (using network distance)
variables. Pairs of independent variables that for every foot closer to a light rail station.
produce a correlation coefficient of 0.6 or Consequently, an average home anywhere in
higher are eliminated from the particular the study area along the light rail line would
regression model wherein a correlation occurs. generally be worth between $990 and $2310
Table 3 provides descriptive statistics for the more than the average home if it were
dependent variables, including the mean, within 1000 feet of a station. The following
maximum, minimum and standard deviation. interpretations of estimated coefficients are

Table 3. Characteristics of variables

Vector Variable Mean deviation Minimum Maximum
Dependent variable
Assessed Property PRICE 70 380.49 52 901.88 1 000.00 1 100 000.00
Value (PRICE)
Independent variables
Rail proximity (Pr) STR_DIST 1 803.34 536.08 129.78 2 610.95
WALK_DIST 2 502.48 803.03 46.03 4936.15
Property AREA 4 804.90 3 083.25 297.00 142 345.00
characteristics (H) AGE 87.81 22.28 6 203
BEDS 4.39 1.37 0 10
BATHRMS 1.87 0.68 1 7
SINGLE a 0.55 0.50 0 1
FIRE 0.30 0.57 0 7
BASEMT a 0.96 0.19 0 1
Locational amenities CBD 17 458.48 7 853.10 1 361.05 30 183.52
(L) NPARK 4 187.41 1 820.60 369.36 8 224.25
DELPRK 8 777.26 3 358.01 1 895.69 22 816.25
EAST a 0.53 0.50 0 1
Neighbourhood PROPCMR 0.04 0.02 0.00 0.56
quality (N) VCMR 0.02 0.02 0.00 0.51
MEDINC1999 31 299.28 20 117.74 0 146 379.00
OCCRTE 20.03 0.21 21.00 1.68
POPRTE 20.06 0.21 21.00 1.62

Notes: aDichotomous variables. All other independent variables are continuous.


Table 4. Estimated model for all stations

Straight-line distance Network distance
Independent Parameter Parameter
Vector variable estimate Probability estimate Probability
Rail proximity (Pr) STR_DIST 22.31 23.50
WALK_DIST 20.99 22.22
Property AREA 4.09 32.84 4.09 32.77
characteristics (H) AGE 233.97 22.02 232.31 21.92
BATHRMS 25054.33 39.10 25085.11 39.13
SINGLE 15399.70 19.06 15485.18 19.14
FIRE 16178.64 23.20 16230.09 23.26
BASEMT 7445.44 4.06 7561.51 4.12
Locational CBD 0.36 6.99 0.38 7.41
amenities (L) NPARK NA NA NA NA
DELPRK 0.32 2.91 0.29 2.67
EAST 228516.69 236.21 228418.35 236.09
Neighbourhood PROPCMR NA NA NA NA
quality (N) VCMR 2291.67 214.94 2287.03 214.72
MEDINC1999 0.36 17.02 0.36 16.95
POPRTE 13436.83 7.21 13943.13 7.46
Model statistics n ¼ 7357; df ¼ 13; n ¼ 7357; df ¼ 13;
adjusted r 2 ¼ 0.71; adjusted r 2 ¼ 0.71;
p ¼ 0.05 p ¼ 0.05

Notes:  p , 0.10, significant at the 0.10 level;  p , 0.05, significant at the 0.05 level;  p , 0.01, significant at the 0.01
level. Variables marked NA are eliminated due to multicollinearity.

drawn from the vectors of independent second most statistically significant indepen-
variables dent variable. For every foot closer to the
CBD, there is a small decrease in property
—Among property characteristics, AREA and value at the rate of $0.36 per foot, indicating
BATHRMS are the most statistically signifi- consistently higher property values in neigh-
cant independent variables and both posi- bourhoods further from downtown.
tively influence property value. For every —Among variables describing neighbour-
square foot increase in the lot area, the hood quality, median income level
property value increases by about $4 (for (MEDINC2000) followed by violent crime
both straight-line and network distance rate (VCMR) are the most statistically sig-
model). For each additional bathroom, the nificant. Median income has only a mar-
property value increases by about $25 000 ginal positive impact on property values:
(for both straight-line and network distance every $100 increase in the median annual
model). household income is associated with a $36
—Among locational characteristics, the EAST increase in property value, suggesting that
dummy variable is the most statistically houses are more likely to sell for a higher
significant. The model suggests that if a prop- price in affluent neighbourhoods. Higher
erty is located on the East Side—which has violent crime rate has a negative impact
the city’s largest share of African Americans on property value: every 1 per cent increase
and low-income households—property in violent crime rate is associated with a
values are likely to decrease by about decrease in property value of about $292
$28 000 (in both straight-line and network dis- (straight-line distance model) and about
tance model). The distance to the CBD is the $287 (network distance model).

The independent variables possess various Individual Station Models

units of measurement, making it difficult to
The all stations model provides an analysis of
compare the relative impact of each indepen-
the cumulative effects of rail proximity on
dent variable on the dependent variable. To
adjoining station area properties for the
overcome this, we calculate the standardised
entire light rail system. The results indicate a
partial regression coefficient, bstandardised, for
marginally positive impact on property
each independent variable using the following
values. However, individual regression
models for each of the 14 stations indicate
 that the effects are not felt evenly throughout
Bstandardised ¼ ½bi (std xi )=(std y) (2)
the system (Almeida, 2004). Summary
where, bi is the slope coefficient of an results are presented in Table 5.
independent variable; std xi is the standard Table 5 also lists the median household
deviation of an independent variable; and income near each station which suggests the
std y is the standard deviation of the dependent relative affluence of station-area neighbour-
variable. hoods. The premiums homeowners will pay
Thus, the partial regression coefficients are for station proximity is greater in high- than
converted to b-values in the all stations model in low-income neighbourhoods, despite the
using the following empirical relationship assumption that many low-income residents
who cannot afford automobiles would pay a
PRICE ¼ b0  0:02(STR DIST) premium on housing to live near a light rail
station. Near some stations, the effects of
þ 0:24(AREA) light rail proximity may be dampened by
 0:01(AGE) þ 0:32 overall neighbourhood effects on property
(BATHRMS) þ 0:15(SINGLE) values (Knaap, 1998).
Proximity to University station, Amherst
þ 0:17(FIRE) þ 0:03(BASEMT) station, Delavan-Canisius College station
þ 0:05(CBD) þ 0:02(DELPARK) and Allen-Medical Campus station has stat-
 0:27(EAST)  0:11(VCMR) istically significant positive effects on prop-
erty values in both models and proximity to
þ 0:14(MEDINC1999) Utica station, Summer-Best station and
þ 0:05(POPRTE) (3) Theater station has statistically significant
negative effects on property values in both
When we standardise the unit of measurement models. LaSalle station shows statistically
for the independent variables to make them significant results in only the straight-line dis-
comparable, three variables—number of bath- tance model and Fountain Plaza shows stat-
rooms (BATHRMS), size of the parcel (AREA) istically significant results in only the
and location on the East Side of Buffalo network model. Two stations—Lafayette
(EAST)—are more influential than rail proxi- Square and Humboldt-Hospital—fail to
mity in predicting property values. The show statistically significant effects in either
b-value for number of bathrooms is about 14 model; at these two stations, there is insuffi-
times greater and the b-value for proximity cient evidence to conclude that a statistically
of the parcel on the East Side is about 12 significant relationship exists between prop-
times greater than the b-value for rail proxi- erty value and accessibility to rail transit.
mity (in both straight-line distance and Interestingly, the second-highest positive
network distance model). Parcels located on price effects on property values are felt at Uni-
the East Side of Buffalo have a negative versity station—the furthest station from the
effect on property values, whereas larger lot central business district—which does not
sizes and number of bathrooms have a positive provide support for the contention that the
impact on property values. amount of commuting time saved is

Table 5. Estimated model for individual stations

Change in property value
(in dollars) per foot closer
N (number of to station
residential Median
properties within household Straight-line Network
Station half a mile) income $ (1999) distance distance
University 697 33 828 þ6.72 þ4.39
LaSalle 1175 22 578 þ3.01 þ0.80
Amherst Street 1510 35 417 þ6.25 þ3.65
Humboldt-Hospital 830 35 667 þ2.22 23.09
Delavan-Canisius College 1020 23 125 þ2.93 þ1.51
Utica Street 946 21 719 230.72 225.74
Summer-Best 365 18 889 242.08 225.51
Allen-Medical Campus 442 23 950 þ13.69 þ4.91
Theater Station 231 18 594 227.26 222.61
Fountain Plaza 77 18 125 þ3.88 þ27.03
Lafayette Square 35 17 035 22.02 20.99
Church Street ,30 Insufficient n for estimation
Seneca Street ,30 Insufficient n for estimation
Erie Canal Harbor ,30 Insufficient n for estimation
All stations 7357 26 295 þ2.31 þ0.99

Notes:  significant at the 95 per cent level or greater.

capitalised into housing values. The neigh- We group the stations with statistically sig-
bourhoods in Buffalo are perhaps just too nificant results by $10 000 intervals into high,
uneven for a traditional pattern (Giuliano, medium, and low annual income, and average
2004) in which housing prices are highest the effect of rail transit proximity on property
near the central business district and experi- values. The results are shown in Table 6. Con-
ence spatial decay moving towards the trary to findings in other metropolitan areas
suburbs. (Bowes and Ihlanfeldt, 2001; Nelson, 1999),

Table 6. Effect of proximity to light rail stations on property values by neighbourhood prosperity
Average change in property
value in dollars (per foot
Median further from station)
Neighbourhood household Number of
income income $ Metro Rail Straight-line Walking
category (1999) stations Stations distance distance
High 30 000 or more 2 Amherst Street, þ6.49 þ4.02
Medium 20 000–30 000 4 Allen-Medical þ2.77 þ4.63
Campus, Utica Street,
College, LaSalle
Low Below 20 000 3 Fountain Plaza, 234.67 27.03
Theater, Summer-

Notes: Certain stations are omitted from this table because model results are not statistically significant at the 95 per cent
level: Lafayette Square and Humboldt-Hospital (both models), Fountain Plaza (straight-line distance model), LaSalle
(network distance model). Average changes in property value are unweighted.

the largest positive effects in Buffalo are felt Robert Cervero’s (1984) decades-old
in high-income neighbourhoods and the warning that, although light rail transit in
largest negative effects are felt in low- Buffalo was perceived as a necessity for
income neighbourhoods. As in many Rustbelt urban revitalisation and had sufficient political
cities, older homes and less in-demand neigh- backing, the lack of a strong regional
bourhoods tend to be closer to the central economy would limit the power of Metro
business district (Quigley, 1979) and housing Rail to revitalise downtown and neighbour-
prices gradually increase as distance from hoods (Knight, 1980).
the central business district increases in defer- The results of the all stations model
ence to traditional bid-rent models (Alonso, suggests that, throughout the system, a
1964; Mills, 1972; Muth, 1969) that place typical home located within one-quarter of a
the highest value at the centre (Burchell mile of a rail station can earn a premium of
et al., 2002; Downs, 2004). $1300– 3000, or 2 – 5 per cent of the city’s
median home value.17 Models for individual
stations suggest that the greatest positive
Actual versus Perceived Proximity to Rail
effect on price is felt at University station,
where residents benefit from rail proximity
There are two groups of residents who benefit in addition to other factors that positively
from proximity to rail stations. One group of influence property values, and the greatest
residents benefits from being geographically negative effect on price is felt at Summer-
closer to a light rail station—measured by Best station, where there is a significant nega-
straight-line distance—although they may not tive premium associated with being close to a
necessarily use or depend on transit. These resi- transit station in addition to other likely
dents or owners have a sense of proximity to factors that negatively influence property
light rail transit and consider this proximity values (Liggett et al., 2001, 2003). An analy-
as a locational advantage to residential proper- sis of controlling factors suggests that three
ties (Bae, 2002). The other group of residents variables used in the models—the number of
benefits from being within convenient bathrooms (BATHRMS ), size of the parcel
walking distance of a light rail station calcu- (AREA) and location on the East Side of
lated along the street network and generally Buffalo (EAST)—are more influential than
includes residents who actually use transit. are rail proximity and other variables in pre-
Our findings suggest that the straight-line dicting property value.18 We thus find
distance models produce more positive support for the hypothesis that proximity to
effects on property values than the network stations increases property values, but the
distance models and that the model results effect in Buffalo is weak, especially compared
have greater statistical significance. A with growing West Coast and Sunbelt cities.
greater premium for straight-line distance Surely Buffalo is unique, though, in having a
than for walking distance suggests that geo- light rail route follow a low-income– middle-
graphical distance to a transit station is income and White –Black/African American
perhaps more important in Buffalo than dividing line.
walking distance. Based on these findings we cannot claim
that amid economic decline and population
loss, light rail transit will unequivocally
increase property values and revitalise
This study adds to the growing literature on depressed neighbourhoods. This is perhaps
the positive relationship between light rail unsurprising, as previous research in Miami
transit and property values, and provides (Gatzlaff and Smith, 1993) and other places
needed evidence for the interrelation demonstrates that properties in neighbour-
between transit access and land values in a hoods experiencing decline show no relative
slow-growth city. We thus find support for benefit in property values. Variation in the

influence of transit on property values by gaining taxpayer support, raising necessary

neighbourhood type in Buffalo echoes find- funds and overcoming political discord
ings in other cities as well (Gatzlaff and (Hess and Lombardi, 2005; Lombardi,
Smith, 1993).19 Clearly, price premium pat- 2006). As a consequence, the 6.4-mile light
terns are not felt widely or evenly throughout rail line, which is only a portion of the
the light rail system as it traverses through intended starter line to the University at
various neighbourhoods. Buffalo Amherst campus, fails to provide
Like many proposals for capital expansion regional accessibility.22 Metro Rail riders
of transit facilities (Anderson, 1970), early have a one-seat transit ride only to desti-
studies for light rail transit in Buffalo justified nations along the corridor; for other desti-
the expenditure by citing indirect or non-user nations in the city—and all destinations in
benefits (New York State Office of Planning the suburbs—Metro Rail riders must transfer
Co-ordination 1969). Today, however, the to buses. While Metro Rail adequately
existence of light rail in Buffalo influences serves an important mobility function for
land values only modestly—and negatively many residents—and particularly those
near three stations. without automobiles—it has little influence
Where access to rail transit is not highly on regional accessibility and travel behaviour
valued, property values do not rise. The analy- (Banister and Berechman, 2000).
sis suggests that transit accessibility does not On the bright side, we consider the lack of
play a large role in property values in high price premiums near Metro Rail stations
Buffalo. We conclude that the access advan- in Buffalo to be an enormous opportunity. In a
tages20 provided by Buffalo’s Metro Rail are region where no prospect to increase access,
not valued to the degree that proximity to attract jobs, workers and residents, and encou-
transit produces price premiums in growing rage economic growth can be ignored, urban
places such as Dallas (Weinstein and planners and municipal officials should
Clower, 2002), Portland (Chen et al., 1998; exploit the existence of reasonably priced
Lewis-Workman and Brod, 1997; Al- properties within many stable neighbourhoods
Mosaind et al., 1993), and San Diego adjacent to light rail stations and Metro Rail’s
(Landis et al, 1995; Cervero and Duncan, relatively speedy service (7-minute headway
2002a) or in certain high-income neighbour- during the peak), low fares and uncongested
hoods in places like Atlanta (Nelson, 1992). travel. The corridor’s universities and col-
This can be attributed to a lack of regional leges, medical laboratories and cultural insti-
accessibility to employment centres (Diaz, tutions, which are solidly supported by
1999) and a lack of development within affordable housing, can help to improve
station areas throughout Buffalo. A fragmen- local and regional prosperity and quality of
ted planning process for Metro Rail in the life.
1960s and 1970s lacked co-ordination
between New York State and the NFTA and
local-level municipal planning; today, Notes
however, every effort should be made to co- 1. For background about residential location
ordinate transport and transit planning with theory, see Giuliano (2004) and Weisbrod
land use. et al. (1980).
2. Observers argue that the nine-year construc-
We find little capitalisation of the light rail tion period—during which the 1.2-mile
system into property values (Skaburskis, Buffalo Place pedestrian mall through the
1982) adjacent to Metro Rail stations and we heart of downtown was impassable for all
attribute this, at least in part, to the limited but pedestrians—dealt a fatal blow to the
extent of service. In the 20 years since light city’s faltering economy, especially down-
town retail enterprises, which are now
rail service began, the region has been nearly non-existent.
unable to agree on how to proceed with a 3. Figure created using population from the US
plan to extend Metro Rail.21 This includes Decennial Census. Population data before

1830 for Buffalo and before 1900 for Erie available for only those parcels that have
County are available from Buffalo and Erie either been purchased or sold. In Buffalo’s
County Public Library (2005). Population depressed economy, it would be difficult to
projection for 2010 for Buffalo from Office amass a large number of sales transactions
of Strategic Planning (2004) and for Erie to study accurately the properties within a
County from New York Statistical Infor- half-mile radius of transit stations. Despite
mation System (2005). this limitation, we used the assessed property
4. The low property values in Buffalo are not data for the entire study; therefore, all prop-
expected to depreciate in the near future, as erties in this research are subject to the same
they might in a high-growth city such as error if any exists. Our goal, however, is not
Boston (Glynn, 2005). to reveal actual assessed or sales prices, but
5. A series of transit planning and environ- to shed light on the magnitude of variations
mental reports sequentially lowered the esti- in property value throughout the light rail
mated number of daily weekday riders from corridor.
92 000 to 46 000 (Richmond, 2005; Voor- 10. While proximity to transit stations produces
hees and Associates, 1976). The number of accessibility benefits, proximity to transit
average weekday unlinked trips peaked at stations can simultaneously produce nega-
29 000 in the early 1990s and has now tive effects, such as noise, visual clutter
dropped by nearly 60 per cent. and nuisance factors. This study, however,
6. The minimum station spacing (0.2 mile) is does not account for any negative effects
between Erie Canal-Harbor Station and arising from being close to the rail line,
Seneca Station and the maximum station which is relevant for the at-grade sections
spacing (1.1 miles) is between Humboldt- of Metro Rail.
Hospital and Amherst Street. 11. Clifton and Handy also assess several spatial
7. Linear regression was performed because analysis methods and report the following
the distribution of data on a scattergram
plot revealed no parabolic or exponential Many accessibility measures make use of
curve. Nevertheless, we experimented with straight-line distances, or buffer distances
both linear regression and linear-logarithmic . . . At the neighborhood scale, the choice
transformation (of the assessed property between using straight-line distances and
value and distance to transit variables); for actual network distances can result in
the majority of station-area models, the different assessments of the transportation
linear model had higher adjusted R 2 environment. This is particularly true for
values. Furthermore, variables describing pedestrian accessibility, where a discre-
the built environment tend to perform pancy of one-quarter mile can influence
better in linear models than in other types the choice of whether to walk or drive to
of specification, compared with variables a destination (Clifton and Handy, 2001,
describing economic characteristics, which p, 22).
perform better in logarithmic models
(Targa and Clifton, 2005; Galea et al., 12. SINGLE is a dichotomous variable coded 1
2005). We therefore limit our discussion in for all single family houses and coded 0 for
this article to linear regression, the most par- all residential dwellings that cannot be
simonious model specification. Although defined as single-family. Similarly,
this article focuses on linear regression, BASEMT is a dichotomous variable coded
other researchers have explored the relation- 1 if a house has either a partial basement
ship between property price and accessibility or a full basement and coded 0 for all
using both additive and multiplicative func- houses that have no basement or a crawl-
tional regression forms (Skaburskis, 1982). space basement.
8. For discussion of hedonic regression theory, 13. The East–West variable is similar to
see Freeman (1979) and Kain and Quigley Nelson’s (1992) variable indicating
(1970a). For examples of hedonic modelling whether properties near MARTA stations
in similar research contexts, see Bajic are located in low-income or high-income
(1985), Ding et al. (2000), and Palmquist neighborhoods.
(1984). 14. All parcels located in a particular census
9. A limitation of this study is the use of block group are assigned the occupancy
assessed property values instead of sale rate and population growth rate of that
prices. Some observers argue that sales block group.
prices provide better data for the property 15. Within the study area, 55 per cent of housing
value study; however, sale prices are units are single family homes, 40 per cent are

two-family homes, 3 per cent are three Portland’s light rail line has also been
family homes and 2 per cent are four or extended three times since its original line
more family homes or buildings. Our opened in 1986 (GAO 2001).
sample thus captures 95 per cent of 22. If Metro Rail were extended to key regional
housing units in the study area. worksites and activity centres—including
16. With only one light rail route in the system, Niagara Falls and the Buffalo-Niagara Inter-
all stations experience the same level of national Airport—the advantages of living
service and service variations by day of near a station would increase, and property
week or time of day are consistent through- price premiums would likely rise above
out the system. those calculated in this study. Light rail
17. The median house value is $59 300 (US extensions would also provide transit acces-
Census 2000) and the mean assessed prop- sibility for a wider labour pool. Federal
erty value of the 7537 parcels in the funds for rail extensions may be less avail-
sample is $70 380 (see Table 3). able than they were decades ago, but the
18. Although the East Side of Main Street NFTA identified several corridors for
(EAST) variable is statistically significant “high-quality transit” capital projects—
in the models, we find inconclusive evidence including bus rapid transit—in its “Strategic
in this study about the effect of race on prop- Transit Assessment” (NFTA 2001). Unfortu-
erty value, as the main focus of the research nately, in a region with budget crises
is physical characteristics of the built looming on nearly all fronts, a new capital
environment. transit project is not currently on the public
19. Negative externality effects of station proxi- agenda, and may not be given serious con-
mity—noise, safety and aesthetics—may sideration for the foreseeable future.
appear near Buffalo’s above-ground stations, Because of the region’s interjurisdictional
but we expect these to have no effect near “turf battles,” lack of tax-sharing agree-
below-ground stations, which are housed in ments, and increasing urban-suburban dis-
station buildings with few immediately adja- parities (Fiscal Policy Institute 2003), many
cent developed residential or commercial suburban municipalities disassociate them-
properties. Nor is increased vehicle and ped- selves from the economic health of the
estrian traffic in station areas assumed to be a central city. Among suburban workers in
detractor. Erie County, 27 per cent have jobs in the
20. In Buffalo, low-income households situated city of Buffalo (Hess, 2005) and therefore
near light rail stations are more likely than should have direct interest in the quality of
high-income households to use transit and public transit, even if they do not regularly
experience travel time saving from im- ride Metro Rail or work in the central
proved service. Higher-income households, business district.
however, are less likely to use transit and
therefore less likely than low-income house-
holds to experience travel time savings from
improved transit service, yet land cost is References
more affected by light rail proximity for
high-income properties. Where residents do ALMEIDA , T. M. (2004) Impact of proximity to
not experience a travel time improvement, light rail rapid transit on station-area property
property prices are not bid higher. The values in Buffalo, New York. MA Thesis, State
average commute time in the Buffalo metro- University of New York.
politan area is 21 minutes, one of the shortest AL -MOSAIND , M. A., DUEKER , K. J. and STRATH-
in the nation for medium-sized cities (Hess, MAN , J. G. (1993) Light rail transit stations and
2005; US Census 2000). The 6.4-mile property values: a hedonic price approach.
length of Metro Rail can itself be traversed Paper presented at the 72th Annual Meeting
end-to-end in 21 minutes—not including of the Transportation Research Board,
walking time and waiting time—suggesting Washington, DC, January.
that light rail offers little time savings advan- ALONSO , W. (1964) Location and Land Use:
tage over an auto trip for many travellers and Toward a General Theory of Land Rents.
for many trips. Cambridge, MA: Harvard University Press.
21. Of the 13 cities that have built light rail ANAS , A. (1982) Residential Location Markets and
systems since 1980, Buffalo is the only one Urban Transportation. New York: Academic
to have built no extensions. Baltimore’s Press.
light rail line has been extended three times ANAS , A. (1983) The effects of transportation on
since its original line opened in 1992 and the tax base and development of cities. Report

No. DOT/OST/P-30/85/005, US Department Report No. 74, Transit Cooperative Research

of Transportation, Washington, DC. Program, Washington, DC.
ANDERSON , A. C. (1970) The effect of rapid transit CAN , A. (1990) Measurement of neighborhood
on property values, Appraisal Journal, 38, pp. dynamics in urban house prices, Economic
59 –68. Geography, 66, pp. 254–272.
ARMSTRONG , R. J. (1994) Impacts of commuter CAPOZZA , D. and HELSLEY , R. (1989) The funda-
rail service as reflected in single-family residen- mentals of land prices and urban growth,
tial property values, Transportation Research Journal of Urban Economics, 26(3), pp. 295–
Record, 1466, pp. 88 –97. 306.
BAE , C.-H. C. (2002) Orenco Station, Portland, CERVANTES , N. and DOLAN , T. J. (2005) Squeez-
Oregon: a successful transit oriented develop- ing homeowners tighter; as municipalities
ment experiment?, Transportation Quarterly, crave revenue, revaluations are hitting residents
56(3), pp. 9–18. harder than businesses, Buffalo News, 27 July,
BAJIC , V. (1985) Housing market segmentation p. A1.
and demand for housing attributes, Journal of CERVERO , R. (1984) Light rail transit and urban
the American Real Estate and Urban Economics development, Journal of the American Planning
Association, 13, pp. 58–75. Association, 50, pp. 133–147.
BANISTER , D. and BERECHMAN , J. (2000) Trans- CERVERO , R. (1994) Rail transit and joint develop-
port Investment and Economic Development. ment: land markets impacts in Washington, DC
London: University College London Press. and Atlanta, Journal of the American Planning
BARKER , W. G. (1998) Bus service and real estate Association, 60(1), pp. 83 –94.
values. Paper presented at the 68th Annual CERVERO , R. and DUNCAN , M. (2002a) Benefits of
Meeting of the Institute of Transportation Engin- proximity to rail on housing markets: experi-
eers. Toronto, August. ences in Santa Clara County, Journal of Public
BENJAMIN , J. D. and SIRMINS , G. S. (1996) Mass Transportation, 5(1), pp. 1–18.
transportation, apartment rent and property CERVERO , R. and DUNCAN , M. (2002b) Transit’s
values, Journal of Real Estate Research, 12(1), value-added effects: light and commuter rail ser-
pp. 1–8. vices and commercial land value, Transpor-
BERECHMAN , J. and PAASWELL , R. (1983) Rail tation Research Record, 1805, pp. 8–15.
rapid transit investment and CBD revitalisation: CERVERO , R. and LANDIS , J. (1995) BART at 20:
methodology and results, Urban Studies, 20(4), land use impacts. Paper presented at the 74th
pp. 471–486. Annual Meeting of the Transportation Research
BOLLINGER , C. and IHLANFELDT , K. (1997) The Board, Washington, DC, January.
impact of rapid rail transit on economic develop- CERVERO , R., FERRELL , C. and MURPHY , S.
ment: the case of Atlanta’s MARTA, Journal of (2002) Transit-oriented development and joint
Urban Economics, 42, pp. 179–204. development in the United States: a literature
BOWES , D. R. and IHLANFELDT , K. R. (2001) Iden- review. Research Results Digest No. 52,
tifying the impacts of rail transit station on prop- Transit Cooperative Research Program,
erty values, Journal of Urban Economics, 50, Washington, DC.
pp. 1–25. CHEN , H., RUFOLO , A. and DUEKER , K. (1998)
BOYCE , D., ALLEN , W. B., MUDGE , R. ET AL . Measuring the impact of light rail systems on
(1972) Impact of rapid transit on suburban resi- single-family home prices: a hedonic approach
dential property values and land development: with GIS applications. Paper presented at the
analysis of the Philadelphia– Lindenwold high- 77th Annual Meeting of the Transportation
speed line. Final Report to the US Department Research Board, Washington, DC, January.
of Transportation, Department of Regional CITY OF BUFFALO (1993) Zoning: charter and
Science, University of Pennsylvania. ordinances of the City of Buffalo. Charter and
BRIGHAM , E. F. (1965) The determinants of resi- Code of the City of Buffalo, Ch. 511. City of
dential land values, Land Economics, 41(4), Buffalo.
pp. 325–334. CLIFTON , K. and HANDY , S. (2001) Limits on
BUCKHAM , T. (2004) City gets unwanted rank on access in low-income neighborhoods and the
US list: hardship score called very high by insti- travel patterns of low-income households.
tute, Buffalo News, 5 September, p. B1. Report No. 167502-1, Center for Transportation
BUFFALO AND ERIE COUNTY PUBLIC LIBRARY Research, University of Texas at Austin.
(2005) Historical population estimates for the DAMM , D., LERMAN , S., LERNER -LAM , E. and
City of Buffalo and Erie County (http://ah.bfn. YOUNG , J. (1980) Responses of urban real
org/h/bflopop.html; accessed 1 June 2005). estate values in anticipation of the Washington
BURCHELL , R. W., LOWENSTEIN , G., DOLPHIN , metro, Journal of Transport Economics and
W. R. ET AL . (2002) The costs of sprawl 2000. Policy, 14, pp. 315– 355.

The impact of rail transport on real estate Bus rapid transit shows promise. GAO,
prices: empirical study of the Dutch housing Washington, DC.
market. Paper presented at the 85th Annual GALEA , S., AHREN , J., RUDENSTINE , S. ET AL .
Meeting of the Transportation Research Board, (2005) Urban built environment and depression:
Washington, DC, January. a multi-level analysis, Journal of Epidemiology
DIAZ , R. B. (1999) Impacts of rail transit on prop- and Community Health, 59, pp. 822–827.
erty values. Paper presented at the Commuter GARGAN , E. A (1984) Will a subway dig Buffalo
Rail/Rapid Transit Conference Sponsored by out?, The New York Times, 10 June, section 4,
the American Public Transportation Associ- p. 7.
ation. Toronto, Ontario May. GARRETT , T. (2004) Light rail transit in America:
DING , C., SIMONS , R. and BAKU , E. (2000) The policy issues and prospects for economic devel-
effect of residential investment on nearby property opment. Federal Reserve Bank of St Louis.
values: evidence from Cleveland, Ohio, Journal of GATZLAFF , D. H. and SMITH , M. (1993) The
Real Estate Research, 19(1), pp. 23–48. impact of the Miami Metrorail on the value of
DOWNS , A. (2004) Still Stuck in Traffic: Coping residences near station locations, Land Econ-
with Peak-hour Traffic Congestion. Washington, omics, 69(1), pp. 54 –66.
DC: Brookings Institution Press. GIULIANO , G. (2004) Land use impacts of transpor-
DUEKER , K. J. and BIANCO , M. J. (1999) Light rail tation investments: highways and transit, in: S.
transit impacts in Portland: the first ten years. HANSON and G. GIULIANO (Eds) The Geography
Paper presented at the 78th Annual Meeting of Urban Transportation, pp. 237–273.
of the Transportation Research Board, New York: Guilford Press.
Washington, DC, January. GLYNN , M. (2005) Area’s affordable homes at low
DUNCOMBE , W. (2002) Estimating the cost of an risk of depreciation, Buffalo News, 13 March,
adequate education in New York. Working p. B7.
Paper No. 44, Center for Policy Research, GOODWIN , R. E. and LEWIS , C. A. (1997) Land
Maxwell School of Citizenship and Public value assessment near bus transit facilities: a
Affairs, Syracuse, NY. case study of selected transit centers in
DUNPHY , R. T. and THE DEPARTMENT OF COMMU- Houston, Texas. Southwest Region University
NITY AND ECONOMIC RESOURCES (1982) Trends Transportation Center, Texas Southern
before Metrorail: a Metrorail before-and-after University.
study report. Metropolitan Washington Council GRASS , R. G. (1992) The estimation of residential
of Governments, Washington, DC. property values around transit station sites in
FAIRBANKS , P. (2004) Key study says sprawl has Washington, DC, Journal of Economics and
hurt city, Buffalo News, 30 May, p. A1. Finance, 16, pp. 139–146.
FAUSOLD , C. J. and LILIEHOLM , R. J. (1999) The GRETHER , D. M. and MIEZKOWSKI , P. (1974)
economic value of open space: a review and syn- Determinants of real estate values, Journal of
thesis, Environmental Management, 23(3), pp. Urban Economics, 1, pp. 127–146.
307–320. HESS , D. B. (2005) Access to employment for
FERGUSON , B. G., GOLDBERG , M. A. and MARK , J. adults in poverty in the Buffalo –Niagara
(1988) The pre-service impacts of the Vancou- region, Urban Studies, 42(7), pp. 1177–1200.
ver advanced light rail transit system on single HESS , D. B. and LOMBARDI , P. A. (2004) Policy
family property values, in: J. M. CLAPP and S. support for and barriers to inner-city transit-
D. MESSNER (Eds) Real Estate Market Analysis: oriented development: a literature review,
Methods and Applications, pp. 78 –110. Transportation Research Record: Journal of
New York: Praeger. the Transportation Research Board, 1887, pp.
FISCAL POLICY INSTITUTE (2003) The state of 26– 33.
working New York 2003: unbalanced regional HESS , D. B. and LOMBARDI , P. A. (2005) Govern-
economies through expansion and recession. mental subsidies for public transit: history,
Fiscal Policy Institute, New York. current issues, and recent evidence, Public
FOLLAIN , J. R. and JIMENEZ , E. (1985) Estimating Works Management and Policy, 10(2), pp.
the demand for housing characteristics: a survey 138–156.
and critique, Regional Science and Urban Econ- HICKLING LEWIS BROD , INC . (2002) Commercial
omics, 15. pp. 77 –107. property benefits of transit: final report.
FREEMAN , A. M. III (1979) The hedonic price Federal Transit Administration, Washington,
approach to measuring demand for neighbor- DC.
hood characteristics, in: D. SEGAL (Ed.) The HUANG , W. (1994) The effects of transportation
Economics of Neighborhood, pp. 193–197. infrastructure on nearby property values: a
New York: Academic Press. review of the literature. Research Report No.

620, Institute of Urban and Regional Studies, LOMBARDI , P. A. (2006) A destination for Buffa-
University of California, Berkeley. lo’s train to nowhere, Buffalo News, 8 January,
IHLANFELDT , K. and BOEHM , T. P. (1987) Govern- pp. H1–H2.
ment intervention in the housing market: an empiri- MALPEZZI , S. (1996) Housing prices, externalities,
cal test of the externalities rationale, Journal of and regulation in US metropolitan areas, Journal
Urban Economics, 22(3), pp. 276–290. of Housing Research, 7(2), pp. 209–242.
KAIN , J. (1988) Choosing the wrong technology: or MANNING , C. (1988) Explaining intercity home
how to spend billions and reduce transit use, price differences, Journal of Real Estate and
Journal of Advanced Transportation, 21, pp. Finance and Economics, 2(2), pp. 131–149.
197–213. McDONALD , J. F. and OSUJI , C. I. (1995) The
KAIN , J. F. and QUIGLEY , J. M. (1970a). Evaluating effect of anticipated transportation improvement
the quality of the residential environment, on residential land values, Regional Science and
Environment and Planning A, 2(1), pp. 23– 32. Urban Economics, 25(3), pp. 261–278.
KAIN , J. F. and QUIGLEY , J. M. (1970b). Measuring MILLS , E. S. (1972) Studies in the Structure of the
the value of housing quality, Journal of the Urban Economy. Baltimore, MD: John Hopkins
American Statistical Association, 65(330), pp. University Press.
532–538. MOHRING , H. (1961) Land values and the measure-
KNAAP , G. (1998) The determinants of residential ment of highway benefits, Journal of Political
property values: implications for metropolitan Economy, 69, pp. 236–249.
planning, Journal of Planning Literature, MUTH , R. F. (1969) Cities and Housing: The
12(3), pp. 267–282. Spatial Patterns of Urban Residential Land
KNAAP , G., HOPKINS , L. and PANT , A. (1996) Does Use. Chicago, IL: University of Chicago Press.
transportation planning matter? Explorations MUTH , R. F. and GOODMAN , A. C. (1989) The
into the effects of planned transportation infra- Economics of Housing Markets. London:
structure on real estate sales, land values, build- Harwood Academic Publishers.
ing permits, and development sequence. NELSON , A. C. (1992) Effects of elevated heavy-
Working Paper, Lincoln Institute of Land rail transit stations on house prices with respect
Policy, Cambridge, MA. to neighborhood income, Transportation
KNIGHT , R. L. (1980) The impact of rail transit on Research Record, 1359, pp. 127– 132.
land use: evidence and a change of perspective, NELSON , A. C. (1999) Transit stations and com-
Transportation, 9, pp. 3–16. mercial property values: a case study with
KOUTSOPOULOS , K. C. (1977) The impact of mass land-use policy implications, Journal of Public
transit on residential property values, Annals of Transportation, 2(3), pp. 77 –93.
the Association of American Geographers, NEW YORK STATE OFFICE OF PLANNING COORDI-
67(4), pp. 564–576. NATION (1969) The Buffalo– Amherst Corridor.
LAKAMP , P. (1999) Transit in a slump: loss of Technical Report, New York State Office of
riders hurts metro bus, rail, Buffalo News, 21 Planning Coordination, Albany, NY.
LANDIS , J., CERVERO , R., GUHATHUKURTA , S. (2005) New York State County population
ET AL . (1995) Rail transit investments, real projections (;
estate values, and land use change: a compara- accessed 31 May 2005).
tive analysis of five California rail transit NFTA (NIAGARA FRONTIER TRANSPORTATION
systems. Research Report No. 48, Institute AUTHORITY ) (2001) Strategic transit assess-
of Urban and Regional Studies, University ment. Prepared for NFTA by Parsons Brinckerh-
of California, Berkeley. off, NFTA, Buffalo.
LEWIS -WORKMAN , S. and BROD , D. (1997) NFTA (2004) Conceptual transit-oriented devel-
Measuring the neighborhood benefits of tail opment plan for Allen-Medical Campus and
transit accessibility, Transportation Research Summer-Best station. Prepared for NFTA by
Record, 1576, pp. 147–153. Chan Krieger, NFTA, Buffalo.
LIGGETT , R., LOUKAITOU -SIDERIS , A. and ISEKI , NFTA (2005) Historical bus ridership data.
H. (2001) Bus stop –environment connection: NFTA, Buffalo.
do characteristics of the built environment corre- OFFICE OF STRATEGIC PLANNING (2004) Queen
late with bus stop crime?, Journal of the Trans- city in the 21st century: Buffalo’s comprehensive
portation Research Board, 1760, pp. 20–27. plan. Office of Strategic Planning, City of Buffalo.
LIGGETT , R., LOUKAITOU -SIDERIS , A. and ISEKI , PALMQUIST , R. (1984) Estimating the demand for
H. (2003) Journeys to crime: assessing the characteristics of housing, Review of Economics
effects of a light rail line on crime in the neigh- and Statistics, 66, pp. 394–404.
borhoods, Journal of Public Transportation, PARSONS BRINCKERHOFF (2001) The effects of rail
6(3), pp. 85–115. transit on property values: a summary of studies.

Project 21439S, Task 7, NEORail II, February, TARGA , F. and CLIFTON , K. J. (2005) Built
Cleveland, OH. environment and nonmotorized travel:
PENDALL , R. (2003) Sprawl without growth: the evidence from Baltimore City using the NHTS,
upstate paradox. Survey Series, Center on Journal of Transportation and Statistics, 8(3),
Urban and Metropolitan Policy, Brookings Insti- pp. 55–70.
PENDALL , R. and CHRISTOPHERSON , S. (2004) PROGRAM ) (1998) Economic impact analysis of
Losing ground: income and poverty in upstate transit investments: guidebook for practitioners.
New York, 1980 –2000. Survey Series, Metropo- Report No. 35, Transportation Research Board,
litan Policy Program, Brookings Institution, Washington, DC.
PICKRELL , D. H. (1992) A desire named streetcar: TRATION ) (1989) Urban rail transit projects:
fantasy and fact in rail transit planning, Journal forecast versus actual ridership and costs. US
of the American Planning Association, 58(2), Department of Transportation, Washington, DC.
pp. 158–176. UNTERMANN , R. K. (1984) Accommodating the
POTEPAN , M. J. (1996) Explaining intrametropoli- Pedestrian: Adapting Towns and Neighbor-
tan variation in housing prices, rents, and land hoods for Walking and Bicycling. New York:
prices, American Real Estate and Urban Econ- Van Nostrand Reinhold Company.
omics Association Journal, 24(2), pp. 219–245. VESSALI , K. V. (1996) Land use impacts of rapid
QUIGLEY , J. M. (1979) What have we learned transit: a review of the empirical literature,
about urban housing markets?, in: P. MIESZ- Berkeley Planning Journal, 11, pp. 71– 105.
KOWSKI and M. STRASZHEIM (Eds) Current VOITH , R. (1993) Changing capitalization of CBD-
Issues in Urban Economics, pp. 391–249. Balti- oriented transportation system: evidence from
more, MD: Johns Hopkins University Press. Philadelphia, 1970 –1988, Journal of Urban
QUIGLEY , J. M. (1985) Consumer choice of dwell- Economics, 33, pp. 361– 376.
ing, neighbourhood, and public services, Regional VOORHEES AND ASSOCIATES (1976) Metro for
Science and Urban Economics, 15, pp. 41–63. Buffalo: transit alternatives for the Buffalo –
RICE CENTER FOR URBAN MOBILITY RESEARCH Amherst corridor. Technical Report, Niagara
(1987) Assessment of changes in property Frontier Transportation Authority, Buffalo.
values in transit areas. Preparedy for the WALLACE MC HARG ROBERTS & TODD (1971) The
Urban Mass Transit Administration, Rice Uni- regional center: a comprehensive plan for
versity, Houston. downtown Buffalo, New York. City of Buffalo.
RICHMOND , J. (2005) Transport of Delight: The WEINBERGER , R. (2000) Commercial property
Mythical Conception of Rail Transit in Los values and proximity to light rail: calculating
Angeles. Akron, OH: University of Akron Press. benefits with a hedonic price model. Paper pre-
ROSEN , S. (1974) Hedonic pricing and implicit sented at the 79th Annual Meeting of the Trans-
markets: product differentiation in pure compe- portation Research Board, Washington, DC,
tition, Journal of Political Economy, 82(1), pp. January.
34– 55. WEINBERGER , R. (2001) Light rail proximity:
RYAN , S. (1999) Property values and transportation benefit or detriment? The case of Santa Clara
facilities: finding the transportation –land use County, California. Paper presented at the 80th
connection, Journal of Planning Literature, Annual Meeting of the Transportation Research
13(4), pp. 412–427. Board, Washington, DC, January.
RYAN , S. (2005) The value of access to highway WEINSTEIN , B. L. and CLOWER , T. L. (2002) An
and light rail transit: evidence for industrial assessment of the DART light rail transit on
and office firms, Urban Studies, 42(4), pp. taxable property valuations and transit-oriented
751–765. development. Center for Economic Develop-
SCHROEDER , T. D. (1982) The relationship of local ment and Research, University of North Texas,
public parks and recreational services to residen- Denton.
tial property values, Journal of Leisure WEISBROD , G., BEN -AKIV , M. and LERMAN , S.
Research, 14(3), pp. 223–234. (1980) Tradeoffs in residential location
SKABURSKIS , A. (1982) Externalities and property decisions: transportation versus other factors,
prices: a test of the Schall hypothesis, Journal of Transportation Policy and Decision-making,
Regional Science, 22(2), pp. 213–223. 1(1), pp. 1 –14.
SMITH , J. J. and GIHRING , T. A. (2004) Financing YINGER , J. (1979) Estimating the relationship
transit systems through value capture: an anno- between location and the price of housing,
tated bibliography. Working paper, Victoria Journal of Regional Science, 19(3), pp. 271–
Transport Policy Institute Victoria, BC, Canada. 290.