Professional Documents
Culture Documents
For SocArXIv
For SocArXIv
Spatial analysis of high school dropout: The role of race, poverty, and
outliers in New York City
1
Departments of Sociology and Comparative Human Development, The University of Chicago
(Chicago, IL); and Institute for the Science and Art of Learning and Teaching, Ateneo de Manila
University (Quezon City, Philippines).
His research intersects organizational sociology, education policy, and quantitative methods. In
particular, he researches how schools organize data and how data organize schools.
Spatial analysis of high school dropout: The role of race, poverty, and
outliers in New York City
ABSTRACT
A major problem in urban areas, dropping out of high school is often related to a
host of student-level factors, like poverty, race, gender, and behavior. Studies
also find that social factors like school context and peer beliefs influence this
phenomenon. However, these studies on social factors are limited by the non-
inclusion of spatial data, which are important given how neighborhood dropout
rates may be associated with adjacent neighborhoods. To address this gap, the
present research suggests a method for analyzing spatial patterns of dropping out
and the role of race, poverty, and outliers. Using exploratory spatial data analysis
of New York City census tracts (n = 2216), the study finds significant spatial
clusters for high school dropouts, racial minority population, and low-income
areas. However, co-location patterns of dropping out and race/poverty suggest
smaller core clusters, and outlier-detection for areas with dropout rates different
from adjacent neighborhoods suggest that higher concentrations of minorities and
poverty do not automatically lead to higher levels of dropout. Implications
include resource allocation and contextualized support for high dropout clusters,
and investigation of outlier neighborhoods.
Keywords: high school dropout; spatial data analysis; New York City; urban
education
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 3
Spatial analysis of high school dropout: The role of race, poverty, and
outliers in New York City
Introduction
education and educational credentials cannot be overstated. Studies show that those
with more advanced educational credentials have better long-term outcomes, based on
wages, earnings, health, and family outcomes (Barrow & Malamud, 2015; Ashenfelter
& Krueger, 1994). However, a constant challenge that has plagued the education system
is the substantial number of students who drop out of high school (Heckman &
LaFontaine, 2010; Orfield, 2004). For this reason, many countries, states, and districts
have instituted ways of helping reduce dropout rates through interventions that focus on
address this problem, however, part of the task is to be aware of risk factors that
increase the likelihood of students’ dropping out, and these may include factors in
personal and school-contextual factors that impact it (T. M. Brown et al., 2019; Wood et
al., 2017; Balkis et al., 2016). In terms of personal factors, certain populations have
been documented to be more likely to dropout, like Blacks and Hispanics (Chapman et
al., 2011), male students (Schulz & Rubel, 2011), those from low-income households
(Kearney & Levine, 2014), and those with mental health or behavioral problems
schools may be associated with high suspension rates (T. Lee et al., 2011; Trinidad,
However, a set of factors that are not as readily studied in the dropout literature
is that of neighborhood-spatial factors that may also affect dropout frequency. One
example is the lack or the presence of economic prospects in a district and how this
affects a student’s decision of finishing high school (Bickel & Papagiannis, 1988). In a
schemas of finishing or foregoing one’s education (Schafer & Hori, 2006). A third
such as deviant and criminal behavior often happening in neighborhoods (Rumberger &
Rotermund, 2012). Given the importance of accounting for spatial patterns and reasons
for dropping out, the present research proposes spatial data analytic procedures to
understand these patterns, interpret the dropout results, and suggest practical
implications.
understanding of dropout patterns by (1) suggesting spatial clusters for dropping out, (2)
providing co-location patterns with explanatory variables, and (3) detecting outliers that
can provide insights into what neighborhoods can do to lessen dropout occurrence. To
do this, I first review the literature on dropout incidence, particularly factors that
contribute to students’ deciding to drop out, or being pushed to drop out. These are
that increase likelihood of high school dropout. The second part outlines the methods
for this research, with an emphasis on the analytic techniques used. The third part
provides the results while the fourth part discusses, in particular, the results of dropout
patterns in New York City, and in general, how to use spatial data analysis with
education data.
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 5
Literature Review
factors that span demographic, socioeconomic, and behavioral reasons. Among the most
frequently studied predictors is that of minority race and ethnicity. High school dropout
rates for Blacks and Hispanics are substantially higher when compared to Whites and
Asians (Belfield & Levin, 2007; Heckman & LaFontaine, 2010; Katsinas, 1989). For
Blacks, one aspect that is often attributed is their higher likelihood of experiencing
have implications on their academic performance (Suh et al., 2014; Kinsler, 2013;
Caton, 2012; Rumberger, 2011). For Hispanics, reasons for dropping out include
difficulties in school, personal problems at home, and the need to economically support
their families (Behnke et al., 2010; Halx & Ortiz, 2011). In particular, immigrant
and language difficulties (Lutz, 2007). For both Blacks and Hispanics, the person’s
In addition to race, gender is also associated with high school dropout. The
graduation rate for females is significantly higher than males in all major racial/ethnic
groups, but particularly for Blacks whose differential is 12.2 percentage points among
20-24 year-olds in 2010 (Murnane, 2013). Although this may seem like an odd new
pattern, Goldin and Katz (2007) show that throughout most of the 20th century more
Research also finds that the family’s socioeconomic position can influence high
school persistence or dropout. A report from the National Center for Education
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 6
Statistics accounts that students living in low-income families were approximately ten
times more likely to drop out of high school as compared to those living in high-income
families (Cataldi et al., 2007). Other studies confirm how family background and lower
socioeconomic status can increase one’s chance of dropping out (Archambault et al.,
2017; Parr & Bonitz, 2015). Often, lack of resources, low parental achievement, and
reduced or levelled expectations also plague these students from lower-income families
(Hao & Yeung, 2015; Trinidad, 2019; Davis-Kean, 2005; MacLeod, 1987; Vowell &
May, 2000).
factors are also associated with increased likelihood of dropping out. A student’s feeling
addition to serious problematic behaviors later in life (Henry et al., 2012). Students’
personal motivation and capacity for self-regulation have important consequences for
predicting academic failure in high school (Casillas et al., 2012). In addition to feelings
and non-cognitive skills necessary for academic success, students’ self-perceptions are
also important, particularly in terms of how much they perceive their efforts will have
an impact on their outcomes (Fall & Roberts, 2012). These research examples highlight
may not always be a completely autonomous act. Contrary to popular thinking that
students decide or are forced to drop out of school because of academic or discipline
failure, there are actually a myriad of reasons for dropping out. Some students are
pushed out of schools because of poor academic performance and behavioral problems
while others are pulled out from school because of pregnancy, the need for a job, or the
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 7
need to care for someone in the family (Doll et al., 2013; Bradley & Renzulli, 2011).
Thus, factors for dropping out are nuanced by the variety in the meaning of dropout.
certain school factors have also been shown to exacerbate the impact of the individual
risk factors for dropping out. Lee and Burkam (2003) find that certain school
organizational factors are associated with reduced dropout rates. These include schools
offering mainly academic rather than nonacademic courses, enrolling fewer than 1,500
students, and having students and teachers with positive relationships (V. E. Lee &
(2013) show that in addition to these school organizational factors, teaching quality and
social capital (i.e., strength of relationships between teachers and students) also have an
certain school factors like feelings of attachment are associated with decreased dropout,
“above and beyond individual characteristics” (Kotok et al., 2016, p. 569). Students’
positive social integration with other students plays a crucial role in preventing them
from feeling alienated and withdrawing academically (Hascher & Hagenauer, 2010).
associated with better educational outcomes (Trinidad, 2018a). On the other end,
negative forms of student interactions, like the perception of teasing or bullying, can
predict increased dropout rates in schools (Cornell et al., 2013). Such hostile school
environments can lead to self-doubt and emotional scars, which are associated with
pursue or drop out of school, supportive relationships with significant adult models, like
teachers and counselors, can also have protective consequences for students (Lessard et
al., 2010; White & Kelly, 2010). On average, schools experience lower dropout rates
when they have an authoritative school climate, or a climate that provides students with
both high demands and high adult supports (Jia et al., 2016). However, schools that
dropping out (Peguero & Bracy, 2015; Pyne, 2019). It is important to understand that
these individual- and school-level factors interact, but do not automatically determine, a
phenomenon, a set of factors that are not as readily studied are neighborhood factors,
safety. More than just the family’s economic situation, the socioeconomic advantage in
a neighborhood also influences the likelihood of graduating from college (Crowder &
South, 2011). Studies find that continued exposure to neighborhood disadvantage has
even more salient effects than just disadvantage experienced at a singular point in time
(Wodtke et al., 2011; Crowder & South, 2011). However, some researchers suggest that
in low-resource settings may not support educational attainment and may be poor role
models to guide and encourage residents to complete school” (Fischer & Kmec, 2004,
p. 520).
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 9
Another set of reasons point to how neighborhood violence and safety also
influence dropout trends. Despite controlling individual, family, and school factors,
high school dropout (Harding, 2009). In a qualitative research among male Latinos in
Los Angeles, Rendón (2014a) finds that urban violence is one of the most salient
so much that violence happens to them but that one’s involvement in peer groups can
and expulsion (Rendón, 2014a). More than the impact of violence on students’
students’ engagement, even for those who are not directly affected by such problems
influencing high school dropout, few published research look at the spatial dynamics
that affect such dropout trends. Among these few, a study of rural Louisiana schools
find mechanisms beyond school-level factors that suggest how wider clusters or
communities can influence dropout figures (Schafer & Hori, 2006). Aside from the fact
that factors beyond the school may affect dropout, using spatial analytical tools may
also capture important, nuanced, and clustered contextual variations that are not
apparent in global analyses of all locations (Tate & Hogrebe, 2015; Edmunds et al.,
may have different outcomes, resources, and historical dropout or poverty “legacies”
This research looks at the spatial patterns of dropouts in New York City, the
largest school district in the United States (Winters & Cowen, 2012). Although the city
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 10
is not representative of urban areas in the country or the world, it does provide an
interesting locus for seeing the impact of spatial factors on dropping out. Given the
city’s large student population and diverse neighborhoods and neighborhood clusters
(Chau & Yager, 2017), demographic and locational patterns can be more carefully
year 2000, prepared by the University of Chicago’s Center for Spatial Data Science.
This dataset had the percentage of the population in a given census tract (alternatively
referred to as neighborhood) meeting a certain criterion. For example, the data had the
percentage of people in a neighborhood who identify themselves as White (or any other
racial/ethnic group). The dataset had 2216 observations, measured as census tracts.
Measures
The dependent variable was the percentage of people over the age of 25 in a particular
census tract that have dropped out of high school. Although it is possible to compute for
the percentage of students ages 16 to 19 who have dropped out of high school, this
variable had a lot of zero values either because no one has yet dropped out in that
neighborhood or because there were not that many young people in the census tract.
Thus, the variable used is the percentage of dropout over age 25.
Two sets of independent variables were crucial for this research. For the
variable for the percentage of racial minorities living in a census tract and a categorical
variable for the majority race in a census tract. The first is the percentage of minority
population in a given tract (i.e., the percentage of the population who were not White).
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 11
The second is a categorical variable where the census tract is designated with the
racial/ethnic category of the majority of its residents. Five groups were created with:
majority White, majority Black, majority Hispanic, majority Black and Hispanic, and
majority mixed. A majority was defined as greater than 50% of the residents identifying
in the particular racial category. However, for those areas where the majority of the
residents were either Black or Hispanic but neither made up the majority, they were
designated as majority Black and Hispanic. Additionally, given the few number of
tracts that have majority Asian and few that have heterogeneously mixed populations,
these have been jointly designed as majority mixed for analytic purposes.
socioeconomic status had also been estimated with the mean of the adult residents’
annual income. Data were kept as US dollar values rather than log values so as to
Data analysis
I used Exploratory Spatial Data Analysis (ESDA) to measure and detect high school
dropout, racial, and economic patterns in New York City. ESDA is a first step into
see if there were spatial components and predictors on the likelihood of people’s
dropping out, several geovisualization approaches were used, like the Local Indicator of
Spatial Association (LISA) clusters, quartile LISA maps, and disaggregated descriptive
statistics. Visualization analyses were done in GeoDa, an open source software for
spatial data analysis (Anselin et al., 2006). To provide a snapshot of the distribution of
the variables, I first mapped the census tracts into color-coded quartiles with darker
shows the location of the five boroughs of New York City (Manhattan, Staten Island,
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 12
Queens, Brooklyn, and Bronx) since I will refer to these boroughs when discussing
populations.
To examine spatial clusters of dropping out, I used LISA maps that mapped out
the adjacent areas that have similar rates for a variable. In this study, the analysis
identified dropout hot spots and cold spots in New York City by testing for statistically
significant associations of dropouts between each census tract and its neighboring tracts.
A technical note is that I used the spatial weights matrix (queen contiguity) to identify a
tract and its contiguous tracts since it may be possible that certain dropout patterns in
Clusters of tracts with higher and lower dropout rates were identified at a
statistical significance level of = 0.05. These correspond to “high-high” (in red) and
“low-low” (in blue) LISA statistics. “High-high” areas have dropout rates above the
mean and a weighted average of the neighboring areas that are also above the mean.
Similarly, “low-low” areas have dropout rates below the mean with contiguous areas
with also low dropout rates. “High-low” and “low-high” LISA outliers were colored
light red and light blue, respectively. These outliers mean that an area had dropout rates
above the mean while surrounded by neighborhoods with low dropout rates (high-low),
and vice versa. Sections not colored do not form part of the core of the clusters that
Aside from doing LISA maps for dropout rates among contiguous census tracts,
similar maps were created for racial minority rates and mean income. Clusters of “high-
high” and “low-low” census tracts provide a picture of which contiguous sections of the
independent variables, I created quartile LISA maps (Anselin, 2019). Using this
method, I assigned the value 1 for tracts in the highest or lowest quartiles of a given
for both dependent and independent variables. For example, I looked for clustered
locations where the highest quartiles of dropout rates and racial minority populations
were present.
Descriptive statistics of cluster cores (dropout hot spots and cold spots) were
calculated to identify characteristics that are shared among clustered census tracts that
have high and low dropout rates. Disaggregating the clusters with high-high and low-
low LISA statistics, I calculated the percentage of census tracts, according to their
majority racial/ethnic composition, mean income, and total number of residents. I also
performed the same demographic analysis for outliers that have high-low and low-high
LISA statistics.
Results
Dropout clusters
Figure 1A presents a map of New York City that have color-coded quartiles, signifying
the areas which have higher or lower dropout percentages in the given census blocks.
Higher dropout percentages (darker shade of orange) were concentrated in the Bronx
and Brooklyn, and lower ones in Manhattan, Queens, and Staten Island.
seen in the Local Moran cluster map in Figure 1B (using queen contiguity, 999
permutations, p < 0.05). In all, 381 tracts were identified as the cores of high-high
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 14
and 454 tracts as cores of low-low clusters. These hot spots in red and cold spots in blue
confirm the clusters in Figure 1A with high and low dropout rates.
However, there were also neighborhoods near these dropout hot spots and cold
spots that did not follow the trend in their contiguous clusters. For example, although
Manhattan had predominantly low dropout rates, some neighborhoods actually had high
dropout rates as depicted in the light red color. This was thus coded an outlier for
having high dropout rates in a location with predominantly low dropout rates. Similarly,
there were neighborhoods in the Bronx, colored light blue, that had low dropout
Brooklyn, the southern side of Queens, and the Bronx had high concentrations of racial
minorities denoted by its stronger shade of orange. For the mean income, Figure 2B
shows an inversion of the previous trends with Manhattan, Staten Island, and Queens
Queens, it has a significantly higher income average when compared to Brooklyn and
the Bronx.
Formal analyses of spatial autocorrelation again show that there are significant
spatial clusters for racial minority populations and income distributions. Figure 3A
shows the Bronx, Brooklyn, and the southern part of Queens all having significantly
more clusters of minority populations while Manhattan, Staten Island, and southern
When compared with Figure 3B, one can see a slight inversion with the Bronx
and Brooklyn (in blue) having lower average income in their clusters while Manhattan
and Staten Island (in red) having higher average income. One will notice, however, that
the inversion is not perfect and that there are far less hot spots and cold spots in Figure
3B than Figure 3A. Both of these observations may denote that average income is not as
minority population rates, respectively. Although there were overlaps, formal analysis
in Figure 4C shows that the binary variables for high dropout rates and high minority
population rates have co-location core clusters of 76 census tracts concentrated in areas
in Brooklyn and the Bronx. However, these co-location patterns in Figure 4C do not
As shown earlier, the high minority concentration in the Queens area did not
seem to have high dropout rates such that the co-location map in Figure 4C did not
show any statistically significant overlap between minority population and dropout rate
A similar analysis was done for co-location of the highest quartile for
neighborhood dropout rates, and the lowest quartile for neighborhood mean income.
Figure 4D reveals 222 locations that form the core clusters (in the Bronx and Brooklyn
boroughs). The number and size of these clusters are markedly larger than the clusters
those clustered census tracts that have high dropout rates and low dropout rates. The
census tracts with high dropout rates were in neighborhoods that were majority Black
and Hispanic (39.95%), majority Black (23.76%), and majority Hispanic (18.02%).
Most of these tracts (96.84%) had mean annual income of less than $40,000. On the
other hand, the clustered census tracts with low dropout rates were in neighborhoods
that were predominantly White (87.72%), and with mean annual income above $40,000
(91.43%). Total residents do not vary as much between the two groups.
When comparing all census tracts with those with high dropout rates, majority
Black neighborhoods comprised 25.98% of census tracts and they accounted for 23.76%
5.21% of all tracts but figured as 18.02% of all high dropout census tracts.
Outliers
Aside from investigating clusters with high and low dropout rates, Table 1 also presents
came from the investigation with Figure 1B, which includes outliers in light red and
light blue colors. Neighborhoods with low dropout rates but have neighboring tracts
with high dropouts (i.e., low-high tracts in light blue) were 31.03% majority White
their mean annual income ($24,364) was similar to the average in high dropout tracts
($22,948).
Neighborhoods with high dropout rates surrounded by tracts with low dropout
rates (i.e., high-low tracts in light red) were 38.89% majority White, 22.22% majority
Black, and 22.22% majority mixed neighborhoods. Interestingly still, their mean annual
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 17
income ($30,879) was actually higher than the average income for those with low-high
tracts ($24,364).
In a society as dependent on knowledge and literacy as the present one, high school
out is a problem that needs to be overcome (Bridgeland et al., 2009). Education scholars
have studied what factors could affect it and found important consequences of personal,
social, and school factors (Rumberger, 2011; Cataldi et al., 2007). More than just these
singular factors, one can find evidence for even more negative effects when there is an
2018b; Lessard et al., 2010). An added set of factors not often studied is an individual’s
neighborhood.
decisions to drop out or stay in school (Doll et al., 2013; Rendón, 2014b). This present
factors. I argue that geospatial cluster visualization and analysis can provide significant
insights into understanding dropout patterns, and that these methods also suggest
Using dropout and demographic data from New York City, I find that dropouts
did not just happen randomly in the city. There were particular places considered
dropout hot spots, where clusters of neighborhoods had more than 35% of their adult
population which have dropped out of high school. Often concentrated in the Bronx and
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 18
upper Brooklyn areas, such clusters suggest possible shared characteristics among these
neighborhoods.
factors are correlated with such dropout statistics. Although research has shown the
socioeconomic status (Archambault et al., 2017; Wood et al., 2017), the present
research suggests that the concentration of racial minority population in an area may
also affect the level of dropout. In particular, majority Hispanic, and majority Black and
Hispanic neighborhoods have figured prominently in these dropout hot spots. Although
the present research cannot test it empirically, one possible reason is the migration of
Hispanic populations to certain clusters and pockets of the city, and the possibility that
they have been unable to finish high school education (Card & Lewis, 2005; Perreira et
opportunities, and thus affect their decision-making (Dupéré et al., 2015; Harding,
2011).
does not necessarily mean that these types of neighborhood clusters automatically have
more dropouts. For example, certain sections in Queens have high minority populations
but do not have the same the high dropout rates as in the Bronx and Brooklyn. Co-
location maps also showed that only 76 out of 2216 census tracts form the core cluster
of high minority and high dropout neighborhoods. For co-location of low income and
high dropout, the number of tracts just increase to 222. These numbers suggest that
although high-minority and low-income neighborhoods are more likely to have higher
evidenced by how the co-location of extreme values for dropout and minority/low-
income neighborhoods also have outliers in terms of dropout percentages; that is, there
are areas which have less dropouts compared to their contiguous census tracts. For
example, certain areas in the Bronx have dropout levels that are no different from that
generalizing how race/ethnicity and poverty can affect neighborhood dropout rate, and
will have to qualitatively investigate why some adjacent neighborhoods with almost the
same socioeconomic and race characteristics have markedly different dropout rates. It is
can have protective effects on their students (Dobbie & Fryer, 2011). Another
possibility is in terms of the social capital among neighbors who are more closely knit
and how their shared experience and expectation can bring about better educational
Although the data concentrated in New York City, the method used here can be
applied to other urban areas. More than simply determining which schools have higher
dropout rates, it may be necessary to see the spatial dimensions and characteristics of
dropout rates. Particularly in urban areas and in neighborhoods with more students at
risk of dropping out, high quality schools can have protective factors and compensatory
effects (Downey & Condron, 2016). In this sense, geospatial analyses can help
determine where the dropout concentrations are, and this determination can lead to
allocating resources for schools that can intervene on the pattern of dropout in an area.
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 20
Methodological contribution
In this research, I highlight the contribution and added value of performing geospatial
analyses of dropout data through (a) quartile, (b) clustered LISA and (c) co-location
maps, and the demographic analysis of these clusters. Using quartile maps can provide
insights into which neighborhoods have high rates of dropout occurrence. Clustered
LISA maps show contiguous locations that have similar high or low dropout rates; this
determination of dropout hot spots or cold spots can lead to greater specificity and
These clustered maps also show outliers in terms of neighborhoods with low dropout
rates when the surrounding ones have high dropout rates, and vice versa. Determining
outliers can prompt qualitative researchers to determine what could distinguish certain
areas from their neighbors. For an area with low dropout rate in a high dropout region,
could certain school factors, like teacher quality and student discipline, be protective
Aside from quartile and LISA maps, I also used Anselin’s (2019) novel
similar quartile levels on the outcome and explanatory variables (e.g., dropout rates and
minority population). Such mapping technique can offer insights into which adjacent
areas have similar characteristics in at least two variables, and how this offers
characteristics of those in high-high, low-low, and outlier clusters (Kolak et al., 2018).
still carries with it some limitations. First, data availability will always pose a problem
for geospatial research and this study is no exception. The use of census data from the
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 21
year 2000 may be unappealing but I used it because of its relative completeness when
compared to other non-census survey data. Moreover, using the percentage of the
population over 25 years old who dropped out includes those who were 26 and 80 years
old, who may have had differences in contexts, reasons for dropping out, and prior
neighborhoods they came from. Second, I was unable to investigate other variables that
may shed more light on and nuance spatially different dropout rates; these possible
variables include crime prevalence, social cohesion, job opportunities, and school
quality. Given that this is an initial attempt at using geospatial analysis for dropout data,
such limitation is an invitation for other researchers to extend the understanding of how
these variables can be spatially clustered and explain spatial dropout patterns. Third,
this present study confines itself to exploratory spatial data analysis and does not yet use
spatial regression analysis (Chi & Zhu, 2008; Anselin, 2007). More than ordinary least
random spatially clustered variables, which may further the insights from the
exploratory spatial data analysis. Such limitations, however, can lead to more research
to be done to fully flesh out the possibilities of doing spatial data analysis with
education data.
Despite the limitations, the research offers both conceptual insights to help
nuance the literature on spatial neighborhood factors affecting dropout rates and
education issues. Conceptually, evidence supports the claim that New York City has
Although such high dropouts are associated with high-minority and low-income
neighborhoods, these are not deterministic. When upper-quartile minority and poverty
variables are intersected with dropout variables, the core clusters are relatively small. In
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 22
addition, outlier detection showed that some high-minority low-income areas can have
Methodologically, the study suggests and encourages the use of exploratory spatial data
analysis techniques that are appropriate for educational issues and problems. Such
resource allocation.
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 23
References
Anselin, L. (2007). Spatial Regression Analysis in R: A Workbook. Center for Spatially
Anselin, L., Syabri, I., & Kho, Y. (2006). GeoDa: An Introduction to Spatial Data
7363.2005.00671.x
Archambault, I., Janosz, M., Dupéré, V., Brault, M.-C., & Andrew, M. M. (2017).
Individual, social, and family factors associated with high school dropout among
https://doi.org/10.1111/bjep.12159
Ashenfelter, O., & Krueger, A. (1994). Estimates of the Economic Return to Schooling
from a New Sample of Twins. The American Economic Review, 84(5), 1157–
1173.
Balkis, M., Arslan, G., & Duru, E. (2016). The School Absenteeism among High
Behnke, A. O., Gonzalez, L. M., & Cox, R. B. (2010). Latino Students in New Arrival
States: Factors and Services to Prevent Youth From Dropping Out. Hispanic
https://doi.org/10.1177/0739986310374025
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 24
Belfield, C. R., & Levin, H. M. (2007). The Educationa Attainment Gap: Who’s
content/uploads/2016/07/pricewepay_chapter.pdf
Bickel, R., & Papagiannis, G. (1988). Post-High School Prospects and District-Level
Pushed or Pulled to Drop Out of High School. Social Forces, 90(2), 521–545.
https://doi.org/10.1093/sf/sor003
Bridgeland, J. M., Dilulio, J. J. J., & Balfanz, R. (2009). The High School Dropout
26.
Brown, B. A. (2010). Social hostility and the “dropout” syndrome: Leadership assisting
https://doi.org/10.1080/00131910903469577
Brown, T. M., Galindo, C., Quarles, B., & Cook, A. L. J. (2019). Self-Efficacy, Dropout
Status, and the Role of In-School Experiences Among Urban, Young Adult
https://doi.org/10.1007/s11256-019-00508-3
Bryan, J., Farmer-Hinton, R., Rawls, A., & Woods, C. S. (2017). Social Capital and
Card, D., & Lewis, E. G. (2005). The Diffusion of Mexican Immigrants During the
http://www.nber.org/papers/w11552.pdf
Casillas, A., Robbins, S., Allen, J., Kuo, Y.-L., Hanson, M. A., & Schmeiser, C. (2012).
Cataldi, E. F., Laird, J., & KewalRamani, A. (2007). High School Dropout and
Completion Rates in the United States: 2007. National Center for Education
https://doi.org/10.1177/0042085912454442
Chapman, C., Laird, J., Ifill, N., & KewalRamani, A. (2011). Trends in High School
https://eric.ed.gov/?id=ED524955
Chau, V., & Yager, J. (2017). Zoning for Affordability: Using the Case of New York to
Chi, G., & Zhu, J. (2008). Spatial Regression Models for Demographic Analysis.
https://doi.org/10.1007/s11113-007-9051-8
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 26
Clewell, B. C., Campbell, P. B., & Perlman, L. (2007). Good Schools in Poor
Cornell, D., Gregory, A., Huang, F., & Fan, X. (2013). Perceived prevalence of teasing
Crowder, K., & South, S. J. (2011). Spatial and temporal dimensions of neighborhood
https://doi.org/10.1016/j.ssresearch.2010.04.013
Daly, B. P., Shin, R. Q., Thakral, C., Selders, M., & Vera, E. (2009). School
Child Achievement: The Indirect Role of Parental Expectations and the Home
https://doi.org/10.1037/0893-3200.19.2.294
De Witte, K., Cabus, S., Thyssen, G., Groot, W., & van den Brink, H. M. (2013). A
Dobbie, W., & Fryer, R. G., Jr. (2011). Are High-Quality Schools Enough to Increase
Achievement among the Poor? Evidence from the Harlem Children’s Zone.
Doll, J. J., Eslami, Z., & Walters, L. (2013). Understanding Why Students Drop Out of
High School, According to Their Own Reports: Are They Pushed or Pulled, or
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 27
Downey, D. B., & Condron, D. J. (2016). Fifty Years since the Coleman Report:
Dupéré, V., Leventhal, T., Dion, E., Crosnoe, R., Archambault, I., & Janosz, M. (2015).
Stressors and Turning Points in High School and Dropout: A Stress Process,
https://doi.org/10.3102/0034654314559845
https://doi.org/10.1007/s11256-014-0285-6
Fall, A.-M., & Roberts, G. (2012). High school dropouts: Interactions between social
https://doi.org/10.1525/sop.2004.47.4.507
Freeman, J., & Simonsen, B. (2015). Examining the Impact of Policy and Practice
205–248. https://doi.org/10.3102/0034654314554431
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 28
Goldin, C., & Katz, L. F. (2007). The Race between Education and Technology: The
http://www.nber.org/papers/w12984
Halx, M. D., & Ortiz, M. (2011). Voices of Latino male high school students on their
Hao, L., & Yeung, W.-J. J. (2015). Parental Spending on School-Age Children:
https://doi.org/10.1353/sof.0.0281
https://doi.org/10.1177/0038040711417008
Hascher, T., & Hagenauer, G. (2010). Alienation from school. International Journal of
https://doi.org/10.1016/j.ijer.2011.03.002
Heckman, J. J., & LaFontaine, P. A. (2010). The American High School Graduation
Rate: Trends and Levels. The Review of Economics and Statistics, 92(2), 244–
262. https://doi.org/10.1162/rest.2010.12366
156–166. https://doi.org/10.1007/s10964-011-9665-3
Jia, Y., Konold, T. R., & Cornell, D. (2016). Authoritative school climate and high
https://doi.org/10.1037/spq0000139
https://doi.org/10.1007/BF01108461
Kearney, M. S., & Levine, P. B. (2014). Income Inequality, Social Mobility, and the
Decision to Drop Out of High School (NBER Working Paper No. 20195).
Kinsler, J. (2013). School Discipline: A Source or Salve for the Racial Achievement
https://doi.org/10.1111/%28ISSN%291468-2354/issues
Kolak, M., Bradley, M., Block, D. R., Pool, L., Garg, G., Toman, C. K., Boatright, K.,
Lipiszko, D., Koschinsky, J., Kershaw, K., Carnethon, M., Isakova, T., & Wolf,
https://doi.org/10.1016/j.healthplace.2018.06.003
Kotok, S., Ikoma, S., & Bodovski, K. (2016). School Climate and Dropping Out of
569–599. https://doi.org/10.1086/687275
Lee, T., Cornell, D., Gregory, A., & Fan, X. (2011). High Suspension Schools and
Dropout Rates for Black and White Students. Education and Treatment of
Children, 2, 167–192.
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 30
Lee, V. E., & Burkam, D. T. (2003). Dropping Out of High School: The Role of School
353–393. https://doi.org/10.3102/00028312040002353
Lessard, A., Poirier, M., & Fortin, L. (2010). Student-teacher relationship: A protective
factor against school dropout? Procedia - Social and Behavioral Sciences, 2(2),
1636–1643. https://doi.org/10.1016/j.sbspro.2010.03.250
Li, G. (2010). Race, Class, and Schooling: Multicultural Families Doing the Hard Work
of Home Literacy in America’s Inner City. Reading & Writing Quarterly, 26(2),
140–165. https://doi.org/10.1080/10573560903547452
018-9458-4
Maynard, B. R., Salas-Wright, C. P., & Vaughn, M. G. (2015). High School Dropouts
https://doi.org/10.1007/s10597-014-9760-5
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 31
Murnane, R. J. (2013). U.S. High School Graduation Rates: Patterns and Explanations.
https://doi.org/10.1257/jel.51.2.370
Parr, A. K., & Bonitz, V. S. (2015). Role of Family Background, Student Behaviors,
https://doi.org/10.1080/00220671.2014.917256
Peguero, A. A., & Bracy, N. L. (2015). School Order, Justice, and Education: Climate,
Perreira, K. M., Harris, K. M., & Lee, D. (2006). Making it in America: High school
https://doi.org/10.1353/dem.2006.0026
https://doi.org/10.1177/0038040718816684
Rendón, M. G. (2014a). “Caught Up”: How Urban Violence and Peer Ties Contribute to
https://doi.org/10.1525/sp.2013.11237
Rendón, M. G. (2014b). Drop Out and “Disconnected” Young Adults: Examining the
Impact of Neighborhood and School Contexts. The Urban Review, 46(2), 169–
196. https://doi.org/10.1007/s11256-013-0251-8
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 32
Rumberger, R. W. (2011). Dropping out: Why students drop out of high school and
https://doi.org/10.1007/978-1-4614-2018-7_24
Schafer, M. J., & Hori, M. (2006). The Spatial Dynamics of High School Dropout: The
Suh, S., Malchow, A., & Suh, J. (2014). Why Did The Black-White Dropout Gap
Tavakolian, H., & Howell, N. (2012). Dropout Dilemma and Interventions. Global
https://doi.org/10.1177/0013124518785444
https://doi.org/10.1016/j.stueduc.2019.01.004
Vowell, P. R., & May, D. C. (2000). Another Look at Classic Strain Theory: Poverty
https://doi.org/10.1111/j.1475-682X.2000.tb00895.x
White, S. W., & Kelly, F. D. (2010). The School Counselor’s Role in School Dropout
https://doi.org/10.1002/j.1556-6678.2010.tb00014.x
Wilson, W. J. (2012). The Truly Disadvantaged: The Inner City, the Underclass, and
Winters, M. A., & Cowen, J. M. (2012). Grading New York: Accountability and
https://doi.org/10.3102/0162373712440039
Wodtke, G. T., Harding, D. J., & Elwert, F. (2011). Neighborhood Effects in Temporal
https://doi.org/10.1177/0003122411420816
Wood, L., Kiperman, S., Esch, R. C., Leroux, A. J., & Truscott, S. D. (2017). Predicting
Figure 1. Dropout Rates in New York City, 2000. (A) Map of New York City that color
codes quartiles of the dropout rates for people above 25 years old. (B) LISA map that
demonstrates contiguous census tracts with significantly high (dark red) or low (dark
blue) dropout rates, or their outliers (light red or blue). Only spatial cluster cores at p <
0.05 are shown.
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 35
Figure 2. Minority Percentage and Income Averages in New York City, 2000. (A) Map
that color codes quartiles of the percentage of minority populations in different census
tracts. (B) Map that color codes quartiles of the mean income in the census tract.
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 36
Figure 3. Minority Percentage and Income Averages LISA Maps of New York City, 2000.
(A) LISA map that demonstrates census tracts with significantly high (dark red) or low
(dark blue) minority populations, or their outliers (light red or blue). (B) LISA map that
demonstrates census tracts with significantly high or low average incomes, or their
outliers. Only spatial cluster cores at p < 0.05 are shown.
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 37
Figure 4. Co-Location Maps for Dropouts and Minority Race/Average Income. (A) Upper
quartile of dropout rate. (B) Upper quartile of minority population rate in a census tract.
(C) Co-location map for clusters with high dropout rates and high-minority populations.
(D) Co-location map for clusters with high dropout rates and low average income per
census tract.
SPATIAL ANALYSIS OF HIGH SCHOOL DROPOUT 38