How Does Light Rail Transit (LRT) Impact Gentrification? Evidence from Fourteen US Urbanized Areas

Abstract

This study examines the relationship between light rail transit (LRT) stations and changes in neighborhood characteristics associated with gentrification using spatial regression analyses with longitudinal data across 14 US urbanized areas (UAs). Overall, we do not find evidence of prevalent gentrification in LRT station areas. An analysis of UA-specific impacts shows heterogeneous outcomes across different UAs, particularly: strong transit-oriented development (TOD) effects accompanied by gentrification in San Francisco and TOD with countergentrification in Portland. Our results highlight that different local and regional planning efforts can lead to different types of changes in transit station neighborhoods.

Keywords

gentrification neighborhood change light rail transit (LRT)transit-oriented development (TOD)

Introduction

Light rail transit (LRT) has seen a surge in popularity in recent decades, largely as a way to promote transit use and to reshape urban areas toward a less auto-dependent form. Besides the environmental benefits from reduced driving, LRT is also expected to stimulate economic activities in station areas. Additionally, LRT station area developments can positively impact the lives of low- and moderate-income residents by improving transit access to regional amenities and work opportunities, providing new jobs in station neighborhoods, and encouraging development in neglected areas (Soursourian 2010). Overall, LRT station areas, often in the form of transit-oriented developments (TODs), serve as desired destinations and regional access points.

However, partly because of these expected benefits, an unfortunate outcome may arise: gentrification (Dawkins and Moeckel 2016; Revington 2015). Researchers note that enhanced locational desirability makes station areas more coveted places to live and, hence, the desirability is likely to increase property values and rents in LRT station neighborhoods (Hess and Almeida 2007). Developers also often target higher-income residents around LRT station areas by developing condominiums and high-rise, upscale housing (Bernick and Cervero 1997; Duncan 2011). As a result, the residents with the most need for and who can benefit the most from improved transit access, namely, low- and moderate-income residents and minorities, may not be able to access or live in the areas.

Nevertheless, the role of LRT investments in triggering gentrification is largely understudied, and existing studies heavily focus on property value changes, but not people who live in LRT neighborhoods. Directly highlighting the impacts of LRT stations on socioeconomic changes in LRT neighborhoods can provide a more substantial understanding of gentrification and overall neighborhood change. It, in turn, can provide direct evidence to inform policy addressing potentially harmful changes (i.e., displacement) stemming from transit.

We thus ask: To what extent is the presence of a light rail station associated with gentrification? We focus on examining socioeconomic characteristics of residents rather than property values in measuring gentrification. To answer this question, we utilize spatial regression analyses to investigate the connection between LRT stations and neighborhood change longitudinally across fourteen US urbanized areas (UA) that built light rail systems in the 1980s and 1990s. Understanding whether or not LRT stations are associated with gentrification will help planners better plan for LRT station areas by identifying who may be impacted by station developments rather than what may be impacted.

Literature Review

We define gentrification as a process of neighborhood change characterized by neighborhood upgrading coupled with residential displacement. We primarily view neighborhood upgrading as changes in the built environment, such as increasing neighborhood housing values (captured in rents and mortgage values), population densities, and fixed capital investments (Clark 1992; Teernstra and Van Gent 2012). Unfortunately, neighborhood upgrading often accompanies the displacement of existing residents. We also identify fixed capital investments related to transit as another potential displacement factor.

In reviewing the literature related to transit and gentrification, we focus on how previous scholars (1) define gentrification, (2) measure gentrification, and (3) account for transit as a cause of gentrification. Doing so informs our definition of gentrification. We also identify the research gap in transit-related gentrification research where scholars primarily focus on the built environment, neglecting socioeconomic changes in transit neighborhoods.

Defining Gentrification

We view displacement as an inherent part of gentrification and not an outcome, separate from the process. Many studies also, while often not explicitly noting displacement, do recognize some form of neighborhood turnover as part of gentrification. Sullivan (2007) defines gentrification as wealthier residents moving into a neighborhood previously occupied by poorer residents—changing the neighborhood’s class composition in the process. Similarly, Slater (2009) emphasizes Peter Marcuse’s (1985) earlier analysis recognizing the inherent detrimental displacement aspect of gentrification:

Gentrification occurs when new residents—who disproportionately are young, white, professional, technical, and managerial workers with higher education and income levels—replace older residents—who disproportionately are low-income, working-class and poor, minority and ethnic group members, and elderly—from older and previously deteriorated inner-city housing in a spatially concentrated manner, that is, to a degree differing substantially from the general level of change in the community or region as a whole. (Marcuse 1985, 198–99)

Marcuse’s (1985) definition covers a few key issues. First, it identifies both the gentrifiers and the gentrified. More importantly, it helps inform needed indicators to measure changes in the two different groups (e.g., by income levels, race, and poverty status). Second, it offers a spatial component to gentrification. It compares change in particular neighborhoods with the “community or region as a whole” (Marcuse 1985, 199). His spatial perspective of gentrification provides the conceptual basis for comparing change in LRT station areas with non-LRT station areas in this study.

The abundance of gentrification studies, meanwhile, take different stances concerning gentrification. Some existing studies discuss whether or not gentrification can produce positive benefits such as social mixing to increase social capital and cohesion (Davidson 2008; Uitermark, Duyvendak, and Kleinhans 2007). Some scholars argue that existing residents are not entirely powerless in the process (Brown-Saracino 2009; Cooley 2010), and there is little evidence of gentrification producing harmful effects such as displacement (Kohn 2013). A recent national-scale empirical study also indicates that gentrification is not necessarily a widespread phenomenon, and neighborhood decline is much more apparent (Landis 2016). Still others counter those views by indicating that displacement is an inherent effect of gentrification processes (Grube-Cavers and Patterson 2014; Smith 2010) and suggest that even if little evidence points to displacement, mitigation measures still should be in place (Freeman 2005).

We view gentrification without displacement as neighborhood or “incumbent” upgrading (Clay 1979), with no change in the social makeup (i.e., the actual residents) of the neighborhood. Incumbent upgrading involves existing residents improving the quality of their neighborhood (Clay 1979) or changing their occupational status from working-class to professional (Atkinson 2000). Our definition here, similar to Marcuse’s (1985), Sullivan’s (2007), and Slater’s (2009), views gentrification as consisting of two neighborhood change components: upgrading and displacement. While researchers may have different approaches to measuring or analyzing upgrading and displacement, not accounting for both does not fully account for gentrification.

Gentrification Cause

Gentrification literature examines both its supply-side and demand-side causes surrounding the production and consumption of neighborhood conditions. The rent gap theory helps explain supply-side activities. Here, gentrification often occurs when land is transformed to meet its potential value. The rent gap is the difference between the actual land value of a land parcel and the parcel’s potential value given that the land parcel had a better or different use (Smith 1987). For Smith (2010), “gentrification is a structural product of the land and housing markets. Capital flows where the rate of return is highest, and the movement of capital to the suburbs along with the continual depreciation of inner-city capital, eventually produces the rent-gap” (94).

Demand-side explanations largely examine who gentrifies—that is, the gentrifiers. Some scholars point toward white, middle-class, urban professionals as those residents moving into gentrifying neighborhoods (Brown-Saracino 2004; Kahn 2007; Stilwell 1993). Other scholars note gentrifiers’ racial diversity (Rose 1984; Taylor 2002). Why gentrifiers gentrify also remains widespread, ranging from social preservation (Brown-Saracino 2009) to establishing artists’ living space (Zukin 2013) to an appeal for cultural diversity and inner city living (Ley 1996). Such differing explanations indicate researcher’s lack of a consensus (Ley 1986). Rather, gentrification unfolds in multiple ways with differing contexts, producers, and consumers shaping whether, how, and where gentrification occurs.

Public investments can also cause gentrification. Researchers have identified that brownfield redevelopments (Bryson 2012; Dillon 2014; Essoka 2010) and charter schools (Davis and Oakley 2013; Hankins 2007) as well as housing programs (Bridge, Butler, and Lees 2012) can potentially catalyze gentrification. Public transportation capital improvements can certainly be added to the list. With transit improvements, such as the opening of a light rail station, undervalued areas potentially gain value—ultimately closing the rent gap where land prices increase to meet their potential value and thereby inducing gentrification. This potential value gain largely stems from the increased accessibility that transit provides, as well as an increase in land use intensity (i.e., increased densities) (Revington 2015). This accessibility increase is then capitalized into land and housing values, potentially causing low-income residents’ displacement from these station areas (Dawkins and Moeckel 2016).

Yet, as Zuk et al. (2015) indicate, few studies have addressed transit’s role in gentrification or how transit investments spur neighborhood decline. A small contingent of transit-related literature explores gentrification (Feinstein and Allen 2011; Kahn 2007; Lin 2002). Lin (2002) specifically examines gentrification and transit in Chicago by analyzing changes in residential property values and concludes that transit access spurs gentrification. However, without accounting for who occupies the housing, it is difficult to weigh whether gentrification or neighborhood upgrading occurs.

Kahn (2007) enhances transit studies by measuring gentrification impacts of rail-based transit between 1970 and 2000 in fourteen US cities. Kahn’s (2007) evidence of gentrification is based on home price dynamics and shares of communities that are college graduates. Kahn found mixed results throughout the cities examined; with some cities experiencing gentrification in communities with increased access to walk and ride stations. However, examining gentrification only in terms of college graduate percentages may be insufficient in understanding gentrification as a whole or who actually occupies the spaces.

Measuring Gentrification

Researchers undertake various techniques to measure gentrification and its inherent displacement. In his study on gentrification-induced displacement in Greater London, Atkinson (2000) used occupational change greater than London’s mean rate as an indicator of gentrification. He additionally used seven different displacement variables: working class, unskilled labor, households privately renting, ethnicity, unemployed, elderly, and lone parent. Atkinson (2000) concluded that gentrification was an active process during the 1980s in Greater London and, where gentrification occurred, he found above-average losses of the displaced groups (displacement variables). An extensive literature review by Zuk et al. (2015) also offers a list of different types of indicators researchers have commonly used to measure gentrification and displacement. Some of these include change in property values and rents, neighborhood investments and/or disinvestments, and demographic changes (Zuk et al. 2015, 37).

Meanwhile, studies of LRT and TOD gentrification primarily use land and/or property value changes as gentrification indicators (Cao and Schoner 2014; Cervero and Landis 1997; Duncan 2010). Areas that are prone to higher property values through increased transit station access will thereby receive greater attention as planners are especially interested in guiding growth and development in these station areas (Hess and Almeida 2007). For instance, in studying the effects light rail planning had on vacant residential property in Portland, Oregon, Knaap, Ding, and Hopkins (2001) found “that plans for light rail investments have positive effects on land values in proposed station areas” (32). Their results indicate that planning for light rail, in this instance, causes housing prices to rise and potentially prices out low-income residents. Similarly, Golub, Guhathakurta, and Sollapuram (2012) found that proximity to Phoenix’s LRT stations in general also positively impacts housing values, whereas positive effects largely begin well before actual LRT operations and accrue throughout the entire implementation process. Meanwhile, multiple studies point to LRT not having much, if any, direct impact on increasing property values (Hess and Almeida 2007; Chatman, Tulach, and Kim 2012).

Overall, the role of LRT investments in triggering gentrification is largely understudied and even just a few existing studies heavily rely on property values as a neighborhood change indicator. Most studies fail to account for what happens to station area residents and neighborhoods’ socioeconomic changes (Golub, Guhathakurta, and Sollapuram 2012; Duncan 2011). Only a few studies include race or income as a secondary variable in examining housing value changes (Debrezion, Pels, and Rietveld 2010; Chatman, Tulach, and Kim 2012). While property value can be a good proxy of accessibility improvement and neighborhood upgrading, exclusive focus on the price variable cannot address displacement, another essential aspect of gentrification. This study thus aims to enhance LRT-related gentrification research by focusing on providing sufficient indicators to account for socioeconomic changes in neighborhoods.

Research Strategy

Our methods chosen here reflect how we define gentrification and the gaps we identify in the current LRT-related gentrification research. Since we view the displacement of older residents as an inherent component of gentrification, our analysis tunes in to sociodemographic changes in the neighborhood unlike previous studies mainly focusing on property values.

To understand how LRT stations impact gentrification and TOD-related neighborhood changes, we conduct a series of spatial autoregressive lag models (also referred to here as SAR or spatial regressions).¹ We use SAR and LRT station areas to account for the spatial component of gentrification. Also, we particularly aim to highlight the relationship between LRT stations and changes in residential characteristics—an underemphasized component of quantitative gentrification studies. In emphasizing the residential characteristics, we do two things. First, we introduce a Neighborhood Change Index (NCI) that combines different neighborhood characteristics focusing on residential socioeconomic characteristics into one indicator. Second, we showcase individual, resident-focused displacement indicators (e.g., race and education, among others) rather than housing characteristics. We detail these indicator variables and the SAR model in the sections that follow.

We use SAR models to regress the changes in residents’ socioeconomic characteristics, including race, income, and education as well as a composite neighborhood change index (NCI) on various census tract–level social and physical attributes, including the presence of an LRT station. We also examine the relationship between LRT stations and TOD-related changes such as increased densities and transit ridership. We expect that the presence of an LRT station is associated with the possible presence of gentrification along with increased public transit use and population density.

Study Areas

Table 1 lists the UAs and their corresponding light rail lines that we examine. We use UAs to delineate the study areas as they comprise both the core city and surrounding municipalities. Such delineation is important as LRT systems generally extend throughout various cities and municipalities. We only select UAs with LRT systems that started light rail operations by or before 2000. Year 2000 provides at least a ten-year period to observe neighborhood changes. We define our study area UAs using 2010 UA boundaries.²

Table 1.

Urbanized Areas with LRT Built before 2000.

Urbanized Area	Year of First LRT Station Opening	LRT Line(s)	System Name
Cleveland	1980	Blue, Green, Waterfront	The Rapid (RTD Rapid Transit)
San Francisco	1980	F-Market & Wharves, J-Church, K-Ingleside, L-Taraval, M-Ocean View, N-Judah, T-Third Street	Muni Metro
San Diego	1981	Blue, Orange	San Diego Trolley
Pittsburgh	1984	Blue-Library, Blue-South Hills Village, Red	The T
Buffalo	1985	Main	Buffalo Metro Rail
Portland	1986	Blue, Green, Red	MAX Light Rail
Sacramento	1987	Blue, Gold, Green	RT Light Rail
San Jose	1987	Alum Rock-Santa Teresa (Blue), Mountain View-Winchester (Green), Ohlone/Chynoweth-Almaden (Orange)	VTA Light Rail
Los Angeles	1990	Blue, Green	Metro Rail
Baltimore	1992	Blue, Yellow, Red	Baltimore Light Rail
St. Louis	1993	Red	MetroLink
Denver	1994	C-Orange, D-Green, E-Purple, F-Red, H-Blue	RTD Light Rail
Dallas	1996	Red, Blue, Green	DART Light Rail
Salt Lake City	1999	Blue	TRAX

We use the census tract as the unit of analysis. The census tract is a bit larger than what is typically considered a neighborhood. However, we believe it approximates an optimal size to analyze socioeconomic changes of residents over time.

Model Specification

To investigate the relationship between the presence of a light rail station and potential gentrification and TOD effects, we use SAR models. We use SAR to capture the spatial dependence that potentially exists among the census tract characteristics. It is widely recognized today that many socioeconomic variables such as income and housing prices at nearby locations are correlated. Using the traditional ordinary least squares (OLS) regression without correcting the spatial autocorrelation problem would result in biased and inefficient estimates (Anselin 2002).³ SAR tests and accounts for the possibility that the dependent variable in a given census tract is correlated to values of the same variable in adjacent census tracts (Anselin 2002; de Smith 2015).

The general form for our SAR model is as follows:

y = α + pWy + β X + ε

where y is our dependent variable (the change in either NCI, Race, Education, Income, or Poverty for our gentrification models and either Population Density, Public Transit use, or all Non–Privately Owned Vehicle usage for our TOD models), α is the constant term parameter, p is the spatial autoregression parameter which is estimated from the data, W is our weighting matrix (Queen weight matrix), X represents our explanatory variables, and ε is the error term. In this SAR specification, the significance test of p tests whether or not the dependent variable is spatially auto-correlated. As will be further described in the next section, independent variables include a light rail transit station dummy, location characteristics of each tract, beginning year demographic characteristics, beginning year housing characteristics, and central city and UA dummy variables.

We conduct two sets of SARs for two pooled data sets where each of pooled data sets represent thirty- and twenty-year change. Our aim here is to examine how station area neighborhoods change over the lifetime of their systems. However, we also ran models examining ten-year change within each set, but we found overall insignificant results—possibly indicating that station area changes, especially unwanted changes, do not accrue until at least a decade after station openings. As such, the immediate decade following station openings could be crucial for planners to enact policies and plans to combat unwanted changes in the years that follow.

Each set of SARs contains eight SAR models using the change in NCI, Race, Education, Income, and Poverty as well as Population Density, Public Transit use, and Non-POV (Non–Privately Operated Vehicle) mode use as the dependent variable. While SAR set 1 uses all UA census tracts in the sample, set 2 only uses tracts we characterize as gentrifiable. The process of neighborhood upgrading and displacement is the most detrimental in low-income and/or minority neighborhoods with initially low housing values and where tenants do not own or have control of the land (Lin 2002; Smith 2010). As such, we try to identify the impacts of light rail stations especially in what we refer to as “gentrifiable” neighborhoods: neighborhoods with initially low levels of socioeconomic characteristics.⁴

Variables

Table 2 lists all variables used in the analysis. We operationalize gentrification as a process of change characterized by neighborhood upgrading coupled with residential displacement. As a process, we measure the change in socioeconomic characteristics in census tracts from the beginning decade year when the first LRT station opened to 2010. For example, considering that Pittsburgh’s light rail stations first opened in 1984, we identify change as occurring between 1980 and 2010, or thirty years. While all UAs have differing opening dates for their LRT systems and stations, the change measured in this way can be pooled into two groups: thirty-year change from 1980 to 2010; and twenty-year change from 1990 to 2010. Doing so provides us with our two pooled data sets.

Table 2.

Description of Variables.

Variable	Description	Data Source	Expected Sign^a
Dependent variables^b
NCI	Neighborhood change index derived from a principal component analysis of socioeconomic variables	LTDB	+
Race	Percent of total population who are white, non-Hispanic	LTDB	+
Education	Percent of population age 25 and over who have at least a 4-year college degree	LTDB	+
Income^c	Log median household income (MHI)	LTDB	+
Poverty^d	Percent poverty rate	LTDB	-
Population density	Log population density (per square mile)	LTDB; US Census TIGER/Line Files	+
Public transit commute	Percent of all commuters who use public transit	NHGIS	+
Non-POV commute	Percent of all commuters who use other modes than privately operated vehicles (POVs)	NHGIS	+
Independent variables
LRT station buffer	Dummy variable: 1 if a census tract’s centroid is inside a half-mile station buffer; otherwise 0	US Census TIGER/Line Files
UALRTs	Interaction terms: LRT station buffer × each of 14 UA dummy variables
Location characteristics
UA dummy	Dummy variables for all UAs in the sample
Core city	Dummy variable: 1 if a census tract is in the core central city; otherwise 0	US Census TIGER/Line Files
Distance to CBD	Log Euclidean distance in meters from census tract centroid to CBD	US Census TIGER/Line Files
Beginning year socioeconomic characteristics
Population density	Population density (percent) in the beginning year (BY)	US Census TIGER/Line Files; LTDB
Log median income	BY log MHI	LTDB
% non-Hispanic white	BY percent of population who are non-Hispanic white	LTDB
% bachelor and above	BY percent of population age 25+ with college degree and above	LTDB
% poverty rate	BY percent poverty rate	LTDB
% professional occupation	BY percent of employed age 16+ with a professional job	LTDB
Housing characteristics
% renter occupied	BY percent renter occupied housing units	LTDB
Log median rent	BY log median rent	LTDB
Log median home price	BY log median housing value	LTDB
Rent gap	BY predicted median rent – Actual median rent	LTDB
% housing 30 yrs & older	BY percent of total occupied housing units that are 30 years or older	LTDB

Note: CBD = central business district; LRT = light rail transit; LTDB = Brown University’s Longitudinal Tract Database; NCI = neighborhood change index; POV = privately owned vehicle; TOD = transit-oriented development; UA = urbanized area.

Expected sign is for the interaction terms of LRT station buffer and UA dummy variables. Thus, they indicate the direction of transit stations’ expected gentrification- and TOD-related impacts on the given dependent variable in each model.

All dependent variables are used as a change. For example, NCI is the difference between 2010 and the beginning year factor scores. The beginning year is defined as the first year of the decade in which the first light rail station was opened.

2010 constant dollars were calculated using the Consumer Price Index for all Urban Consumers (CPI-U) (United States Department of Labor’s Bureau of Labor Statistics 2015).

The LTDB defines the poverty rate as the number of persons in poverty divided by the number of persons for whom poverty status is determined.

Dependent Variables: Neighborhood Changes

We have eight different dependent variables for our SAR models that, when taken together, measure neighborhood change: a neighborhood change index (NCI), Race, Education, Income, Poverty, Population Density, Public Transportation Use, and Non-Privately Owned Vehicle (Non-POV) Use. With the first five variables, we aim to measure the extent of neighborhood change associated with gentrification. Unlike previous studies, we focus on analyzing socioeconomic status of residents before and after building LRT stations rather than property values. While it is difficult to fully account for displacement—especially why residents left a neighborhood—without tracking the moves of older residents, the changes in these socioeconomic variables—especially race and education—will serve as second best proxy indicators. For Race, we measure change in non-Hispanic white population percentage (hereby referred to as white) because gentrification often manifests itself as racial transition. Previous studies often refer to an increase in white population percentage as a possible indicator for gentrification (Kahn 2007; Smith 2010).

The latter three variables, while still tied to neighborhood upgrading, measure the extent of neighborhood change more so associated with Transit-Oriented Development (TOD) benefits. Promoting sustainable transportation modes by concentrating high-density and mixed use developments in main transit corridors are widely known goals of TOD and can indicate improved public transportation capital improvements related to neighborhood upgrading. We try to examine in what UAs the TOD effects are significant.

Building from previous studies measuring neighborhood change and gentrification (Ley 1986; Meligrana and Skaburskis 2005), we develop NCI as a composite variable derived from the socioeconomic status of residents (absent from Ley’s index) and housing characteristics (absent from Meligrana and Skaburskis’s index). We develop our NCI in two steps. First, we use principal components analysis (PCA) to derive census tract–level socioeconomic indices for both beginning year (1980 or 1990) and 2010. PCA is a commonly used data reduction method that derives a small number of summary variables as linear combinations (principal components) of a large number of original variables (Shaw and Wheeler 1994). We derive just one standardized principal component without factor rotation from housing and rent values, race, income, education, poverty, and professional occupation variables for each year and UA. All input variables except poverty rate are positively loaded to the index. In the second step, we develop our NCI by calculating the difference between 2010 and beginning year principal components. An increase in our NCI represents neighborhood upgrading potentially with displacement.

Independent Variables

The interaction terms of the LRT station buffer and UA dummy variables (i.e., UALRT) are the key variables that will reveal how the presence of LRT stations affects neighborhood socioeconomic changes. Our LRT station dummy variable indicates whether or not a census tract’s centroid is within a half-mile, straight-line radius of a light rail station. Our UA dummy variables indicate whether or not a census tract is within a given UA. While UA dummy variables will show UA-specific changes in the given dependent variable for nonstation neighborhoods, interaction terms of the LRT station dummy with the UA dummy variables are designed to capture UA-specific station effects on neighborhood change.⁵ We believe that the impacts of LRT stations on nearby neighborhoods should be different from one UA to another because the role of LRT and policy contexts are all different across UAs.

In addition, the right-hand side of our SAR models includes all variables we expect to affect neighborhood changes, as listed in Table 2. For each census tract’s location characteristics, we assume that its presence in the core, central city (CC) and its distance to the central business district (CBD) influence the likelihood of neighborhood upgrading or gentrification. We also include neighborhood beginning year socioeconomic characteristics such as Income, Race, Education, Poverty, Professional Employment, and Population Density. Finally, the characteristics of housing stock in the neighborhood are also very important determinants of neighborhood change. We include Rent Gap, Tenure Type, and Percent old housing stock (Housing Age) as well as median Rent and Home Values.

The Rent Gap variable aims to capture development opportunities in disinvested neighborhoods by measuring whether or not rent values for the census tract are undervalued—and can be more prone to gentrification—or overvalued (Landis 2016, 23). We follow Landis’s (2016) approach to operationalizing this variable where it is essentially the difference between actual median rents and predicted rents⁶ in the decade year prior to light rail station openings.

Data Sources

Most of the census tract–level socioeconomic variables derive from Brown University’s Longitudinal Tract Database (LTDB) (US 2010 Project 2014). The LTDB provides decennial year census data normalized to 2010 census tract boundaries.⁷ We also collect public transportation data from the National Historical Geographic Information System (NHGIS) database (Minnesota Population Center 2011) and use the LTDB crosswalk file to convert the data. To determine the urbanized areas and core cities at the census tract level, we accessed the US Census Bureau’s TIGER/Line Shapefiles for 2010 Urbanized Areas and its Shapefiles for Places. We retrieved the station point and line shapefiles from a variety of federal, state, metropolitan/regional, and local sources: US Department of Transportation (USDOT 2013), Utah Automated Geographic Reference Center (2013), Los Angeles County Metropolitan Transportation Authority (2016), Metropolitan Transportation Commission (2014), North Central Texas Council of Governments Transportation Department (n.d.), Regional Transportation District (n.d.), SANDAG (n.d.), Trimet (2014), Pittsburgh (2015), and Sacramento (2014).

We also utilize ArcGIS software to (1) create a Euclidean buffer radius of a half-mile around every light rail station, (2) create census tract centroids, and (3) determine the distance from each census tract centroid to its given CBD.⁸ We use the half-mile buffer because researchers have found that the half-mile catchment area works best in predicting ridership as a function of station area population, as well as its being a generally standard area for measuring LRT station impacts (Guerra, Cervero, and Tischler 2012). Mirroring Kahn (2007), we use the Geographic Reference Manual: 1982 Economic Censuses to determine the CBD boundaries (US Bureau of the Census 1983), converting them into 2010 census tract boundaries with the LTDB.

Set 1 (Full Sample) Results: LRT Impacts on Gentrification and TOD

We limit our discussion to the coefficients of interaction terms between the LRT station dummy and UA dummy variables that are summarized in Tables 3 and 4. Full model results are presented in Appendix Tables A3 and A4. The coefficients of interaction terms indicate to what extent (statistically) the presence of an LRT station influences gentrification and TOD-related change in station area neighborhoods for a given UA compared to nonstation area neighborhoods. For example, the interaction term between the LRT station and Denver (DenverLRT) was statistically significant in the Race (i.e., white population percentage) model, with an estimated coefficient of 0.0432. It indicates that Denver’s LRT station neighborhoods experienced a 4 percent relative increase in white population compared to nonstation tracts.

Table 3.

Summary Results for the Full Sample (Set 1 SARs).

	NCI		Race		Education		Income		Poverty		PD		PT		Non-POV
	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value
1980 data set
San Francisco	0.0911	0.0888*	0.0426	0.0002***	0.0147	0.1772	0.3074	0.002***	−0.0284	0.0005***	0.277	0***	0.0377	0.0002***	0.0693	0***
Cleveland	−0.0923	0.3166	−0.0865	0***	−0.0042	0.8243	−0.1199	0.4846	−0.0081	0.5646	−0.1082	0.3521	0.0589	0.0008***	0.0281	0.3128
San Diego	−0.012	0.8725	−0.0307	0.0575*	−0.013	0.3942	0.0037	0.9789	0.0228	0.0463**	−0.0461	0.6254	0.009	0.5266	−0.0098	0.6653
Pittsburgh	−0.1424	0.1177	−0.0016	0.9369	−0.016	0.3894	−0.4312	0.0109**	0.0037	0.7926	−0.0726	0.5272	−0.0345	0.0457**	−0.0644	0.019**
Buffalo	−0.0121	0.9233	−0.0506	0.0618*	0.0324	0.2053	−0.391	0.0943*	0.002	0.9166	0.0095	0.9522	−0.0282	0.238	−0.212	0***
Portland	−0.2766	0.0011***	−0.0295	0.1053	−0.0229	0.1836	−0.0728	0.6439	0.0397	0.0021***	0.1617	0.1298	0.0378	0.0187**	0.0083	0.745
Sacramento	−0.0528	0.5883	0.0355	0.0911*	0.0031	0.8754	−0.0688	0.7041	0.0028	0.8506	−0.3471	0.0048***	0.0246	0.1845	0.0223	0.4487
San Jose	−0.0213	0.8008	0.0248	0.1742	0	0.9987	0.1596	0.3101	−0.0128	0.3208	−0.0741	0.4874	0.0063	0.6954	0.0233	0.3618
Rho		0.558		0.7771		0.5882		0.1713		0.3766		0.4623		0.4249		0.3751
Log likelihood		−2676.45		3141.469		3506.339		−4978.198		4723.878		−3537.477		3857.855		2069.278
AIC (SAR)		5414.9		−6220.9		−6950.7		10018		−9385.8		7137		−7653.7		−4076.6
AIC (OLS)		6357.9		−3358.1		−5949.9		10123		−9080.5		7939.2		−7234.4		−3767.7
1990 data set
Los Angeles	−0.0502	0.2634	−0.0308	0.0057***	−0.0221	0.0357**	−0.1465	0.1724	0.0161	0.058*	−0.0329	0.6469	0.0027	0.6809	0.0089	0.364
Baltimore	0.0491	0.4487	0.0313	0.0516*	0.0163	0.2812	−0.2714	0.0799*	0.0156	0.2004	−0.0799	0.4403	−0.0061	0.526	−0.0175	0.219
St. Louis	0.0874	0.3792	−0.014	0.5706	0.0454	0.0512*	0.3075	0.1958	0.0354	0.059*	0.0368	0.817	−0.0322	0.029**	−0.0392	0.0727*
Denver	0.2554	0.0086***	0.0432	0.0732*	0.0254	0.2628	0.1342	0.5634	−0.0175	0.3408	−0.2111	0.1738	−0.0116	0.421	−0.0251	0.2383
Dallas	−0.0748	0.3619	0.0046	0.8198	0.0109	0.5711	0.0649	0.7405	−0.0285	0.0653*	−0.1111	0.3965	−0.0185	0.1272	−0.052	0.0039***
Salt Lake City	−0.117	0.2627	0.0217	0.4023	−0.0322	0.1878	0.2702	0.2791	0.0434	0.0276**	−0.2498	0.1344	−0.0124	0.4232	−0.0379	0.0984*
Rho		0.5388		0.7387		0.5064		0.1156		0.2495		0.4466		0.2961		0.2807
Log likelihood		−2595.398		5276.421		5827.39		−7477.225		7186.533		−5248.621		8563.747		6299.979
AIC (SAR)		5244.8		−10499		−11601		15008		−14319		10551		−17073		−12546
AIC (OLS)		6448.4		−7132.4		−10649		15039		−14140		11394		−16831		−12321

Note: This table only presents the coefficients of interaction terms between LRT station dummy and UA dummy variables. Full results including control variables are reported in Table A3. AIC = Akaike information criterion; LRT = light rail transit; OLS = ordinary least squares; PD = Population Density; Non-POV = Non–Privately Owned Vehicle; PT = Public Transit; SAR = spatial autoregressive lag model; UA = urbanized area.

*p < .10, **p < .05, ***p < .01.

Table 4.

Summary Results for the Gentrifiable Sample (Set 2 SARs).

	NCI		Race		Education		Income		Poverty		PD		PT		Non-POV
	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value	Coeff.	P Value
1980 data set
San Francisco	0.0612	0.4607	0.0814	0***	0.0285	0.0853*	0.4699	0.0056***	−0.034	0.0133**	0.3718	0.0005***	0.0379	0.0392**	0.069	0.0166**
Cleveland	0.0688	0.6548	−0.0217	0.5332	0.0113	0.7129	−0.2002	0.5237	−0.044	0.0846*	−0.2735	0.1659	0.0786	0.0209**	0.0611	0.2514
San Diego	−0.0603	0.5428	−0.0309	0.1676	−0.0209	0.2893	0.0491	0.8081	0.0406	0.0133**	−0.0242	0.8494	0.0099	0.6523	−0.0165	0.6303
Pittsburgh	0.1124	0.5921	−0.0132	0.7817	−0.0141	0.7351	−0.425	0.3211	−0.0161	0.6426	−0.2843	0.2917	−0.0391	0.399	0.0052	0.9428
Buffalo	0.1782	0.3582	−0.0633	0.1486	0.0854	0.027**	−0.5693	0.1502	−0.0433	0.177	−0.1305	0.5997	−0.0762	0.0755*	−0.3522	0***
Portland	−0.3714	0.0035***	−0.046	0.1097	−0.0279	0.2697	−0.0718	0.7818	0.0427	0.0424**	0.2247	0.1681	0.054	0.0549*	0.0004	0.9931
Sacramento	−0.1369	0.3426	0.049	0.1324	−0.015	0.6014	−0.1449	0.6222	0.0102	0.669	−0.485	0.0089***	0.0276	0.3858	0.0023	0.9627
San Jose	0.0399	0.7226	0.0316	0.2137	0.0181	0.4185	0.1661	0.4685	−0.0063	0.7354	−0.0424	0.7684	0.0044	0.8608	0.0154	0.6928
Rho		0.4452		0.4891		04562		0.1542		0.2182		0.3389		0.3427		0.2365
Log likelihood		−1391.752		1105.158		1326.075		−2552.985		1677.664		−1790.672		1173.108		430.03
AIC		2845.5		−2148.3		−2590.2		5168		−3293.3		3643.3		−2284.2		−798.06
1990 data set
Los Angeles	−0.0235	0.6675	−0.0056	0.6726	−0.0099	0.3725	0.1171	0.391	0.0257	0.0381**	0.1113	0.1418	−0.001	0.9288	0.0099	0.5372
Baltimore	0.0181	0.8542	0.0127	0.5984	0.0225	0.262	0.114	0.6439	0.0478	0.0324**	0.0208	0.8793	0.0107	0.6105	0.0159	0.5821
St. Louis	0.099	0.4814	0.0415	0.2271	0.0495	0.0833*	0.5898	0.0932*	0.0473	0.1376	0.1982	0.3094	−0.0788	0.0084***	−0.1043	0.0112**
Denver	0.394	0.0009***	0.108	0.0002***	0.0409	0.0891*	0.0864	0.7702	−0.018	0.5018	−0.1644	0.3164	−0.0179	0.4764	−0.0346	0.3164
Dallas	−0.0302	0.7963	0.0046	0.8198	0.0002	0.992	−0.2289	0.4331	−0.0001	0.9961	−0.1111	0.3965	−0.0185	0.1272	−0.1165	0.0007***
Salt Lake City	−0.0269	0.8376	0.0217	0.4023	−0.0396	0.1378	0.2431	0.4582	0.0542	0.0682*	−0.2498	0.1344	−0.0124	0.4232	−0.0379	0.0984*
Rho		0.3216		0.4433		0.3612		0.0965		0.1310		0.2893		0.1674		0.1840
Log likelihood		−944.5905		1726.129		2098.39		−2674.921		1918.968		−1565.392		2038.219		1426.629
AIC		1943.2		−3398.3		−4142.8		5403.8		−3783.9		3184.8		−4022.4		−2799.3

Note: This table only presents the coefficients of interaction terms between LRT station dummy and dummy variables. Full results including control variables are reported in Table A4. AIC = Akaike information criterion; LRT = light rail transit; OLS = ordinary least squares; PD = Population Density; Non-POV = Non–Privately Owned Vehicle; PT = Public Transit; SAR = spatial autoregressive lag model; UA = urbanized area.

p < .10, **p < .05, ***p < .01.

We also categorize our regression results into four different neighborhood change typologies in LRT station areas based on the significant coefficients: Gentrification and TOD (relative increases in both indicators), Countergentrification and TOD (relative decreases in gentrification indicators and relative increases in TOD indicators), Gentrification and Counter- or no TOD (relative increases in gentrification indicators and relative decreases in TOD indicators), and Neighborhood Decline (relative decreases in both indicators).

Table 3 summarizes results for full sample models (Set 1). It shows mixed results. Station areas in the UAs from the 1980–2010 (thirty-year change) data set, with the exception of San Francisco and Sacramento, exhibited varying countergentrification (or decline) related aspects of neighborhood change. Changes in Race were most pronounced throughout the UAs–with relative decreases in white populations occurring in Cleveland’s (–8 percent), San Diego’s (–3 percent), and Buffalo’s (–5 percent) station tracts. Results for San Francisco (4 percent) and Sacramento (2 percent) indicate relative increases.

A few places stood out in the 1980–2010 sample. San Francisco exhibited both strong gentrification and TOD-related changes. San Francisco’s census tracts with LRT stations performed significantly better than nonstation tracts in increasing Population Density and retaining commute mode shares by Public Transit and all Non-POV modes. However, these positive TOD effects were accompanied by gentrification. Station tracts exhibited substantial relative increases in Income (31 percent) and NCI (9 percent), with a moderate relative increase in white population, and a decrease in Poverty.

Portland, meanwhile, exhibited a relatively strong decrease in its NCI and a slight increase in Poverty. In fact, out of all UAs for both 1980 and 1990 sets, Portland’s LRT stations had the largest impacts on NCI change: –28 percent. This countergentrification change came with positive TOD effects shown in the Public Transit commute share model. While LRT stations also had positive impacts on Population Density, the coefficient was not statistically significant. These positive TOD effects imply that the countergentrification in Portland’s LRT station areas was not due to neighborhood decline but more so because more residents with the most need for transit were able to occupy light rail station areas.

None of the 1990 UAs experienced a positive TOD-related change in LRT station neighborhoods. Rather, St. Louis, Dallas, and Salt Lake City showed negative impacts on Public Transit and Non-POV mode commute uses. However, there are some noteworthy results. Denver’s LRT station coefficient was especially striking as it indicates a 26 percent relative increase in NCI when compared to nonstation tracts. Similarly, we also found a relative increase of 4 percent in white population for Denver. Such results combined with little TOD impact could indicate that Denver’s light rail station areas possibly experience gentrification related neighborhood change without gaining much on the sustainability front. Conversely, results for Los Angeles indicate neighborhood decline without TOD impact. Los Angeles’s white population and percentage of educated residents in station tracts relatively decreased by 3 percent and 2 percent, respectively, and the poverty rate relatively increased by 2 percent. Los Angeles’s results indicate a continued decline in the areas served by LRT lines.

Set 2 Results: Do Gentrifiable Tracts Experience More Change?

Table 4 summarizes results for Set 2 models with only gentrifiable tracts that are in the bottom two quintiles in each UA as measured by the NCI. The full results are listed in Appendix Tables A5 and A6. Results for Set 2 show similar patterns to Set 1’s results with only minor differences. Overall, Set 2’s significant coefficients were of a greater magnitude than Set 1, indicating that the gentrifiable tracts experienced greater changes in regard to gentrification and TOD. For example, results for Portland’s NCI indicated that station tracts experienced a decrease by 37 percent relative to nonstation tracts, compared to a 28 percent decrease in Set 1. Similarly, St. Louis shows larger coefficients for Public Transit use, Non-POV transit use, and income changes for Set 2 than Set 1.

The directions of changes in notable UAs are not much different from the whole sample results: strong TOD effects accompanied by gentrification in San Francisco, TOD effects with countergentrification in Portland, significant gentrification with little evidence of TOD in Denver, and further decline of station areas without TOD impact in Los Angeles. The only notable change in Set 2’s result is that gentrification-related change is more pronounced in St. Louis, with statistically significant coefficients in both education and income models.

Neighborhood Change Typology in LRT Station Areas

We further summarize our results to discover LRT station impact typologies using the scheme described in Table 5. We count the number of models where light rail station variable coefficients are statistically significant for each UA. Based on these counts, we describe the sets as possibly exhibiting strong to weak (counter-) gentrification and (counter-) TOD-related neighborhood changes. If all significant coefficients are consistently counter our expectations, we refer to these as countergentrification (or decline) and counter-TOD. Table 6 summarizes the type of gentrification or TOD-related change for each UA.

Table 5.

Gentrification and TOD Typology Scheme.

Gentrification or Countergentrification (-C)		TOD or Counter-TOD (-C)
Level of Change	No. of statistically significant coefficients	Level of Change	No. of statistically significant coefficients
Strong	4 to 5	Strong	3
Moderate	2 to 3	Moderate	2
Weak	1	Weak	1
None	0	None	0
Mixed	Significant coefficients for both gentrification and countergentrification present	Mixed	Significant coefficients for both TOD and counter-TOD present

Table 6.

Gentrification and TOD Impacts in Fourteen UAs.

	Set 1: Full Sample		Set 2: Gentrifiable Tracts
	Gentrification	TOD	Gentrification	TOD
San Francisco	Strong	Strong	Strong	Strong
Cleveland	Weak-C	Weak	Weak	Weak
San Diego	Moderate-C	None	Weak-C	None
Pittsburgh	Weak-C	Moderate-C	None	None
Buffalo	Moderate-C	Weak-C	Weak	Moderate-C
Portland	Moderate-C	Weak	Moderate-C	Weak
Sacramento	Weak	Weak-C	None	Weak-C
San Jose	None	None	None	None
Los Angeles	Moderate-C	None	Weak-C	None
Baltimore	Mixed	None	Weak-C	None
St. Louis	Mixed	Moderate-C	Moderate	Moderate-C
Denver	Moderate	None	Moderate	None
Dallas	Weak	Weak-C	None	Moderate-C
Salt Lake City	Weak-C	Weak-C	None	None

As noted above, we characterize four possible interpretations of our results:

Gentrification and TOD: Occurs with significant positive coefficients for both Gentrification and TOD indicators. We expect this result where LRT stations triggered new developments with increased transit access, but also priced-out low-income and minority households;

Countergentrification and TOD: Occurs with significant negative and positive coefficients for Gentrification and TOD indicators, respectively. We view this as the best possible outcome whereby LRT station areas have possibly attracted low-income and minority households who have more needs for transit;

Gentrification and Counter- or no TOD: Occurs with significant positive and negative coefficients for Gentrification and TOD indicators, respectively. We view this as the worst possible outcome whereby LRT stations may have triggered new developments involving displacement and attracted high income households who do not often use transit; and

Neighborhood Decline: Occurs with significant negative coefficients for both Gentrification and TOD-related change indicators. We view this as indicative of continued neighborhood decline. By continued neighborhood decline, we refer to station areas as a whole experiencing greater negative change indicators than the given UA census tracts for the time periods examined.

We identify the urbanized areas that fit into these categories in Table 7. We initially expected that the presence of an LRT station prompts TOD type changes and possibly leads to gentrification. As shown in Table 7, however, our analysis of fourteen UAs reveals a much more complicated picture of neighborhood change, with only San Francisco (and Cleveland’s gentrifiable tracts) confirming our expectation. Perhaps, the people with the most need for improved transit access and transit developments (i.e., minority, low-income residents) were not primary beneficiaries of these benefits in San Francisco. Meanwhile, Denver, Sacramento, Dallas, and St. Louis, along with Buffalo’s gentrifiable tracts, all confirmed our expectation for gentrification, but not for TOD-related change. This possibly indicates that gentrification occurred in transit station areas, but without the benefits commonly associated with successful TODs (increased densities and public transit use, bicycle use, and walking).

Table 7.

Gentrification and TOD Impact Typology.

	All Tracts (Set 1)	Gentrifiable Tracts (Set 2)
Gentrification and TOD	San Francisco	San Francisco, Cleveland
Countergentrification and TOD	Cleveland, Portland	Portland
Gentrification and counter- or no TOD	Denver, Sacramento, Dallas	Denver, St. Louis, Buffalo
Neighborhood decline	San Diego, Buffalo, Los Angeles, Pittsburgh, Salt Lake City	San Diego, Los Angeles, Baltimore

On the other hand, Portland’s LRT neighborhoods experienced TOD, but without generating gentrification problems. Neighborhood changes in the past thirty years actually occurred in the direction of countergentrification, possibly indicating that the residents with the most need for improved transit developments received access to them. Other UAs such as Pittsburgh, Salt Lake City, San Diego, Los Angeles, and Baltimore, though, experienced neighborhood decline. For these areas, the people with the most need for transit possibly have become further marginalized.

Discussion

Using spatial regression analyses of longitudinal data specifically focusing on neighborhood-level socioeconomic characteristics across fourteen US urbanized areas, we aimed to address the following: To what extent is the presence of a light rail station associated with gentrification? Overall, we find no evidence of prevalent gentrification in LRT station areas. An analysis of UA-specific impacts gives a complicated story of possible (counter-) gentrification and TOD-related changes, implying that the impacts of LRT stations can vary depending on local and regional contexts and planning efforts.

For San Francisco and Denver, in particular, our results reveal that light rail station areas have become relatively occupied by whiter, richer, and better-educated residents. Such changes are key indicators of gentrification occurring. On the other hand, our results indicate that for areas such as Portland, Los Angeles, and Buffalo, station areas are characterized by gaining relatively less white and educated populations as well as having relatively greater poverty rates than tracts without stations. Such changes indicate that these station areas either are further declining or increasingly occupied by the actual residents needing improved transit access. The former looks the case for LA and Buffalo while improved transit access is largely enjoyed by the residents with low socioeconomic status in Portland.

More importantly, our results highlight that efforts by local and regional planners for more inclusive developments can maximize the benefits of TOD. LRT stations in both San Francisco and Portland had significant TOD impacts. Portland, for instance, is an often studied region (Bae 2002; Knaap, Ding, and Hopkins 2001) that specifically focuses on developing around light rail stations and seeks to maintain equitable TOD plans. Our results indicating countergentrification could be largely due to sustained efforts by local and regional planners to ensure equitable access to transit. Portland’s Metropolitan Regional Council, Metro, has a specific TOD Program for transportation-related developments. It is mainly aimed at developers, providing incentives to them to develop in an economically feasible way around transit (Metro 2015). Even still, Metro’s TOD Strategic Plan prepared by the Center for Transit Oriented Development (CTOD) expressly discusses TOD equity stating: “one of the key challenges that future TOD implementation will need to address is fostering new transit oriented housing that is affordable to the workforce” (CTOD 2011, 27). The Portland region also has strong support for TOD and affordable housing from the State of Oregon. In 1995, Oregon enacted ten-year property tax abatements for multifamily and affordable housing that can easily access major transit facilities (Cervero, Ferrell, and Murphy 2002, 48). Such policies and legislation may help explain why our results suggest that Portland experienced countergentrification and TOD impacts related to light rail. In this regard, we recommend that planners develop and implement equitable, inclusive, and affordable housing policies in TOD and LRT plans to better mitigate station area gentrification.

Meanwhile, the San Francisco Bay Area’s Transit-Oriented Affordable Housing (TOAH) Fund aims to develop affordable housing especially near transit lines (Bay Area TOAH 2015). However, the region’s overall high housing prices and real estate market may dominate any equitable efforts and strongly influence the gentrification-related residential changes that our results suggest. Also, St. Louis’s decline in public transit use in its station areas could be a result of its light rail station design and lack of TOD planning. Operating along an existing rail line that was once used for industrial purposes, the light rail line mainly runs below street-grade. As a once industrially used rail line, some of the station’s immediately surrounding areas are still undeveloped or inadequately developed for commercial and residential purposes. Additionally, even though St. Louis’s light rail line began operations in 1993, the region did not develop TOD or station area plans until nearly two decades later. Such a lack of planning may have led to a decline in public transit use in LRT station areas. For these reasons, we recommend that planners aggressively plan station areas to better ensure that equitable outcomes result from station area developments—well before LRT operations begin and consistently throughout the life of the LRT systems.

Overall, this study provides a necessary component to understanding transit’s role in neighborhood change; but it has limitations. While our analysis of socio-demographic attributes in LRT neighborhoods offers useful insights into gentrification related changes, it cannot present hard evidence of displacement. Therefore, our findings point to more future research being warranted that combines both quantitative methods and more localized qualitative approaches. A deeper understanding of the regional policies that go into planning transit and developing around transit, analyzing design components of the station areas, studying transit infrastructure and its placement, and surveying the historical aspects and lived experiences of existing and future neighborhood residents will be vital components in future research. Future research can also focus on the decision-making processes that influence inclusive TOD spaces in specific regions.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Supplemental Material

Supplemental material for this article is available with the manuscript on the JPER website.

Author Biographies

Dwayne Marshall Baker is a postdoctoral fellow in the Department of City Planning at the University of Manitoba. His research interests include transportation policy and planning, spatial analysis and GIS, and inclusivity.

Bumsoo Lee is an associate professor in the Department of Urban and Regional Planning at the University of Illinois at Urbana-Champaign. His research interests include urban spatial organization, sustainable infrastructure and development, and land use and transportation interactions.

References

Anselin

2002. “Under the Hood: Issues in the Specification and Interpretation of Spatial Regression Models.” Agricultural Economics 27 (3): 247–67.

Arribas-Bel

Sanz-Gracia

2014. “The Validity of the Monocentric City Model in a Polycentric Age: US Metropolitan Areas in 1990, 2000 and 2010.” Urban Geography 35 (7): 980–97.

Atkinson

2000 “Measuring Gentrification and Displacement in Greater London.” Urban Studies 37 (1): 149–65.

Bae

C-H. C.

2002. “Orenco Station, Portland, Oregon: A Successful Transit Oriented Development Experiment?” Transportation Quarterly 56 (3): 9–18.

Bay Area TOAH. 2015. Bay Area TOAH Fund. http://bayareatod.com/. Accessed October 28.

Bernick

Cervero

1997. Transit Villages in the 21st Century. New York: McGraw-Hill.

Bivand

Hauke

Kossowski

2013. “Computing the Jacobian in Gaussian Spatial Autoregressive Models: An Illustrated Comparison of Available Methods.” Geographical Analysis 45 (2): 150–79.

Bivand

Lewin-Koh

2015. Maptools: Tools for Reading and Handling Spatial Objects. http://CRAN.R-project.org/package=maptools.

Bivand

Piras

2015. “Comparing Implementations of Estimation Methods for Spatial Econometrics.” Journal of Statistical Software 63 (18): 1–36.

10.

Brambor

Clark

W. R.

Golder

2006. “Understanding Interaction Models: Improving Empirical Analyses.” Political Analysis 14 (1): 63–82.

11.

Bridge

Butler

Lees

2012. Mixed Communities: Gentrification by Stealth? Bristol, UK: Polity Press.

12.

Brown-Saracino

2004. “Social Preservationists and the Quest for Authentic Community.” City & Community 3 (2): 135–56.

13.

Brown-Saracino. 2009. A Neighborhood that Never Changes: Gentrification, Social Preservation, and the Search for Authenticity. Chicago: University of Chicago Press.

14.

Bryson

2012. “Brownfields Gentrification: Redevelopment Planning and Environmental Justice in Spokane, Washington.” Environmental Justice 5 (1): 26–31.

15.

Cao

X. J.

Schoner

2014. “The Influence of Light Rail Transit on Transit Use: An Exploration of Station Area Residents along the Hiawatha Line in Minneapolis.” Transportation Research Part A: Policy and Practice 59:134–43.

16.

CTOD (Center for Transit Oriented Development). 2011. Transit-oriented development strategic plan: metro TOD program. http://www.oregonmetro.gov/sites/default/files/tod_final_report.pdf.

17.

Cervero

Landis

1997. “Twenty Years of the Bay Area Rapid Transit System: Land Use and Development Impacts.” Transportation Research Part A: Policy and Practice 31 (4): 309–33.

18.

Cervero

Ferrell

Murphy

2002. “Transit-Oriented Development and Joint Development in the United States: A Literature Review.” TCRP Research Results Digest 52.

19.

Chatman

D. G.

Tulach

N. K.

Kim

2012. “Evaluating the Economic Impacts of Light Rail by Measuring Home Appreciation: A First Look at New Jersey’s River Line.” Urban Studies 49 (3): 467–87.

20.

Clark

1992. “On Blindness, Centrepieces and Complementarity in Gentrification Theory.” Transactions of the Institute of British Geographers 17 (3): 358–62.

21.

Clay

P. L.

1979. Neighborhood Renewal: Middle-Class Resettlement and Incumbent Upgrading in American Neighborhoods. Lexington, MA: Lexington Books.

22.

Cooley

2010. “Moving On Out: Black Pioneering in Chicago, 1915-1950.” Journal of Urban History 36 (4): 485–506.

23.

Davidson

2008. “Spoiled Mixture: Where Does State-Led ‘Positive’ Gentrification End?” Urban Studies 45 (12): 2385–2405.

24.

Davis

Oakley

2013. “Linking Charter School Emergence to Urban Revitalization and Gentrification: A Socio-spatial Analysis of Three Cities.” Journal of Urban Affairs 35 (1): 81–102.

25.

de Smith

M J

. 2015. STATSREF: Statistical Analysis Handbook: A Web-Based Statistics Resource. Winchelsea, UK: The Winchelsea Press.

26.

Dawkins

Moeckel

2016. “Transit-Induced Gentrification: Who Will Stay, and Who Will Go?.” Housing Policy Debate 26 (4-5): 1-18.

27.

Debrezion

Pels

E. A. J. H.

Rietveld

2010. “The Impact of Rail Transport on Real Estate Prices: An Empirical Analysis of the Dutch Housing Market.” Urban Studies 48 (5): 997–1015.

28.

Dillon

2014. “Race, Waste, and Space: Brownfield Redevelopment and Environmental Justice at the Hunters Point Shipyard.” Antipode 46 (5): 1205–21.

29.

Duncan

2010. “The Impact of Transit-Oriented Development on Housing Prices in San Diego, CA.” Urban Studies 48 (1): 101–27.

30.

Duncan

2011. “The Synergistic Influence of Light Rail Stations and Zoning on Home Prices.” Environment and Planning A 43 (9): 2125-2142.

31.

Essoka

J. D.

2010. “The Gentrifying Effects of Brownfields Redevelopment.” Western Journal of Black Studies 34 (3): 299–315.

32.

Feinstein

B. D.

Allen

2011. “Community Benefits Agreements with Transit Agencies: Neighborhood Change along Boston’s Rail Lines and a Legal Strategy for Addressing Gentrification.” Transportation Law Journal 38 (85): 85–113.

33.

Freeman

2005. “Displacement or Succession?: Residential Mobility in Gentrifying Neighborhoods.” Urban Affairs Review 40 (4): 463–91.

34.

Gardner

T. K.

Marlay

2013. “Whither the Jobs: Identifying Concentrations of Employment in Metropolitan Areas.” Urban Geography 34 (6): 795–814.

35.

Golub

Guhathakurta

Sollapuram

2012. “Spatial and Temporal Capitalization Effects of Light Rail in Phoenix: From Conception, Planning, and Construction to Operation.” Journal of Planning Education and Research 32 (4): 415–29.

36.

Grube-Cavers

Patterson

2014. “Urban Rapid Rail Transit and Gentrification in Canadian Urban Centres: A Survival Analysis Approach.” Urban Studies 52 (1): 1–17.

37.

Guerra

Cervero

Tischler

2012. “Half-Mile Circle: Does It Represent Transit Station Catchments?.” Transportation Research Record: Journal of the Transportation Research Board (2276): 101–9.

38.

Hankins

K. B.

2007. “The Final Frontier: Charter Schools as New Community Institutions of Gentrification.” Urban Geography 28 (2): 113–28.

39.

Hess

D. B.

Almeida

T. M.

2007. “Impact of Proximity to Light Rail Rapid Transit on Station-Area Property Values in Buffalo, New York.” Urban Studies 44 (5/6): 1041–68.

40.

Kahn

M. E.

2007. “Gentrification Trends in New Transit-Oriented Communities: Evidence from 14 Cities that Expanded and Built Rail Transit Systems.” Real Estate Economics 35 (2): 155–82.

41.

Kissling

W. D.

Carl

2008. “Spatial Autocorrelation and the Selection of Simultaneous Autoregressive Models.” Global Ecology and Biogeography 17 (1): 59–71.

42.

Knaap

G. J.

Ding

Hopkins

L. D.

2001. “Do Plans Matter?: The Effects of Light Rail Plans on Land Values in Station Areas.” Journal of Planning Education and Research 21 (1): 32–39.

43.

Kohn

2013. “What Is Wrong with Gentrification?” Urban Research & Practice 6 (3): 297–310.

44.

Landis

2016. “Tracking and Explaining Neighborhood Socioeconomic Change in US Metropolitan Areas between 1990 and 2010.” Housing Policy Debate 26 (1): 1–51.

45.

Lee

2007. “‘Edge’ or ‘Edgeless’ Cities? Urban Spatial Structure in US Metropolitan Areas, 1980 to 2000.” Journal of Regional Science 47 (3): 479–515.

46.

Ley

1986. “Alternative Explanations for Inner-City Gentrification: A Canadian Assessment.” Annals of the Association of American Geographers 76 (4): 521–35.

47.

Ley

1996. The New Middle Class and the Remaking of the Central City. Oxford: Oxford University Press.

48.

Lin

2002. “Gentrification and Transit in Northwest Chicago.” Transportation Quarterly 56 (4): 175–91.

49.

Logan

J. R.

Stults

B. J.

2014. “Interpolating US Decennial Census Tract Data from as Early as 1970 to 2010: A Longitudinal Tract Database.” Professional Geographer 66 (3): 412–20.

50.

Los Angeles County Metropolitan Transportation Authority. 2016. Metro: Download GIS Data. http://developer.metro.net/introduction/gis-data/download-gis-data/.

51.

Marcuse

1985. “Gentrification, Abandonment, and Displacement: Connections, Causes, and Policy Responses in New York City.” Washington University Urban Law Journal; Journal of Urban and Contemporary Law 28:195–240.

52.

Mazerolle

M. J.

2004. Appendix 1: Making Sense out of Akaike’s Information Criterion (AIC): Its Use and Interpretation in Model Selection and Inference from Ecological Data.” Mouvements et Reproduction des Amphibiens en Tourbières Perturbées 174–90.

53.

McMillen

2013. “McSpatial: Nonparametric Spatial Data Analysis.” http://CRAN.R-project.org/package=McSpatial.

54.

Meligrana

Skaburskis

2005. “Extent, Location and Profiles of Continuing Gentrification in Canadian Metropolitan Areas, 1981-2001.” Urban Studies 42 (9): 1569–92.

55.

Metro. 2015. “Guide to Transit-Oriented Development.” http://www.oregonmetro.gov/tools-partners/guides-and-tools/guide-transit-oriented-development. Accessed November 3, 2016.

56.

Metropolitan Transportation Commission. 2014. “Metropolitan Transportation Commission Open Data.” http://www.mtc.ca.gov/maps_and_data/GIS/data.htm.

57.

Minnesota Population Center. 2011. National Historical Geographic Information System: Version 2.0. Minneapolis, MN: University of Minnesota 2011. http://www.nhgis.org.

58.

North Central Texas Council of Governments Transportation Department. n.d. Rail Station Access: Bicycle and Pedestrian Needs Assessment: GIS Data. http://www.nctcog.org/trans/sustdev/bikeped/routes_to_rail/GIS_web.html.

59.

Pittsburgh. 2015. “Geographic Information Systems Data Download: GIS Data Page: Transportation Data.” http://pittsburghpa.gov/dcp/gis/gis-data-new.

60.

R Core Team. 2015. R: A Language and Environment for Statistical Computing. https://www.R-project.org/.

61.

Regional Transportation District. n.d. “Regional Transportation District GIS Data Download.” http://maps.rtd-denver.com/GisDatadownload/datadownload.aspx.

62.

Revington

Nick

. 2015. “Gentrification, Transit, and Land Use: Moving Beyond Neoclassical Theory.” Geography Compass 9 (3): 152–63.

63.

Rose

1984. “Rethinking Gentrification: Beyond the Uneven Development of Marxist Urban Theory.” Environment and Planning D: Society and Space 2 (1): 47–74.

64.

Sacramento. 2014. “Geographic Information Systems Data.” http://www.cityofsacramento.org/GIS/Data.

65.

SANDAG. n.d. “GIS Downloads.” http://www.sandag.org/index.asp?subclassid=100&fuseaction=home.subclasshome.

66.

Shaw

Wheeler

1994. Statistical Techniques in Geographical Analysis. London: David Fulton.

67.

Slater

2009. “Missing Marcuse: On Gentrification and Displacement.” City 13 (2/3): 292–311.

68.

Smith

1987. “Gentrification and the Rent Gap.” Annals of the Association of American Geographers 77 (3): 462–65.

69.

Smith

2010. “Toward a Theory of Gentrification: A Back to the City Movement by Capital, Not People.” In The Gentrification Reader, edited by Lees

Loretta

Slater

Tom

Wyly

Elvin

, 85–96. London: Routledge.

70.

Soursourian

2010. “Equipping Communities to Achieve Equitable Transit-Oriented Development.” Community Investments 22 (2): 22–27.

71.

Stilwell

1993. Reshaping Australia: Urban Problems and Policies. Sydney: Pluto Press.

72.

Sullivan

D. M.

2007. “Reassessing Gentrification: Measuring Residents’ Opinions Using Survey Data.” Urban Affairs Review 42 (4): 583–92.

73.

Taylor

M. M.

2002. Harlem: Between Heaven and Hell. Minneapolis: University of Minnesota Press.

74.

Teernstra

A. B.

Van Gent

W. P. C.

2012. “Puzzling Patterns in Neighborhood Change: Upgrading and Downgrading in Highly Regulated Urban Housing Markets.” Urban Geography 33 (1): 91–119.

75.

Trimet. 2014. “Trimet Geospatial Data.” http://developer.trimet.org/gis/.

76.

Uitermark

Duyvendak

J. W.

Kleinhans

2007. “Gentrification as a Governmental Strategy: Social Control and Social Cohesion in Hoogvliet, Rotterdam.” Environment and Planning A 39 (1): 125–41.

77.

United States Bureau of the Census. 1983. Geographic Reference Manual: 1982 Economic Censuses. Washington, DC: US Department of Commerce, Bureau of the Census.

78.

United States Bureau of the Census. 2010. “2010 TIGER/Line Shapefiles: Census Tracts.” https://www.census.gov/cgi-bin/geo/shapefiles/index.php?year=2010&layergroup=Census+Tracts.

79.

United States Department of Labor, Bureau of Labor Statistics. 2015. “Consumer Price Index.” Accessed May 12. bls.gov/cpi/#tables.

80.

United States Department of Transportation. 2013. “Polyline: Fixed-Guideway Transit Facilities.” http://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/national_transportation_atlas_database/2013/polyline.html.

81.

US 2010 Project . 2014. Spatial Structures in the Social Sciences (S4), Brown University, http://www.s4.brown.edu/us2010/Researcher/Bridging.htm. Accessed 1 July, 2014.

82.

Utah Automated Geographic Reference Center. 2013. “Transit: Bus, Commuter, and Light Rail.” http://gis.utah.gov/data/sgid-transportation/transit/.

83.

Zhang

2013. “Comparison of Spatial and Non-spatial Logistic Regression Models for Modeling the Occurrence of Cloud Cover in North-eastern Puerto Rico.” Applied Geography 37: 52-62.

84.

Zuk

Bierbaum

Ariel H.

Chapple

Karen

Gorska

Karolina

Loukaitou-Sideris

Anastasia

Ong

Paul

Thomas

Trevor

. 2015. “Gentrification, Displacement and the Role of Public Investment: A Literature Review.” Federal Reserve Bank of San Francisco.

85.

Zukin

2013. “The Creation of a ‘Loft Lifestyle.’” In The Gentrification Debates: A Reader 2013, edited by Brown-Saracino

Japonica

, 175–84. New York, NY: Routledge, 2013.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.58 MB