Content uploaded by Michael S. Barton
Author content
All content in this area was uploaded by Michael S. Barton on Jan 25, 2021
Content may be subject to copyright.
Article
Urban Studies
2016, Vol. 53(1) 92–111
ÓUrban Studies Journal Limited 2014
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0042098014561723
usj.sagepub.com
An exploration of the importance of
the strategy used to identify
gentrification
Michael Barton
Louisiana State University, USA
Abstract
Urban scholars have described the importance of gentrification in major cities across the USA
since the 1970s. While there is consensus that gentrification shaped social and physical aspects of
neighbourhoods, scholars have yet to agree on how gentrified neighbourhoods should be identi-
fied. Owing to the lack of consensus, gentrification was measured in a variety of ways, which
greatly influenced the neighbourhoods studied in previous research and potentially the findings of
research that assessed the importance of gentrification for other neighbourhood outcomes. The
current study contributes to this debate by applying and comparing two census-based strategies
for identifying gentrified neighbourhoods with a qualitative neighbourhood selection strategy
derived from The New York Times to New York City neighbourhoods for the span of years from
1980 to 2009. Results confirm that each of the strategies identified different neighbourhoods and
that qualitative strategies for identifying gentrified neighbourhoods may overlook areas that expe-
rienced similar changes to those more widely recognised as gentrified. Given these findings, addi-
tional analyses assessed which census-based neighbourhood selection strategy better
represented the neighbourhoods perceived by The New York Times, a major media outlet that
shaped discourse on gentrification in the USA, as having experienced gentrification.
Keywords
gentrification, identifying gentrified neighbourhoods, neighbourhood change, New York City, The
New York Times
Received June 2014; accepted November 2014
Introduction
Gentrification has been studied for nearly 50
years, but scholars have yet to come to con-
sensus on how to define the process or iden-
tify gentrified neighbourhoods (Hackworth,
2007; Lees et al., 2008). While the specific
definition varied, gentrification was often
understood as ‘the process by which higher-
income households displace lower income
[households] of a neighborhood, changing
the essential character and flavor of that
neighborhood’ (Kennedy and Leonard,
2001: 5). With this definition in mind, a vari-
ety of methods was used to identify gentrified
Corresponding author:
Michael Barton, Sociology Department, Louisiana State
University, 139 Stubbs Hall, Baton Rouge, LA 70803, USA.
Email: mbarto3@lsu.edu
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
neighbourhoods. An underexplored feature of
previous research was that the strategy used
to identify gentrified neighbourhoods poten-
tially had important implications for whether
gentrification was associated with other neigh-
bourhood outcomes.
When identifying gentrified neighbour-
hoods, qualitative studies typically identified
a single or a small group of neighbourhoods
that gentrified. In contrast, quantitative
studies typically used a threshold strategy
where neighbourhoods were identified as
gentrifiable if they featured a particular char-
acteristic or characteristics at the beginning
of a decade and gentrified if they experienced
a change in the characteristic or characteris-
tics at a later time. Recent research has
sought to bridge the gap between qualitative
and quantitative neighbourhood selection
strategies by incorporating non-census based
measures of gentrification, but such efforts
have yet to incorporate perceptions of neigh-
bourhood change into large-scale measure-
ment of gentrification.
The current study seeks to bridge this gap
by illustrating how neighbourhoods identi-
fied by census-based strategies described by
Bostic and Martin (2003) and Freeman
(2005) relate to neighbourhoods recognised
as gentrified by The New York Times
between 1980 and 2009. This particular
media source was selected because analyses
conducted by Brown-Saracino and Rumpf
(2011: 293) found that high-profile newspa-
pers, such as The New York Times, were
more likely to report on gentrification than
other local papers in large cities such as New
York City. Results of descriptive and bivari-
ate analyses indicate that the strategy used to
identify gentrified neighbourhoods has
important implications and that there is
much room for improvement in bridging the
gap between qualitative and quantitative
strategies for identifying gentrified neigh-
bourhoods. The discussion section explores
the importance of these results for research
on the association of gentrification and other
neighbourhood outcomes. While the current
study is a case study of New York City, the
results have implications for the study of
gentrification in other large American cities
as well.
What makes a neighbourhood
gentrified?
The title of an often-cited piece by
Beauregard (1986), ‘The chaos and complex-
ity of gentrification’, succinctly summarises
the current state of the definition of gentrifi-
cation. Most scholars acknowledge that gen-
trification was first identified in London
during the 1960s by Glass (1964: xviii) who
used the term to describe the ‘invasion’ of
members of the middle and upper classes
into traditionally working-class neighbour-
hoods, resulting in the displacement of
incumbent residents and a change of the
social character of the neighbourhood. This
definition can be broken down into two
interrelated components. First, gentrifica-
tion raises the economic level of a neigh-
bourhood population. Second, gentrification
changes the ‘social character’ or culture of
neighbourhoods. These components were
important because they shaped the defini-
tions of gentrification that followed.
The primary cause of the ‘chaos’ that fol-
lowed Glass’ (1964) conception has to do
with methodological differences, as concep-
tualisations of gentrification were often
determined by whether qualitative or quanti-
tative data were used. Qualitative data tend
to be much richer, which allowed for the uti-
lisation of complex, multidimensional defini-
tions. A prime example of this was Hamnett
(1984: 284), who defined gentrification as
‘simultaneously a physical, economic, social
and cultural phenomenon’ that involved ‘the
invasion by middle-class or higher-income
groups of previously working-class neigh-
borhoods or multi-occupied ‘‘twilight areas’’’
Barton 93
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
and the physical renovation or rehabilitation
of deteriorated housing stock. While this was
one of the most comprehensive definitions of
gentrification to date, it was also one of the
more complex, as it required the simulta-
neous measurement of physical, economic,
social and cultural neighbourhood changes.
Finding data that allowed for the simulta-
neous measurement of all of these aspects
has been an arduous task because of limited
quantitative information on cultural changes
for a large sample of neighbourhoods.
Quantitative studies, in contrast, typically
relied on census data and therefore defined
gentrification through demographic changes.
Simple versions of this strategy focused
solely on changes to the socio-economic
composition of neighbourhood residents.
For example, Atkinson (2000: 149) defined
gentrification as the ‘process of class succes-
sion and displacement in areas broadly char-
acterized by working-class and unskilled
households’. More complex quantitative
conceptualisations recognised the multidi-
mensionality of gentrification. For example,
Clark (2005: 258) defined gentrification as a
process that resulted in an increased propor-
tion of higher socio-economic status residents
in combination reinvestment of the built envi-
ronment. Clark’s (2005) definition captured
more of the spirit Glass’ (1964) definition as it
recognised that gentrification changed the
composition of neighbourhood populations
and the physical appearance of neighbour-
hoods, both of which were important because
they influenced the perception of the neigh-
bourhood among residents and outsiders.
While the definitions used in quantitative stra-
tegies were easier to operationalise, they often
did not include references to changes in the
‘social character’ or local culture.
In summary, scholars have yet to come to
a consensus concerning the definition of gen-
trification. Qualitative studies often defined
the process in regards to changes to the eco-
nomic and racial composition as well as the
character of neighbourhoods, while quanti-
tative studies emphasised demographic
changes. While not the focus of the current
study, this definitional debate is important
to recognise as the conceptualisation of gen-
trification influenced the strategy used to
identify gentrified neighbourhoods.
Identification of gentrified
neighbourhoods
The lack of consensus concerning the concep-
tualisation of gentrification allowed research-
ers to identify gentrified neighbourhoods in a
variety of ways. Qualitative studies typically
identified a single neighbourhood or small
group of neighbourhoods that experienced
gentrification. Such research has identified
New York City neighbourhoods such as
Clinton Hill (Freeman, 2006), Harlem
(Freeman, 2006; Maurrasse, 2006), the Lower
East Side (Mele, 2000; Smith, 1996; Zukin
et al., 2009), and Williamsburg (Curran,
2007) as gentrified. Outside of New York
City, this strategy has identified gentrified
neighbourhoods in Chicago (Boyd, 2008),
Philadelphia (Anderson, 1990) and London
(Hamnett and Whitelegg, 2006).
When identifying gentrified neighbour-
hoods, qualitative research frequently
selected neighbourhoods that experienced
cultural changes because of shifts in demo-
graphics of neighbourhood populations.
Much of this research emphasised the transi-
tion from racial or ethnic neighbourhood
cultures to middle-class, white culture
(Anderson, 1990; Maurrasse, 2006). This
cultural shift was often associated with
changes in local businesses and housing. For
example, Maurrasse (2006) and Zukin et al.
(2009) described examples of gentrification
replacing traditional ‘mom and pop’ stores
and restaurants with chain stores and restau-
rants. Similarly, many former manufactur-
ing sites in gentrified neighbourhoods were
converted to loft-style condominiums and
94 Urban Studies 53(1)
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
apartments (Curran, 2007; Hamnett and
Whitelegg, 2006).
Quantitative studies often used a thresh-
old strategy where neighbourhoods were
identified as gentrifiable if they featured a
particular characteristic or characteristics at
the beginning of a decade and gentrified if
the characteristic changed in particular way.
A simple version of this strategy was used
by Atkinson (2000) who identified neigh-
bourhoods that experienced gains in the pro-
portion of workers employed in professional
occupations between two time points. More
complex versions of this strategy include the
multivariate strategies described by Freeman
(2005) and Bostic and Martin (2003), which
are replicated in the current study.
Quantitative research has increasingly
sought to break away from relying solely on
census indicators of neighbourhood change.
Hammel and Wyly (1996) and Wyly and
Hammel (1998, 1999) were the forerunners
in this area. The authors identified gentrified
neighbourhoods through their familiarity
with Minneapolis-St. Paul and consultation
of published materials (previous research
and newspaper articles) before conducting
fieldwork to determine the quality and extent
of gentrification in identified areas (Hammel
and Wyly, 1996). After the success of their
initial application, the authors replicated this
strategy in 22 other large American cities
and produced a data base of gentrified cen-
sus tracts (Wyly and Hammel, 1999). While
this remains the richest source of informa-
tion on gentrification available, it was lim-
ited in that it only explored neighbourhood
changes during the 1990s and did not include
New York City, which attracted a great deal
of media and scholarly attention (Brown-
Saracino and Rumpf, 2011; Zukin et al.,
2009).
More recently, Hwang and Sampson
(2014) drew upon a combination of census
indicators and systematic social observation
of Chicago streets using Google Street View
to identify gentrified neighbourhoods. The
use of Google Street View allowed the
researchers to track relatively fine-grain
changes in neighbourhoods between 2007
and 2009. Similar to Hammel and Wyly’s
(1996) data base, this strategy suffered from
a limited study period, which was important
given the gradual nature of gentrification.
Further, Hwang and Sampson (2014: 22)
note that their strategy emphasised physical
forms of reinvestment and potentially suf-
fered from subjectivity in the identification
of street-level indicators of gentrification.
The primary advantage to the strategies
described by Hammel and Wyly (1996) and
Wyly and Hammel (1998, 1999) and Hwang
and Sampson (2014) was that they were able
to confirm that the identified neighbour-
hoods experienced gentrification as opposed
to more natural forms of improvement such
as incumbent upgrading. The primary disad-
vantage was that these strategies were
extremely labour-intensive to replicate. Less
labour-intensive efforts supplemented census
indicators with non-census based measures
of gentrification such as coffee shops
(Papachristos et al., 2011) or mortgage lend-
ing information (Kreager et al., 2011). Such
variables allowed for more refined measure-
ment of gentrification, but were limited in a
few respects. For example, the location of
coffee shops used by Papachristos et al.
(2011) was determined by unmeasurable
influences such as city planning efforts, indi-
vidual tastes and residential preferences.
Additionally, coffee shops were clustered in
the central business district, which meant
that a potentially large number of neigh-
bourhoods were excluded. Similarly, the
home mortgage measure used by Kreager et
al. (2011) did not allow for the distinction of
gentrification from other types of invest-
ment, such as incumbent upgrading.
To summarise, previous research used a
variety of strategies to identify gentrified
neighbourhoods. Relying only on qualitative
Barton 95
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
methods ensured that some gentrified neigh-
bourhoods were overlooked, while relying
only on quantitative methods identified
neighbourhoods that underwent naturally
occurring improvements (incumbent upgrad-
ing) as well as those that improved because
of gentrification. Recent efforts to bridge
this gap incorporated non-census based
information sources, but a measure of the
perceptions of neighbourhood change has
yet to be included in large-scale measure-
ment of gentrification. The current study
contributes to research in this area by show-
ing that the strategy used to identify gentri-
fied neighbourhoods can lead to very
different understandings of where gentrifica-
tion occurred. Specifically, the current study
describes how gentrified neighbourhoods
identified through a simple content analysis
of The New York Times compared with mul-
tivariate census-based strategies developed
by Bostic and Martin (2003) and Freeman
(2005). The New York Times was selected
because Brown-Saracino and Rumpf (2011:
239) reported that more prolific newspapers
such as The New York Times were more
likely to discuss gentrification than other
local papers in large cities such as New York
City.
Current study
The current study explores the importance
of the strategy used to identify gentrified
neighbourhoods by comparing New York
City neighbourhoods identified through
replications of census-based strategies
described by Bostic and Martin (2003) and
Freeman (2005) with neighbourhoods identi-
fied by The New York Times between 1980
and 2009. In doing so, the current study
explored two questions. First, does the mea-
surement of gentrification influence which
neighbourhoods were identified as gentri-
fied? Second, how do neighbourhoods iden-
tified by quantitative strategies relate to
neighbourhoods identified by qualitative
research and The New York Times? The dis-
cussion section explores the importance of
the results for research on the association of
gentrification with other neighbourhood
outcomes.
Analysis strategy
The current study is an exploratory analysis
and as such draws upon descriptive and
bivariate analyses. The first part of the anal-
yses provides descriptive statistics on the
number of gentrifiable and gentrified tracts
identified by the census strategies. The sec-
ond part of the analyses describes how the
distribution of gentrified neighbourhoods
identified by each strategy and determined
which census-based neighbourhood selection
strategy better matched with The New York
Times. The final section of the analyses con-
sists of exploratory spatial data analysis that
help explain the differences between the
neighbourhoods identified by The New York
Times and the census-based strategies.
Units of analysis
The units of analysis in the current study are
the 188 populated neighbourhood areas
identified by the New York City
Department of City Planning (NYCDCP).
Each neighbourhood area conformed to
Census 2000 boundaries, contained about 11
census tracts and was associated with a col-
loquial neighbourhood name (New York
City Department of City Planning, 2011).
Similar units of analysis were used by the
Project on Human Development in Chicago
Neighbourhoods researchers (Sampson,
2012). Neighbourhood areas were used as
the units of analysis to facilitate the compar-
ison of the results of the Bostic and Martin
(2003) and Freeman (2005) strategies to The
New York Times. A more refined unit of
analysis such as census tracts would have
96 Urban Studies 53(1)
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
been preferable given that gentrification
typically occurred in part of a neighbour-
hood as opposed to a whole neighbourhood
(Boyd, 2008; Freeman, 2006), but was not
possible given the limited geographic identi-
fying information provided in most of The
New York Times articles.
Data and measures
The New York Times
The New York Times was selected to repre-
sent perceptions of neighbourhood change
in New York City because it is a nationally
recognised news source with a strong
emphasis on events in New York City. An
additional reason for selecting The New
York Times over other newspapers based in
New York City was that Brown-Saracino
and Rumpf (2011) reported that more proli-
fic newspapers such as The New York Times
were more likely to report on gentrification
than other local papers in large cities such
as New York City. Specifically, articles were
collected from the LexisNexis data base
published between 1 January 1980 and 31
December 2009 that contained the terms
‘gentrification’ and ‘New York City’. The
decision to use the search term ‘gentrifica-
tion’ mirrors that of Brown-Saracino and
Rumpf (2011). Articles had to discuss gen-
trification as currently occurring or having
recently occurred in a specific neighbour-
hood to be considered. This strategy pro-
duced 759 documents, each of which was
coded according to the year and month of
publication and the neighbourhood identi-
fied. The final data set identified whether
neighbourhoods gentrified during the 1980s
(1980–1989), the 1990s (1990–1999) and
2000s (2000–2009).
As recognised by Sampson (2012: 189), a
limitation of using media sources to identify
temporal changes in neighbourhoods was
that such sources tend to report on ‘news-
worthy’ events while overlooking similar
events in less noteworthy areas. While
Sampson did not find this was an issue for
Chicago Tribune reporting, it remains
unclear whether this was an issue for The
New York Times reporting. A similar criti-
cism of focusing on more noteworthy areas
might be levelled against the qualitative gen-
trification research as such research focused
on a small sample of neighbourhoods. Even
with this potential limitation, The New York
Times was the best source of information on
the distribution of gentrification in New
York City that could be located that covered
the entire city for the period 1980 to 2009.
Census-based strategies for
identifying gentrified
neighbourhoods
The Bostic and Martin (2003) and Freeman
(2005) strategies were selected because
descriptive analyses conducted by Freeman
(2009) compared neighbourhoods identified
through a replication of strategies described
by Freeman (2005) and Hammel and Wyly
(1996), on which Bostic and Martin (2003)
was based. The current study uses the Bostic
and Martin (2003) strategy to address
Freeman’s (2009: 2094) critique that
Hammel and Wyly’s (1996) strategy was
based upon a limited number of MSAs. In
contrast to the Hammel and Wyly (1996)
strategy, the Bostic and Martin (2003) strat-
egy was developed with and applied to a
national sample of census tracts. This made
it an ideal choice for comparison with the
Freeman (2005) strategy, which was also
developed with and applied to a national
sample of census tracts.
The first step of the replication process
required that the Bostic and Martin (2003)
and Freeman (2005) strategies each be
applied to tract-level data from New York
City. The Neighbourhood Change Database
(NCDB) was used to collect information
from the 1980, 1990 and 2000 Censuses to
Barton 97
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
ensure uniformity of tract boundaries during
the study period. Tract-level census informa-
tion for the 2005–2009 period was down-
loaded from the American Community
Survey and conformed to Census 2000
boundaries (US Census Bureau, 2013). The
proceeding sections describe the steps fol-
lowed to replicate each of these strategies.
Replication of Bostic and Martin. Bostic and
Martin (2003) utilised tract-level data for the
50 largest metropolitan areas and identified
gentrified tracts in two ways. The first
strategy replicated Hammel and Wyly (1996)
and classified tracts that featured a median
income that was less than 50% of the
median income for the MSA at a particular
census point as gentrifiable.
1
Gentrified
tracts were those considered gentrifiable at
the earlier time point and that had changed
to non-gentrifiable at the later time point.
Replication of this strategy with New York
City census tracts identified 48 gentrified
tracts during the 1980s, 110 tracts during the
1990s and 60 tracts during the 2000s.
Bostic and Martin (2003: 2431) stated
that identifying gentrified tracts with a single
variable was ‘almost certain to fail’. Given
this limitation, the authors drew upon the
nine census indicators highlighted by
Hammel and Wyly (1996) to develop a mul-
tivariate strategy for identifying gentrified
tracts. For each decade, tracts that featured
population at the start and end of each
decade were ranked according to how much
the proportion with college degrees, family
income, home-ownership rates, proportion
aged 30 to 44, proportion white non-family
households, proportion managerial and
administrative workers, and the proportion
with some college increased and how much
the percent poverty and black decreased.
The average of the rankings was computed
and compared with the number of tracts
identified by the univariate strategy. For
example, since the first strategy identified 48
tracts as switching from gentrifiable to non-
gentrifiable between 1980 and 1990, the 48
tracts with the lowest average rank values
were coded as gentrified. This was repeated
for each decade. While both strategies
described by Bostic and Martin (2003) were
replicated, analyses in the current study
emphasise the multivariate strategy to facili-
tate the comparison with the Freeman
(2005) strategy.
Replication of Freeman. For the Freeman
(2005) strategy, gentrifiable tracts were
populated, featured a median income that
was less than the median for the city and
contained a proportion of housing built
within the past 20 years lower than the pro-
portion found at the median for the city.
2
Gentrified neighbourhoods featured the
characteristics of gentrifiable neighbour-
hoods and experienced an increase in educa-
tional attainment greater than the median
for the city and an increase in real housing
prices during each intercensal period.
Results
Comparison of the tract-level selection
strategies
Table 1 shows the number of gentrifiable
and gentrified tracts identified by the Bostic
and Martin (2003) and Freeman (2005) stra-
tegies. As described earlier, Bostic and
Martin (2003) identified the number of
tracts that featured a median income that
was less than 50% of the median income for
the MSA (or city in the current study) at
Time 1 (1980 for example) and a median
income that was greater than 50% of the
median income for the MSA (or city) at
Time 2 (1990 for example) as the base num-
ber of tracts for their multivariate strategy.
Using this rule, 202 tracts were identified as
gentrifiable in 1980, 268 tracts were consid-
ered gentrifiable in 1990 and 176 tracts were
98 Urban Studies 53(1)
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
gentrifiable in 2000. These counts were con-
siderably lower than the Freeman strategy,
which identified 400 tracts as gentrifiable in
1980, 582 tracts in 1990 and 576 tracts in
2000.
In addition to identifying a smaller num-
ber of gentrifiable tracts than the Freeman
strategy, the results in Table 1 also indicate
that the Bostic and Martin strategy identi-
fied fewer gentrified tracts. Specifically, the
Freeman strategy identified 132 gentrified
tracts during the 1980s, while the Bostic and
Martin strategy identified 48 gentrified
tracts. During the 1990s, the Freeman strat-
egy identified 240 gentrified tracts where the
Bostic and Martin strategy identified 110
gentrified tracts. Similarly, during the 2000s,
the Freeman strategy identified 281 gentri-
fied tracts where the Bostic and Martin
strategy identified 60.
Comparison of the neighbourhood
selection strategies
Replications of the Bostic and Martin (2003)
and Freeman (2005) strategies identified
gentrified tracts, but the units of analysis for
the comparison of the census-based strate-
gies with The New York Times are the 188
populated neighbourhood areas defined by
the NYCDCP. The shift in unit of analysis
from census tracts to DCP neighbourhoods
was necessary because The New York Times
frequently discussed neighbourhoods as
experiencing gentrification as opposed to
specific parts, or in this case specific census
tracts, of a neighbourhood. This was impor-
tant because previous research found gentri-
fication typically occurred in part of a
neighbourhood instead of the whole neigh-
bourhood (Boyd, 2008; Freeman, 2006).
Therefore, neighbourhood areas were classi-
fied as gentrified according to the Bostic and
Martin (2003) or Freeman (2005) strategies
if they contained at least one gentrified tract
during a given decade.
3
It is important to
note that direct comparison of the propor-
tion of tracts identified in Table 1 and pro-
portion of neighbourhoods identified in
Table 2 should not be made because the
tract-level analyses identify the number of
neighbourhood parts (census tracts) that
gentrified, while the neighbourhood level
analyses identify the number of neighbour-
hoods that gentrified.
The first question addressed in this study
was whether the measurement of gentrifica-
tion influenced which neighbourhoods were
identified as gentrified. Results in Table 2
identify the number of neighbourhoods
identified as gentrified by each strategy dur-
ing the 1980s, 1990s and 2000s. As expected,
each strategy identified a different number
of gentrified neighbourhoods. These results
indicate that, with the exception of the
1980s, the strategy used by The New York
Times was the most restrictive, while the
Freeman strategy was consistently the most
Table 1. Number of gentrifiable and gentrified tracts by identification strategy and year
Census strategy 1980–1990 1990–2000 2000–2005/2009
Bostic and Martin
Gentrifiable 202 268 176
Gentrified 48 110 60
Freeman
Gentrifiable 400 582 576
Gentrified 132 240 281
Note: N = 2135.
Barton 99
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
expansive. The importance of and potential
reasons for these differences are highlighted
in the remainder of the discussion of the
results.
The second research question, which
census-based strategy better matched with
The New York Times, was addressed by
computing crosstabulations for each decade.
Results of these analyses are presented in
Table 3. The phi coefficients in Table 3 have
a similar interpretation to Pearson correla-
tion coefficients, but are used to assess
bivariate relationships when at least one
variable is categorical. Results for the 1980s
show that the association of The New York
Times with Bostic and Martin (phi =
0.333***) was higher than the association
with the Freeman strategy (phi = 0.178*).
Similarly, the results for the 1990s identified
a stronger association with Bostic and
Martin (0.299**) than the Freeman
(0.248**) strategy. This pattern continued
into the 2000s as results identified a phi
coefficient of 0.511*** with Bostic and
Martin and 0.302*** with Freeman. A likely
cause of the lower association of The New
York Times with the Freeman strategy was
because the Freeman strategy identified a
substantially larger number of neighbour-
hoods as gentrified for each decade as this
increased the potential for differences to be
found between The New York Times and
Freeman strategies.
Results of exploratory spatial data analy-
sis suggest an alternative reason for the dif-
ference in agreement between the census-
based strategies and The New York Times.
Figure 1 displays the spatial distribution of
neighbourhoods identified by each strategy
during the 1980s. These maps show that all
three strategies identified neighbourhoods
throughout the five boroughs, but the neigh-
bourhoods identified by The New York
Times were concentrated in Brooklyn,
Manhattan and Queens. In comparison,
the neighbourhoods identified by the
Table 3. Crosstabulations of census strategies for identifying gentrified neighbourhoods and The New York
Times.
Decade Strategy Association with The New York Times
1980–1990 Bostic and Martin Phi = 0.333***
Freeman Phi = 0.178*
1990–2000 Bostic and Martin Phi = 0.299**
Freeman Phi = 0.248**
2000–2005/2009 Bostic and Martin Phi = 0.511**
Freeman Phi = 0.302***
Notes: N = 188. *p\0.05; **p\0.01; ***p\0.001.
Table 2. Number of neighbourhoods identified as gentrified.
Strategy 1980–1990 1990–2000 2000–2005/2009
Bostic and Martin 28 61 35
Freeman 69 104 97
The New York Times 44 27 32
Note: N = 188.
100 Urban Studies 53(1)
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
census-based strategies were more dispersed
among the five boroughs, which helps to
explain the relatively low associations with
The New York Times. Further, these results
suggest that an important reason for the
stronger association between the Bostic and
Martin and The New York Times strategies
was a greater proportion of agreements
in Manhattan and Brooklyn. The greater
dispersion of neighbourhoods identified by
the census-based strategies suggests that
Sampson’s (2012) critique that media
sources may emphasise changes in particular
areas applied to The New York Times.
Whether this was intentional or not was
unable to be determined.
The results for the 1990s (Figure 2) pres-
ent a similar pattern as those for the 1980s.
Specifically, these maps show that The New
York Times was more likely to identify neigh-
bourhoods in Brooklyn and Manhattan. The
New York Times also identified a few neigh-
bourhoods in The Bronx and Queens, but
did not identify any neighbourhoods in
Staten Island. In comparison, the neighbour-
hoods identified by the census-based strate-
gies were dispersed throughout the city. Also
similar to the 1980s, the maps for the 1990s
suggest that the greater proportion of agreed
upon neighbourhoods in Brooklyn and
Manhattan by the Bostic and Martin strat-
egy was a contributor to the stronger associa-
tion with The New York Times.
The maps in Figure 3 show that, similar
to the previous decades, the neighbourhoods
identified by The New York Times during
the 2000s were concentrated in Brooklyn
and Manhattan. Further, these results indi-
cate that even though the Freeman strategy
agreed with The New York Times about a
larger number of neighbourhoods, the asso-
ciation of The New York Times with the
Bostic and Martin strategy was stronger
because of the greater proportion of agree-
ments in Brooklyn, particularly the north-
western section, and Manhattan.
Overall, the results indicate that the strat-
egy used to identify gentrified neighbour-
hoods can have important consequences as
the qualitative and quantitative strategies
for identifying gentrification produced sub-
stantially different sets of neighbourhoods.
Further, results showed that the Bostic and
Martin strategy better matched with The
New York Times than the Freeman strategy.
The stronger associations with the Bostic
and Martin strategy was likely the result of
its more restrictive nature and a greater pro-
portion of neighbourhoods identified in
Brooklyn and Manhattan. Finally, it is
interesting to note that the associations of
the census-based strategies with The New
York Times increased over time. While there
the association of the Bostic and Martin
strategy with The New York Times declined
slightly during the 1990s, a comparison of
the coefficients for the 1980s and 1990s with
the 2000s shows a dramatic increase in the
convergence of each strategy. In compari-
son, the increased association with the
Freeman strategy increased steadily over
time. Overall, however, these findings sug-
gest that the strategy used by The New York
Times staff became more expansive.
Discussion and conclusion
The current study contributed to research
on the identification of gentrified neighbour-
hoods by comparing two census-based stra-
tegies with a qualitative strategy derived
from The New York Times. Three findings
stand out in the results. First, the number
and geographic distribution of gentrified
neighbourhoods identified by each strategy
varied greatly. Second, the results suggest
that media sources may overlook changes in
less popular neighbourhoods that underwent
similar transformations to those identified as
gentrified by census-based strategies for iden-
tifying gentrified neighbourhoods. Third,
while the more restrictive strategy described
Barton 103
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
by Bostic and Martin (2003) better matched
with The New York Times, there was still a
great deal of room for improvement as the
associations of both census-based strategies
with The New York Times were moderate at
best.
Freeman’s (2009) descriptive comparison
of tracts identified through replication of
strategies described by Freeman (2005) and
Hammel and Wyly (1996) highlighted sub-
stantial differences in the number of gentri-
fied tracts identified, but did not test
whether these differences resulted in signifi-
cantly different outcomes because the
Hammel and Wyly strategy was developed
with data from a limited number of MSAs.
The current study addressed this critique by
comparing neighbourhoods identified
through replication of strategies described
by Freeman (2005) and Bostic and Martin
(2003), both of which were developed with
and applied to national samples of census
tracts. Descriptive results show that not only
was the sheer number of neighbourhoods
identified by each strategy different, but also
that the geographic distribution of sampled
neighbourhoods varied. These are simple
findings, but have potentially important
implications for research that assesses the
influence of gentrification on other neigh-
bourhood outcomes such as residential dis-
placement and crime.
Much of the gentrification research has
discussed the potential for gentrification-
related displacement, but only a few studies
empirically assessed this issue and findings
differed. For example, Atkinson (2000) iden-
tified gentrified neighbourhoods as those
that featured an increase in the proportion
of neighbourhood residents employed in
professional occupations and found that
neighbourhoods that experienced an
increase in this population also experienced
displacement. In contrast, Freeman and
Braconi (2004: 43) identified gentrified
neighbourhoods based upon their familiarity
with recent trends in New York City and
found that disadvantaged households were
less likely to be displaced in gentrified neigh-
bourhoods than in non-gentrified neigh-
bourhoods. There is support in the literature
for each of these strategies, but for the pur-
poses of the current discussion it is impor-
tant to note that results of an analysis that
selected a sample based upon changes in a
single demographic found gentrification was
positively associated with residential displa-
cement, while results of a study that selected
neighbourhoods based upon personal
knowledge found a negative association.
Each of these strategies was somewhat lim-
ited as the Atkinson (2000) strategy relied
upon a single variable, which Bostic and
Martin (2003: 2431) suggested was ‘almost
certain to fail’ as a means of identifying gen-
trified neighbourhoods, while the Freeman
and Braconi (2004) strategy potentially over-
looked areas that experienced changes to
those they identified as gentrified. It remains
unclear whether Atkinson (2000) or
Freeman and Braconi (2004) assessed other
gentrification measurements or whether
alternative measurements might have pro-
duced different findings about the extent of
residential displacement. Such information
is important as the findings of these studies
and others that assessed this relationship
were used to inform policies that influenced
the availability of affordable housing in
American cites (Newman and Wyly, 2006).
Assessments of the association of gentrifi-
cation with crime also produced mixed
results. Early assessments failed to control
for omitted variable bias and identified gen-
trified neighbourhoods solely through
changes in census indicators, but recent
research overcame both limitations by using
sophisticated statistical techniques and
incorporating non-census measures.
Papachristos et al. (2011) included a cultural
measure of gentrification, coffee shops, and
reported a negative association of
108 Urban Studies 53(1)
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
gentrification with violent crime. In contrast,
Kreager et al. (2011) included a property-
based measured of gentrification, the num-
ber of home mortgages distributed, and
reported a negative association of gentrifica-
tion with property crime rates, but no asso-
ciation with violent crime. Comparisons of
these studies was facilitated by the use of
statistical techniques that controlled for
unmeasured variable bias, but remain tenta-
tive given differences in the cities studied by
each. A speedy resolution to this debate is
important given that local governments have
increasingly looked to gentrification as a
means of reducing the prevalence of neigh-
bourhood social problems such as crime.
In addition to having implications for the
neighbourhood selection process in research
on the influence of gentrification on other
neighbourhood outcomes, the results also
suggest Sampson’s (2012) critique that media
sources may under identify the extent of tem-
poral neighbourhood change processes such
as gentrification because of a focus on ‘news-
worthy’ neighbourhoods applied to The New
York Times. This was highlighted in maps
that showed The New York Times was more
likely to identify gentrified neighbourhoods
in Brooklyn and Manhattan, particularly
during the 1990s and 2000s, while overlook-
ing a large number of neighbourhoods in
other boroughs that underwent similar trans-
formations. Sampson (2012) did not find this
was an issue for Chicago Tribune reporting,
but future research should explore whether
this was an issue for The New York Times.
This is especially important given Brown-
Saracino and Rumpf’s (2011: 239) finding
gentrification was disproportionately dis-
cussed in major newspapers such as The New
York Times, Chicago Tribune and Los
Angeles Times than local newspapers.
The critique of focusing on a small sam-
ple of popular neighbourhoods also applies
to qualitative research of gentrification in
New York City as much of this literature
emphasised changes in the Brooklyn neigh-
bourhoods of Park Slope (Carpenter and
Lees, 1995; Freeman and Braconi, 2004) and
Williamsburg (Curran, 2007; Zukin et al.,
2009) and the Manhattan neighbourhoods
of Harlem (Freeman, 2006; Zukin et al.,
2009) and the Lower East Side (Mele, 2000;
Smith, 1996). Valuable insights were gained
from these studies, but expanding the scope
of gentrification research to other neigh-
bourhoods such as those identified by the
census-based strategies that were not also
identified by The New York Times in
the current study may illuminate underex-
plored facets of the gentrification process.
Quantitative case studies of understudied
neighbourhoods would help to address this
issue by identifying specific changes that
occurred. A better solution, however, would
require qualitative data collection by research-
ers and journalists, as such information would
provide more detailed information on how
the gentrification process occurred and was
perceived by neighbourhood residents.
To bridge the gap between qualitative and
quantitative strategies for identifying gentri-
fied neighbourhoods, the current study
assessed which of two census-based strate-
gies better matched with neighbourhoods
identified by The New York Times. While
these results indicate that the more restrictive
measurement strategy described by Bostic
and Martin (2003) was a better match, there
remained a substantial amount of disagree-
ment. This was not entirely surprising as
gentrification scholars have yet to come to a
consensus about the definition of gentrifica-
tion, let alone where it occurred.
Exploratory data analyses suggest that
the primary source of disagreement was
related to The New York Times staff’s
emphasis on changes in Brooklyn and
Manhattan as the geographic distribution of
neighbourhoods identified by both census-
based strategies were more dispersed. It is
possible, however, that The New York Times
Barton 109
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
used other terms to describe changes in
improving neighbourhoods such as urban
renewal, urban revitalisation or urban regen-
eration, which were often treated as syno-
nyms for gentrification (Kennedy and
Leonard, 2001: 5).
Future research could continue to bridge
the gap between qualitative and quantitative
measurement strategies by expanding the
scope of the content analysis discussed in the
current study to include other search terms
such as urban renewal, urban revitalisation
or urban regeneration. The incorporation of
these terms may identify additional neigh-
bourhoods and in turn increase the consen-
sus among The New York Times and the
census-based strategies. Further, it remains
unclear whether the findings of the current
study are generalisable to other cities. To
address this, content analyses of media
sources for other large cities such as those
included in the Hammel and Wyly (1996)
and Wyly and Hammel (1998, 1999) data
base might be conducted to determine how
well the identified neighbourhoods relate to
sources that shaped the perception of gentri-
fication in those cities.
Acknowledgements
Thank you to Samantha Friedman, Steven
Messner, Glenn Deane, Joe Gibbons, Sarah
Becker, Heather Rackin, and Sam Stroope for
their insightful feedback on early drafts.
Funding
This research received no specific grant from any
funding agency in the public, commercial, or not-
for-profit sectors.
Notes
1. The median income for the city was used
instead of the MSA because of the focus on
New York City.
2. Freeman used metropolitan statistical areas,
so this language was changed to match the
focus on a single city.
3. Variables were also created that reflected the
percent of neighbourhoods that gentrified
according to each census strategy and the
number of times a neighbourhood was men-
tioned by The New York Times for a given
decade. The results of analyses that used these
variables produced substantively similar
results to those presented in the current study.
These results are available upon request.
References
Anderson E (1990) Streetwise: Race, Class, and
Change in an Urban Community. Chicago, IL:
The University of Chicago Press.
Atkinson R (2000) Measuring gentrification and
displacement in Greater London. Urban Stud-
ies 37: 149–165.
Beauregard RA (1986) The chaos and complexity
of gentrification. In: Smith N and Williams P
(eds) Gentrification of the City. Boston, MA:
Allen and Unwin, pp. 35–55.
Bostic RW and Martin RW (2003) Black home-
owners as a gentrifying force? Neighbourhood
dynamics in the context of minority home-
ownership. Urban Studies 40: 2427–2449.
Boyd M (2008) Defensive development: The role
of racial conflict in gentrification. Urban
Affairs Review 43: 751–776.
Brown-Saracino J (2004) Social preservationists
and the quest for authentic community. City &
Community 3: 135–156.
Brown-Saracino J and Rumpf C (2011) Diverse
imageries of gentrification: Evidence from
newspaper coverage in seven U.S. cities, 1986–
2006. Journal of Urban Affairs 33: 289–315.
Carpenter J and Lees L (1995) Gentrification in
New York, London and Paris: An interna-
tional comparison. International Journal of
Urban and Regional Research 19: 286–303.
Clark E (2005) The order and simplicity of gentri-
fication – A political challenge. In: Atkinson
R and Bridge G (eds) Gentrification in a Global
Context. New York: Routledge, pp. 256–264.
Curran W (2007) ‘From the frying pan to the
oven’: Gentrification and the experience of
industrial displacement in Williamsburg,
Brooklyn. Urban Studies 44: 1427–1440.
Freeman L (2005) Displacement or succession?
Residential mobility in gentrifying neighbor-
hoods. Urban Affairs Review 40: 463–491.
110 Urban Studies 53(1)
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
Freeman L (2006) There Goes the ‘Hood’: Views
of Gentrification from the Ground Up. Philadel-
phia, PA: Temple University Press.
Freeman L (2009) Neighbourhood diversity, met-
ropolitan segregation and gentrification: What
are the links in the US? Urban Studies 46:
2079–2101.
Freeman L and Braconi F (2004) Gentrification
and displacement: New York City in the
1990s. Journal of the American Planning Asso-
ciation 70: 39–52.
Glass R (1964) Introduction: Aspects of change.
In: Glass R, Hobsbawm EJ, Pollins H, et al.
(eds) London: Aspects of Change. London:
Macgibbon and Kee, pp. xiii–xlii.
Hackworth J (2007) The Neoliberal City: Govern-
ance, Ideology, and Development in American
Urbanism. Ithaca, NY: Cornell University
Press.
Hammel DJ and Wyly EK (1996) A model for
identifying gentrified areas with census data.
Urban Geography 17: 248–268.
Hamnett C (1984) Gentrification and residential
location theory: A review and assessment. In:
Herbert DT and Johnston RJ (eds) Geography
and the Urban Environment: Progress in
Research and Applications. London: Wiley, pp.
283–319.
Hamnett C and Whitelegg D (2006) Loft conver-
sion and gentrification in London: From
industrial to postindustrial land use. Environ-
ment and Planning 39: 106–124.
Hwang J and Sampson RJ (2014) Divergent path-
ways of gentrification: Racial inequality and
the social order of renewal in Chicago neigh-
borhoods. American Sociological Review 79:
726–751.
Kennedy M and Leonard P (2001) Dealing with
neighborhood change: A primer on gentrifica-
tion and policy changes. Discussion Paper Pre-
pared for the Brookings Institution Center on
Urban and Metropolitan Policy.
Kreager DA, Lyons CJ and Hays ZR (2011)
Urban revitalization and Seattle crime, 1982–
2000. Social Problems 58: 615–639.
Lees L, Slater T and Wyly E (2008) Gentrification.
New York: Routledge.
Maurrasse DJ (2006) Listening to Harlem: Gentri-
fication, Community, and Business. New York:
Routledge.
Mele C (2000) Selling the Lower East Side: Cul-
ture, Real Estate, and Resistance in New York
City. Minneapolis, MN: University of Minne-
sota Press.
New York City Department of City Planning
(2011) BYTES of the Big Apple. Available at:
http://www.nyc.gov/html/dcp/html/bytes/appl
byte.shtml.
Newman K and Wyly EK (2006) The right to stay
put, revisited: Gentrification and resistance to
displacement in New York City. Urban Studies
43: 23–57.
Papachristos AV, Smith CM, Scherer ML, et al.
(2011) More coffee, less crime: The relation-
ship between gentrification and neighborhood
crime rates in Chicago, 1991 to 2005. City &
Community 10: 215–240.
Sampson RJ (2012) Great American City: Chi-
cago and the Enduring Neighborhood Effect.
Chicago, IL: The University of Chicago Press.
Smith N (1996) The New Urban Frontier: Gentri-
fication and the Revanchist City. New York:
Routledge.
US Census Bureau (2013) American Community
Survey, 2005–2009. Available at: http://factfin
der2.census.gov/faces/nav/jsf/pages/index.xhtml
(accessed 30 April 2013).
Wyly EK and Hammel DJ (1998) Modeling the
context and contingency of gentrification.
Journal of Urban Affairs 20: 303–326.
Wyly EK and Hammel DJ (1999) Islands of decay
in seas of renewal: Housing policy and the
resurgence of gentrification. Housing Policy
Debate 10: 711–771.
Zukin S, Trujillo V, Frase P, et al. (2009) New
retail capital and neighborhood change: Bou-
tiques and gentrification in New York City.
City & Community 8: 47–64.
Barton 111
at LSU Libraries on December 15, 2015usj.sagepub.comDownloaded from
