The Effects of Owner-Occupied Housing on Student Outcomes: Evidence from NYC

Abstract

The view of owning a home as integral to the “American dream” is enshrined in numerous policies designed to promote homeownership. Whether or not these policies are worth their cost is unclear and depends, in part, on the extent to which owner-occupied housing (OOH) confers socially important benefits. Yet identifying the effects of OOH is complicated, not only due to standard concerns about selection, but also because OOH tends to be located in neighborhoods with better amenities (including schools) and is often synonymous with living in a single-family home. In this paper we use rich, longitudinal student-level data to examine whether students in OOH have better academic and health outcomes than those in renter occupied housing (ROH). We address concerns about selection using student fixed effects and a rich set of individual, building, and neighborhood controls. We find that that there is notable variation in both the characteristics and size of OOH and the types of students who live in OOH in NYC. While raw differences show that students who live in OOH have better outcomes—they are less likely to be chronically absent, obese, or overweight and have higher standardized test scores—much of this disparity is explained by differences in the students who select into OOH. In models where we account for selection into OOH and building type with rich controls and student fixed effects, we find small positive effects of moving into OOH on attendance and math scores with no consistent evidence of any impacts of OOH on BMI or obesity, suggesting that policies that promote homeownership might be oversold.

Introduction

According to a 2016 survey from the Pew Research Center, 81 percent of adults agree that “buying a home is the best long-term investment in the U.S” and 72 percent of renters indicated a desire to buy a house in the future (Fry and Brown, 2016). The view of homeownership as integral to the “American dream” is also enshrined in policies that are designed to promote homeownership, like those that allow homeowners to deduct mortgage payments from their taxes or others that offer down payment assistance for first-time homebuyers. Such policies are expensive, with the mortgage tax deduction alone costing an estimated $30 billion in forgone tax revenue in 2020 (Congressional Research Service, 2020). Whether or not these policies are worth their cost is unclear and depends, in part, on the extent to which owner-occupied housing (OOH) confers socially important benefits, such as higher earnings, better health, economic mobility or improvements in children’s academic outcomes. Indeed, theory suggests several reasons to believe that homeownership could be beneficial for both adults and children, through mechanisms such as wealth accumulation and residential stability, as well as the potential for OOH to promote neighborhood investment and cohesion. Yet identifying the effects of OOH is complicated, not only due to standard concerns about selection, but also because OOH tends to be located in neighborhoods with better amenities (including schools) and is often synonymous with single-family homes. Thus, it is difficult to disentangle whether differences in the outcomes of homeowners are driven by ownership or differences in these other factors.

Prior work has used various approaches to address issues of selection, yet we have made little progress on isolating the effects of OOH per se from characteristics of neighborhoods or housing on children. In this paper, we begin to fill this gap by examining the effects of living in OOH on children’s academic and health outcomes in New York City (NYC) using detailed longitudinal data and student fixed effects.

NYC is an ideal context to study the effects of owner-occupied housing for a number of reasons. First, there is wide variation in its housing stock, such that micro-neighborhoods are characterized by a mixture of owner-occupants and renters, single and multi-family buildings, etc. Furthermore, there are a large number of owners who live in multi-family buildings and renters who live in single-family homes. This allows us to disentangle the effects of OOH from other factors such as neighborhoods and building characteristics. Rich, longitudinal student-level data also allow us to address selection in multiple ways, including student fixed effects. Further, the broad variation in neighborhoods, which range from high density tourist destination neighborhoods in Manhattan to low density “bedroom communities” in Staten Island, allow us to explore heterogeneity across neighborhoods and gain insight into how – or where – these results might apply. In this paper we examine whether students in OOH have better academic and health outcomes than those in renter-occupied housing (ROH). We then explore the extent to which disparities are explained by differences in neighborhood and housing characteristics, or the types of students who live in OOH.

Briefly, we find that that there is notable variation in both the characteristics and size of OOH and the types of students who live in OOH in NYC. A substantial fraction of students in OOH live in medium/large apartment buildings and over 40 percent of students living in OOH in a given year are poor, as defined by eligibility for free or reduced-price lunch. While raw differences show that students who live in OOH have better outcomes—they are less likely to be chronically absent, obese, or overweight and have higher standardized test scores—much of this disparity is explained by differences in the neighborhoods in which OOH is located and the students who select into OOH and building types. In models where we account for these differences with rich controls and student fixed effects, we find small positive effects of moving into OOH on attendance and math scores, particularly for students moving into OOH in larger buildings. There is little consistent evidence of any impacts of moving into OOH on BMI or obesity or evidence that moving out of OOH is deleterious to student outcomes. These results are robust to alternative specifications including models that include both student and census tract fixed effects. Further, effects disappear in placebo models where students are assigned a random “OOH move-in” year prior to the year they actually move in or “OOH move-out” year after the year they actually move out.

Why might homeownership matter for children’s outcomes?

Theory

To better understand how OOH might affect student outcomes, regardless of whether the student’s parents or other adults in the household are themselves owner occupants, we draw upon a standard education production function:

$${\text{A}}_{\text{it}}=\text{g}\left({\text{F}}_{\text{it}},{\text{S}}_{\text{it}},{\text{P}}_{\text{it}},{\text{H}}_{\text{it}},{\text{O}}_{\text{it}},{\text{I}}_{\text{it}}\right)$$

where the outcome of student i in year t is a function of family characteristics (F), such as income, education or wealth; school inputs (S), such as class size or spending; peer characteristics (P), such as poverty status, home language, or prior academic performance of classmates; housing and neighborhood characteristics (H), including square footage, heating, ventilation and air conditioning (HVAC), and density; other external influences (O), such as residential stability; and individual student characteristics related to achievement (I), such as English language proficiency, home language, or disability. In this framework, owner-occupancy can affect student outcomes because it affects (or captures) changes in one of these underlying variables. That is, owner-occupancy may affect family characteristics, the quality of school inputs or peers, the characteristics of housing or neighborhoods, or the characteristics of the students themselves.

To begin, owner-occupancy may affect family characteristics (F) – either because it serves as a means of wealth accumulation (Boehm & Schlottmann, 2008; Herbert et al., 2013; Wainer & Zabel, 2020) or through increased employment and earnings of parents (Hausman, Ramot-Nyska, and Zussman, 2020). To the extent that increased wealth and earnings are used to invest in better educational or health resources, owner-occupancy may improve student outcomes such as attendance, performance, and BMI. Further, the well-documented clustering of OOH families means children living in OOH may be exposed to higher performing peers (P) than those living in ROH. Similarly, if homeownership increases residential stability, clustering of OOH families may also mean that students living in OOH attend school with more stabile peers. Previous evidence on the negative spillovers of mobility on non-mobile students (Whitesell et al., 2016), suggests this could lead to increased performance and attendance.

Turning to housing and neighborhoods (H), if OOH units are more likely to be larger or offer backyard space, students living in OOH may have better access to a quiet place to study or outdoor areas that facilitate physical activity. To the extent that owner occupants are more likely to engage in regular maintenance and upgrades of their properties, living in OOH may improve housing quality, which could affect student academic and health outcomes through a variety of pathways including ability to concentrate, improved air quality, or ventilation (Currie & Yelowitz, 2000; Leventhal & Newman, 2010). Furthermore, prior evidence suggests that owner occupants may be more likely to support spending on amenities because it will ultimately be capitalized into housing prices (Fischel, 2001; Hilber & Mayer, 2009). Therefore, in addition to better schools (S), students living in OOH may have better access to parks and open spaces, which could reduce BMI, obesity, and overweight. Owner-occupancy may also increase social cohesion--feelings of connectedness and solidarity--in the neighborhood, which may yield a variety of benefits for children including better health and/or higher educational attainment (Ferguson, 2006). Finally, owner-occupancy may reduce neighborhood crime rates (Glaeser & Sacerdote, 1999). This may also improve student academic performance (Schwartz et al., 2021; Sharkey et al., 2014; Sharkey, 2010) and health outcomes, as crime is associated with higher BMI and obesity (Laurito et al., 2022; Theall, et al., 2019; Yu & Lippert, 2016).

Lastly, homeownership may affect student outcomes by increasing residential stability (O) due to increased residential satisfaction or higher moving costs (Warner & Sharp, 2016; Speare, 1974; Landale & Guest, 1985; Aaronson, 2000). Reducing residential mobility, which has been shown to negatively affect student performance (Scanlon & Devine, 2001; Rumberger, 2002; Cordes et al., 2019), and/or reducing school mobility, which has also been shown to be detrimental to student outcomes (Schwartz et. al, 2017), may improve student outcomes. That said, it is possible that the reduced mobility due to homeownership has negative effects if, say, families are prevented from moving to better-matched housing, neighborhoods, or schools by housing market declines, for example. While evidence is thin, Been et al., (2021) suggest that families with high loan-to-value ratios -for whom moving costs would likely be higher- are more likely to move (Been et al., 2021).

Existing Evidence of Impacts on Children

A small body of existing literature explores the effects of homeownership on educational outcomes, which finds mixed results. Early work was somewhat positive. Using the Panel Study of Income Dynamics (PSID), Green and White (1997) find that parental homeownership is associated with a lower high school dropout rate and lower teen birth rate (age 17 or younger) after controlling for a rich set of child and family characteristics. Further, they find no evidence of selection using a bivariate probing model, although Barker and Miller (2009) later argue that the instrument used for this test—the value of housing CPI in the year the family last moved divided by the national average 30-year mortgage rate in that year--may not capture the full economic costs of moving and that therefore the results in Green and White (1997) might still suffer from selection bias.

Also using the PSID, but accounting for selection by instrumenting for homeownership using annual state average homeownership rates by race and income quintile, Aaronson (1999) finds positive effects of parental homeownership on graduation outcomes, although these effects entirely driven by greater residential stability among homeowners. These findings were confirmed by Galster et al (2007) who use a complex set of 13 different instruments meant to capture the relative affordability of owning and unobserved characteristics of the housing stock that might drive differences in homeownership and find significant effects of parental homeownership on graduation and attainment that are reduced and/or become insignificant after controlling for residential stability. Harkness and Newman (2003) also use the PSID to explore the effects of homeownership from ages 11 to 15 on a variety of later life outcomes, including teen births (females only), idleness, graduation, attainment, and earnings and whether effects differ for children growing up in low-income families. Using annual changes in the state’s per capita highway investment, ratio of renter to owner costs in the census region, metropolitan area or county ration of median rent to median property value, and state homeownership rate as instruments for homeownership, they find that homeownership during childhood leads to higher attainment and earnings and also decreases the likelihood of welfare receipt for children growing up in low-income families with no benefits for those from higher income families. Finally, using the National Longitudinal Survey of Youth (NLSY), Haurin, Parcel, and Haurin (2002) instrument for homeownership using house price and an indicator of whether the household was down payment constrained, and find that living in an owned home increases math and reading scores and decreases behavior problems.

Although each of these studies relies on a somewhat different set of instruments, the intuition behind the instruments across all papers is roughly the same—the effects of homeownership are identified from cross-sectional variation in the relative “ease” and affordability of homeownership experienced by households in the sample. There are two overarching considerations when interpreting this body of evidence. The first consideration is the strength of the instruments and plausibility of the exclusions restrictions, which determines the extent to which these estimates can be interpreted as causal. Here, the underlying logic is reasonable--students are more likely to live with parents who are homeowners in locations where housing is relatively more affordable—and conditional on background characteristics housing affordability may be otherwise uncorrelated with outcomes. However, in all cases the instruments are relatively weak or no F-statistics are reported, warranting some caution in interpretation. The second consideration is that these estimates represent the local average treatment effect among households for whom the instruments predict homeownership—those for whom homeownership is relatively more affordable--which may not represent the broader population of homeowners. Thus, estimates from these analyses may not shed light on the effects of homeownership or OOH in places with tight housing markets, such as NYC.

Later U.S. based work has been less positive. Barker and Miller take (2009) use the NLSY and Early Childhood Longitudinal Study (ECLS) to compare changes in outcomes from students who transition from own-to-rent to outcomes from students who transition from rent-to-own controlling for a variety of student and family background characteristics. They find no effects of homeownership on the cognitive environment, behavior, reading recognition, reading comprehension or test scores among children. Finally, using the PSID and NLSY, Holupka and Newman (2012) employ a matching approach based on background characteristics, region of residence, metropolitan area house prices, and whether the child lived in a nonmetro area for more than half of childhood. They find little evidence that homeownership affects children’s cognitive performance, behavior, or health.¹ While both papers control for a broad set of observed characteristics, they may not fully account for unobserved differences between children with different levels of exposure to homeownership, and neither study includes detailed controls for neighborhood characteristics, which are an important pathway through which homeownership may be linked to student outcomes.

More recent work drawing on data from Norway is intriguing. Aarlan et. al (2021) use data from 1990–2014 and a family fixed effects approach to explore the effects of housing tenure on educational attainment in Norway. They find that compared to other siblings in their household, youth who lived one more year in a home owned by their parents were more likely to complete high school by age 21 and more likely to enroll in college. Further, these effects appear to be due to home ownership rather than dwelling type, mobility, or neighborhood. As the authors note, the primary appeal of a family fixed effects approach is that it accounts for all family characteristics that siblings share, such as family structure and general tastes for education. However, there are also two key drawbacks. First, this approach can only identify effects among families with two or more children. Second, while models control for birth order and birth year, there may be other systematic unobserved differences between siblings who are exposed to different years of homeownership that explain some of the observed effects.

To summarize, almost all of the U.S. studies draw on data on children growing up in the 1970s and 1980s. Given changes in the mix of homeowners over the past several decades, the foreclosure crisis, etc., results may well be different for more recent cohorts. Furthermore, while these studies control for a wide range of individual differences, some using more detailed information on parental education, family structure, and parental wealth than we use here (but are accounted for with student fixed effects), and many account for selection using an IV approach that identifies homeownership based on variation in the relative affordability of homeownership, most prior work does not account for the differences in neighborhood and housing stock between owner-occupants and renters. While Aarlan et. al (2021) address several of these issues with more recent data and controls for housing and neighborhood characteristics, they do so in the Norwegian context, where educational systems and housing policy is markedly different from the U.S Further, none of these studies explore the potential impacts of homeownership on childhood BMI, overweight, or obesity, nor do they employ student or child fixed effects to address unobserved individual differences. We do so here.

Empirical Strategy

There are two primary challenges to identifying the effects of OOH on student outcomes. First, differences in outcomes between students living in OOH and ROH may reflect differences between the students/families choosing OOH and those choosing ROH. Those in OOH may have higher tastes for education or health that lead them to seek lower density housing and higher quality neighborhoods, or they may be wealthier, more patient or prefer residential stability. Indeed, as shown in Table 1 and is confirmed by Mohanty and Raut (2009), owner-occupants differ from renters in myriad ways. Second, differences in outcomes between students in OOH and ROH may reflect differences in the neighborhoods surrounding OOH and ROH. Indeed, as shown in Table 1, OOH is located in neighborhoods with lower density, better amenities (including schools), and higher median income, all of which are likely to be positively related to health and academic outcomes. Thus, differences in outcomes are likely to overstate the effect of OOH per se on student outcomes because they also capture these neighborhood effects. To address these issues, we estimate the following

$$Y_{it}=\beta 1MOVEIN\_OOH{H}_{it}+\beta 2MOVEOUT\_OO{H}_{it}+{\mathit{\beta }}_{\mathit{3}}\mathit{B}\mathit{U}\mathit{I}\mathit{L}\mathit{D}\mathit{S}\mathit{I}\mathit{Z}{\mathit{E}}_{ht}+{\text{α}}_{\text{g}}+{\text{γ}}_{\text{t}}+{\text{η}}_{\text{b}}+{\text{θ}}_{\text{i}}+{\epsilon }_{\text{ighbt}}$$ (1)

Where Y is an academic or health outcome for student i in in year t, MOVEIN_OOH is an indicator equal to all years after a student moves into OOH from ROH, MOVEOUT_OOH is an indicator equal to all years after a student moves into ROH from OOH, α are grade effects, γ are year effects, η are borough effects, θ are student fixed effects, and ε is the error term, clustered at the school-year level because schools administer and collect both test score and Fitnessgram data.² The coefficients of interest in this model are β₁ and β₂, which are identified by the within student variation in housing tenure – put differently, they are only identified by students observed in both OOH and ROH. We separate the effects of moving into OOH and moving out of OOH for two reasons. First, due to the time period of the study, which spans the great recession and volatile housing market, moves out of OOH could be driven by foreclosure, which could lead to a spurious negative relationship between changes in housing tenure and student outcomes. Second, it is possible that moves into and out of OOH are driven by different processes and effects may differ. While our data do not include detailed measures of parental/family structure or financial circumstances, the use of student fixed effects accounts for unobserved characteristics of families that do not vary over our analysis period, such as preferences for education or average wealth, that may drive selection into OOH, housing type, and student outcomes The disadvantage is that, by construction, this group may differ from the full set of students in NYC public schools.

Table 1.

Characteristics by Tenure, AY 2012, Grades 4–8

	All Students	Owner Occupied	Renter Occupied
Housing Characteristics
Single family	0.15	0.57	0.05
2–4 family	0.39	0.01	0.48
5–19 family	0.10	0.04	0.12
20+ family	0.36	0.39	0.35
Building age	72.79	63.01	75.10
# of stories	4.26	5.37	4.00
Elevator	0.22	0.29	0.21
Sq. ft. per res. unit	1,326.95	2,327.03	1,088.39
Value per sq. foot ($100s)	116.50	598.88	1.49
Owner occupied	0.19	1.00	0.00
Neighborhood Characteristics
% White	38.3	50.0	35.5
% Black	27.0	21.9	28.2
% Asian	14.1	16.1	13.7
% Hispanic	32.9	20.1	36.0
% Below poverty	7.4	5.8	7.8
% Owner occupied	31.4	49.0	27.1
% in majority (> 50%) owner occupied tract	23.1	49.8	16.7
Median HH income	50,937.0	66,337.5	47,266.6
Population Density	57,481.4	42,971.2	60,942.6
Borough
Manhattan	0.09	0.11	0.08
Bronx	0.20	0.10	0.23
Brooklyn	0.31	0.18	0.34
Queens	0.33	0.44	0.30
Staten Island	0.07	0.17	0.05
Student Characteristics
White	0.17	0.34	0.13
Black	0.25	0.22	0.26
Hispanic	0.39	0.20	0.44
Asian/other	0.19	0.24	0.18
Female	0.50	0.50	0.50
Poor	0.67	0.46	0.72
LEP	0.11	0.04	0.13
SWD	0.11	0.10	0.11
Student outcomes
Attendance (0–1)	94.9	96.0	94.6
Chronic Absenteeism	0.14	0.08	0.15
zELA	0.16	0.51	0.08
zMath	0.18	0.49	0.11
zBMI	0.65	0.49	0.68
Obese	0.22	0.18	0.23
Overweight	0.42	0.36	0.43
Observations	200,284	38,574	161,710

Note: All differences between OOH and renter housing significant at the 0.01 level, except female.

To explore the stability of our estimates, we re-estimate equation (1) adding controls for time-varying student characteristics including poverty (as measured by eligibility for free or reduces price lunch), limited English proficiency, and receipt of special education services; building characteristics including building size (i.e., single-family home, 2–4 family), number of stories, and building age; tract characteristics including racial composition, percent of units occupied by owners, median household income of renters, median household income of owners, percent of residents with more than a high school diploma, poverty rate, unemployment rate, and population density; and school characteristics including ELA proficiency, math proficiency, pupil-teacher ratio, percent of students eligible for free or reduced price lunch, racial composition, log per pupil expenditures.³ These additional controls should help to capture any changes in family circumstances leading to changes in OOH that are not accounted for by student fixed effects. To the extent that our estimates are stabile with the addition of these controls, it will lend confidence that estimates are not picking up these changing circumstances.

Given that most of the work on homeownership has focused on single-family homes (SFH), we then explore heterogeneity in the effect of OOH by building size. To do so, we replace individual MOVEIN_OOH and MOVEOUT_OOH indicators with a series of indicators capturing moves into and out of OOH by building types (single-family, 2–3 family, 5–19 family, and 20+ family), yielding a set of β₁ and β₂ that capture the disparities between students moving into an out of OOH and those living in rental housing for each of these housing types. More specifically, move into OOH, single family, captures the effect of moving into OOH SFH from ROH, while move out of OOH, single family captures the effect of moving out of OOH SFH into ROH.

Data, Sample, and Measures

Data

We use detailed longitudinal student-level data from 2007–2012 provided by the NYC Department of Education (NYCDOE), which we link with building-level data from the NYC Department of Finance (NYCDOF), property-level tax exemption data from the Department of Taxation, census-tract characteristics from the American Community Survey (ACS) 5-year estimates for 2012, and school characteristics from the New York State School Report Card (SRC). Student-level data from the NYCDOE include race/ethnicity and eligibility for free or reduced price lunch, program participation (limited English proficiency, receipt of special education services, etc.), attendance, and performance on standardized tests given in grades 3–8. These also include data on student height and weight, body mass index (BMI) and indicators for obese and overweight and performance on a series of fitness tests based upon the annual @FITNESSGRAM. Importantly for this paper, these also contain a building-block-lot identifier (BBL), which identifies the building where a student lives. We use student BBLs to link data on building characteristics from the NYCDOF Real Property Assessment Database, including number of residential units, building age, gross square footage, current value, etc. We also use BBLs to link students with property-level data on the New York State (NYS) School Tax Relief (STAR) program, which we use as our primary measure of owner-occupancy in single family homes. The NYS STAR credit is a school tax relief program for residents that own their primary residence, but earn less than $500,000.⁴ Since the STAR credit can only be applied to an individual’s primary residence and most eligible owners register for the credit, this is a good measure of owner-occupancy.⁵ Further, the distribution of household income in NYC over this period suggests the income limit is unlikely to be binding for most families with children in public schools. For example, according to the 3-year ACS estimates for 2010–2012, the mean household income in the highest income quintile ranged from $123,950 in the Bronx to $393,956 in Manhattan, and the median income of families with children under 18 ranged from $18,901 in the Bronx to $31,789 in Queens.

Sample

Our sample includes students in grades 4–8 from academic years (AYs) 2006–07 to 2011–12 who live in residential buildings and have at least two years of Fitnessgram and test score data, required to estimate models with lagged outcomes and/or student fixed effects. In addition, we exclude students living in public housing (approximately 9 percent of students in grades 4–8) from our main results because long waiting lists, low exit rates, subsidized rent, and ancillary services may make them different than students living in OOH in unobserved ways. Finally, we exclude 815 observations with biologically implausible BMI and 2,664 students who are observed both moving into and out of OOH.

Measures

We focus on three key weight measures: BMI normalized by gender and age to mean zero and standard deviation 1, and indicators of obesity (BMI greater than or equal to 95^th percentile for age) and overweight (BMI between 85^th and 95^th percentile for age). For simplicity, we present only results for zBMI in the main set of tables and include results for overweight/obesity, which are largely similar, in Appendix B. Key academic outcomes are attendance rate (percent of days enrolled that a student is present), chronic absenteeism (absent more than 10% of days enrolled), and reading and math test scores standardized by grade and year to mean of 0 and standard deviation 1. Attendance and chronic absenteeism are important measures of academic outcomes as chronic absenteeism has been linked to increased high school dropout, and delinquency and substance abuse in later adolescence and early adulthood (Wang & Fredricks, 2014; Henry et al, 2012; Hirshfield & Gasper, 2011). Further, under the Every Student Succeeds Act (ESSA), the federal law governing school accountability, New York and many other states use chronic absenteeism as a school accountability metric, such that attendance and chronic absenteeism are policy-relevant metrics.

The key to our empirical work is identifying students living in OOH which we do at the building level. Thus, students living in OOH includes two types. First, the set of students living in a building that receives the STAR credit for owner occupants and, second, the set of students student living in a condominium or coop building. Thus, if a STAR credit is claimed for a multi-family building because one of the units is the primary residence of an owner, all students living that building are classified as living in OOH.⁶ Further, all residents of condos and coops (2.1 and 5.2 percent of our observations, respectively) are defined as living in OOH, since 50 percent of condo and 70 percent of coop units in the NYC metropolitan area were owner-occupied according to the 2009 American Housing Survey.⁷ To be clear, students living in OOH consist of two groups: students living with their owner-occupant parents/guardians and students living with their renter-occupant parents/guardians in a building where are least one other at least one other resident is the owner-occupant. We classify all other buildings as ROH and students living in those buildings as living in rental housing.

The implication is that, in buildings other than single-family homes, the estimated effects of OOH will capture a weighted average of the effects of owner-occupancy per se and the effects of living in an owner-occupied building. The meaning of this for our results is unclear and depend on both the share of owner-occupants in OOH buildings and the effect of owner-occupancy relative to the effects of living in OOH. If, for example, the effects of owner-occupancy are larger than those for living in an owner-occupied buildings then our results are likely to be underestimates of the effects of owner-occupancy. Further, to the extent that we misclassify buildings as OOH—for example, if there are condos in our sample with no owner-occupants-- this will bias our results towards zero.⁸

Next, we explore a number of building characteristics that might be associated with health or academic outcomes including building age (a proxy for quality), whether the building has an elevator, number of stories, and square footage per residential unit. We also include measures of what we refer to as building type, which is based on the number of residential units: single-family homes (1 residential unit), 2–4 family buildings (2–4 residential units), 5–19 family buildings (5–19 residential units), and 20+ family buildings (20 or more residential units). This building measure is meant to distinguish between SFHs, small, medium, and large buildings.

Finally, we include a set of controls for tract and school characteristics. Tract characteristics come from 2012 ACS and include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, and population density. School characteristics come from the SRC and include ELA proficiency, math proficiency, pupil-teacher ratio, percent of students eligible for free or reduced price lunch, racial composition, and log per pupil expenditures.

Results

What are the characteristics of students in OOH?

Despite the popular conception of NYC as a city of renters living in large, multi-family buildings, the housing stock is diverse: nearly one in five students (19 percent) lives in OOH. and the majority (64 percent) live in buildings with fewer than 20 residential units. More specifically, 15 percent live in single-family homes, 39 percent in 2–4 family homes, and 10 percent live in medium-sized, 5–19 family buildings (Table 1). Compared to NYC as a whole, which was 32 percent White, 29 percent Hispanic, 22 percent Black, and 14 percent Asian, the average student in our sample lives in a neighborhood that is roughly representative although somewhat disproportionately White (38.3 percent of residents) and Black (27.0 percent of residents). Further, the average student lives in a neighborhood where 31.4 percent of housing is owner-occupied and 23.1 percent of students live in neighborhoods where over 50 percent of units are owner occupied. The majority of our students live in Brooklyn or Queens and roughly two-thirds are either Black (25 percent) or Hispanic (39 percent). Average test scores are 0.16 standard deviations (sds) in ELA and 0.18 sds in math, indicating that our sample is somewhat positively selected. This may reflect the exclusion of students without at least two years of outcome data who may be more mobile and/or lower performing as well as the exclusion of students in public housing.⁹ Finally, roughly two thirds of our students are overweight or obese - nearly a quarter are obese and another 42 percent overweight.

There are key differences in these characteristics by housing tenure. While more than half of students (57 percent) in OOH live in single-family homes, only 5 percent in ROH do so. Further, few students in OOH live in small multifamily buildings while almost half (48 percent) of students in ROH do so. On average, students in OOH live in buildings with more square feet per residential unit and higher building valuation per square foot. That said, roughly equal shares of students in OOH and ROH live in large apartment buildings (39 percent vs. 35 percent).

Students in OOH live in neighborhoods with notably higher percentages of White residents (50.0 percent versus 35.5 percent) and owner-occupied housing units (49.0 percent versus 27.1 percent), as well as a lower percentage of residents living below the poverty line, higher median household incomes, and lower population density. Similarly, while almost half of students in OOH live in majority owner-occupied neighborhoods, this is true for only 16.7 percent of students in ROH.

Similarly, a higher share of students living in OOH are White (34 percent versus 11 percent) and Asian (24 percent versus 16 percent), while a lower share are Hispanic (20 percent versus 44 percent). Further, lower shares of students in OOH are limited English proficient or eligible for special education services. Notably, even though a lower share of students in OOH are poor compared to students in ROH, nearly half of students in OOH qualify for free or reduced price lunch. While this contradicts popular perceptions that only the wealthy live in OOH in NYC, the city itself includes many low-income, low density neighborhoods as well as the high income, high density neighborhoods so often featured in popular press.

Taken together, the differences in housing characteristics, neighborhoods, and individual characteristics between students living in OOH and those living in rental housing suggest that students who live in OOH are likely to have better outcomes. They live in housing likely to offer more space to sleep, study and play, and in neighborhoods likely to offer higher quality educational and health amenities. Further, they are disproportionately White and not low-income. Thus, our findings of better outcomes among students in OOH across a variety of dimensions are unsurprising. They have higher attendance and test scores, lower zbmi, and lower prevalence of obesity or overweight.

These differences highlight the importance of accounting for potential selection into OOH, and for neighborhoods and housing characteristics in estimating the effects of OOH.

How much movement is there between housing types?

While the descriptive statistics in Table 1 help to paint a portrait of who lives in OOH, they do not indicate the extent to which students move between OOH and ROH, or between smaller and larger buildings. Understanding this movement sheds light on the identifying variation in student fixed effects models. To explore the number and kinds of moves between types of housing, we focus on the sample of students who make residential moves at any point between 2007 and 2012 and compare their housing type in the current year (t) to housing type in the subsequent year (t+1). During our sample period, we observe 89,658 students making slightly over 100,000 moves. We first examine transitions from OOH to ROH and then explore transitions between housing types.

Although slightly less than two-thirds (63.1%) of moves among students in OOH are from OOH into ROH, this still leaves slightly over one third of students who move from one owner-occupied building to another (Table 2). That is, the majority of students movers in OOH are moving out of OOH. In contrast, the vast majority of moves among students in ROH to other ROH (90.0%), while only 10.0 percent move into OOH. That is, the majority of student movers in ROH stay in ROH. In terms of absolute numbers, however, there are slightly more moves from ROH to OOH (9,133) than from OOH to ROH (7,317). Since moves into OOH conflate both moves into owner-occupancy and moves by renters into an owner-occupied building, these patterns likely reflect a combination of the challenges of purchasing a home and a lower supply of units in owner-occupied buildings.

Table 2:

Housing Ownership Transition Table, Academic Years 2007 – 2012, Student Moves

Number (Row %) (Total %)	Owner Occupied Housing in Year t+1	Renter Occupied Housing in Year t + 1	Total
Owner Occupied Housing in Year t	4,277 (36.9%) (4.2%)	7,317 (63.1%) (7.1%)	13,217 (100.0%) (11.3%)
Renter Occupied Housing in Year t	9,133 (10.0%) (8.9%)	81,935 (90.0%) (79.8%)	89,450 (100.0%) (88.7%)
Total	13,410 (13.1%)	89,252 (86.9%)	102,667 (100.0%)

Notes: The sample includes only students with at least 2 years of attendance, test scores, and weight outcomes from 2007 – 2012 and are observed in 2 consecutive years. Each observation represents a student’s residential move, which is defined as the student living in a different building in year t+1 than in year t. A student may be included more than once if they moved multiple times during the sample period.

There are 89,658 unique students and roughly 87% of these students moved only once.

The most common transition for students in all housing types, with the exception of 20+ family buildings, is into a 2–4 family building (Table 3). For example, while slightly over one-third of students in SFH move to another SFH, almost half (47.5%) move to a 2–4 family building, and a similar share of students in 5–19 family homes (45.2%) also move to a 2–4 family building. While students in 2–4 family buildings are the most likely to move to another 2–4 family building (60.5%), the next most common move is to a 20+ family building (17.3%). Among students in large buildings, about half (51.5%) move into another large building, but even among this group of students, almost a third move into a 2–4 family building.

Table 3:

Housing Type Transition Table, Academic Years 2007 – 2012, Student Moves

Number (Row %) (Total %)		Housing Type in Year t +1
		SFH	2–4 Family	5–19 Family	20+ Family	Total
Housing Type in Year t	SFH	3,698 (33.6%) (3.6%)	5,229 (47.5%) (5.1%)	467 (4.2%) (0.5%)	1,626 (14.8%) (1.6%)	11,020 (100.0%) (10.7%)
	2–4 Family	6,208 (13.2%) (6.0%)	28,472 (60.5%)	4,205 (8.9%)	8,155 (17.3%)	47,040 (100.0%)
	5–19 Family	779 (6.4%)	5,527 (45.2%)	2,485 (20.3%)	3,435 (28.1%)	12,226 (100.0%)
	20+ Family	2,365 (7.3%) (2.3%)	10,088 (31.2%) (9.8%)	3,257 (10.1%) (3.2%)	16,666 (51.5%) (16.2%)	32,376 (100.0%) (31.5%)
	Total	13,050 (12.7%)	49,316 (48.0%)	10,414 (10.1%)	29,882 (29.1%)	102,662

Notes: The sample includes only students with at least 2 years of attendance, test scores, and weight outcomes from 2007 – 2012 and are observed in 2 consecutive years. Each observation represents a student’s residential move, which is defined as the student living in a different building in year t+1 than in year t. A student may be included more than once if they moved multiple times during the sample period. There are 89,658 unique students and roughly 87% of these students moved only once.

Appendix Table A1 shows transitions from OOH to ROH by housing type, which is the source of identifying variation for our student fixed effects models. Of the 9,133 moves from ROH into OOH, the largest group is students moving from renter occupied 2–4 family homes into OOH SFH (3,156), followed by students moving from ROH 20+ family buildings to OOH buildings of the same size (1,636), students moving from ROH 2–4 family homes to OOH 20+ family buildings (1,508), and students moving from ROH 20+ unit buildings to OOH SFH (812). Therefore, while some of the moves into OOH involved substantial changes in building size, many of moves from rental housing to OOH are also made to buildings of a similar size.

Effects of OOH on Outcomes

Unlike the large raw differences shown in Table 1, estimates of the OOH premium are small or insignificant (Table 4, Panel A), which suggests that much of the observed benefit of OOH may be due to unobserved characteristics of students and families who select into OOH. Specifically, although there is a benefit to moving into OOH for attendance (0.265pp), chronic absenteeism (−0.7pp), and math scores (0.013 sd), these effects are relatively small and all other estimates are statistically insignificant and close to zero. Further, if there was a large effect of OOH per se, we might expect to see large reductions in outcomes among students who move out of OOH, which we do not observe. Indeed, we find little evidence of harm, as all estimates are insignificant with the exception of ELA scores, where we find that moving out of OOH may confer a slight benefit. These estimates are insensitive to controls for student characteristics (Panel B), building characteristics (Panel C), and tract characteristics (Panel D). When we control for school characteristics, the effects of moving into OOH on math scores become insignificant, effects on attendance attenuate slightly, and the R-squared increases slightly (Panel E).¹⁰ However, these models may be overcontrolling, as the degree of school choice in NYC means that schools are not always tied directly to neighborhoods or housing. Therefore, our preferred specification includes controls for student, building, and tract characteristics, although we do not rule out that some of the small observed benefit to OOH may be due to differences in school quality experienced by students in OOH and ROH.

Table 4.

Owner Occupied Housing and student outcomes, AY 2007–2012, Grades 4–8

	(1) Attendance	(2) Chronic Absenteeism	(3) zELA	(4) zMath	(5) zBMI
Panel A: Building Size
Move into OOH	0.265***	−0.007**	0.000	0.013**	−0.003
	(0.043)	(0.003)	(0.008)	(0.006)	(0.007)
Move out of OOH	−0.013	−0.001	0.017**	−0.002	−0.015
	(0.050)	(0.004)	(0.008)	(0.007)	(0.008)
R 2	0.784	0.668	0.783	0.835	0.846
Panel B: Building Size and Student Characteristics
Move into OOH	0.264***	−0.007**	0.001	0.015**	−0.004
	(0.043)	(0.003)	(0.008)	(0.006)	(0.007)
Move out of OOH	−0.015	−0.001	0.019**	−0.000	−0.015
	(0.050)	(0.004)	(0.008)	(0.007)	(0.008)
R 2	0.784	0.668	0.784	0.835	0.847
Panel C: Building Size, Student, and Building Characteristics
Move into OOH	0.262***	−0.007**	−0.000	0.015**	−0.005
	(0.043)	(0.003)	(0.008)	(0.006)	(0.007)
Move out of OOH	−0.013	−0.001	0.021**	−0.000	−0.015
	(0.050)	(0.004)	(0.008)	(0.007)	(0.008)
R 2	0.784	0.668	0.784	0.835	0.847
Panel D: Building Size, Student, Building, and Tract Characteristics
Move into OOH	0.247***	−0.006	−0.001	0.015**	−0.004
	(0.044)	(0.003)	(0.008)	(0.006)	(0.007)
Move out of OOH	0.000	−0.002	0.020**	−0.000	−0.015
	(0.050)	(0.004)	(0.008)	(0.007)	(0.008)
R 2	0.784	0.668	0.784	0.835	0.847
Panel E: Building Size, Student, Building, Tract, and School Characteristics
Move into OOH	0.262***	−0.007**	−0.002	0.008	−0.001
	(0.043)	(0.003)	(0.007)	(0.006)	(0.007)
Move out of OOH	0.006	−0.003	0.017**	−0.008	−0.012
	(0.051)	(0.004)	(0.008)	(0.007)	(0.008)
R 2	0.785	0.669	0.785	0.839	0.847
N	1,240,042	1,240,042	1,240,042	1,240,042	1,218,967

Standard errors clustered by School and Year, and are shown in parentheses,

^***p<0.01

^**p<0.05

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. All models include controls for building size (single family, 2–4 family, and 5–19 family), year, and grade effects. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data. School characteristics include ELA proficiency, math proficiency, pupil-teacher ratio, percent of students eligible for free or reduced price lunch, racial composition, log per pupil expenditures, and controls for missing data. Missing data on school and tract characteristics are replaced with the average for the borough in that year. Attendance ranges from 0 to 100.

Estimated effects are also largely consistent across boroughs, which vary widely in terms of density and housing stock (Appendix Table A3). We find that students moving into OOH have higher attendance and math scores in all boroughs except the Bronx and Staten Island, and lower rates of chronic absenteeism in Brooklyn. Among students moving out of OOH, we observe higher attendance and test scores in Manhattan and higher ELA scores and lower zBMI in Queens. The only evidence of deleterious effects of moving out of OOH is lower attendance in Staten Island and math scores in the Bronx. All other estimates are insignificant.

Robustness

Even with a rich set of controls and student fixed effects, estimates may still be biased by unobserved, time-varying characteristics of students and families that precipitate changes in housing tenure, such as shocks to income or household structure. To test this concern, we conduct a placebo analysis where we estimate the “impact” of moving into and out of OOH using a student fixed effects model with randomly generated “OOH entry” dates in the “pre-OOH” period for students who move into OOH and randomly generated “OOH exit” dates in the “post-OOH” period for students who move out of OOH. Specifically, for students who we observe moving into OOH, we drop all observations they actually move into OOH and then randomly generate a random “OOH entry” year. For students who we observe moving out of OOH, we drop all observations before they actually move out of OOH and then randomly generate a random “OOH exit” year. Thus, we estimate these effects of moving into and out of OOH using only observations of students while they actually live in ROH. We then re-estimate our student fixed effects model using these random entry (move into OOH) and exit (move out of OOH) years. As shown in Table 5, these placebo estimates are statistically insignificant. This suggests that our estimates are indeed identifying the effects of OOH rather than changes in family circumstances that may accompany changes in tenure.¹¹

Table 5.

OOH and Student Outcomes, AY 2007–2012, Grades 4–8, Student Fixed Effects, Placebo Effects

	(1) Attendance	(2) Chronic Absenteeism	(3) zELA	(4) zMath	(5) zBMI
Main estimates
Move in OOH	0.247*** (0.044)	−0.006 (0.003)	−0.001 (0.008)	0.015** (0.006)	−0.004 (0.007)
Move out of OOH	0.000 (0.050)	−0.002 (0.004)	0.020** (0.008)	−0.000 (0.007)	−0.015 (0.008)
R 2	0.784	0.668	0.784	0.835	0.847
Obs.	1,240,924	1,240,924	1,240,924	1,240,924	1,219,803
Placebo Estimates
Placebo Move into OOH	0.070 (0.067)	−0.006 (0.006)	−0.014 (0.014)	0.004 (0.011)	−0.016 (0.013)
Placebo Move out of OOH	0.043 (0.078)	0.001 (0.007)	−0.001 (0.018)	0.018 (0.014)	0.009 (0.017)
R 2	0.788	0.675	0.789	0.838	0.849
Obs.	1,197,912	1,197,912	1,197,912	1,197,912	1,177,608

Standard errors clustered by School and Year, and are shown in parentheses,

^***p<0.01

^**p<0.05

Notes: For students who ever move into OOH, placebo Move in OOH equals 1 in a randomly selected year before a student is actually observed in OOH and in all years after. Observations for students ever moving into OOH are dropped in the years they actually live in OOH. For students who ever move out of OOH, placebo Move out OOH equals 1 in a randomly selected year after a student actually moves out of OOH and in all years after. Observations for students ever moving out of OOH are dropped in the years they actually do not live in OOH. All models include controls for building size (single family home, 2–4 family, and 5–19 family), student characteristics, building characteristics, tract characteristics, year, and grade effects. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. Attendance ranges from 0 to 100.

As a second robustness check, we re-estimate our models with both student and census tract fixed effects simultaneously (Appendix Table A4) to address the possibility that some of the changes we observe in outcomes is explained by unobserved changes in neighborhood of residence rather than OOH. In these models, we drop controls for neighborhood characteristics as they do not vary over time within census tracts. General patterns remain the same, although the magnitude of coefficients is slightly smaller and the point estimate for the effect moving into OOH on math scores, while positive, is no longer statistically significant.

As a third robustness check, we re-estimate our results including students in public housing as part of the sample. While students who live in public housing may differ from those in OOH in unobserved ways, several of the mechanisms through which owner-occupancy is theorized to improve student outcomes are shared with features of public housing—namely predictable monthly payments and residential stability. Our results for OOH are largely similar (Appendix Table A5). Further, we find no relationship between public housing and most student outcomes, except for a decrease in chronic absenteeism, which suggests that common features between OOH and public housing are unlikely to explain the small effects of OOH on attendance and math scores.

Finally, one explanation for our small or insignificant findings is that estimates are primarily identified based on students who have been living in OOH for relatively short periods of time. We explore this possibility by examining whether the effects of moving into OOH differ for students who have lived in OOH for 2, 3, or 4 years compared only one year (Table 6). In these models, we find some evidence that the effects of OOH are cumulative. For example, improvements in attendance are increasing with the number of years in OOH and, while we find small average effects of moving into OOH on chronic absenteeism, there does appear to be an effect among students who have lived in OOH for two or three years.¹² However, as with our main estimates, effects are relatively small, even among students who have lived in OOH for multiple years. Alternatively, we examine whether the effects of OOH differ based on the grade in which a student first moves into OOH (Appendix Table A6). To the extent that dosage is important, we should expect to see larger effects of OOH among students who move into OOH in earlier grades. We find some evidence that this is the case, particularly for attendance, where effects are largest among students who move into OOH before 8^th grade, although again, the magnitude of these effects is small.

Table 6.

OOH and Student Outcomes, AY 2007–2012, Grades 4–8, Dosage

	(1) Attendance	(2) Chronic Absenteeism	(5) zBMI
Move into OOH	0.135***	−0.002	−0.003
	(0.046)	(0.003)	(0.008)
OOH for 2 years	0.371***	−0.013***	−0.004
	(0.057)	(0.004)	(0.010)
OOH for 3 years	0.336***	−0.012***	−0.013
	(0.073)	(0.006)	(0.014)
OOH for 4 years	0.491***	−0.000	−0.015
	(0.161)	(0.013)	(0.031)
Move out of OOH	0.012	−0.003	−0.016***
	(0.051)	(0.004)	(0.008)
R 2	0.784	0.668	0.847
N	1,240,418	1,240,418	1,219,269

Standard errors are clustered by School and Year, and is reported in parentheses

^**p < 0.05

^***p < 0.01

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. OOH for 2 years is equal to 1 if a student moved into OOH and has lived in OOH for two years as of year t. All models include controls for building size (single family home, 2–4 family, and 5–19 family), year, and grade effects. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. School characteristics include ELA proficiency, math proficiency, pupil–teacher ratio, percent of students eligible for free or reduced price lunch, racial composition, log per pupil expenditures, and controls for missing data on school characteristics. Missing data on school characteristics are replaced with the average for a student’s residence borough in that year. Models also include an indicator for OOH for 5+ years, which is suppressed because it is only identified by students who were retained in grade. Attendance ranges from 0 to 100.

Do effects differ by building size?

Given descriptive patterns of relationships between housing size and student outcomes coupled with the tendency of prior literature to conflate the effects of OOH with the effects of living in SFH, we explore whether the effects of OOH vary based on building type, where 20+ family ROH serve as the reference (Table 7). Although we find a benefit to moving into OOH SFH for attendance (0.142 pp), all other estimates for moving into OOH SFH are statistically insignificant and close to zero. Instead, we find that there are small academic benefits of moving into OOH in large, 20+ family buildings, increasing attendance (0.445 pp), reducing chronic absenteeism (1.9 pp), and increasing math scores (0.040). Moving into OOH in 2–4 or 5–19 buildings has little effect on academic performance, with the exception of math scores for students moving into OOH 5–19 family buildings and we observe no effect of moving into OOH of any size on zBMI, obesity, or overweight.¹³

Table 7.

OOH and Student Outcomes by Building Size, AY 2007–2012, Grades 4–8, Student Fixed Effects

	(1) Attendance	(2) Chronic Absenteeism	(3) zELA	(4) zMath	(5) zBMI
Move into OOH
Single Family	0.17 7*** (0.056)	0.001 (0.004)	−0.015 (0.009)	−0.006 (0.009)	−0.012 (0.010)
2–4 Family	0.019 (0.330)	−0.016 (0.028)	−0.027 (0.075)	−0.012 (0.070)	0.011 (0.089)
5–19 Family	0.546** (0.222)	−0.027 (0.015)	0.045 (0.034)	0.115*** (0.030)	−0.011 (0.031)
20+ Family	0.445***	−0.019***	0.017	0.040***	0.002
	(0.072)	(0.005)	(0.013)	(0.010)	(0.012)
Move out of OOH
Single Family	−0.172***	0.001	0.000	−0.019**	0.000
	(0.066)	(0.005)	(0.011)	(0.009)	(0.011)
2–4 Family	0.313	−0.022	0.150***	0.080**	−0.081**
	(0.237)	(0.018)	(0.046)	(0.037)	(0.040)
5–19 Family	0.763***	−0.032	0.049	0.056	0.017
	(0.234)	(0.018)	(0.035)	(0.030)	(0.039)
20+ Family	0.196**	−0.005	0.027**	0.007	−0.038***
	(0.085)	(0.006)	(0.013)	(0.011)	(0.013)
Building size
Single Family	0.080	−0.012***	0.014	−0.002	0.010
	(0.060)	(0.004)	(0.008)	(0.007)	(0.009)
2–4 Family	−0.002	−0.002	−0.000	0.003	0.017***
	(0.047)	(0.003)	(0.006)	(0.005)	(0.006)
5–19 Family	0.083	0.000	0.004	0.009	0.008
	(0.053)	(0.004)	(0.006)	(0.006)	(0.007)

Standard errors are clustered by School and Year, and is reported in parentheses

^**p < 0.05

^***p < 0.01

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. All models include controls for building size (single family home, 2–4 family, and 5–19 family), year, and grade effects. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. School characteristics include ELA proficiency, math proficiency, pupil-teacher ratio, percent of students eligible for free or reduced price lunch, racial composition, log per pupil expenditures, and controls for missing data on school characteristics. Missing data on school characteristics are replaced with the average for a student’s residence borough in that year. Attendance ranges from 0 to 100.

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. OOH for 2 years is equal to 1 if a student moved into OOH and has lived in OOH for two years as of year t. All models include controls for building size (single family home, 2–4 family, and 5–19 family), year, and grade effects. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. School characteristics include ELA proficiency, math proficiency, pupil-teacher ratio, percent of students eligible for free or reduced price lunch, racial composition, log per pupil expenditures, and controls for missing data on school characteristics. Missing data on school characteristics are replaced with the average for a student’s residence borough in that year. Models also include an indicator for OOH for 5+ years, which is suppressed because it is only identified by students who were retained in grade. Attendance ranges from 0 to 100.

The effects of moving out of OOH are more mixed—here we find small deleterious effects of moving out of OOH SFH for attendance (−0.172pp) and math scores (−0.019 sd). However, moving out of larger OOH appears to have no, or small positive effects. For example, students moving out of OOH 20+ family buildings experience small increases in attendance (0.196pp) and ELA scores (0.017 sd), and small reductions in zBMI (0.038sd). Among students in ROH, the estimated effects of building size are largely insignificant, although living in SFH decreases chronic absenteeism by 1.2 pp.

Given the period we are studying, one possible explanation for the negative estimates of moving out of OOH SFH on attendance and test scores is that these moves out of OOH are precipitated by a foreclosure. To explore this possibility, we re-estimated this model controlling for whether a student’s building in t-1 received a lis pendens (Appendix Table A4), and results are unchanged.

While raw differences suggest that students in OOH and lower density housing have better outcomes than those living in ROH or high-density housing, this likely overstates the OOH premium. Accounting for differences in housing and individual selection into housing using student fixed effects, yields dramatically smaller estimates, and ultimately suggest small or no benefits to OOH and lower-density housing for academic outcomes and no benefits for zBMI.

Discussion and Conclusion

Contrary to the popular conception of OOH as a free-standing single-family home inhabited by middle-income families, there is significant variation across NYC in both the characteristics OOH itself and the students who live there. Over 40 percent of OOH is in large buildings with at least 20 units, and 45 percent of students who live in OOH qualify for free or reduced price lunch, meaning that their household income is at or below 185 percent of the federal poverty level. There is also significant mobility by housing tenure and housing size. Students who live in OOH have significantly better health and academic outcomes than their peers, but this disparity is largely driven by differences in the neighborhoods where OOH is located and the characteristics of students living in OOH.

In student fixed effects models that better account for individual selection into housing, we find small benefits to moving into OOH for academic outcomes and little evidence of deleterious consequences for moving out of OOH, which are robust to alternative specifications. More specifically, time-varying student, building, tract, and school characteristics explain little of the relationship between OOH and outcomes above and beyond what is explained by student fixed effects.

The benefits to moving into OOH appear to be concentrated among students moving into OOH in large buildings, while there is some evidence that moving out of OOH SFH has small negative effects on attendance and math scores. Taken together, these results suggest a small benefit to OOH and lower density housing for academic outcomes, but no consistent benefit for zBMI, obesity or overweight. These findings are similar across boroughs, including lower density Queens and Staten Island, where the percentage of students living in OOH is higher than the other boroughs at approximately 26 and 45 percent, respectively. This suggests that our results could be generalizable to other locations, particularly small cities and inner ring suburbs.

Our results add to a small body of work that finds limited positive or no impacts of homeownership or OOH on short- and medium-term outcomes such as test scores, behavior, and health (Haurin, et al., 2002; Barker and Miller, 2009). Further, they are largely consistent with prior work finding that homeownership improves educational attainment. While test scores may be a poor proxy for later life outcomes, chronic absenteeism is associated with an increased risk of dropout. Thus, our findings of reduced chronic absenteeism among students living in larger OOH buildings is consistent with prior work that finds homeownership increases graduation. However, it should be noted that similar to our findings, most of the positive impacts identified in prior literature on homeownership are relatively small in magnitude.

It is worth noting that the time period of our analysis (2007–2011) is characterized by both the great recession and a volatile housing market. This does, indeed, present challenges for assessing generalizability. Yet our results combined with those of previous literature exploring the effects of homeownership in earlier decades, suggest that policies that promote homeownership may not confer large academic or health benefits to children and might be oversold. Although further work might explore the extent to which these benefits are concentrated among owner-occupants themselves rather than renter-occupants in OOH.

• There is notable variation in the characteristics and size of owner-occupied housing

• (OOH) and the types of students who live in OOH in New York City

• Raw differences show that students who live in OOH have better academic and health outcomes, including higher attendance and test scores, and lower BMI

• In models that account for selection into OOH and building type with rich controls and student fixed effects, we find small positive effects of moving into OOH on attendance and math scores.

• There is little consistent evidence of any impacts of moving into OOH on BMI or obesity or evidence that moving out of OOH is deleterious to student outcomes

APPENDIX

Appendix A

Table A1:

Housing Type Transition Table, Academic Years 2007 – 2012, Student Moves

Number (Row %) (Total %)		Owner Occupied Housing in Year t				Renter Occupied Housing in Year t
		SFH	2–4 Family	5–19 Family	20+ Family	SFH	2–4 Family	5–19 Family	20+ Family	Total
Owner Occupied Housing in Year t-1	SFH	1,959 (26.4%) (1.9%)	15 (0.2%) (0.0%)	31 (0.4%) (0.0%)	434 (5.8%) (0.4%)	805 (10.8%) (0.8%)	3,334 (44.9%) (3.2%)	229 (3.1%) (0.2%)	616 (8.3%) (0.6%)	7,423 (100.0%) (7.2%)
	2–4 Family	8 (8.2%) (0.0%)	0 (0.0%) (0.0%)	2 (2.1%) (0.0%)	10 (10.3%) (0.0%)	15 (15.5%) (0.0%)	47 (48.5%) (0.0%)	3 (3.1%) (0.0%)	12 (12.4%) (0.0%)	97 (100.0%) (0.1%)
	5–19 Family	25 (5.2%) (0.0%)	2 (0.4%) (0.0%)	20 (4.2%) (0.0%)	63 (13.2%) (0.1%)	14 (2.9%) (0.0%)	183 (38.2%) (0.2%)	57 (11.9%) (0.1%)	115 (24.0%) (0.1%)	479 (100.0%) (0.5%)
	20+ Family	528 (10.1%) (0.5%)	10 (0.2%) (0.0%)	68 (1.3%) (0.1%)	1,085 (20.8%) (1.1%)	293 (5.6%) (0.3%)	1,588 (30.4%) (1.5%)	308 (5.9%) (0.3%)	1,338 (25.6%) (1.3%)	5,218 (100.0%) (5.1%)
Renter Occupied Housing in Year t-1	SFH	715 (15.0%) (0.7%)	7 (0.1%) (0.0%)	14 (0.3%) (0.0%)	199 (4.2%) (0.2%)	571 (12.0%) (0.6%)	2,467 (51.8%) (2.4%)	236 (5.0%) (0.2%)	550 (11.6%) (0.5%)	4,759 (100.0%) (4.6%)
	2–4 Family	3,830 (8.3%) (3.7%)	55 (0.1%) (0.1%)	230 (0.5%) (0.2%)	1,755 (3.8%) (1.7%)	2,958 (6.4%) (2.9%)	27,423 (59.1%) (26.7%)	3,805 (8.2%) (3.7%)	6,328 (13.6%) (6.2%)	46,384 (100.0%) (45.2%)
	5–19 Family	432 (3.8%) (0.4%)	15 (0.1%) (0.0%)	84 (0.7%) (0.1%)	475 (4.1%) (0.5%)	363 (3.2%) (0.4%)	5,182 (45.1%) (5.0%)	2,233 (19.4%) (2.2%)	2,703 (23.5%) (2.6%)	11,487 (100.0%) (11.2%)
	20+ Family	1,012 (3.8%) (1.0%)	27 (0.1%) (0.0%)	151 (0.6%) (0.1%)	1,806 (6.7%) (1.8%)	764 (2.8%) (0.7%)	8,382 (31.3%) (8.2%)	2,642 (9.9%) (2.6%)	12,036 (44.9%) (11.7%)	26,820 (100.0%) (26.1%)
	Total	8,509 (8.3%)	131 (0.1%)	600 (0.6%)	5,827 (5.7%)	5,783 (5.6%)	48,606 (47.3%)	9,513 (9.3%)	23,698 (23.1%)	102,667

Notes: The sample includes only students with at least 3 years of attendance, test scores, and weight outcomes from 2007 – 2012 and are observed in 2 consecutive years. Each observation represents a student’s residential move, which is defined as the student living in a different building in year t+1 than in year t. A student may be included more than once if they moved multiple times during the sample period. There are 71,206 unique students and roughly 84% of these students moved only once.

Table A2.

Variable Definitions

Variable Names	Definition
Outcomes
Attendance rate	percent of days attended (0–100)
Chronic abs.	=1 if attendance rate is below 90 percent
Zread	State reading score standardized by grade and year, with mean 0 and SD = 1
Zmath Zbmi	State math score standardized by grade and year, with mean 0 and SD = 1 BMI standardized by…
Obese	=1 if BMI is 95th percentile or greater for gender and age
Overweight	=1 if BMI is 85th-95th percentile for gender and age
Student Characteristics
Asian	=1 if student is Asian or other race
Black	=1 if student is Black
White	=1 if student is White
Female	=1 if student is female
Poor	=1 if student is eligible for free or reduced price lunch
LEP	=1 if student is classified as limited English proficient
SPED	=1 if student receives special education services
Age at Fitnessgram	Age at time of Fitnessgram test (in years)
Housing Characteristics
OOH	=1 if student lives in owner occupied housing. A building is designated as OOH if it is a condo or coop, or at least one occupant claims a STAR exemption.
Move into OOH	=1 in all years after a student moves into OOH from ROH
Move out of OOH	=1 in all years after a student moves out of ROH from OOH
SFH	=1 if student lives in a building with only one residential unit
2–4 family	=1 if student lives in a building with 2–4 residential units
5–19 family	=1 if student lives in a building with 5–19 residential units
20+ family	=1 if student lives in a building with 20 or more residential units
Building Age	Age of building
Elevator	=1 if building has an elevator as defined by either building class or data from NYC Department of Buildings
Sq. ft./res. Unit	=total square footage of building divided by number of residential units
Stories	Total number of stories in the building
Tract Characteristics
% white, non-Hispanic	Percent of tract residents identified as White, non-Hispanic
% Black, non-Hispanic	Percent of tract residents identified as Black, non-Hispanic
% Asian, non-Hispanic	Percent of tract residents identified as Asian, non-Hispanic
% other race, non-Hispanic	Percent of tract residents identified as another race, non-Hispanic
% Hispanic	Percent of tract residents identified as Hispanic
% of units occupied	Percent of housing units that are occupied
% of units owner occupied	Percent of housing units that are owner occupied
Median household income, owners	Median household income of owner-occupants
Median household income, renters	Median household income of renters
Median value	Median value of owner occupied properties
Median gross rent	Median gross rent
% with more than HS diploma	Percent of population over age 25 with more than a high school diploma or equivalent
Unemployment rate	Unemployment rate for civilization population in labor force 16 years and over
Poverty rate, under 18	Poverty status for children under 18
Poverty rate, 18–64	Poverty status for population age 18 to 64
Poverty rate, 65 and over	Poverty status for population age 65 and over
Population Density	Number of people per square mile
School characteristics
Proficiency rate, ELA	Percent of students scoring proficient or above on ELA exam
Proficiency rate, math	Percent of students scoring proficient or above on math exam
Pupil-teacher ratio	Number of students/number of full-time equivalent teachers
Percent poor	Percent of students eligible for free or reduced price lunch
Percent White	Percent of students who are White
Percent Black	Percent of students who are Black
Percent Hispanic	Percent of students who are Hispanic
Percent Asian/other race	Percent of students who are Asian or some other race
Log of per pupil expenditures	Log of spending per pupil

Table A3.

OOH and Student Outcomes by Borough, AY 2007–2012, Grades 4–8, Student Fixed Effects

	(1) Attendance	(2) Chronic Absenteeism	(3) zELA	(4) zMath	(5) zBMI
Move into OOH
Manhattan	0.490*** (0.173)	−0.017 (0.011)	−0.009 (0.032)	0.058** (0.024)	0.032 (0.029)
Bronx	0.257	−0.011	−0.019	−0.026	0.003
	(0.145)	(0.010)	(0.020)	(0.016)	(0.024)
Brooklyn	0.479***	−0.029***	−0.019	0.032**	−0.005
	(0.081)	(0.006)	(0.014)	(0.013)	(0.015)
Queens	0.229***	−0.000	0.018	0.020**	−0.014
	(0.062)	(0.004)	(0.011)	(0.010)	(0.012)
Staten Island	0.016	0.015	−0.025	−0.022	−0.003
	(0.120)	(0.008)	(0.020)	(0.020)	(0.019)
Move out of OOH
Manhattan	0.665***	−0.024	0.061	0.065**	−0.046
	(0.184)	(0.014)	(0.039)	(0.030)	(0.028)
Bronx	−0.138	−0.008	−0.009	−0.055***	−0.012
	(0.174)	(0.012)	(0.020)	(0.020)	(0.024)
Brooklyn	0.126	−0.006	0.012	−0.003	0.017
	(0.103)	(0.007)	(0.016)	(0.013)	(0.016)
Queens	0.007	0.001	0.030***	0.004	−0.037***
	(0.073)	(0.005)	(0.012)	(0.010)	(0.012)
Staten Island	−0.275**	0.001	−0.018	−0.020	0.004
	(0.112)	(0.010)	(0.020)	(0.018)	(0.020)
Borough
Manhattan	−0.082	0.005	−0.056***	−0.059***	0.029
	(0.122)	(0.008)	(0.014)	(0.013)	(0.017)
Bronx	−0.294**	0.013	−0.001	−0.002	0.024
	(0.128)	(0.008)	(0.012)	(0.012)	(0.015)
Queens	−0.172**	0.011**	−0.014	−0.010	−0.014
	(0.083)	(0.005)	(0.009)	(0.009)	(0.012)
Staten Island	−0.707***	0.034***	−0.048***	−0.060***	0.019
	(0.143)	(0.009)	(0.016)	(0.016)	(0.018)
R 2	0.784	0.668	0.784	0.835	0.847
N	1,240,418	1,240,418	1,240,418	1,240,418	1,219,269

Standard errors are clustered by School and Year, and is reported in parentheses

^**p < 0.05

^***p < 0.01

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. All models include controls for building size (single family home, 2–4 family, and 5–19 family), student characteristics, building characteristics, tract characteristics, year, and grade effects. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. Attendance ranges from 0 to 100.

Table A4.

Owner Occupied Housing and student outcomes, AY 2007–2012, Grades 4–8, Student and Tract Fixed Effects

	(1) Attendance	(2) Chronic Absenteeism	(3) zELA	(4) zMath	(5) zBMI	(6) Obese	(7) Overweight
Panel A: Building Size
Move into OOH	0.234***	−0.005	−0.002	0.011	−0.007	0.004	0.001
	(0.044)	(0.003)	(0.008)	(0.006)	(0.007)	(0.003)	(0.003)
Move out of OOH	0.045	−0.004	0.016**	−0.005	−0.010	0.004	−0.000
	(0.052)	(0.004)	(0.008)	(0.007)	(0.008)	(0.003)	(0.004)
R 2	0.785	0.669	0.785	0.836	0.847	0.793	0.798
N	1,240,042	1,240,042	1,240,042	1,240,042	1,218,967	1,218,967	1,218,967

Standard errors clustered by School and Year, and are shown in parentheses,

^***p<0.01

^**p<0.05

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. All models include controls for building size (single family, 2–4 family, and 5–19 family), student characteristics, building characteristics, year, and grade effects. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Attendance ranges from 0 to 100.

Table A5.

OOH, NYCHA, and Student Outcomes by Building Size, AY 2007–2012, Grades 4–8, Student Fixed Effects

	(1) Attendance	(2) Chronic Absenteeism	(3) zELA	(4) zMath	(5) zBMI	(6) Obesity	(7) Overweight
Move into OOH	0.304***	−0.008**	0.000	0.017***	−0.006	0.003	0.002
	(0.043)	(0.003)	(0.007)	(0.006)	(0.007)	(0.003)	(0.003)
Move out of OOH	−0.001	−0.003	0.019**	−0.001	−0.017**	0.001	−0.003
	(0.051)	(0.004)	(0.008)	(0.007)	(0.008)	(0.003)	(0.004)
NYCHA	0.099	−0.011**	−0.006	−0.004	0.004	−0.002	0.004
	(0.078)	(0.005)	(0.008)	(0.008)	(0.009)	(0.004)	(0.004)
R 2	0.783	0.670	0.785	0.835	0.847	0.794	0.798
N	1,370,236	1,370,236	1,370,236	1,370,236	1,345,544	1,345,544	1,345,544

Standard errors are clustered by School and Year, and is reported in parentheses

^**p < 0.05

^***p < 0.01

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. NYCHA is an indicator equal to 1 in years that a student lives in public housing. All models include controls for building size (single family home, 2–4 family, and 5–19 family), student characteristics, building characteristics, tract characteristics, year, and grade effects. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. Attendance ranges from 0 to 100.

Table A6.

OOH and Student Outcomes by Grade of Move, AY 2007–2012, Grades 4–8, Student Fixed Effects

	(1) Attendance	(2) Chronic Absenteeism	(3) zELA	(4) zMath	(5) zBMI	(6) Obesity	(7) Overweight
Move into OOH
Grade 5	0.26 8***	−0.001	−0.021	0.019	−0.006	0.010	0.004
	(0.066)	(0.005)	(0.013)	(0.012)	(0.015)	(0.005)	(0.006)
Grade 6	0.462***	−0.023***	0.003	0.013	−0.008	0.002	−0.008
	(0.073)	(0.006)	(0.014)	(0.011)	(0.014)	(0.005)	(0.006)
Grade 7	0.286***	−0.010	0.003	0.018	−0.017	−0.003	0.010
	(0.083)	(0.006)	(0.013)	(0.011)	(0.013)	(0.005)	(0.006)
Grade 8	0.015	0.011	−0.003	0.003	0.009	0.005	0.000
	(0.084)	(0.006)	(0.013)	(0.012)	(0.012)	(0.005)	(0.006)
Move out of OOH
Grade 5	0.111	−0.012	0.009	−0.010	−0.026	0.004	0.002
	(0.088)	(0.007)	(0.015)	(0.012)	(0.015)	(0.006)	(0.007)
Grade 6	0.059	0.003	0.029	−0.014	−0.013	0.001	0.000
	(0.099)	(0.008)	(0.015)	(0.013)	(0.015)	(0.006)	(0.007)
Grade 7	−0.253**	0.013	0.036**	−0.008	0.013	0.005	0.005
	(0.112)	(0.008)	(0.017)	(0.013)	(0.015)	(0.006)	(0.008)
Grade 8	−4.233	0.184	0.252	−0.183	0.075	0.001	0.099
	(2.603)	(0.115)	(0.286)	(0.152)	(0.095)	(0.038)	(0.070)
R 2	0.784	0.668	0.784	0.835	0.847	0.792	0.798
N	1,240,042	1,240,042	1,240,042	1,240,042	1,218,967	1,218,967	1,218,967

Standard errors are clustered by School and Year, and is reported in parentheses

^**p < 0.05

^***p < 0.01

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. All models include controls for building size (single family home, 2–4 family, and 5–19 family), student characteristics, building characteristics, tract characteristics, year, and grade effects. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. Attendance ranges from 0 to 100.

Table A7.

OOH and Student Outcomes by Building Size, AY 2007–2012, Grades 4–8, Student Fixed Effects, Controlling for Foreclosure

	(1) Attendance	(2) Chronic Absenteeism	(3) zELA	(4) zMath	(5) zBMI
OOH*Move In*
Single Family	0.177***	0.001	−0.015	−0.006	−0.012
	(0.056)	(0.004)	(0.009)	(0.009)	(0.010)
2−4 Family	0.019	−0.016	−0.027	−0.012	0.011
	(0.330)	(0.028)	(0.075)	(0.070)	(0.089)
5−19 Family	0.546**	−0.027	0.045	0.115***	−0.011
	(0.222)	(0.015)	(0.034)	(0.030)	(0.031)
20+ Family	0.445***	−0.019***	0.017	0.040***	0.002
	(0.072)	(0.005)	(0.013)	(0.010)	(0.012)
OOH*Move Out*
Single Family	−0.172***	0.001	0.000	−0.019**	0.000
	(0.066)	(0.005)	(0.011)	(0.009)	(0.011)
2−4 Family	0.313	−0.022	0.150***	0.080**	−0.081**
	(0.237)	(0.018)	(0.046)	(0.037)	(0.040)
5−19 Family	0.763***	−0.032	0.049	0.056	0.017
	(0.234)	(0.018)	(0.035)	(0.030)	(0.039)
20+ Family	0.196**	−0.005	0.027**	0.007	−0.038***
	(0.085)	(0.006)	(0.013)	(0.011)	(0.013)
Building size
Single Family	0.080	−0.012***	0.014	−0.002	0.010
	(0.060)	(0.004)	(0.008)	(0.007)	(0.009)
2−4 Family	−0.001	−0.002	−0.000	0.003	0.017***
	(0.047)	(0.003)	(0.006)	(0.005)	(0.006)
5−19 Family	0.083	0.000	0.004	0.009	0.008
	(0.053)	(0.004)	(0.006)	(0.006)	(0.007)
Foreclosure, t−1	−0.024	0.001	−0.004	−0.002	0.002
	(0.038)	(0.003)	(0.004)	(0.004)	(0.005)
R 2	0.784	0.668	0.784	0.835	0.847
N	1,240,042	1,240,042	1,240,042	1,240,042	1,218,967

Standard errors are clustered by School and Year, and is reported in parentheses

^**p < 0.05

^***p < 0.01

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. Foreclosure t-1 equals 1 if the building a student lived in in year t-1 received a lis pendens. All models include controls for building size (single family home, 2–4 family, and 5–19 family), student characteristics, building characteristics, tract characteristics, year, and grade effects. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. Attendance ranges from 0 to 100.

Appendix B –. Full Estimates for Table 4, Panel E and results for Overweight and Obesity

Table B1.

Owner Occupied Housing and student outcomes, Full Results, AY 2007–2012, Grades 4–8, Student Fixed Effects

	(1) Attendance	(2) Chronic Absenteeism	(3) zELA	(4) zMath	(5) zBMI	(6) Obesity	(7) Overweight
Move into OOH	0.262***	−0.007**	−0.002	0.008	−0.001	0.004	0.003
	(0.043)	(0.003)	(0.007)	(0.006)	(0.007)	(0.003)	(0.003)
Move out of OOH	0.006	−0.003	0.017**	−0.008	−0.012	0.005	−0.001
	(0.051)	(0.004)	(0.008)	(0.007)	(0.008)	(0.003)	(0.004)
Student characteristics
FRPL	0.07 1***	−0.003***	0.006	0.009**	−0.006	−0.001	0.000
	(0.019)	(0.001)	(0.003)	(0.004)	(0.005)	(0.001)	(0.002)
LEP	0.096***	−0.009***	−0.176***	−0.157***	0.006	0.003	0.005
	(0.034)	(0.002)	(0.006)	(0.006)	(0.007)	(0.002)	(0.003)
SPED	−0.269***	0.015***	0.056***	0.031***	0.013**	0.004	0.006**
	(0.043)	(0.003)	(0.005)	(0.005)	(0.006)	(0.002)	(0.003)
Age in years	−0.164***	0.005**	−0.025***	−0.026***	−0.179***	−0.054***	−0.066***
	(0.034)	(0.002)	(0.006)	(0.006)	(0.009)	(0.003)	(0.003)
Housing characteristics
Single Family	0.091	−0.010**	0.012	−0.004	0.003	−0.003	−0.001
	(0.056)	(0.004)	(0.008)	(0.007)	(0.008)	(0.003)	(0.004)
2−4 Family	−0.001	−0.001	−0.002	0.004	0.015**	0.000	0.007**
	(0.046)	(0.003)	(0.006)	(0.005)	(0.006)	(0.003)	(0.003)
5−19 Family	0.072	0.001	0.002	0.009	0.006	0.000	0.007*
	(0.052)	(0.004)	(0.006)	(0.006)	(0.007)	(0.003)	(0.003)
Building age	−0.000	0.000	−0.000	0.000	−0.000	0.000	0.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Number of stories	−0.001	0.000	0.001**	0.000	0.000	0.000	−0.000
	(0.004)	(0.000)	(0.001)	(0.001)	(0.001)	(0.000)	(0.000)
Elevator in building	−0.005	−0.001	0.005	0.015***	−0.002	−0.000	0.002
	(0.047)	(0.003)	(0.006)	(0.005)	(0.006)	(0.003)	(0.003)
Gross sq. footage per res. unit	−0.000	0.000	0.000	−0.000	−0.000	0.000**	−0.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Tract characteristics
% white, non−Hispanic	0.001	−0.000	−0.001***	−0.000	0.000	0.000	0.000**
	(0.002)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
% Black, non−Hispanic	−0.002	−0.000	−0.001***	−0.000	0.001	0.000**	0.000**
	(0.002)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
% Asian, non−Hispanic	0.001	−0.000	−0.001***	0.000	0.000	0.000	0.000
	(0.002)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
% other race, non−Hispanic	0.001	−0.000	−0.002***	−0.000	0.001	0.000	0.001**
	(0.004)	(0.000)	(0.001)	(0.001)	(0.001)	(0.000)	(0.000)
% Hispanic	0.001	−0.000	−0.001***	−0.000	0.000	0.000	0.000
	(0.002)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
% of units occupied	−0.000	0.000	0.000	0.000	0.001	0.000	0.000
	(0.003)	(0.000)	(0.000)	(0.000)	(0.001)	(0.000)	(0.000)
% of units owner occupied	0.002**	−0.000	0.000	−0.000	−0.000	−0.000	−0.000
	(0.001)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Median HH inc., owners	−0.000	0.000	0.000	0.000***	−0.000	0.000	−0.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Median HH inc., renters	0.000**	−0.000	0.000	0.000	−0.000	−0.000	−0.000**
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Median Value	0.000	0.000	0.000	0.000	−0.000	0.000	−0.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Median Gross Rent	−0.000**	0.000	0.000	−0.000	0.000	−0.000	−0.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
% more than HS	0.002	−0.000	0.000	−0.000	−0.000	−0.000	0.000
	(0.002)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Unemployment rate	−0.002	0.000	0.000	−0.001	−0.001	−0.000	−0.000
	(0.008)	(0.001)	(0.001)	(0.001)	(0.001)	(0.000)	(0.000)
Poverty rate, under 18	0.001	−0.000	0.000	0.000	−0.000**	−0.000	−0.000
	(0.001)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Poverty rate, 18−64	0.002	0.000	0.000	−0.000	0.001	−0.000	−0.000
	(0.003)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Poverty rate, 65+	−0.002	−0.000	−0.000	−0.000	0.000	−0.000	−0.000
	(0.001)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Population Density	−0.000	−0.000	−0.000	−0.000	0.000	0.000	0.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
School characteristics
Prof. rate, ELA	0.065	−0.005	0.52 7***	−0.313***	0.084*	0.03 0***	0.017
	(0.180)	(0.010)	(0.033)	(0.038)	(0.043)	(0.011)	(0.014)
Prof. rate, math	2.055***	−0.073***	0.032	1.084***	−0.138***	−0.032***	−0.047***
	(0.174)	(0.009)	(0.026)	(0.028)	(0.039)	(0.010)	(0.012)
Pupil−teacher ratio	0.022***	−0.001***	−0.014***	−0.016***	−0.001	−0.001	−0.000
	(0.008)	(0.000)	(0.001)	(0.002)	(0.002)	(0.000)	(0.001)
% poor	0.002*	−0.000	0.001***	0.000**	−0.000	−0.000	0.000
	(0.001)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
% White	−0.009	0.001	−0.001	−0.004	−0.005	−0.001	−0.001
	(0.016)	(0.001)	(0.002)	(0.002)	(0.003)	(0.001)	(0.001)
% Black	−0.000	0.000	0.000	−0.003	−0.005*	−0.001	−0.002
	(0.016)	(0.001)	(0.002)	(0.002)	(0.003)	(0.001)	(0.001)
% Hispanic	0.003	−0.000	0.001	−0.002	−0.005	−0.001	−0.001
	(0.016)	(0.001)	(0.002)	(0.002)	(0.003)	(0.001)	(0.001)
% Asian/other race	−0.009	0.000	−0.002	−0.006**	−0.004	−0.001	−0.001
	(0.016)	(0.001)	(0.002)	(0.002)	(0.003)	(0.001)	(0.001)
Log PPE	0.209***	−0.014***	0.003	0.020***	0.050***	0.011***	0.014***
	(0.030)	(0.002)	(0.005)	(0.006)	(0.007)	(0.002)	(0.002)
Constant	95.117***	0.162	−0.176	−0.200	2.487***	0.794***	1.073***
	(1.660)	(0.091)	(0.214)	(0.231)	(0.321)	(0.075)	(0.102)
R-squared	0.785	0.669	0.785	0.839	0.847	0.793	0.798
N	1,240,042	1,240,042	1,240,042	1,240,042	1,218,967	1,218,967	1,218,967

Table B2.

Owner Occupied Housing and student outcomes, AY 2007–2012, Grades 4–8, Student Fixed Effects

	(1) Obese	(2) Overweight
Panel A: Building Size
Move into OOH	0.003	0.003
	(0.003)	(0.003)
Move out of OOH	0.005	−0.005
	(0.003)	(0.004)
R 2	0.792	0.797
Obs.	1,219,803	1,219,803
Panel B: Building Size and Student Characteristics
Move into OOH	0.003	0.002
	(0.003)	(0.003)
Move out of OOH	0.005	−0.006
	(0.003)	(0.004)
R 2	0.793	0.798
Obs.	1,219,803	1,219,803
Panel C: Building Size, Student, and Building Characteristics
Move into OOH	0.003	0.002
	(0.003)	(0.003)
Move out of OOH	0.005	−0.005
	(0.003)	(0.004)
R 2	0.793	0.798
Obs.	1,219,803	1,219,803
Panel D: Building Size, Student, Building, and Tract Characteristics
Move into OOH	0.003	0.002
	(0.003)	(0.003)
Move out of OOH	0.005	−0.004
	(0.003)	(0.004)
R 2	0.793	0.798
Obs.	1,219,803	1,219,803

Standard errors are clustered by School and Year, and is reported in parentheses

^**p < 0.05

^***p < 0.01

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. All models include controls for building size (single family home, 2–4 family, and 5–19 family), year, and grade effects. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract

Table B3.

OOH and Student Outcomes, AY 2007–2012, Grades 4–8, Student Fixed Effects, Placebo Effects

	(1) Obese	(2) Overweight
Main estimates
Move in OOH	0.003	0.002
	(0.003)	(0.003)
Move out of OOH	0.005	−0.004
	(0.003)	(0.004)
R 2	0.793	0.798
Obs.	1,219,803	1,219,803
Placebo Estimates
Placebo Move into OOH	0.002	−0.008
	(0.005)	(0.006)
Placebo Move out of OOH	0.005	0.001
	(0.006)	(0.007)
R 2	0.797	0.801
Obs.	1,177,608	>1,177,608

Standard errors clustered by School and Year, and is shown in parentheses,

^***p<0.01

^**p<0.05

^*p<0.1

Notes: For students who ever move into OOH, placebo Move in OOH equals 1 in a randomly selected year before a student is actually observed in OOH and in all years after. Observations for students ever moving into OOH are dropped in the years they actually live in OOH. For students who ever move out of OOH, placebo Move out OOH equals 1 in a randomly selected year after a student actually moves out of OOH and in all years after. Observations for students ever moving out of OOH are dropped in the years they actually do not live in OOH. All models include controls for building size (single family home, 2–4 family, and 5–19 family), student characteristics, building characteristics, tract characteristics, year, and grade effects. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. School characteristics include ELA proficiency, math proficiency, pupil-teacher ratio, percent of students eligible for free or reduced price lunch, racial composition, log per pupil expenditures, and controls for missing data on school characteristics. Attendance ranges from 0 to 100.

Table B4.

OOH and Student Outcomes by Building Size, AY 2007–2012, Grades 4–8, Student Fixed Effects

	(1) Obesity	(2) Overweight
OOH*Move In*
Single Family	−0.001	0.001
	(0.004)	(0.004)
2−4 Family	0.027	−0.011
	(0.022)	(0.027)
5−19 Family	0.006	−0.010
	(0.012)	(0.015)
20+ Family	0.010**	0.003
	(0.005)	(0.006)
OOH*Move Out*
Single Family	0.011***	−0.001
	(0.004)	(0.005)
2−4 Family	0.026**	−0.048**
	(0.013)	(0.021)
5−19 Family	0.018	−0.030
	(0.014)	(0.017)
20+ Family	−0.009	0.003
	(0.005)	(0.006)
Building size
Single Family	0.001	−0.000
	(0.003)	(0.004)
2−4 Family	0.001	0.007**
	(0.003)	(0.003)
5−19 Family	0.001	0.006
	(0.003)	(0.004)
R 2	*0.793	0.798
N	1,218,967	1,218,967

Standard errors are clustered by School and Year, and is reported in parentheses

^**p < 0.05

^***p < 0.01

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. All models include controls for building size (single family home, 2–4 family, and 5–19 family), year, and grade effects. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. Attendance ranges from 0 to 100

Table B5.

OOH and Student Outcomes by Building Size, AY 2007–2012, Grades 4–8, Student Fixed Effects, Controlling for School Characteristics

	(1) Obesity	(2) Overweight
OOH*Move In*
Single Family	−0.000	0.003
	(0.004)	(0.004)
2−4 Family	0.026	−0.011
	(0.022)	(0.027)
5−19 Family	0.007	−0.007
	(0.012)	(0.015)
20+ Family	0.010**	0.004
	(0.005)	(0.006)
OOH*Move Out*
Single Family	0.013***	0.000
	(0.004)	(0.005)
2−4 Family	0.029**	−0.045**
	(0.013)	(0.021)
5−19 Family	0.018	−0.030
	(0.014)	(0.017)
20+ Family	−0.008	0.003
	(0.005)	(0.006)
Building size
Single Family	0.001	−0.000
	(0.003)	(0.004)
2−4 Family	0.001	0.007**
	(0.003)	(0.003)
5−19 Family	0.001	0.006
	(0.003)	(0.004)
R 2	*0.793	0.798
N	1,218,967	1,218,967

Standard errors are clustered by School and Year, and is reported in parentheses

^**p < 0.05

^***p < 0.01

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. All models include controls for building size (single family home, 2–4 family, and 5–19 family), year, and grade effects. Student Characteristics include poverty, eligibility for special education, limited English proficiency, and age. Building characteristics include building age, elevator, square footage per residential unit, value per square foot, and total number of stories. Tract characteristics include racial composition, percent of units occupied, percent of units owner occupied, median household income of renters, median household income of owners, gross median rent, median value of owner occupied units, percent of residents with more than a high school diploma, poverty rate for children under 18, poverty rate for adults 18–64, poverty rate for adults 65 or over, unemployment rate, population density, and controls for missing data on tract characteristics. Missing data on tract characteristics are replaced with the average for the borough in that year. Attendance ranges from 0 to 100

Table B6.

OOH and Student Outcomes, AY 2007–2012, Grades 4–8, Dosage

	(1) Obese	(2) Overweight
Move into OOH	0.003	0.002
	(0.003)	(0.004)
OOH for 2 years	0.002	0.000
	(0.004)	(0.004)
OOH for 3 years	−0.001	−0.011
	(0.005)	(0.007)
OOH for 4 years	0.007	−0.003
	(0.012)	(0.015)
Move out of OOH	−0.021	0.084
	(0.067)	(0.180)
R 2	0.793	0.798
N	1,218,967	1,218,967

Standard errors are clustered by School and Year, and is reported in parentheses

^**p < 0.05

^***p < 0.01

Notes: OOH equals 1 if a student lives in a building that received a STAR credit, a condo, or a COOP. Move into OOH equals 1 in all years after a student moves into OOH from ROH. Move out of OOH equals 1 in all years after a student moves into ROH from OOH. OOH for 2 years is equal to 1 if a student moved into OOH and has lived in OOH for two years as of year t. All models also include borough, grade, and year effects, controls for student characteristics (poor, special education status, limited English proficiency, and age), controls for other building characteristics (building age, elevator, squared footage, value, and total number of stories). Models in Panel B also include an indicator for OOH for 5+ years, which is suppressed because it is only identified by students who were retained in grade. Models in Panel C also include an indicator for OOH, Grade 4, which is suppressed because it is only identified by students who were retained in fourth grade. Coefficient for OOH for 5+ years is suppressed because it is only identified by students who have been retained in grade. Attendance ranges from 0 to 100.