Urban Studies Is neighbourhood destiny? Exploring Urban Studies … · Is neighbourhood destiny? Exploring the link between neighbourhood mobility and student outcomes Sarah A Cordes

Article

Urban Studies1–18� Urban Studies Journal Limited 2015Reprints and permissions:sagepub.co.uk/journalsPermissions.navDOI: 10.1177/0042098014563469usj.sagepub.com

Is neighbourhood destiny? Exploringthe link between neighbourhoodmobility and student outcomes

Sarah A CordesNew York University, USA

Amy Ellen SchwartzNew York University, USA

Leanna StiefelNew York University, USA

Jeffrey ZabelTufts University, Economics, USA

AbstractThe notion that children from ‘good’ neighbourhoods are destined for success while those from‘bad’ neighbourhoods are destined for failure has considerable popular appeal. Residential loca-tion is strongly linked to school quality, access to educated adults, exposure to violence, etc.There is, however, surprisingly little evidence on the link between the neighbourhood in which achild begins school and later schooling outcomes. Understanding early neighbourhood experi-ences is important for determining whether students are ‘stuck’ in neighbourhoods of disadvan-tage. It is also critical for determining the extent to which students who begin their schoolingcareers in disadvantaged neighbourhoods are destined for poor schooling outcomes, and conver-sely, whether changing neighbourhood context improves student performance. In this study,therefore, we document how students’ early neighbourhood and schooling experiences arerelated to later success in school, and explore how changing neighbourhood and school contextsexplain differences in academic outcomes. Using data from New York City (NYC), we constructa panel containing all students enrolled as first graders in NYC public schools in 1996–1997, fol-lowing them through academic years 2007–2008, which would be their 12th grade year if theymade standard academic progress (annual one-grade promotion). Far from supporting the sim-plistic story of ‘dead-end’ neighbourhoods, our analyses describe a situation where students frompoor neighbourhoods actually move more often than their peers in less disadvantaged

Corresponding author:

Sarah Cordes, New York University, Wagner Graduate School of Public Service, 665 Broadway, Suite 805, NY 10012, USA.

Email: [email protected]

at Bobst Library, New York University on May 13, 2015usj.sagepub.comDownloaded from

http://usj.sagepub.com/

neighbourhoods and are more likely to experience changes in neighbourhood and school quality,with 45.7% of neighbourhood moves from the poorest neighbourhoods being made to signifi-cantly higher quality neighbourhoods.

Keywordsmobility, neighbourhood change, neighbourhood quality, school quality, student outcomes

Received March 2014; accepted November 2014

Introduction

In 2011, New Jersey Governor ChrisChristie voiced a commonly held beliefabout the role of neighbourhoods in the lifecourse of children: ‘How do we change thissystem so that destiny is not determined byzip code .?’ (Christie, 2011). The notionthat children born into ‘good’ neighbour-hoods are destined for success while thoseborn into ‘bad’ neighbourhoods are destinedfor failure has considerable intuitive appeal.Residential location is strongly linked toschool quality, access to educated adult rolemodels, exposure to violence and crime, etc.There is, however, surprisingly little evidenceon the link between origin neighbourhoods– that is, the neighbourhood in which a childbegins school – and later schooling out-comes. Does origin neighbourhood (orschool) matter in the long run? Does theeffect disappear with residential (or school)moves? That is, is it possible to move toopportunity and thereby improve lateroutcomes?

Considerable research examines theimportance of neighbourhood and school,and, separately, the impact of early child-hood experiences on academic performance.Little work, however, considers thesetogether, parsing the contribution of earlyneighbourhood and schools, so there is littleevidence on how origin neighbourhoodmatters in the long run. We begin to fill thisgap, using rich longitudinal data on NYCpublic school students, which allows us to

track their progress and mobility from firstgrade through high school. We explore therelationships between origin neighbourhoodand schools, subsequent residential andschool moves, and, ultimately, performancein school.

Understanding early neighbourhoodexperiences is important for determiningwhether students are ‘stuck’ in neighbour-hoods of disadvantage. Such understandingis also critical for determining the extent towhich students who begin their schoolingcareers in disadvantaged neighbourhoodsare destined for poor schooling outcomes,and conversely, whether changing the neigh-bourhood context can have beneficial effectson student performance. In this study,therefore, we focus on documenting howstudents’ early neighbourhood and school-ing experiences are related to later success inschool, as well as the extent to which chang-ing neighbourhood and school contexts, par-ticularly changing socio-economic contexts,explain differences in student academicoutcomes.

Context

A long-standing line of research exploreshow neighbourhoods shape academic perfor-mance, with consistent findings that studentsliving in poor neighbourhoods experienceworse outcomes than their peers (Brooks-Gunn et al., 1993; Duncan, 1994; Ellen andTurner, 1997; Halpern-Felsher et al., 1997).While this research is largely correlational

2 Urban Studies



and does not necessarily provide causal esti-mates of neighbourhood effects, neighbour-hoods may shape student outcomes for avariety of reasons. First, neighbourhood ofresidence determines access to public goodsand institutions that may contribute to aca-demic success. These might include publicschools or libraries, cultural institutions suchas museums, or community organisationssuch as the YMCA that provide extra-curricular activities (Tiebout, 1956;Weinstein, 2008). Second, neighbourhoodsinfluence the set of adults available to serveas role models/mentors to students and toprovide access to networks. Third, neigh-bourhoods determine the students’ peergroup – both in and outside of school –which may have implications for perfor-mance (Sin, 2011; Crane, 1991; Hanushek etal., 2003; Jencks and Mayer, 1990; Leventhaland Brooks-Gunn, 2000; Weinstein, 2008).Fourth, the neighbourhood environmentmay influence student performance moredirectly. As an example, neighbourhoodcrime, violence or noise may hamper aca-demic success (Lacoe, 2013; Sharkey et al.,2013). Finally, ecological perspectives positthat neighbourhoods may influence childrenmore indirectly through parents’ behaviourand family practices (Leventhal and Brooks-Gunn, 2000).

At the same time, there is considerableevidence that early childhood experiencesmatter to success in the long run. Severalrecent studies conclude that early schoolexperiences, such as kindergarten class size(Chetty et al., 2010; Heckman, 2006) or thequality of third grade teachers (Chetty et al.,2011) affect later academic and non-academic outcomes, such as test scores, col-lege attendance and wages.

Importantly, virtually all of the existingresearch on neighbourhoods focuses on theshort-term impacts of neighbourhoods.There is little evidence on the long-termeffect of neighbourhoods experienced in

early childhood, and few studies disentanglethe effects of origin neighbourhood fromcurrent neighbourhood and/or personal andfamily characteristics, such as race, incomeor nativity (Duncan et al., 1994; Holzer etal., 2008). One exception to this is theMoving to Opportunity study, which exam-ines the effects of moving public housing res-idents into lower poverty neighbourhoods inan experimental context (Kling et al., 2007).This study does not, however, tell us any-thing about how moving to lower povertyneighbourhoods is related to later outcomesfor children who do not live in public hous-ing or those families who choose to move toa lower poverty neighbourhood absent afinancial incentive. We address this gap byexamining how mobility out of origin neigh-bourhoods is related to student performanceboth in the short and long term.

Data

We draw on a rich longitudinal database con-taining individual-level data for a completecensus of students attending NYC publicschools in grades 1–12 from academic years(AY) 1996–1997 through 2007–2008 (roughly1,000,000 observations per year). Every stu-dent record contains demographic, pro-gramme and academic information, includingschool attended, birthplace, race, gender, lan-guage ability, poverty status (i.e. eligibility forfree or reduced price lunch), attendance, par-ticipation in special education and languageprogrammes, standardised test scores andgraduation outcomes. Importantly, these dataalso contain the zip code of the student’s resi-dence in each academic year, which identifiesstudents’ neighbourhoods. Thus, we observestudents’‘origin’ neighbourhoods at firstgrade and track neighbourhood moves overtheir schooling careers.

Using this data, we construct a panel con-taining all students enrolled as first gradersin NYC public schools in 1996–1997,

Cordes et al. 3



excluding those students living in zip codeswith fewer than 25 students1 as well as stu-dents who are ever in ungraded special edu-cation. This leaves us with a total of 81,502first graders, who lived in 169 different zipcodes and attended 685 different schools infirst grade. We trace this group of 81,502students for 11 additional years, throughwhat would be their 12th grade year if theymade standard academic progress (annualone-grade promotion).2 Of course, whilesome students make standard academicprogress (SAP) and successfully reach 12thgrade ‘on time’ others are retained once (ortwice) and are enrolled in an earlier grade in

2007–2008. Finally, many students leave theNYC public schools either temporarily orpermanently prior to 2007–2008.

As shown in Table 1, students included inour final sample differ from those who areexcluded. First graders included in our sam-ple are much more likely to be female,Asian, white and reduced or full price lunchstudents than students who are excluded.Conversely, they are less likely to be black,Hispanic, ESL, LEP, SPED, native born,free lunch eligible or to speak a languageother than English at home than studentswho are excluded. Much of this is reflectiveof the decision to exclude students in

Table 1. Descriptive characteristics, 1st grade class, 1996–1997.

All Included Excluded

Female 49.2 50.7 32.7Asian/other 10.3 11.0 2.9Black 34.2 33.3 44.1Hispanic 39.9 39.3 45.7White 15.4 16.1 7.1Non-English at home1 44.3 55.1 62.0ESL2 21.4 20.8 27.1LEP2 17.0 16.3 25.0SPED3 6.5 5.0 21.9Native born 89.5 89.1 94.5Free lunch 77.1 76.0 89.8Reduced lunch 6.2 6.5 2.8Full price lunch 12.5 13.3 3.7Missing lunch info 4.2 4.3 3.7SAP student4 26.1 24.9 0.2Retained5 15.8 17.3 0.1Ungraded SPED student6 8.3 0.0 99.2Exiter7 49.7 54.2 0.5N 88,991 81,502 7489

Notes: 1The percentage of students who speak a language other than English at home.2English as a Second Language (ESL) and limited English proficient (LEP) students are classified based on results from the

Language Assessment Battery and receive additional support services including placement in bilingual education or a

free-standing ESL programme.3Those students who are ever classified as eligible for special education services. These students are placed in graded

classrooms but receive additional supports.4Those students who are continuously enrolled for all 12 years of the sample and who are in 12th grade in 2007–2008.5Those students who are ever held back in a grade.6Students whose disabilities are severe enough that they are educated in self-contained classrooms.7Students who are enrolled in first grade in 1996–97 but exit NYC public schools either temporarily or permanently at

some point during the 12 year period.

4 Urban Studies



ungraded special education from our sam-ple.3 Of the original first grade cohortincluded in our sample, by 2007–2008,17.3% are retained at least once, 54.2% exitNYC public schools either temporarily orpermanently during the period, and the rest,22.8%, remain on track to graduate highschool ‘on time’. As is evident from thesedescriptive statistics, the NYC public schoolpopulation is relatively disadvantaged as theextremely wealthy tend to enroll their chil-dren in private schools. Our results, there-fore, largely speak to the implications ofneighbourhood and school moves for thelower middle class and poor students whomake up the majority of the public schoolpopulation.

Neighbourhood boundaries are measuredusing zip codes, which we select for two rea-sons.4 First, in urban areas such as New YorkCity, the geographic area defined by zip codestends to be quite small and provides a goodcompromise between census tracts, which areextremely fine grained, and either communitydistricts or boroughs, which are quite largeand heterogeneous.5 Second, while we haveinformation on census tract of residence start-ing in AY 2004–2005, we have information onzip code of residence for all years of the data,which allows us to use a consistent measure ofneighbourhood over time. Neighbourhoodquality is captured by poverty rate for familieswith children under 18 from the 2000 Census.Specifically, we divide neighbourhoods intodeciles based on this measure, with ‘higherpoverty’ neighbourhoods in the higher decilesand ‘lower poverty’ neighbourhoods in thelower deciles.6

School quality is defined separately forelementary and middle schools using studenttest scores. For elementary schools, schoolquality is measured by the standardised read-ing and math scores for students in grades 3–5 in 1996–1997 and 1997–1998, averagedacross both subjects and years.7 This

corresponds to the years that sample stu-dents were in grades 1 and 2, respectively, sothat school quality measures do not containstudents’ own test scores. Analogously, mid-dle school quality is captured by the standar-dised reading and math scores for studentsin grades 6–8 in 1999–2000 and 2000–2001,averaged across both subjects and years.This corresponds to the years that samplestudents were in grades 4 and 5. We use suchan aggregate measure rather than a changein scores between grades or years (value-added) since this analysis is focused on iden-tifying how overall school quality is relatedto performance.8

Portraits of neighbourhoods andmobility

How much do neighbourhoodcharacteristics vary?

Students in NYC public schools experiencevariation in the poverty they experience intheir neighbourhoods; the poverty rate offamilies with children under the age of 18ranges from a low of 1.3% in the SoutheastBronx to 54.5% in Hunts Point and MottHaven in the Bronx. Additionally, a non-trivial 15.5% of neighbourhoods exhibitextreme poverty with rates above 40%(Figure 1).

In addition to the wide variation experi-enced by students in terms of environmentalinputs such as poverty, students also experi-ence a wide range of outcomes by neighbour-hood. Among students who are still enrolledin NYC public schools during what shouldbe their 10th grade year, an average of 62%graduate from high school by 2007–2008.9

This average graduation rate masks signifi-cant variation, however, as the rate rangesfrom a low of 35% in Central Harlem to ahigh of 91% in lower Manhattan with17.2% of neighbourhoods graduating lessthan half of 10th graders.

Cordes et al. 5



How tight is the link between neighbourhood andschool? As discussed previously, a key way inwhich neighbourhood shapes schooling out-comes is that districts often assign studentsto schools through residentially based atten-dance zones.

In an average neighbourhood, studentsattend 22 different schools – and the range iswide (Figure 2). In some neighbourhoods,the link between neighbourhood and schoolis tight, and all of the first graders attend thesame schools. In neighbourhoods at theother extreme, students attend as many as76 distinct schools, and even the most widelyattended elementary school contains only8% of the first grade public school residents.This variation may reflect factors such asgeography and proximity to transportationas well as the variations in the size and char-acteristics of the schools and, importantly,

the school choices made by families. Thus,for many students, neighbourhood does notunilaterally determine school attended – sug-gesting that zip code may not, indeed, be‘destiny’ for New York City school children.

Neighbourhood mobility. In order to determinewhether and to what extent origin neigh-bourhoods matter in the long run andwhether any influence of origin neighbour-hoods disappears once students make schoolmoves, we first examine the extent to whichstudents ever leave their origin neighbour-hoods. As shown in Figure 3, New YorkCity includes both high mobility neighbour-hoods and neighbourhoods where studentsmake relatively few residential moves.10

We explore the neighbourhood-level corre-lates of this mobility by regressing the

0.0

1.0

2.0

3D

ensi

ty

0 20 40 60 80 100Percent of Families Below Fed. Pov. Line

Figure 1. Percent poor students and families by neighbourhood, AY 1996–1997. Density is over the totalnumber of neighbourhoods (169), where neighbourhood is defined as students’ zip code of residence.Percent of families below the federal poverty line is based on the 2000 Census.

6 Urban Studies



neighbourhood average student residentialmobility rate on neighbourhood poverty andpublic school characteristics including race,free lunch eligibility and nativity. Amongthese, the only significant predictor ofneighbourhood mobility rates is poverty(Table 2). Specifically, in neighbourhoodswith poverty rates greater than 40%, themobility rate is between 5.2 to 8.0 percentagepoints higher than in neighbourhoods withpoverty rates below 20%.11 Furthermore,for each 10 percentage point increase in thenumber of students eligible for free lunch,the neighbourhood mobility rate increasesby approximately 0.05 percentage points.This suggests that rather than being dead-end neighbourhoods where students are‘stuck’, high poverty neighbourhoods exhibitparticularly high student mobility, althoughthese results do not indicate whether such

students are able to escape to higher qualityneighbourhoods.

Student residential mobility: How much dostudents move between neighbourhoods?Are students ‘stuck’ in poverty?

As just discussed, high poverty neighbour-hoods exhibit higher mobility than low pov-erty neighbourhoods. In this section, weanalyse characteristics and patterns of stu-dents who are mobile across neighbourhoodsto understand which types of students moveand to where.

To document significant changes in neigh-bourhood quality, neighbourhood poverty isdivided into deciles. Changes in neighbour-hood quality are calculated by taking the dif-ference between neighbourhood quality ofstudent i in year t and neighbourhood

05

1015

2025

Per

cent

of Z

ip C

odes

0 20 40 60 80Number of Different Schools Attended by First Graders in Zip

Figure 2. Number of distinct schools attended by first graders by neighbourhood, AY 1996–1997.Neighbourhood is defined as students’ zip code of residence. Number of schools calculated as the distinctnumber of schools attended by students living in a particular neighbourhood.

Cordes et al. 7



quality in that student’s first grade year. Astudent is then classified as having experi-enced a significant change in neighbourhoodquality if the level of neighbourhood povertychanges by two or more deciles between thestudent’s first grade neighbourhood and thestudent’s current neighbourhood. Accordingto this definition, 19% of students in firstgrade experienced a significant increase inneighbourhood quality and only 10% expe-rienced a significant decrease.12

Next we examine what types of neigh-bourhoods movers come from and wherethey go. If students move from one neigh-bourhood to another without a concomitantchange in neighbourhood quality, we mightnot expect large effects on performance. Wetherefore examine the relationship betweenthe likelihood of making a significant movein neighbourhood quality and the initialquality of a student’s first grade

neighbourhood. To do so, we calculate thepercentage of students who move to betteror worse neighbourhoods by the decile ofstudents’ first grade neighbourhood (wherelower deciles correspond to lower poverty or‘better’ neighbourhoods). By construction,students living in the lowest and highest dec-iles of neighbourhood poverty are unable toexperience significant increases anddecreases in neighbourhood quality, respec-tively, because they can only ‘move down’(if starting in the highest quality neighbour-hoods) or ‘move up’ (if starting in the lowestquality neighbourhoods), whereas studentsin average quality neighbourhoods can makesignificant moves in either direction.13

Consistent with this, students from lowquality (high poverty) neighbourhoods aremore likely to experience significantincreases in neighbourhood quality com-pared with students in higher quality (low

Figure 3. Distribution of neighbourhood mobility rates. Mobility rate is calculated as the percent of firstgraders in a given neighbourhood (zip code) who make any moves to a different neighbourhood betweengrades 1 and 8.

8 Urban Studies



poverty) neighbourhoods. As shown inTable 3, Panel A, 45.7% of students in thehighest two deciles of poverty move to betterneighbourhoods at some point but only7.7% of students in the third and fourth dec-iles ever make such moves. In contrast, stu-dents from high quality neighbourhoods aremuch less likely to experience decreases inneighbourhood quality than their peers fromlow quality neighbourhoods. Only 11.5% of

students who make moves to significantlyworse neighbourhoods come from the lowesttwo deciles of neighbourhood poverty,whereas 26.5% of movers to worse neigh-bourhoods come from the seventh andeighth deciles of neighbourhood poverty.Contrary to the theory that students arestuck in bad neighbourhoods, this suggeststhat there is considerable mobility out ofhigh poverty neighbourhoods, although

Table 2. Regression estimates, neighbourhood mobility rate and neighbourhood characteristics.

Variables Neighbourhood average of % students moving between 1st and

5th grade 8th grade 10th grade(1) (2) (3)

Neighbourhood pov . 40%1 0.057*** 0.052*** 0.080***(0.018) (0.020) (0.028)

Neighbourhood pov . 20%1 20.011 20.026 0.014(0.015) (0.021) (0.023)

% of public school students:Free lunch eligible 0.003*** 0.005*** 0.004***

(0.001) (0.001) (0.001)Reduced price lunch eligible 20.002 20.000 0.002

(0.002) (0.002) (0.003)Black 0.010 0.008 0.000

(0.013) (0.019) (0.038)Hispanic 0.009 0.007 0.001

(0.013) (0.019) (0.039)Asian/other 0.011 0.010 0.002

(0.013) (0.019) (0.038)White 0.009 0.007 0.002

(0.013) (0.019) (0.039)Native born 20.001 20.002 20.004*

(0.001) (0.002) (0.002)Do not speak English at home 20.001 20.002 20.004*

(0.001) (0.001) (0.002)No. of students (in 100s) 20.001*** 20.001*** 20.001**

(0.000) (0.000) (0.000)Constant 20.737 20.367 0.559

(1.327) (1.897) (3.765)Observations 169 169 169R2 0.612 0.627 0.171

Notes: Robust standard errors in parentheses, ***p \ 0.01, **p \ 0.05, *p \0.1.1Neighbourhood Pov greater than 40% is an indicator equal to one if the poverty rate of households with children under

18 is greater than 40% in a particular zip code. Neighbourhood Pov greater than 20% is measured analogously.

Thresholds of 20% and 40% were chosen based on designations used by the Census, which labels neighbourhoods with

poverty rates greater than 20% as ‘poverty areas’ and those with poverty rates above 40% as an ‘extreme poverty

areas’. The percentage of public school students with a given characteristic are based on the entire population of public

schools students (grades 1–8) living in a particular zip code. All percentages are calculated based on a neighbourhoods

public school population in AY 1996–1997.

Cordes et al. 9



among all mobile students, there is a mixtureof moves to both significantly better and sig-nificantly worse neighbourhoods.

Student school mobility: How much isthere? Do students move to better orworse schools?

One of the main mechanisms through whichresidential mobility is expected to influenceschool performance is by providing access to

higher quality schools. Of the initial firstgrade cohort, almost all students (98%)change schools at least once during their K-12 academic career, but not all of thesemoves result in exposure to schools of signif-icantly different quality.

Like changes in neighbourhood quality,changes in school quality are calculated bytaking the difference between school qualityfor student i in year t and school quality inthat student’s first grade year. Since the

Table 3. Panel A. Percentage of students who ever experience significant increases or decreases inneighbourhood quality by decile of neighbourhood quality in 1997.

Decile Ever better(1)

Ever worse(2)

Frequency Percent Frequency Percent

1 137 2.02 653 9.53 101 0.9 646 9.44 444 6.8 1246 18.15 816 6.9 1209 17.66 1252 10.6 1156 16.87 1656 14.0 894 13.08 2144 18.2 927 13.59 2381 20.210 3006 25.5Total 11,800 100.0 6868 100.0

Panel A. Percentage of students who ever experience significant increases or decreases in neighbourhoodquality by decile of neighbourhood quality in 1997.

Decile Ever better Ever worse

Frequency Percent Frequency Percent

1 3298 17.32 3481 18.33 2881 15.1 1407 6.54 3137 16.5 2755 12.85 2087 11.0 3369 15.66 1685 8.9 4664 21.67 1650 8.7 2989 13.98 807 4.2 2279 10.69 2467 11.410 1635 7.6Total 19,026 21,565 100.0

Notes: Students are recorded as having ever moved to a better (worse) neighbourhood or school if they experience a

two decile increase (decrease) in neighbourhood or school quality at any point between 1997 and 2008.

10 Urban Studies



measure of school quality used in this analy-sis is constant within most schools over time,students will only experience a change inschool quality when they change schools.14

Additionally, because any school change islikely to be accompanied by at least somesmall change in school quality we focus ouranalysis on ‘significant’ changes in schoolquality.15

Significant changes in school quality aredefined analogously to significant changes inneighbourhood quality. That is, as a changein school quality of two or more deciles sincefirst grade. According to this definition,23.4% of students experience a significantincrease in school quality, and 26.5% experi-ence a significant decrease at some point intheir schooling careers. Note that an increaseor decrease in school quality might not bethe result of a single move, but rather mayoccur gradually over multiple moves toslightly better or worse schools. The currentanalysis does not distinguish between theseoutcomes, as it is ultimately focused ondetermining the likelihood of a student‘escaping’ to a significantly better or worseschool conditional on the quality of theschool he or she attended in first grade.

Table 3, Panel B presents informationabout the likelihood that students experiencea significant change in school quality basedon the initial quality of their first gradeschools. Using this measure, students frombottom (top) decile schools are, by construc-tion, unable to make moves to significantlyworse (better) schools. It is not surprisingthen that 35.6% of moves to better qualityschools occur among students in the lowesttwo deciles of school quality, and only12.9% occur among students in the seventhand eighth deciles of school quality. Movesto lower quality schools display a much dif-ferent pattern, however – students who startout in high quality schools are actuallyequally likely to make moves to lower

quality schools (19.0% ) as their peers whostart in lower quality schools (19.3%).Therefore, while students in the lowest decileof school quality have the highest likelihoodof moving to a significantly better school,the reverse is not true – students at both thebest and worst schools have high likelihoodsof moving to a significantly worse school.First grade schools are not ‘destiny’.

Neighbourhood change andschooling outcomes: Methods

In the previous section, we saw the diverseorigin neighbourhood experiences of firstgraders in NYC public schools and docu-mented high levels of both neighbourhoodand school mobility. Now we examine howthese diverse origin neighbourhood experi-ences combine with neighbourhood andschool mobility patterns to influence schooloutcomes. We do so using regression modelsto explain the influence of neighbourhoods(and schools) on both short- and long-termstudent outcomes including test scores, pro-gression through school and high schoolgraduation. As mobility is likely an endo-genous process, the models outlined beloware not causal, and are not to be interpretedas such. Rather, they are meant to describeand compare the outcomes of movers andnon-movers and to examine the extent towhich moving from one’s origin neighbour-hood is associated with a change in aca-demic performance and attainment.

To determine the relationship betweenneighbourhood and school mobility andshort-term outcomes, we estimate the follow-ing model:

Pisng =b0 +b1MOVEig +b2MOVEUPig

+b3MOVEDNig +b4MOVESCHLig

+b5SCHLUPig +b6SCHLDN +b7Iig

+ai + eisng ð1Þ

Cordes et al. 11



where Pisng is short-term performance of stu-dent i in school s living in current neigh-bourhood n and in current grade g; MOVEis an indicator of whether the student movedneighbourhoods in g; MOVEUP is an indi-cator of whether this move was to a signifi-cantly higher quality neighbourhood ascompared with the student’s origin neigh-bourhood; MOVEDN is an indicator ofwhether this move was to a significantlylower quality neighbourhood; MOVESCHLis an indicator of whether a student movedschools in g; SCHLUP is an indicator ofwhether this move was to a significantlyhigher quality school; SCHLDN is an indi-cator of whether this move was to a signifi-cantly worse school; Iig is a vector ofobservable time-varying individual charac-teristics including free lunch eligibility, spe-cial education, limited English proficiency(LEP), and English as a Second Language(ESL); and ai are student fixed effects. Notethat the inclusion of student fixed effectscontrols for unobserved time invariant stu-dent characteristics, such as motivation,ability or past home inputs as well as originneighbourhoods and origin schools, so we canidentify how changing neighbourhoods con-tributes to student achievement beyondwhere students originate.

To determine the relationship betweenneighbourhood and school mobility and long-term outcomes, we estimate the followingmodel:

Pisng =b0 +b1MOVEig +b2MOVEUPig

+b3MOVEDNig +b4MOVESCHLig

+b5SCHLUPig +b6SCHLDN

+b7Xi3 +b8Iig +us1 +hn1 + eisng

ð2Þ

where Pisng is now long-term performance orattainment of student i in school s living incurrent neighbourhood n and currently ingrade g; MOVE is an indicator of whether

the student ever moved neighbourhoodsbetween first grade and grade g; MOVEUPis an indicator of whether a student everexperienced a significant increase in neigh-bourhood quality between first gradeand grade g; MOVEDN is an indicatorof whether a student ever experienced asignificant decrease in neighbourhoodpoverty between first grade and grade g;MOVESCHL is an indicator for whether astudent ever moved schools between firstgrade and grade g; SCHLUP is an indicatorof whether a student ever experienced a sig-nificant increase in school quality betweenfirst grade and grade g; SCHLDN is an indi-cator of whether a student ever experienceda significant decrease in school qualitybetween first grade and grade g; Xi3 is a vec-tor of test scores as measured in third grade;Ii1 is a vector of observable individual char-acteristics measured in first grade includingnativity, ESL, LEP, receipt of special educa-tion services, race, gender, free lunch eligibil-ity and whether the student speaks alanguage other than English at home. Themodel includes controls for third grade testscores to capture student ability as well asschool and family inputs into a student’slearning experience up through 3rd grade.16

The remaining terms, us1 and hn1, are firstgrade school and origin neighbourhoodfixed effects, respectively, which allow us toestimate the relationship between long-termachievement and moving neighbourhoods,conditional on students’ origin school andneighbourhood experiences.17

Using this model we examine a number ofperformance measures including 8th gradetest scores, New York State high schoolexam scores (Regents exams),18 and highschool graduation outcome at the end of 12years.19 We also examine measures of attain-ment, including the number of years it takesto get to different grade levels (longer isworse) and the likelihood of attaining

12 Urban Studies



standard academic progress over the fullperiod of our analysis.

Neighbourhood and schoolchange and schooling outcomes:Results

Before turning to our regression estimates,we first examine the descriptive statistics forall short- and long-term student perfor-mance measures (Table 4). Note that allaverage 3rd- through 8th-grade z-scores areabove the grade-wide averages of zero,because our sample includes only studentswho remain in NYC public schools overtime (which tends to be a positively selectedgroup). Additionally, there is considerablevariation in these outcomes, as evidenced bythe wide ranges between the 10th and 90thpercentiles. For example, standardised

scores on 4th-grade ELA range from 21.10to 1.24, and the number of years it takes astudent to reach 10th grade ranges from 9 to11 years.

Panel A in Table 5 (columns 1 and 2) pre-sents estimates from models of short-termperformance (equation 1). The results showthat, there is no relationship between neigh-bourhood changes and test scores in theshort-term (Panel A), but a consistently pos-itive (negative) relationship between perfor-mance and making a move to a higher(lower) quality school. Making any schoolmove is negatively related to performance,even if it is to a similar quality school.Students who move to lower quality schoolsperform the worst (0.051 to 0.061 sds lowerthan their peers who do not change schools),and students who move to higher qualityschools experience no change in

Table 4. Descriptve statistics, short-term and long-term student performance outcomes.

Variable name Mean SD 10th percentile 90th percentile N

Short-term performance outcomes3rd grade Math 0.05 0.95 21.08 1.07 70,0573rd grade ELA 0.03 0.95 21.13 1.02 67,4274th grade Math 0.03 0.95 21.11 1.21 65,1824th grade ELA 0.02 0.96 21.10 1.24 65,2205th grade Math 0.03 0.94 21.03 1.05 62,2775th grade ELA 0.02 0.94 21.04 1.16 62,1466th grade Math 0.06 0.94 21.03 1.10 58,9476th grade ELA 0.02 0.98 21.09 1.13 58,8917th grade Math 0.06 0.93 20.99 1.10 56,9097th grade ELA 0.03 0.96 21.14 1.21 56,6898th grade Math 0.06 0.91 21.02 1.12 53,7548th grade ELA 0.04 0.97 20.98 1.27 54,504

Long-term performance outcomes8th grade Math 0.06 0.91 21.02 1.12 53,7548th grade ELA 0.04 0.97 20.98 1.27 54,504Regents Math 0.14 0.89 20.97 1.32 37,798Regents English 0.17 0.82 20.75 1.12 37,715Graduate 0.55 0.50 0 1 47,170

Long-term attainment outcomesYears to 5th grade 4.20 0.46 4 5 44,131Years to 8th grade 7.25 0.54 7 8 44,239Years to 10th grade 9.52 0.89 9 11 47,170SAP student 0.43 0.49 0 1 47,170

Cordes et al. 13



performance for math and a slight decreasein reading performance, where the negativerelationship between moving and readingscores is significantly mitigated by increasesin school quality.

In the long term, however, both neigh-bourhood and school moves are related toperformance (see Table 5, Panel B). In par-ticular, students who move to a similar qual-ity neighbourhood without moving schoolsperform 0.026–0.030 SDs lower on both 8thgrade exams and Math Regents than theirpeers who never move. Students who moveto lower quality neighbourhoods fare consis-tently worse: students who move to a worseneighbourhood and remain in the sameschool score 0.024–0.091 SDs below theirpeers who never move neighbourhoods.Moves to better quality neighbourhoodsappear to have mixed results, resulting ineither similar or worse outcomes than movesmade to similar quality neighbourhoods.Changes in school quality, however, have amore consistent relationship with studentperformance, such that moving to betterschools helps performance and moving toworse schools harms performance. In fact,moving to better schools can be protective ofthe harmful effect of moving neighbour-hoods, at least when moves are made toneighbourhoods of similar quality.20

Finally, results for attainment (Panel C)show that most moves of any kind prolongthe time to 5th, 8th and 10th grades.21

Perhaps the protection that moving to a bet-ter school provides for residential movers isat the expense of progressing in a timely wayto 10th grade, as students may need addi-tional time to catch up to classmates in thesebetter schools, for example.

In summary, residential mobility tends tobe negatively related to schooling outcomes,but moving to a significantly better quality

school is positively related to student out-comes and may mitigate/overcome the addi-tional negative impacts of moving to lowerquality neighbourhoods.

Summary and conclusions

There is a high degree of heterogeneityamong origin neighbourhoods of NYC pub-lic school students, but there is no obviousset of characteristics that define a ‘bad’ ori-gin neighbourhood. Moreover, there is alarge amount and wide range of mobilityexperienced among NYC public school stu-dents. Far from supporting the simplisticstory of ‘dead-end’ neighbourhoods, ouranalyses describe a situation where not onlydo students from poor neighbourhoodsmove, these students actually move moreoften than their peers in less disadvantagedneighbourhoods. Furthermore, studentsfrom lower quality neighbourhoods andschools are more likely to experience changesin neighbourhood and school quality, withschool quality changes more likely to be pos-itive than negative.

There is little evidence that moving tohigher quality neighbourhoods is related toimproved performance; rather, there is con-sistent evidence that moving neighbour-hoods is associated with decreases in bothshort- and long-term student performanceand attainment, especially among studentswho move to lower quality neighbourhoods.Moreover, there is evidence that movingneighbourhoods is not all about schools, aschanges in neighbourhoods and neighbour-hood quality continue to have statisticallysignificant relationships with long-term stu-dent performance, even when controlling forchanges in school quality. This is not toimply that schools do not matter; in fact,moving to a higher quality school has

14 Urban Studies



Tab

le5.

Reg

ress

ion

resu

lts,

per

form

ance

,nei

ghbourh

ood

and

schoolm

obili

ty,g

rades

3–8.

Pan

elA

.Sh

ort

-ter

mac

hie

vem

ent

Pan

elB.L

ong-

term

achie

vem

ent

Pan

elC

.Lo

ng-

term

atta

inm

ent

Mat

hR

eadin

g8th

grad

eM

ath

8th

grad

eR

eadin

g

Mat

hR

egen

tsEng

lish

Reg

ents

Gra

dua

tion

Yrs

to5th

Yrs

to8th

Yrs

to10t

hSA

Pst

uden

t

Var

iable

s(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)(1

0)

(11)

Move

nei

ghbourh

oods

20.0

07(0

.005

)0.0

04

(0.0

05)

20.0

26***

(0.0

07)

20.0

24***

(0.0

07)

20.0

30***

(0.0

10)

20.0

05

(0.0

09)

20.0

31***

(0.0

05)

0.0

40***

(0.0

06)

0.0

08

(0.0

06)

0.0

29***

(0.0

10)

20.0

49***

(0.0

05)

Move

tobet

ter

nei

ghbourh

ood

0.0

05(0

.008

)2

0.0

05

(0.0

08)

20.0

38***

(0.0

10)

20.0

38***

(0.0

10)

0.0

01

(0.0

12)

20.0

05

(0.0

11)

20.0

05(0

.006

)0.0

00

(0.0

08)

0.0

59***

(0.0

08)

0.0

46***

(0.0

12)

20.0

17***

(0.0

06)

Move

tow

ors

enei

ghbourh

ood

0.0

10(0

.009

)0.0

12

(0.0

10)

20.0

27**

(0.0

12)

0.0

06(0

.013

)2

0.0

53***

(0.0

15)

20.0

41***

(0.0

14)

20.0

66***

(0.0

08)

0.0

05

(0.0

10)

0.0

89***

(0.0

10)

0.1

40***

(0.0

15)

20.0

66***

(0.0

08)

Move

schools

20.0

31***

(0.0

02)

20.0

23***

(0.0

03)

20.1

54**

(0.0

71)

20.0

89(0

.089

)2

0.1

30

(0.0

94)

20.1

24

(0.0

81)

20.0

38(0

.049

)0.0

06

(0.0

11)

20.0

02

0.2

35***

0.2

32***

(0.0

57)

(0.0

87)

(0.0

49)

Move

tobet

ter

school

0.0

32***

(0.0

04)

0.0

15***

(0.0

05)

0.0

63***

(0.0

09)

0.0

80***

(0.0

09)

0.0

79***

(0.0

11)

0.0

69***

(0.0

10)

0.0

18***

(0.0

06)

0.0

76***

(0.0

08)

0.0

17**

(0.0

07)

0.0

06

(0.0

11)

0.0

53***

(0.0

06)

Move

tow

ors

esc

hool

20.0

20***

(0.0

04)

20.0

38***

(0.0

05)

20.1

03***

(0.0

09)

20.1

03***

(0.0

09)

20.0

72***

(0.0

11)

20.0

48***

(0.0

10)

20.0

39***

(0.0

06)

0.0

95***

(0.0

08)

0.0

44***

(0.0

07)

0.0

72***

(0.0

11)

20.0

03

(0.0

06)

Obse

rvat

ions

367,1

26

364,8

77

53,

718

54,4

70

37,7

8437,

702

47,1

53

44,1

1844,

219

47,1

5247,1

53R

20.7

90

0.7

67

0.4

87

0.5

09

0.3

73

0.3

60

0.2

12

0.2

11

0.2

230.1

94

0.2

07

No.ofst

uden

ts71,0

67

71,2

9653,

718

54,4

70

37,7

8437,

702

47,1

53

44,1

1844,

219

47,1

5247,1

53

Not

es:R

obust

stan

dar

der

rors

inpar

enth

eses

,***p

\0.0

1,**p

\0.0

5,*p

\0.1

.

All

model

sin

Pan

elA

incl

ude

studen

tfix

edef

fect

s.A

studen

tis

reco

rded

as‘c

hangi

ng’

nei

ghbourh

oods

inan

yye

arth

atth

eym

ake

are

siden

tial

move

toa

diff

eren

t

nei

ghbourh

ood.M

ovi

ng

toa

bet

ter

nei

ghbo

urh

ood

isco

ded

1in

the

year

that

ast

uden

tm

akes

am

ove

toa

zip

code

that

istw

oor

more

dec

iles

low

erin

the

pove

rty

dis

trib

ution.

Movi

ng

toa

wors

enei

ghbourh

ood

isco

ded

as1

inth

eye

arth

ata

studen

tm

akes

am

ove

toa

nei

ghbo

urh

ood

that

istw

oor

more

dec

iles

hig

her

inth

epove

rty

dis

trib

ution.

Reg

ress

ions

inPan

els

Ban

dC

also

incl

ude

contr

ols

for

studen

tch

arac

teri

stic

sas

mea

sure

din

first

grad

e,1st

grad

esc

hoola

nd

ori

gin

nei

ghbourh

ood

effe

cts

and

indic

ators

for

whet

her

schoolq

ual

ity

mea

sure

sar

eim

pute

d.R

egen

tste

stsc

ore

regr

essi

ons

incl

ude

indic

ators

for

the

grad

ein

whic

ha

studen

tto

ok

the

Reg

ents

exam

.

Cordes et al. 15



positive effects on short- and long-term stu-dent performance, and in some cases theseare large enough in magnitude to swamp theadditional negative effects of moving to aworse neighbourhood.

These results, although not causal, haveimportant implications. Students and fami-lies who make moves to higher qualityneighbourhoods are unlikely to be an aver-age group, but are likely different in impor-tant respects (e.g. more concerned abouttheir children’s education outcomes, moreambitious, more informed, etc.). This is par-ticularly true given that our measures ofmoving to better neighbourhoods andschools are non-marginal, considering onlysubstantial changes in quality (two or moredeciles). Nonetheless, these families do man-age to move themselves out of their originschools in ways that improve outcomes forstudents, even when they begin in poorneighbourhoods and low quality schools.Thus, there are possibilities for improvingthe life chances of central city students, sothat neighbourhood is not destiny. Providingenough ‘good’ schools, helping familieschoose the good ones when they move neigh-bourhoods, and even allowing them to stayin good schools if they must move neigh-bourhoods are all mechanisms that couldimprove student outcomes. It should benoted that the findings presented here use arelative measure of school quality. Definingan absolute measure of ‘good’ schools is out-side the scope of this paper, but nonethelessa valuable area for future research.

Funding

This work was supported by the SpencerFoundation [grant number 201100049]

Notes

1. These zip codes contain portions that fallboth inside and outside the city limits, butthe majority of their area falls outside the

city, and are therefore not appropriate toinclude in this analysis.

2. This includes students who are repeatingfirst grade. Results are not sensitive toexcluding them.

3. According to the New York State report carddata glossary (http://data.nysed.gov/glossar-y.php?report=reportcards), ‘Ungraded stu-dents are those assigned to a class that is notorganized on the basis of grade grouping andhas no standard grade designation. Thisincludes both regular and special classes that

have no grade designations. Such a class maycontain students of different ages who areidentified according to level of performancein one or more areas of instruction, ratherthan according to grade level or age level’.Students in ungraded special education typi-cally do not take standardised exams.

4. Zip codes have been used by others in neigh-bourhood studies (see for example Datcher,1982; Ku et al., 1993).

5. This is well documented in other work onNew York City (see for example Ellen et al.,2002).

6. We experimented with alternative measuresof neighbourhood quality, including a neigh-bourhood advantage index based on percentof families with children under 18 NOT liv-ing in poverty, percent of 8th grade studentspassing the ELA exam in AY 1999–2000,and percent of students passing the mathexam in AY 1999–2000. Results are similar.

7. Standardised scores have citywide means of0 and standard deviations of 1 based onscores of all students in a grade in a year.

8. Average test scores are commonly used as asummary measure of multiple dimensions ofschool quality including the quality of peers,level of resources and quality of the schoolitself.

9. Using all first grade students to calculategraduation rates is problematic because wedo not have graduation information on stu-dents who exit NYC public school students.

That is, many students who exit NYC publicschools to attend private schools and schoolsin another district may very well graduate,but we do not have a record of this, so thatour graduation rate calculated from all first

16 Urban Studies



grade students is almost certainly an under-estimate. Focusing attention on studentswho are still enrolled by their 10th gradeyear should mitigate this problem. There are47,170 students who fall into this category.

10. This distribution widens as students progressthrough school (data not shown).

11. Thresholds of 20% and 40% were chosenbased on designations used by the Census,which labels neighbourhoods with povertyrates greater than 20% as ‘poverty areas’and those with poverty rates above 40% as

an ‘extreme poverty areas’. See http://www.census.gov/population/socdemo/stat-briefs/povarea.html.

12. In results not shown here, we also examinegentrification/decline over time in a neigh-bourhood for those students who do notmove, defining a changes in neighbourhoodquality by differences in poverty between the2000 and 2010 Census. We find no change inoutcomes among such students and there-fore do not address this in the remainder ofthe paper. We feel this is justified by the rel-atively small number of students who experi-ence such changes, approximately 156 ofwhom live in neighbourhoods that deterio-rate over time and 4312 of whom live inneighbourhoods that improve.

13. Thus ‘ever better’ moves from the lowestpoverty deciles (1 and 2) and ‘ever worse’moves from the highest poverty decides (9and 10) are left blank in Table 3.

14. Because measures of school quality are cal-culated separately for elementary and middleschool grades, schools that contain both ele-mentary and middle grades will have schoolquality measures that change depending onwhat grade the student is in. The majority ofschools, however, do not fit this case and wewould expect such ‘changes’ in quality to besmall within a single school.

15. In most cases, one would expect that ‘stan-dard’ moves to a feeder school (e.g. a studentin a K-5 school moving to a 6–8 school in

6th grade) would not involve large changesin quality because such a move is likely to beto a school of similar quality. Rather, themeasure below is designed to capture eitherstrategic moves to better schools, or reactive

moves that accompany unforeseen eventssuch as a residential move or school closure,that result in substantive changes in schoolquality.

16. Note that unlike the short-term models, wecannot include student fixed effects in thelong-term models since we observe studentsat one point in time (8th grade or 12 years inNYC public schools) for these models. Weuse 3rd grade test scores because this is theearliest grade that performance is measuredin NYC.

17. We also estimate models including interac-tions between all neighbourhood and schoolmove variables. The results are qualitativelysimilar to those from the more parsimoniousmodel and for ease of interpretation, weinclude only the latter here. Results frommodels including interaction terms are avail-able from the authors upon request.

18. These are called Regents exams after thegoverning body responsible for supervisionof all educational activities within NewYork State (The New York State Board ofRegents).

19. For models where the dependent variable isRegents test scores, we include indicators ofthe grade the student was in when he/shetook the Regents as additional controls.

20. Models are also estimated using the fullsample of students (including exiters) andalso using a Heckman-style correction forselection of students out of the sample. Inorder to perform the Heckman-style correc-tion, we use data on housing from the 1990Census. Specifically, we include percentageof households living in the same house asfive years ago, percentage of households inthe same MSA as five years ago, and home-ownership rate. Results from these alterna-tive specifications are similar and availablefrom authors upon request.

21. The coefficients on all move variables areeither in the direction of worse attainmentor are statistically insignificant.

References

Brooks-Gunn J, Duncan GJ, Klebanov PK, et al.

(1993) Do neighborhoods influence child and

Cordes et al. 17



adolescent development? American Journal of

Sociology 99: 353–395.Chetty R, Friedman JN, Hilger N, et al. (2010)

How does your kindergarten classroom affect

your earnings? Evidence from project STAR.National Bureay of Economic Research Work-ing Paper No. 16381.

Chetty R, Friedman JN and Rockoff JE (2011).The Long-term Impacts of Teachers: Teacher

Value-added and Student Outcomes in Adult-

hood. No. w17699. National Bureau of Eco-nomic Research.

Christie C (2011) Speech given at town hall meet-ing, Fair Lawn, NJ, June. Available at: http://stage.mrctv.org/videos/chris-christie-destiny-should-not-be-determined-zip-code.

Crane J (1991) The epidemic theory of ghettosand neighborhood effects on dropping outand teenage childbearing. American Journal of

Sociology 96: 1226–1259.Datcher L (1982) Effects of community and fam-

ily background on achievement. The Review of

Economics and Statistics 64: 32–41.Duncan GJ (1994) Families and neighbors as

sources of disadvantage in the schooling deci-sions of white and black adolescents. American

Journal of Education 103: 20–53.Duncan GJ, Brooks-Gunn J and Klebanov PK

(1994) Economic deprivation and early child-hood development. Child Development 65:296–318.

Ellen IG and Turner MA (1997). Does neighbor-hood matter? Assessing recent evidence. Hous-

ing Policy Debate 97: 833–866.Ellen IG, Schill MH, Susin S, et al. (2002) Build-

ing homes, reviving neighborhoods: Spilloversfrom subsidized construction of owner-occupied housing in New York City. In: Retsi-nas NP and Belsky ES (eds) Low Income

Homeownership: Examining the Unexamined

Goal. Washington, DC: Brookings InstitutionPress, pp. 447–77.

Halpern-Felsher B, Connell JP, Spencer MB, etal. (1997) Neighborhood and family factorspredicting educational risk and attainment in

african american and white children and ado-lescents. Neighborhood Poverty 1: 146–173.

Hanushek EA, John FK, Markman JM, et al.(2003) Does peer ability affect student

achievement? Journal of Applied Econometrics

18: 527–544.Heckman J (2006) Skill formation and the eco-

nomics of investing in disadvantaged children.Science 312: 1900–1902.

Holzer HJ, Schanzenbach DW, Duncan GJ, et al.(2008) The economic costs of childhood pov-erty in the United States. Journal of Childrenand Poverty 14: 41–61.

Jencks C and Mayer SE (1990) The social conse-quences of growing up in a poor neighbor-hood. In: Lynn LE and McGeary MGH (eds)

Inner-city Poverty in the United States.Washington, DC: The National AcademiesPress, pp. 111–186.

Kling JR, Liebman JB and Katz LF (2007)Experimental analysis of neighborhood effects.Econometrica 75(1): 83–119.

Ku L, Sonenstein FL and Pleck JH (1993) Neigh-borhood, family, and work: Influences on thepremarital behaviors of adolescent males.Social Forces 72(2): 479–503.

Lacoe J (2013) Too Scared to Learn? The Aca-

demic Consequences of Feeling Unsafe at

School. Institute for Education and Social Pol-icy Working Paper #02–13.

Leventhal T and Brooks-Gunn J (2000) Theneighborhoods they live in: The effects ofneighborhood residence on child and adoles-cent outcomes. Psychological Bulletin 126:309–337.

Sharkey P, Schwartz AE, Ellen IG, et al. (2013)High Stakes in the Classroom, High Stakes on

the Street: The Effects of Community Violence

on Students’ Standardized Test Performance.Institute for Education and Social PolicyWorking Paper #03–13.

Sin SCJ (2011) Neighborhood disparities in accessto information resources: Measuring and map-ping US public libraries’ funding and servicelandscapes. Library & Information Science

Research 33(1): 41–53.Tiebout CM (1956) A pure theory of local expen-

ditures. The Journal of Political Economy 64:416–424.

Weinstein MG (2008) In and out of school: Out-

of-school time programs, nonprofits, title I and

the distribution of education resources in new

york city. PhD Thesis, New York University.

18 Urban Studies



Documents

Urban Studies Is neighbourhood destiny? Exploring Urban Studies … · Is neighbourhood destiny? Exploring the link between neighbourhood mobility and student outcomes Sarah A Cordes