University of Southern California Dissertations and Theses

Redlining, neighborhood change, and individual outcomes: an exploration of how space shapes the landscape of inequality from past to present

(USC Thesis Other)

Redlining, neighborhood change, and individual outcomes: an exploration of how space shapes the landscape of inequality from past to present

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content REDLINING, NEIGHBORHOOD CHANGE, AND INDIVIDUAL OUTCOMES: AN
EXPLORATION OF HOW SPACE SHAPES THE LANDSCAPE OF INEQUALITY FROM
PAST TO PRESENT
by
Rebecca Brooks Smith
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PUBLIC POLICY AND MANAGEMENT)
May 2024

ii
Table of Contents
Abstract.......................................................................................................................................... iv
Chapter 1. Introduction ................................................................................................................... 1
1. Neighborhoods, housing, and individual well-being .......................................................... 1
2. Historical processes and neighborhood trajectories............................................................ 4
3. Contributions of the current work....................................................................................... 6
4. Overview of findings .......................................................................................................... 7
References................................................................................................................................. 10
Chapter 2. People and Place: Does the reason for redlining impact the long-term
trajectory of neighborhoods.......................................................................................................... 13
1. Introduction....................................................................................................................... 13
2. Background ....................................................................................................................... 15
3. Data and Methodology...................................................................................................... 19
4. Methodology ..................................................................................................................... 24
5. Results............................................................................................................................... 29
6. Discussion......................................................................................................................... 39
7. Conclusion ........................................................................................................................ 40
References................................................................................................................................. 42
Appendix................................................................................................................................... 45
Chapter 3. Historical Neighborhood Conditions and Modern-Day Disparities in
Asthma Prevalence and Air Quality: An Exploration of Redlining and Industrial
Impact……....... ............................................................................................................................ 46
1. Introduction....................................................................................................................... 46
2. Literature Review.............................................................................................................. 49
3. Data ................................................................................................................................... 52
4. Statistical Analysis............................................................................................................ 56
5. Results............................................................................................................................... 59
6. Discussion......................................................................................................................... 65
7. Conclusion ........................................................................................................................ 69
References................................................................................................................................. 70
Chapter 4. Do publicly funded neighborhood investments impact individual level
health-related outcomes? A longitudinal study of two neighborhoods in
Pittsburgh, PA from 2011 to 2018................................................................................................. 74
1. Introduction....................................................................................................................... 74
2. Literature Review.............................................................................................................. 76
3. Methods............................................................................................................................. 81

iii
4. Analysis............................................................................................................................. 90
5. Results............................................................................................................................... 92
6. Discussion....................................................................................................................... 102
7. Conclusion ...................................................................................................................... 107
References............................................................................................................................... 108
Appendix..................................................................................................................................112
Chapter 5. Conclusion..................................................................................................................113
1. Implications of findings...................................................................................................115
2. Policy Recommendations.................................................................................................119
References............................................................................................................................... 121
References ................................................................................................................................. 122

iv
Abstract
This dissertation is comprised of three essays that analyze both modern-day and historical
neighborhood level processes with the goal of peeling back the myriad of characteristics that
make up a neighborhood and pinpointing how specific mechanisms link to outcomes. Chapters 2
and 3 dive into redlining maps created by the Home Owner's Loan Corporation (HOLC) to both
test whether these maps are indicative of systematic mortgage discrimination and to suggest
ulterior ways this data may be useful to link past neighborhood characteristics with modern-day
outcomes. Specifically, chapter 2 explores heterogeneity in homeownership rates and racial
composition according to a D-graded areas initial population composition. Chapter 3 focuses on
how HOLC grades compare to past industrial activity as predictors of future asthma prevalence
and air quality. Chapter 4 departs from a historical lens by analyzing the impacts of economic
development on individual-level health-related outcomes within two historically disadvantaged
neighborhoods. These papers suggest nuanced ways of understanding the relationship between
neighborhood-level characteristics and individual outcomes, while reaffirming that such
characteristics are integral to our understanding of current population-level disparities.

1
Chapter 1. Introduction
1. Neighborhoods, housing, and individual well-being
Where one lives has a significant impact on daily life, shaping a variety of experiences that
can range in intensity from the length of one’s commute to whether one feels safe to go for a
walk. These experiences not only dictate daily patterns and exposures, but they play a role in
shaping the economic and physical well-being of individuals. These exposures begin within the
home. For example, low-quality, older housing stock can increase contact with poorly regulated
heating, pests, and toxins such as lead, all of which have been linked to adverse health outcomes
(D'Alessandro & Appolloni, 2020; Swope & Hernández, 2019).
The scope of a home’s impact extends beyond its physical structure to include the
surrounding neighborhood. The concept of neighborhood effects, now driving a body of
scholarship seeking to understand disparities in population health and economic well-being, is
the idea that each neighborhood is made up of influential characteristics that have the power to
fundamentally shape one’s life course. Examples of such characteristics include the norms of
behavior and sense of community established by those living within the neighborhood, the
connections to potential job opportunities, and the access to institutions such as schools and
parks, to name a few (Wilson, 2012).
A variety of studies have tested this concept of neighborhood effects, using both
randomized controlled trials and quasi-experimental analysis techniques. The Moving to
Opportunities (MTO) project is one such study which randomized offerings of housing vouchers
to households living in public housing. Results show that in the short-term, those who moved
experienced improvements in adult mental health, but there was no significant difference in

2
socioeconomic outcomes among treatment and control groups (Katz et al., 2001; Kling et al.,
2007). In the longer term, Chetty et al. (2016) find that for children within the MTO study who
were exposed to living in lower poverty areas for longer were more likely to attend college and
to have significantly higher incomes as adults. These children who received a voucher also have
lower hospitalization rates over time (Pollack et al., 2019). Another study which looks at
outcomes for households whose locations were randomized across public housing units, finds
that children living within neighborhoods with lower social problems indices have on average
better health outcomes than children living in neighborhoods with higher social problems indices
(Santiago et al., 2014). More recently, it was found that children with asthma who moved to lowpoverty neighborhoods have seen improvements in asthma symptoms, without such
improvements being linked to changes in exposures to indoor allergens (Pollack et al., 2023).
While the studies just described suggest that where one lives matters, they do not connect
specific neighborhood characteristics to specific outcomes, calling into question the underlying
causal mechanisms impacting individuals. When a person moves, a bundle of characteristics
change, making it difficult to understand what the actual intervention is that is being measured
(Arcaya et al., 2016). To try to pinpoint potential areas of intervention, studies have examined
how specific characteristics such as social cohesion and feelings of safety (Kim, 2010; Ross &
Mirowsky, 2001; Ruijsbroek et al., 2016), physical disorder (South et al., 2015), walkability
(Barile et al., 2017), and greening interventions (Moyer et al., 2019) impact a variety of health
and health-related outcomes.
Partly because of their influence on the individual, neighborhoods are increasingly
becoming a topic through which to better understand observed racial and ethnic disparities due in
part to the non-random nature of residential mobility. Perhaps contrary to belief, where a person

3
lives is not always based on a match between supply, demand, and an ability to pay. Historically,
minority households in general, and black households in particular, have been limited in their
housing choices due to discriminatory real-estate practices, intimidation and violence by white
communities, racially restrictive covenants forbidding sales to black homebuyers, and an overall
lack of the same economic opportunities from which other racial groups benefited (Massey &
Denton, 1988; Rothstein, 2017). This has resulted in a segregated housing landscape, with black
households historically segregated at rates about twice as high as Asian and Hispanic populations
(Fischer, 2003). This segregation is not simply due to income differences. For black families,
increases in socioeconomic status do not translate to decreased spatial isolation as is seen for
other racial/ethnic groups (Massey & Denton, 1988).
This segregated housing landscape has led to disparate exposure to neighborhood
conditions at a population level, with black households being particularly subjected to areas of
higher poverty. Logan (2011) finds that on average, affluent black households live in
neighborhoods with only slightly higher poverty rates than what is experienced by the poorest
white households (14 -15% versus 12-13% respectively). Such patterns of spatial segregation
have consequences beyond just poverty exposure. Black and minority populations are twice as
likely as non-Hispanic white populations to live in a census block group with air pollution levels
above the 90th percentile (Liu et al., 2021), and black individuals are significantly more likely to
live within a mile of a polluting facility (Mohai et al., 2009). Black and minority populations also
have greater exposure to low-quality housing stock which can contribute to health problems,
particularly asthma exacerbations (Bailey et al., 2017; Bryant-Stephens et al., 2021). One study
finds that each 10 percent increase in the proportion of white residents living in a neighborhood
is associated with a decrease of 3.4 reported in-home asthma triggers (Lemire et al., 2022).

4
2. Historical processes and neighborhood trajectories
One important question is why have neighborhoods turned out so differently? Scholars
are increasingly using the concept of structural racism as a lens through which to understand
these geographic patterns of segregation and population-level disparities. Structural racism refers
to the utilization of systems to reinforce racial discrimination and to manipulate the distribution
of resources to favor certain groups over others (Bailey et al., 2017). An example of such a
system is that of denying favorable housing finance to certain neighborhoods, a practice known
as redlining. Unlike segregationist processes such as white flight and intimidation tactics which
occurred at the individual level, redlining is of particular interest as it was a federally sanctioned
policy.
The federal government entered the housing finance market in the 1930s with the
creation of two New Deal agencies: the Home Owner’s Loan Corporation (HOLC) and the
Federal Housing Administration (FHA) (Fishback et al., 2022). The purpose of the two were
quite different. The HOLC was a temporary program tasked with refinancing existing troubled
loans while the FHA was designed to revamp the loan insurance system and to incentivize new
housing construction projects (Fishback et al., 2022). The HOLC itself did not have particularly
discriminatory practices. Indeed, it issued loans to black households in a greater than
proportionate share to the overall number of black homeowners in some cities (Michney &
Winling, 2020). The FHA, on the other hand, was highly discriminatory. Not only did the FHA
make extremely few loans to black borrowers themselves, but they refused to insure mortgages
in neighborhoods with even the potential of racial change (Fishback et al., 2022). This behavior
also extended to new construction projects. In some cases, the FHA made the adoption of racially

5
restrictive property deeds a requirement for new developments to qualify for FHA financing
(Rothstein, 2017).
The two organizations are linked in the literature not only for their similar creation
timeline, but because they are both known for having created maps of cities throughout the
United States. These maps demarcated neighborhoods according to the relative risk of lending
within their borders. Neighborhoods were given a grade A through D, with A signifying the least
credit risk and D signifying the most. The risk scores are of particular interest as they incorporate
the prevailing practices of the real estate industry, practices which equated the presence of
minority households with lower long-term housing values (Light, 2010). While the HOLC maps
have been digitized and are readily available for analysis, the FHA maps have been long
destroyed (Xu, 2022). Because of this, the HOLC maps are used as a proxy to understand the
impacts of redlining, even though the HOLC itself had already made approximately 90% of its
loans before creating its maps (Michney, 2022). Debate still exists as to whether the HOLC
grades are an accurate representation of future area-based credit restrictions.
Scholars have increasingly used HOLC maps to understand how redlining has impacted
neighborhood-level trajectories - and hence the outcomes of individuals living in these
neighborhoods - over time. In the long run, living on the lower graded side of a D-C boundary is
associated with a decrease in educational attainment and annual real wages (Aaronson et al.,
2023) as well as lower homeownership rates and a higher black population share (Aaronson,
Hartley, et al., 2021). Even as late as 2018, almost 8 decades after grading, populations living in
formerly D-graded areas seem to have worse outcomes at the margin such as a higher prevalence
of single parent households and teen births as well as lower credit scores (Aaronson, Faber, et al.,
2021). Regarding health-related outcomes, living in a formerly D-graded area is associated with

6
significantly higher rates of emergency department visits due to asthma (Nardone et al., 2020)
and higher median levels of air pollution (Jung et al., 2022).
3. Contributions of the current work
This dissertation provides a detailed and nuanced understanding of the role that
neighborhood-level development plays in individual and area-level outcomes, interrogating the
role of both historic and modern-day influences. This work first contributes to the ongoing
discussion of the impacts of redlining by testing the extent to which the HOLC maps are
indicative of systematic mortgage discrimination as well as by proposing new ways of utilizing
HOLC data to analyze the connections between past area-level characteristics and current
disparities. The first two studies make use of information in the HOLC area description files,
which accompany the maps and provide details regarding each graded neighborhood’s
characteristics at the time of grading. These files allow for graded areas to be categorized
according to characteristics beyond just the grade given, making it possible to test potential
mechanisms beyond mortgage discrimination that may be related to observed outcomes.
Diverging from this historical lens, the third paper analyzes the impact of modern-day area-level
investments on individual-level health-related outcomes. This study uses longitudinal survey
data to contribute to the neighborhood effects literature in two primary ways. The first is through
its targeting of the impacts of a specific neighborhood-level intervention as opposed to defining
the intervention as the overall neighborhood. The second is that by focusing on the impacts of
investments within a low-income neighborhood, it departs from a larger body of neighborhood
effects literature which can be stigmatizing in its discussion of inherently “good” versus “bad”
neighborhoods. Instead, it focuses on the dynamic qualities of all neighborhoods, allowing for a

7
more nuanced discussion of neighborhood effects more broadly, and place-based investments in
particular.
4. Overview of findings
The first study in this dissertation analyzes the potential heterogeneity in outcomes that
may exist among HOLC D-graded areas by categorizing them according to their population
composition at the time of grading. In doing so, it investigates the extent to which HOLC maps
may be viewed as a proxy of mortgage discrimination. If HOLC maps are indeed indicative of
discriminatory lending behavior, we should see a similar direction of impact across all D-graded
areas, regardless of initial population composition. Outcome variables of interest are
homeownership rates and the black population percent in 1960 and 2000. Nighborhoods are
given the label “Any Black”, “Immigrant”, or “Neither” depending on the population mix noted
in the accompanying area description file.
This analysis presents several key findings. Notably, receiving a D grade is not a
universal predictor of either lower homeownership rates or higher percentages of black
households once D-graded areas with different initial populations are allowed to have divergent
pathways. D-graded neighborhoods with immigrant, but no black, residents do not consistently
show lower homeownership rates or higher levels of black residents than C-graded areas. On the
other hand, D-graded neighborhoods with any black residents at the time of grading have higher
percentages of black residents than C-graded areas in both 1960 and 2000, showing the
persistence of racial concentration over time. These results suggest that racial and ethnic
composition is a more salient predictor of future area-level outcomes than the specific HOLC
grade given.

8
The second study further explores the utility of the HOLC data by focusing on another
neighborhood-level characteristic that is described in the area description files: industrial activity.
Recent scholarship has linked HOLC graded areas to disparities in asthma outcomes (Nardone et
al., 2020; Schuyler and Wenzel, 2022) and poor air quality exposure (Lane et al., 2022). This
study uses the area description files to distinguish neighborhoods experiencing industry in some
capacity from those that were not. Observed long-term outcomes in these neighborhoods, which
are coded irrespective of grades, are then compared to observed outcomes in differently graded
areas. Outcome variables of interest include asthma prevalence, fine particulate matter
concentrations (PM2.5), and nitrogen dioxide concentrations (NO2).
Findings reveal that in this case, having received a D grade is a stronger indicator of
future asthma prevalence and NO2 concentrations than having been described as having
industrial activity at the time of grading. However, all outcome variables of interest are still
significantly predicted by the indicator of industrial activity, suggesting that differences in
industry siting may be capturing some, but not all, of the underlying differences in C versus Dgraded areas. One reason that discrepancies may exist, particularly in asthma outcomes, is that
industrial activity is only one of many contributing factors. Others include indoor environmental
and may speak to the differences in overall housing quality that occurred over time within the
two neighborhood types.
The third paper shifts from a historic to modern-day perspective and analyzes the impact
of economic investment in the form of residential, commercial, business, or recreational new
construction or renovation projects. It focuses on health-related outcomes within two historically
disadvantaged neighborhoods in the City of Pittsburgh using three distance exposure measures
from an individual’s place of residence: the neighborhood level, investments between ½ mile and

9
1 mile, and investments withing ½ mile. It uses longitudinal survey data occurring between 2011
and 2018, therefore departing from a cross-sectional analysis which cannot capture dynamic
impacts. Additionally, it tracks one specific neighborhood-level intervention and uses individual
and year fixed effects as well as sub-neighborhood-level controls to account for other changes
that may have been taking place concurrently.
Findings show a significant negative relationship between investments and food
insecurity that is largely driven by commercial and residential investments, suggesting that
investments may be altering socio-economic status which is a primary driver of food insecurity.
The distance-level analyses provide additional insight. For example, while at the neighborhood
level, there is a positive relationship between investments and perceived safety, this relationship
becomes negative for individuals living within a ½ mile of commercial investments. Such
findings suggest the nuanced relationship that can exist between specific investment types and
specific outcome variables, answering a call to better understand how changes to the built
environment impact long-term residents.

10
References
Aaronson, D., Faber, J., Hartley, D., Mazumder, B., & Sharkey, P. (2021). The long-run effects of
the 1930s HOLC “redlining” maps on place-based measures of economic opportunity and
socioeconomic success. Regional science and urban economics, 86, 103622.
Aaronson, D., Hartley, D., & Mazumder, B. (2021). The effects of the 1930s HOLC “redlining”
maps. American Economic Journal: Economic Policy, 13(4), 355-392.
Aaronson, D., Hartley, D., Mazumder, B., & Stinson, M. (2023). The long-run effects of the
1930s redlining maps on children. Journal of Economic Literature, 61(3), 846-862.
Arcaya, M. C., Tucker-Seeley, R. D., Kim, R., Schnake-Mahl, A., So, M., & Subramanian, S.
(2016). Research on neighborhood effects on health in the United States: a systematic
review of study characteristics. Social Science & Medicine, 168, 16-29.
Bailey, Z. D., Krieger, N., Agénor, M., Graves, J., Linos, N., & Bassett, M. T. (2017). Structural
racism and health inequities in the USA: evidence and interventions. The lancet,
389(10077), 1453-1463.
Barile, J. P., Kuperminc, G. P., & Thompson, W. W. (2017). Resident characteristics and
neighborhood environments on health‐related quality of life and stress. Journal of
Community Psychology, 45(8), 1011-1025.
Bryant-Stephens, T. C., Strane, D., Robinson, E. K., Bhambhani, S., & Kenyon, C. C. (2021).
Housing and asthma disparities. Journal of Allergy and Clinical Immunology, 148(5),
1121-1129.
Chetty, R., Hendren, N., & Katz, L. F. (2016). The effects of exposure to better neighborhoods on
children: New evidence from the moving to opportunity experiment. American Economic
Review, 106(4), 855-902.
D'Alessandro, D., & Appolloni, L. (2020). Housing and health. An overview. ANNALI DI
IGIENE MEDICINA PREVENTIVA E DI COMUNITÀ, 32(5 Supple 1), 17-26.
Fischer, M. J. (2003). The relative importance of income and race in determining residential
outcomes in US urban areas, 1970-2000. Urban affairs review, 38(5), 669-696.
Fishback, P., Rose, J., Snowden, K. A., & Storrs, T. (2022). New evidence on redlining by federal
housing programs in the 1930s. Journal of Urban Economics, 103462.
Jung, K. H., Pitkowsky, Z., Argenio, K., Quinn, J. W., Bruzzese, J.-M., Miller, R. L., . . .
Lovinsky-Desir, S. (2022). The effects of the historical practice of residential redlining in

11
the United States on recent temporal trends of air pollution near New York City schools.
Environment international, 169, 107551.
Katz, L. F., Kling, J. R., & Liebman, J. B. (2001). Moving to opportunity in Boston: Early results
of a randomized mobility experiment. The quarterly journal of economics, 116(2), 607-
654.
Kim, J. (2010). Neighborhood disadvantage and mental health: The role of neighborhood
disorder and social relationships. Social science research, 39(2), 260-271.
Kling, J. R., Liebman, J. B., & Katz, L. F. (2007). Experimental analysis of neighborhood
effects. Econometrica, 75(1), 83-119.
Lemire, E., Samuels, E. A., Wang, W., & Haber, A. (2022). Unequal Housing Conditions And
Code Enforcement Contribute To Asthma Disparities In Boston, Massachusetts: Study
examines housing conditions, code enforcement, and asthma disparities in Boston,
Massachusetts. Health Affairs, 41(4), 563-572.
Light, J. S. (2010). Nationality and neighborhood risk at the origins of FHA underwriting.
Journal of Urban History, 36(5), 634-671.
Liu, J., Clark, L. P., Bechle, M. J., Hajat, A., Kim, S.-Y., Robinson, A. L., . . . Marshall, J. D.
(2021). Disparities in air pollution exposure in the United States by race/ethnicity and
income, 1990–2010. Environmental health perspectives, 129(12), 127005.
Logan, J. R. (2011). Separate and unequal: The neighborhood gap for blacks, Hispanics and
Asians in metropolitan America (US 2010 Project, July Report). In.
Massey, D. S., & Denton, N. A. (1988). The dimensions of residential segregation. Social forces,
67(2), 281-315.
Michney, T. M. (2022). How the city survey’s redlining maps were made: a closer look at
HOLC’s mortgagee rehabilitation division. Journal of Planning History, 21(4), 316-344.
Michney, T. M., & Winling, L. (2020). New perspectives on new deal housing policy: explicating
and mapping HOLC loans to African Americans. Journal of Urban History, 46(1), 150-
180.
Mohai, P., Lantz, P. M., Morenoff, J., House, J. S., & Mero, R. P. (2009). Racial and
socioeconomic disparities in residential proximity to polluting industrial facilities:
evidence from the Americans' Changing Lives Study. American journal of public health,
99(S3), S649-S656.
Moyer, R., MacDonald, J. M., Ridgeway, G., & Branas, C. C. (2019). Effect of remediating
blighted vacant land on shootings: a citywide cluster randomized trial. American journal
of public health, 109(1), 140-144.

12
Nardone, A., Casey, J. A., Morello-Frosch, R., Mujahid, M., Balmes, J. R., & Thakur, N. (2020).
Associations between historical residential redlining and current age-adjusted rates of
emergency department visits due to asthma across eight cities in California: an ecological
study. The Lancet Planetary Health, 4(1), e24-e31.
Pollack, C. E., Blackford, A. L., Du, S., Deluca, S., Thornton, R. L., & Herring, B. (2019).
Association of receipt of a housing voucher with subsequent hospital utilization and
spending. JAMA, 322(21), 2115-2124.
Pollack, C. E., Roberts, L. C., Peng, R. D., Cimbolic, P., Judy, D., Balcer-Whaley, S., . . . Davis,
M. F. (2023). Association of a housing mobility program with childhood asthma
symptoms and exacerbations. Jama, 329(19), 1671-1681.
Ross, C. E., & Mirowsky, J. (2001). Neighborhood disadvantage, disorder, and health. Journal of
health and social behavior, 258-276.
Rothstein, R. (2017). The color of law: A forgotten history of how our government segregated
America. Liveright Publishing.
Ruijsbroek, A., Droomers, M., Hardyns, W., Groenewegen, P. P., & Stronks, K. (2016). The
interplay between neighbourhood characteristics: The health impact of changes in social
cohesion, disorder and unsafety feelings. Health & place, 39, 1-8.
Santiago, A. M., Galster, G., Lucero, J., Ishler, K., Lee, E. L., Kypriotakis, G., & Stack, L.
(2014). Opportunity neighborhoods for Latino and African American children. Available
at SSRN 2563141.
South, E. C., Kondo, M. C., Cheney, R. A., & Branas, C. C. (2015). Neighborhood blight, stress,
and health: a walking trial of urban greening and ambulatory heart rate. American journal
of public health, 105(5), 909-913.
Swope, C. B., & Hernández, D. (2019). Housing as a determinant of health equity: A conceptual
model. Social Science & Medicine, 243, 112571.
Wilson, W. J. (2012). The truly disadvantaged: The inner city, the underclass, and public policy.
University of Chicago Press.
Xu, W. (2022). Legacies of institutionalized redlining: A comparison between speculative and
implemented mortgage risk maps in Chicago, Illinois. Housing Policy Debate, 32(2),
249-274

13
Chapter 2. People and Place: Does the reason for redlining impact the long-term
trajectory of neighborhoods1
1. Introduction
The Home Owner’s Loan Corporation (HOLC) and its associated maps have become
increasingly used in the effort to study the impacts of mortgage discrimination. These maps
detail the assessments of local appraisers from across the country who graded urban
neighborhoods according to their relative credit risk (Greer, 2013). In addition, these maps are
notorious for how the grading process interacted with race and ethnicity, both for their
descriptions of marginalized populations and for the high concentration of black and immigrant
residents in the lowest graded areas (Fishback et al., 2022; Greer, 2013; Light, 2010). For
example, Greer (2013) finds “a perfect overlap geographically of the presence of even a single
non-white resident…and a red or hazardous HOLC grade” when analyzing the process of
redlining in Chicago. Scholars argue that this reflects local appraisers inheriting the “racial
common sense” of the real-estate industry which viewed black and immigrant communities as
risks to long-term property values (Taylor, 2019; Rothstein, 2017). Despite the interaction
between race, ethnicity, and redlining, very little is known about how the presence of certain
population groups interacted with the practice of redlining to either exacerbate or mitigate its
effects. This is an important question as it not only addresses potential heterogeneity in the
distribution of the impacts of receiving a D-grade, but it also addresses how policy interacts with
social perception to either exacerbate or mitigate disadvantage. Despite both immigrant and
black populations being associated with increased credit risk by the HOLC maps, these two
populations have very different social histories in the United States. For example, disparities in
1 Co-authored with Gary Painter

14
trends of discrimination can be seen following WWII as white immigrant groups were able to
access certain benefits and programs that black individuals were barred from (Brodkin, 1998).
Additionally, while immigrant groups did form ethnic enclaves in some cases, they did not
experience segregation and ghettoization to a level comparable to that which was experienced by
black populations (Massey & Denton, 1993).
Given the increasing use of the HOLC maps as a tool to understand discriminatory
lending practices as well as to understand their intersection with marginalized population groups,
the goals of this research are two-fold. We examine the uses and limitations of the HOLC maps
as a proxy for mortgage discrimination as well as investigate how the existing discrimination
towards marginalized populations interacted with HOLC redlined areas. To do this, we
investigate the potentially heterogenous trajectories of D-graded neighborhoods with either
immigrant and/or black residents mentioned as living in the area at the time of grading by
investigating the evolution of homeownership rates and the percentage of black residents in both
1960 and 2000, using a sample of C-graded neighborhoods as a comparison group.
If the grades given by the Home Owner’s Loan Corporation have sufficient overlap with
the subsequent lending decisions of influential actors such as the Federal Housing Association
(FHA), then we should see a similar direction of impact across all D-graded areas. If insufficient
overlap exists, the direction of impact may be inconsistent or nonexistent, suggesting that lenders
were making decisions across areas that diverged from those made by the HOLC (i.e., providing
loans within D-graded areas or denying loans in C-graded areas).
The results of this analysis confirm previous national studies which find an association
between areas assigned a D grade vs. a C grade and area outcomes. However, when
decomposing areas graded D because of the presence of immigrant populations and not black

15
populations, we do not find long term negative impacts of the D grade for such areas. At the
same time, we find using a national sample that D-graded neighborhoods that had any black
residents at the time of grading have lower homeownership rates in both 1960 and 2000 than Cgraded neighborhoods as well as a higher percentage of black residents in both years. When
stratifying by region, we find no difference between the homeownership rates of C and D-graded
areas that contained any black residents at the time of grading but do find a statistically higher
black population percentage in D-graded areas than C-graded areas across all regions.
2. Background
The Home Owner’s Loan Corporation and the Federal Housing Administration
The Home Owner’s Loan Corporation (HOLC) and the Federal Housing Administration
(FHA) both produced maps that graded neighborhoods according to their perceived credit risk;
however, the purpose of each organization as well as the realized use of each set of maps was
markedly different. The HOLC was originally established with the goal of addressing the failing
home finance industry through refinancing troubled loans. It wasn’t until the HOLC had already
closed approximately 90% of its loans that it began creating its now infamous maps (Michney,
2022). In contrast to the HOLC, the FHA was designed to be a long-term actor within the
housing sector whose mission was to fundamentally reform mortgage markets (Fishback et al.,
2022). The FHA’s own set of maps directly dictated its lending practices as the FHA refused to
guarantee any loans in D-graded neighborhoods, which increased the cost of homeownership and
construction in these areas (Xu, 2022).
Because the HOLC did not utilize its maps in its own lending activity, the only way that
the maps could be causally related to patterns of mortgage discrimination is if they were
subsequently utilized by either private or public lenders. Given the confidential nature of the

16
HOLC maps, it is unlikely that private lenders had access to them; however, there is
documentation that the FHA was given copies of the HOLC’s finished maps (Michney, 2022;
Hillier, 2005). The degree to which the HOLC maps may have influenced the FHA’s decisionmaking is debated. Fishback et al. (2022) note that the FHA was engaged in lending even before
the HOLC created its maps, and that in several cities, the FHA was already making fewer loans
in areas later graded C or D by the HOLC. Hillier (2005) additionally states that the FHA had
already created a risk system of its own before the HOLC started its map-making process.
Even if the HOLC maps did not directly influence how the FHA assigned grades to
neighborhoods, data derived from the HOLC maps could still be useful in understanding the
landscape of discriminatory real estate practices if sufficient overlap exists between the two sets
of maps. Indeed, it is plausible that the two organizations graded neighborhoods similarly as they
focused on many of the same neighborhood characteristics when making their determinations.
For example, both the FHA and the HOLC favored areas with either zoning regulations or
racially restrictive covenants already in place (Greer, 2012). Additionally, both were instructed
to rely on the expertise of local real estate agents when making their assessments, implying an
overlap of source material that contributed to the final grading decision (Michney 2022; Light
2010). If in fact there exists a degree of overlap between the HOLC maps and “other private or
public maps that shared common neighborhood grades and borders,” then they could plausibly
serve as a proxy for discriminatory lending (Aaronson et al., 2022).
Given that the FHA maps for every city except Chicago were destroyed, directly testing
the overlap between the grades given by the HOLC and the FHA is difficult. Xu (2022), studying
the relationship between the HOLC and the only existing FHA map of Chicago, notes that the
overlap of grades is substantial but not exact. Using Xu’s publicly available data (Xu, 2022), we

17
determine that approximately 75% of areas graded D by the HOLC in Chicago were also graded
D by the FHA (of the areas graded D by the FHA, just over 80% were also graded D by the
HOLC).2 While by no means a perfect overlap, this example implies that a high degree of
similarity exists between the grading decisions of the two entities in Chicago. The question
remains as to whether such overlap exists in other graded cities.
One goal of this work is to explore the merits and limitations of utilizing data associated
with the HOLC maps to understand discriminatory loaning practices. If the HOLC maps are a
proxy for discriminatory lending, we should see D-graded areas exhibiting the same direction of
impact in relation to comparable C-graded areas for outcomes related to mortgage availability.
By exploring heterogeneity of outcomes across D-graded areas using weighted C-graded areas as
the reference group, we can assess to what extent an area graded C or D by the HOLC implies
systematic access to, or denial from, subsequent housing finance.
Population Composition and Neighborhood Desirability
As mentioned, one neighborhood characteristic that both the Home Owner’s Loan
Corporation and the Federal Housing Administration were particularly concerned with is that of
population composition. Drawing from the prevailing wisdom of the real estate industry that
“differences in neighborhood values and conditions were ones that different races and
nationalities inherently produced” (Light, 2010), both organizations assessed a neighborhood’s
foreign born and black population as well as whether a non-white population shift was
anticipated when assigning grades (Light 2010; Greer, 2012; Taylor, 2019; Rothstein, 2017).
While population composition was neither the only characteristic used nor always the most
2 Xu (2022) assigns grades at the census tract level. Tracts that have 25% or more area overlap with a HOLC or
FHA graded neighborhood is given the grade of that overlapped neighborhood. Tracts with less than 25% overlap
are excluded from the analysis

18
influential when assigning grades (Fishback et al. 2020), several scholars have found that the
presence of black and/or immigrant residents independently predicted a worse grade (Hillier
2005; Light, 2010).
Anecdotal evidence for this relationship exists within the HOLC area description files
themselves, which describe the characteristics of an area being graded. In Chicago, IL, one
appraiser makes the statement that “the area is well laid out and improvements are attractive, but
the proximity of negro families on Spruce St. at the southern edge precludes the district from a
better rating.”3 Another appraiser, also in Chicago, states, “this is a comparatively new
neighborhood of mediocre housing, adversely affected by the marked infiltration of low-class
Italian and Sicilian.”4 Population composition was certainly impacting the appraisers’ valuations,
with foreign-born and black residents being viewed as detrimental to long-term value.
The level of discrimination faced by black and immigrant residents outside of these
valuations, however, was vastly different and only continued to diverge over time. Once again
using Chicago as an example, the top three regions of origin of foreign-born populations living in
D-graded areas around 1940 are Eastern Europe, Southern Europe, and Western Europe, and
were therefore largely white (see Appendix Table 1).5 Examples of disparities in the urban
experience of white immigrant versus black individuals are numerous (Eriksson & Ward, 2019;
Shertzer et al., 2016; Hershberg et al., 1979; Massey & Denton, 1993), such as white immigrant
groups gaining access to benefits and programs that black individuals were barred from postWWII (Brodkin, 1998).
3 Mapping Inequality database, Chicago B2
4 Mapping Inequality database, Chicago C106
5 Data drawn from Mapping Inequality database

19
These disparities were particularly prevalent when it came to housing access and
residential mobility. As black populations moved north, white residents made a concerted effort,
often involving intimidation and outright violence, to confine black residents to racially
homogenous neighborhoods (Massey & Denton, 1993). The real estate industry assisted in this
effort through the practice of blockbusting, which involved “turning” a neighborhood for profit
by buying cheaply from white residents fearing the arrival of black neighbors and then selling at
a much higher price to black families whose housing options were limited (Akbar et al., 2022;
Gotham, 2002; Massey & Denton, 1993).
It is unclear whether the differing layers of social and housing discrimination faced by
each population heterogeneously impacted D-graded neighborhoods where these groups were
living. By analyzing the trajectory of D-graded areas based on initial population composition, we
provide important evidence on the salience of the D-grade vs. other factors that might lead to
differences in neighborhood outcomes.
3. Data and Methodology
Description of data used
Our outcomes of interest in this study are the percentage of black residents and the
homeownership rate in an area. As black residents had few housing choices outside of the most
economically disadvantaged neighborhoods during the creation of the HOLC maps (Fishback et
al., 2020), measuring the change in the percentage of black residents across the lowest graded
neighborhoods allows us to understand the dynamics of racial filtering in these areas. We focus
on homeownership rate as it represents a successful transaction within the housing market from
loan acquisition to an accepted offer and has been used in several other papers to assess the

20
impacts of mortgage discrimination (Faber, 2020; Aaronson et al., 2021; Appel & Nickerson,
2016; Krimmel, 2020).
We measure the outcome variables of interest from two years: 1960 and 2000. We chose
these years as they mark the periods before and after the passage of the 1968 Fair Housing Act.
Data from 1960 is the product of a period in the housing market where discriminatory lending
practices were commonplace and segregation was moving towards its 1970 peak (Xu, 2022;
Cutler et al., 1999). Conversely, data from 2000 represents an era where overt discriminatory
housing policies had been legally banned (Xu, 2022).
The data used in the empirical analysis come from three sources. The 1960 and 2000
outcome data as well as neighborhood control variables from 1940 are drawn from the Census.6
We use the publicly available tabulated HOLC area description file data from Markley (2022) to
derive the percentage breakdown of black and immigrant residents. From the University of
Richmond’s Mapping Inequality dataset, we retrieve shapefiles containing the original HOLCgraded neighborhood boundaries (Nelson et al., 2021) which include graded neighborhoods from
128 cities across 28 states. After merging with census data, however, our full sample includes
1,781 observations from 17 states and 40 cities. This decrease in observations is due to
limitations of the 1940 census, which only includes data from 74 counties across 29 states.
Despite the decrease in overall observations, our sample still provides adequate regional
variance, with 717 graded neighborhoods in the Midwest, 497 in the Northeast, 231 in the South,
and 336 in the West.
6 The 1940, 1960 and 2000 census data are available from the National Historic Geographic Information System
(NHGIS).
2 We will use the term area and neighborhoods interchangeably in the narrative.

21
Census tract boundaries do not perfectly align with the boundaries of the HOLC
neighborhoods. To correct this mismatch in geographic overlap, we calculate the percentage of
each census tract that lies within a HOLC boundary and use these values to weight the census
data. We then aggregate this data at the HOLC boundary level. Following Aaronson et al., 2021,
we exclude any census tract that has a less than 15 percent area overlap with a HOLC
neighborhood.
Categorization of D-graded neighborhoods by population composition
To analyze how D-graded areas interact with race and ethnicity over time, we create an
indicator of population composition by isolating neighborhoods that had any black households at
the time of grading from neighborhoods where appraisers only mentioned immigrants as living
in the area, but no black households. Neighborhoods with any combination of black and
immigrant population we call “Any Black”, while neighborhoods where only an immigrant
population is mentioned, we call “Immigrant”. We also isolate D-graded neighborhoods that had
neither black nor foreign-born populations at the time of grading. These we refer to as “Neither”.
Given the concerted and often violent reaction of neighborhoods to even one black
household crossing the “color line” (Massey & Denton, 1993; Taylor, 2019), we hypothesize that
a D-graded area with even one black family would face layers of marginalization and
discrimination in excess to those seen by neighborhoods with only foreign-born or white native
residents. This strict classification allows us to explore this hypothesis. However, since this
breakdown inherently masks more complex relationships that may exist between outcomes and
relative population compositions, we additionally explore whether areas with a relatively greater
percentage of black and/or immigrant residents have different outcomes as an additional test.

22
We use data from the area description files to create these indicators. While census data
has more consistent methods of acquisition, area description file data offers several advantages
for this study. Since we are concerned with the initial population composition of graded
neighborhoods, the mismatch of borders between census data and the HOLC graded
neighborhoods becomes problematic. To use census data, all indicators must be weighted
according to their percent overlap with the HOLC graded neighborhoods, meaning that it is
possible only a portion of the census tract falls inside of the HOLC boundary. Since we are
unable to determine the exact geographic distribution of households within the census tract, it is
difficult to know which households fall within the HOLC-graded neighborhood. This becomes a
problem as we are interested in how even very small percentages of disadvantaged populations
interact with neighborhood trajectory. By using data from area description files, we know
whether appraisers consider a minority household to be inside or outside of the HOLC boundary.
Additionally, the HOLC assessments largely represent the opinions of local real estate
professionals (Michney, 2022), meaning that even if the local appraisers are incorrect in their
estimations of population percentage, it is likely that real estate officials shared this assumption
and treated neighborhoods accordingly.
The maps in Figures 1a and 1b give an example of how foreign-born and black residents
are distributed across neighborhoods within New York City, New York. Here we see D-graded
neighborhoods with high concentrations of foreign-born residents (60-100%) geographically
spread across the city. D-graded neighborhoods with the highest percentage of black households
(70-100%) are much more clustered.

23
Figure 1a: Distribution of foreign born residents across New York and New Jersey
Figure 1b: Distribution of black residents across New York and New Jersey

24
4. Methodology
In this paper, we are interested in the interaction between population composition and the HOLC
credit assessment to analyze the long-term trajectory of D graded neighborhoods with different
demographic characteristics. A common approach in the literature to assess the impact of
receiving a D grade has been to compare outcomes in D graded areas to outcomes in non-D
graded areas, with C graded areas used most often as the comparison group. Summary statistics
shown in Table 1 indicate that there are sufficient differences in initial characteristics at the time
of grading between C and D graded areas in the Northeast and Midwest to suggest that C graded
neighborhoods are not an appropriate counterfactual. The implication of these differing baseline
characteristics is that C and D graded neighborhoods were on different trajectories at the time of
grading. Any relationship between these graded neighborhoods and our outcomes of interest
could therefore be attributed to differences in initial characteristics rather than to the grade or
demographic composition.

25
Table 1: Regional descriptive statistics by HOLC grade
Notes: Data taken from 1940 Census: Population & Housing Data [Tracts & NY Health Areas:
Major Cities & Surrounds] and tabulated area description file (ADF) data from Markley (2022);
Census data weighted according to percent overlay with HOLC-graded neighborhood boundaries
and aggregated at the HOLC neighborhood level. Neighborhoods are categorized according to
corresponding area description files taken from the University of Richmond’s Mapping
Inequality database; Distance from city center calculated using point data from Open Street
Maps
To address this selection on observable characteristics, we use inverse propensity score
weighting. Because local appraisers did not have a definitive guide on how to translate
neighborhood characteristics to a grade, it is very possible that certain C and D graded areas
Northeast Midwest South West
C-graded
Mean
D-graded
Mean
C-graded
Mean
D-graded
Mean
C-graded
Mean
D-graded
Mean
C-graded
Mean
D-graded
Mean
Total 302 195 447 270 123 109 239 112
Percent
Black ADF
0.52% 11.55% 0.38% 15.46% 0.36% 46.74% 0.18% 5.01%
Percent
Immigrant
ADF
24.98% 47.63% 21.39% 35.70% 0.28% 1.72% 4.28% 26.07%
Percent
Owner
1940
35.98% 26.26% 46.73% 40.37% 38.12% 30.87% 47.01% 42.10%
Median
Rent 1940
32.12 24.99 26.16 17.37 19.20 12.28 25.74 20.64
Percent
Vacant
1940
7.22% 8.30% 2.87% 3.06% 3.32% 2.83% 5.65% 5.28%
Percent
Repairs
1940
5.38% 9.48% 7.39% 13.30% 7.95% 13.11% 4.90% 11.28%
Male
Schooling
1940
(years)
7.85 6.97 8.06 6.61 8.58 6.82 9.85 8.58
Dist from
City Center
(miles)
9.54 7.71 10.33 8.32 4.76 4.29 13.57 12.82

26
were characteristically indistinguishable and their difference in grade was simply due to an
arbitrary decision made by the grader. Using inverse propensity score weighting will ensure that
areas graded D that are characteristically the most similar to C graded areas and areas graded C
that are characteristically the most similar to D graded areas receive the greatest weight in the
analysis.
We perform two different analyses – the first is on a national sample and the second
stratifies by region to control for systematic regional differences. The necessity of this can be
seen by looking at the distribution of neighborhood types across region. Table 2 shows how the
overall number of D neighborhood types varies, reflecting each region’s different
sociodemographic history. For example, there are only 3 D graded immigrant neighborhoods in
the South, while in the Northeast, there are 101 D graded immigrant areas, encompassing about
46% of the total D graded areas in the Northeast.
Table 2: Count and percentage of HOLC C- and D-graded neighborhood type by region
Cneighborhood
D-any Black
neighborhood
D-immigrant
neighborhood
D-neither
neighborhood
Region total Region total Region total Region total
Midwest 454 145 92 33
Northeast 338 105 101 10
South 123 81 3 25
West 239 64 42 6
Notes: Counts of neighborhoods by region taken from The University of Richmond’s Mapping
Inequality Database.
Inverse propensity score weighting
To determine propensity scores, we predicted the likelihood of receiving a D grade
(treatment = 1) based on the following data: percentage of black residents, percentage of
immigrant residents, median rent, average years of male schooling, percentage of units that are

27
vacant, the homeownership rate, the percentage of units needing repairs, and the distance to city
center. The distribution of the inverse propensity score weights for the Northeast and Midwest
are shown in Figures 2 through 5.7
Figure 2: Distribution of inverse propensity score weights in the Northeast
Figure 3: Distribution of inverse propensity score weights in the Midwest
7 To prevent outliers from having an undue influence on the regression, they have been removed. In the Northeast
and Midwest, this included areas with weights greater than 10 (removes 10 areas in the Northeast and 7 in the
Midwest). In the South and West, this included areas with weights greater than 5 (removes 11 areas in the South and
23 in the West)

28
Figure 4: Distribution of inverse propensity score weights in the South
Figure 5: Distribution of inverse propensity score weights in the West
Table 3 compares regional C and D neighborhood characteristics when weighted by their
inverse propensity scores. As can be seen in Table 1, in an unweighted sample, C neighborhoods
are typically farther from the city center, have male populations with slightly higher mean years
of schooling, have higher homeownership rates, and have a lower percentage of units needing

29
repairs. Additionally, C-graded neighborhoods have a lower immigrant population percentage
and a negligible black population percentage. In Table 3 we see that these differences, though
not eradicated, are minimized when using inverse propensity score weighting, especially in the
Midwest. For example, in an unweighted sample, the difference between homeownership rates in
the Midwest is about 6 percentage points while in the weighted sample, the difference is less
than 2 percentage points. In the Northeast, we also see differences between C and D graded areas
minimized. The difference in percent ownership drops from just under 10 percentage points in
the unweighted sample to just under 6, while the difference in median rent drops from about 7
units to about 4.5.
Inverse propensity score weighting is less effective in producing a comparable sample in
the South and the West, where differences between C and D graded areas remain high. These
results imply that differences between characteristics of C and D graded neighborhoods at the
time of grading were much more pronounced in the West and the South than in the Northeast
and Midwest. Thus, results in the West and the South are more subject to questions of bias.
5. Results
Descriptive statistics
Table 4 shows neighborhood characteristics for four different combinations of grade and
initial population classification. Focusing on demographics, there is predictably much variation
among the neighborhood types. D neighborhoods that were noted as having any black residents
by appraisers have a much higher percentage of black residents than C neighborhoods, whose
total percentage is only 0.38%. This is in contrast to the percent immigrant difference between D
neighborhoods noted as having an immigrant, but no black, population and C graded
neighborhoods. Here, we see that while D graded immigrant neighborhoods have the highest

30
percentage of immigrant residents, C immigrant neighborhoods still have an immigrant
population of over 16%.8
In other words, while immigrant groups seem to be dispersed across D
neighborhood types and even represented in C graded areas, black residents are not. Instead,
there are heavily concentrated in fewer areas that are almost exclusively D graded.
Table 4: Neighborhood characteristics by HOLC grade and population categorization
C- neighborhood D-any black
neighborhood
D-immigrant
neighborhood
D-neither
neighborhood
Total 1154 395 238 76
Percent Black ADF 0.38 (1.81) 31.37 (32.32) 0.00 (0.00) 0.00 (0.00)
Percent Immigrant
ADF
16.90 (21.96) 29.38 (27.49) 47.57 (31.12) 0.00 (0.00)
Percent Owner 1940 42.44 (17.85) 32.06 (17.27) 35.84 (20.54) 44.05 (20.34)
Median Rent 27.01 (11.37) 18.24 (9.72) 20.90 (9.46) 19.73 (11.40)
Percent Vacant 1940 4.68 (5.96) 4.25 (5.01) 5.92 (6.65) 4.93 (5.42)
Percent Repairs 1940 6.26 (6.34) 12.68 (10.59) 10.19 (9.52) 11.77 (10.97)
Male Schooling 1940 8.37 (2.18) 6.71 (2.20) 7.17 (2.05) 7.98 (2.50)
Dist from City Center 10.22 (6.88) 7.15 (6.40) 10.14 (7.88) 7.93 (6.17)
Notes: Data taken from 1940 Census: Population & Housing Data [Tracts & NY Health Areas:
Major Cities & Surrounds] and tabulated area description file (ADF) data from Markley (2022);
Census data weighted according to percent overlay with HOLC-graded neighborhood boundaries
and aggregated at the HOLC neighborhood level. Neighborhoods are categorized according to
corresponding area description files taken from the University of Richmond’s Mapping
Inequality database; Distance from city center calculated using point data from Open Street
Maps
D graded areas with any black residents at the time of grading also have populations with
the lowest average years of male schooling, are closest to the center of the city, have the lowest
ownership rates and rents, and have the highest percentage of units needing repairs. D
8
In results not shown, we investigated whether we could detect differences between European and Latin American
immigrants. The results were qualitatively similar although sample sizes for Latin American immigrants were too
small for statistically significant estimates.

31
neighborhoods noted as having immigrant but no black residents present at the time of grading
are on average about as far from the city center as C graded neighborhoods. D immigrant
neighborhoods have lower homeownership rates, rents, and average years of schooling of
residents than do C graded areas.
Table 5: Outcome variables by HOLC grade and population categorization
C- neighborhood D-any black
neighborhood
D-immigrant
neighborhood
D-neither
neighborhood
Total 1154 395 238 76
Percent Owner 1960 54.55 (23.48) 42.41 (21.93) 45.35 (26.67) 57.67 (25.07)
Percent Owner 2000 45.15 (19.21) 35.25 (17.52) 41.55 (20.50) 45.83 (20.24)
Percent Black 1960 8.46 (18.43) 33.05 (31.77) 7.83 (19.20) 16.10 (23.62)
Percent Black 2000 29.24 (33.75) 45.45 (33.45) 24.11 (32.77) 46.50 (32.84)
Notes: Data taken from 1940 Census: Population & Housing Data [Tracts & NY Health Areas:
Major Cities & Surrounds]; Census data weighted according to percent overlay with HOLCgraded neighborhood boundaries and aggregated at the HOLC neighborhood level.
Neighborhoods are categorized according to corresponding area description files taken from the
University of Richmond’s Mapping Inequality database
Table 5 shows the evolution of the percentage of black residents and of homeownership
rates absent any attempt to control for initial neighborhood characteristics over time. In 1960,
neighborhoods initially identified as having any black residents still have the highest percentage
of black residents at 33.05% while C neighborhoods and D immigrant neighborhoods only have
a black population share of 8.46% and 7.83% respectively. By 2000, however, the share of black
residents in D any black neighborhoods rises by about 12 percentage points while the percentage
jumps by about 17 percentage points in D graded immigrant neighborhoods and 21 percentage
points in C graded neighborhoods.
Homeownership rates in 1960 are highest in C graded areas at 54.55% and lowest in D
any black neighborhoods at 42.41%. D immigrant neighborhoods have a homeownership rate
about 9 percentage points lower than C graded areas. By 2000, however, these numbers shift as

32
homeownership rates decrease across the board. D immigrant neighborhoods are now only about
4 percentage points lower than C graded neighborhoods while D any black neighborhoods are
about 10 percentage points lower than C graded neighborhoods.
Results: National Sample
C-graded versus D-graded areas
Table 6 displays regression output for C versus D graded neighborhoods using inverse
propensity score weighting. As is evident comparing graded neighborhoods across the nation, we
note that D neighborhoods have lower homeownership rates than C neighborhoods in 1960 and
2000 as well as a higher percentage of black residents in both 1960 and 2000. These initial
results are consistent with previous literature which implies that simply receiving a D grade as
opposed to a C grade adversely impacts homeownership rates for decades post-grading.
Table 6: National IPW regression by HOLC grade
Dependent variable:
Percent Owner
(1960)
Percent Owner
(2000)
Percent Black
(1960)
Percent Black
(2000)
(1) (2) (3) (4)
HOLC Grade D -0.031*** (0.006) -0.013** (0.006) 0.027*** (0.010) 0.032** (0.016)
Median Rent (1940) -0.001 (0.0004) 0.002*** (0.0004) -0.003*** (0.001) 0.0002 (0.001)
Male Schooling (1940) -0.005*** (0.002) -0.011*** (0.002) -0.009*** (0.003) -0.023*** (0.005)
Percent Owner (1940) 1.132*** (0.018) 0.814*** (0.019) -0.265*** (0.032) -0.002 (0.050)
Percent Vacant (1940) 0.192*** (0.052) 0.337*** (0.054) -0.157* (0.092) -0.487*** (0.144)
Percent Repairs
(1940)
0.063* (0.038) 0.139*** (0.040) 0.178*** (0.068) -0.042 (0.105)
Distance from City
Center
-0.0002 (0.0005) 0.002*** (0.0005) -0.002** (0.001) -0.003** (0.001)
Constant 0.118*** (0.013) 0.110*** (0.014) 0.368*** (0.024) 0.543*** (0.037)
Observations 1,832 1,832 1,832 1,832
Notes: Reference category is all C-graded neighborhoods. Standard errors are in parentheses.
*p<0.1; **p<0.05; ***p<0.01
C-graded versus D-graded areas grouped by population composition

33
When we examine heterogeneity in D-graded areas according to initial demography
rather than grouping these neighborhood types together (Table 7), we find important differences
for neighborhoods that do not include black residents. Here, neighborhoods indicated as having
any black residents living in them at the time of grading still have lower homeownership rates
than C graded areas in both 1960 and 2000, while neighborhoods where only immigrant
populations were identified as living at the time of grading no longer have lower homeownership
rates than C graded neighborhoods by 2000. Additionally, D any black neighborhoods have
higher black populations than C graded areas in both 1960 and 2000 whereas D immigrant
neighborhoods have lower concentrations of black populations in 1960, and a black population
that is statistically indistinguishable from C graded areas by 2000.
Table 7: National IPW regression by population categorization
Dependent variable:
Percent Owner
(1960)
Percent Owner
(2000)
Percent Black
(1960)
Percent Black
(2000)
(1) (2) (3) (4)
AnyBlack -0.027*** (0.008) -0.023*** (0.009) 0.130*** (0.014) 0.074*** (0.023)
Immigrant -0.042*** (0.007) -0.012 (0.007) -0.039*** (0.012) -0.028 (0.020)
Neither -0.008 (0.010) 0.002 (0.011) 0.015 (0.018) 0.117*** (0.028)
Median Rent (1940) -0.0005 (0.0004) 0.002*** (0.0004) -0.002*** (0.001) 0.001 (0.001)
Male Schooling
(1940) -0.006*** (0.002) -0.011*** (0.002) -0.010*** (0.003) -0.025*** (0.005)
Percent Owner (1940) 1.130*** (0.018) 0.807*** (0.019) -0.223*** (0.032) -0.001(0.050)
Percent Vacant
(1940) 0.190*** (0.052) 0.327*** (0.054) -0.092 (0.090) -0.481*** (0.143)
Percent Repairs
(1940) 0.057 (0.038) 0.144*** (0.040) 0.103 (0.066) -0.084 (0.105)
Distance from City
Center -0.0001 (0.0005) 0.002*** (0.0005) -0.002** (0.001) -0.002* (0.001)
Constant 0.118*** (0.013) 0.113*** (0.014) 0.343*** (0.023) 0.538*** (0.037)
Observations 1,832 1,832 1,832 1,832
Notes: Reference category is all C-graded neighborhoods. Standard errors are in parentheses.
*p<0.1; **p<0.05; ***p<0.01

34
Results: regional
C-graded versus D-graded areas grouped by population composition
Focusing first on the Northeast (Table 8), D graded neighborhoods marked as having any
black residents present do not have a statistically significant difference in homeownership rates
from C graded areas in either 1960 or 2000. These neighborhoods do have a greater percentage
of black residents than C graded areas in 1960, but this difference is no longer significant by
2000. D immigrant neighborhoods have lower homeownership rates than C graded areas in
1960, but surprisingly, have higher homeownership rates in 2000. D immigrant neighborhoods
have lower percentages of black residents than C graded areas in both 1960 and 2000.
In the Midwest (Table 9), we once again see that D graded neighborhoods with any black
residents at the time of grading do not have statistically significant differences in homeownership
rates from C graded areas. These neighborhoods do, however, have larger percentages of black
individuals in both 1960 and 2000. D immigrant neighborhoods again have lower
homeownership rates than C graded areas in 1960, with this difference remaining marginally
significant by 2000. Comparing percentages of black residents between D immigrant
neighborhoods and C graded neighborhoods, there is no difference in either year.
In the South (Table 10), due to the low sample size of D immigrant neighborhoods, we
only focus on D graded areas indicated as having any black individuals at the time of grading.
Here, we see that D graded neighborhoods with any black residents have significantly higher
percentages of black residents in both 1960 and 2000 than C graded areas, but we see no
statistically significant differences in homeownership rates in either year.

35
Finally, in the West (Table 11), D graded areas with any black residents at the time of
grading have lower homeownership rates than C graded areas in 1960 as well as a higher
percentage of black residents in 1960. The difference in both homeownership rates and black
residents dissipates by 2000. For D graded areas with immigrant but no black individuals at the
time of grading, there are fewer black residents than C graded areas in 1960.
Additional analysis: graded versus ungraded
Faber (2020) uses the strategy of comparing graded and non-graded metropolitan areas to
identify the impact of redlining on segregation. Below we employ a similar strategy within
graded metropolitan areas to determine trajectories of ungraded census tracts to those receiving
C and D grades. Since the HOLC appraisers did not follow census tract boundaries when
defining neighborhoods for grading, in many graded cities, we observe a number of tracts that
appear in the 1940 census data but that do not appear in the HOLC maps. This boundary
mismatch is demonstrated in Figure 6 showing a map of Union City, NJ, Manhattan, NY, and
their surroundings. All yellow areas are tracts in the 1940 census which were ungraded by the
HOLC. In other metropolitan areas, the majority of ungraded areas were in less developed or
suburban areas. Because ungraded areas were not part of the HOLC maps, they were not
influenced by any credit restrictions that the HOLC maps may have overlapped with. We
therefore were interested in using these areas as a control group for our graded C and D sample.
Figure 6: HOLC graded versus ungraded areas in New York and New Jersey.

36
We restrict the sample to include only cities that include both graded and ungraded tracts.
Cities range from being about 30% graded to about 60% graded. The results in Table 12
demonstrate that all D graded areas show lower homeownership rates than ungraded areas in
both 1960 and 2000. Only D graded areas that had an immigrant population at the time of
grading show no difference from ungraded areas in the percentage of black residents across both
1960 and 2000. In Table 13, we additionally compare C graded areas to ungraded tracts and find
lower homeownership rates in 1960 and 2000. We also find a greater percentage of black
residents living in C graded areas than ungraded areas in both decades. We attribute these results
to the possibility that most ungraded areas were later developed into suburban areas that were
largely restrictive to black residents. The difference between the trajectory of graded and

37
ungraded tracts to better understand the processes occurring in both warrants subsequent
research.
Table 12: National IPW regression results by population categorization, D-graded HOLC
neighborhoods versus ungraded census tracts
Dependent variable:
Percent Owner
(1960)
Percent Owner
(2000)
Percent Black
(1960)
Percent Black
(2000)
(1) (2) (3) (4)
AnyBlack -0.059*** (0.010) -0.077*** (0.011) 0.173*** (0.016) 0.148*** (0.021)
Immigrant -0.073*** (0.012) -0.047*** (0.012) -0.019 (0.019) -0.021 (0.024)
Neither -0.011 (0.018) -0.041** (0.019) 0.049* (0.029) 0.194*** (0.038)
Median Rent (1940) -0.001 (0.0005) -0.0005 (0.001) -0.002* (0.001) -0.0004 (0.001)
Male Schooling
(1940) -0.011*** (0.002) -0.007*** (0.002) -0.004 (0.003) -0.015*** (0.004)
Percent Owner (1940) 1.128*** (0.024) 0.777*** (0.026) -0.203*** (0.039) -0.082 (0.050)
Percent Vacant (1940) -0.021 (0.073) 0.131* (0.078) -0.249** (0.118) -0.637*** (0.152)
Percent Repairs
(1940) 0.019 (0.032) 0.090*** (0.035) 0.166*** (0.052) 0.016 (0.068)
Distance from City
Center 0.003*** (0.001) 0.004*** (0.001) -0.004*** (0.001) -0.005*** (0.001)
Constant 0.181*** (0.016) 0.190*** (0.017) 0.262*** (0.025) 0.482*** (0.033)
Observations 1,166 1,166 1,166 1,166
Notes: Reference category is ungraded census tracts. Standard errors are in parentheses. *p<0.1;
**p<0.05; ***p<0.01
Table 13: National IPW regression results, C-graded HOLC neighborhoods versus ungraded
census tracts
Dependent variable:
Percent Owner
(1960)
Percent Owner
(2000)
Percent Black
(1960)
Percent Black
(2000)
(1) (2) (3) (4)
C-grade -0.026*** (0.007) -0.043*** (0.008) 0.041*** (0.009) 0.082*** (0.017)
Median Rent (1940) -0.002*** (0.0003) 0.00001 (0.0004) -0.002***
(0.0005) -0.002** (0.001)

38
Male Schooling
(1940) -0.009*** (0.002) -0.007*** (0.002) -0.001 (0.002) -0.016*** (0.004)
Percent Owner (1940) 1.139*** (0.019) 0.798*** (0.022) -0.199*** (0.025) -0.090** (0.046)
Percent Vacant (1940) 0.123** (0.058) 0.092 (0.065) -0.123 (0.076) -0.275** (0.137)
Percent Repairs
(1940) 0.060* (0.033) 0.152*** (0.037) 0.157*** (0.043) -0.079 (0.078)
Distance from City
Center 0.003*** (0.0005) 0.003*** (0.001) -0.003*** (0.001) -0.006*** (0.001)
Constant 0.166*** (0.014) 0.165*** (0.015) 0.213*** (0.018) 0.511*** (0.033)
Observations 1,577 1,577 1,577 1,577
Notes: Reference category is ungraded census tracts. Standard errors are in parentheses. *p<0.1;
**p<0.05; ***p<0.01
Robustness check: C-graded versus D-graded areas grouped by population percentage
categories
To determine whether our analysis is masking potentially important relationships
between our outcome variables of interest and the relative concentration of immigrant and/or
black residents, we create indicators based on the population percentage of D graded areas. To
do this, we divide D graded areas into tertiles based on the regional distribution of the black
population and the regional distribution of the immigrant population. This allows each increase
in quantile to be relative to a region’s existing population. Additionally, categorizing
distributions in this way as opposed to using a continuous population measure allows us to see if
a non-linear relationship exists between population concentration and our outcome variables of
interest. This method does not allow us to distinguish the overlap between black and immigrant
presence within an area and is therefore asking a fundamentally different question than our main
analysis.
We initially note that the black population outside of the South was relatively low and
not widely dispersed. In the Northeast, for example, only about 17% of all D graded

39
neighborhoods had a black population greater than 20% and only about 7% had a majority black
population. The only consistent relationship that emerges is that areas with the highest tertile of
black residents have statistically greater populations of black residents than C graded areas in
both 1960 and 2000 across all regions. This statistically greater black population, however, is not
exclusively confined to these areas. For example, in the Midwest, even areas with no black
residents at the time of grading have statistically higher percentages of black residents in 2000.
There is no other consistent relationship between either immigrant or black population
percentages and our outcome variables of interest. Additionally, we see minimal differences in
direction and/or significance of impact between these and our original results.
6. Discussion
After creating a comparable C and D graded sample using inverse propensity score weighting
and analyzing heterogeneity of D graded areas by initial population composition, we do not find
consistent differences between C and D graded areas. This inconsistency is especially prevalent
when allowing for regional differences.
Most importantly, there is little evidence that D graded areas with noted immigrant, but
no black, residents have lower homeownership rates or higher levels of black residents than C
graded areas. The most consistent relationship in the analysis is between D graded with any
black residents and the future concentration of black residents. In all regions, D graded areas
with any black residents in 1940 have higher percentages of black residents in 1960 than C
graded areas. In the Northeast, Midwest, and South, this difference remains in 2000.
This analysis provides several insights regarding how we understand and utilize the
HOLC redlining maps. The differences in regional outcomes suggest different appraisal
processes regionally which warrant greater examination, especially when attempting to use the

40
HOLC maps as a proxy for patterns of mortgage discrimination. For example, we observe little
impact on homeownership rates across the nation. These findings are consistent with work
suggesting that effects along the C-D border are more muted, and in some cases reversed, than
the effects found along other margins, such as the B-C border (Aaronson et al., 2021). We posit
that such discrepancies may be occurring because of a mismatch of overlap between the HOLC
and the FHA grades given. Geographically, C and D graded areas are more likely to be located
towards the center of the city, an area that FHA largely avoided in favor of outlying A and B
graded areas where new construction was taking place (Fishback et al., 2022). Such an
interpretation is also supported by our analysis of graded and ungraded areas, showing that both
C and D graded areas have lower homeownership rates and percentages of black residents than
comparable ungraded areas.
7. Conclusion
In this analysis, we explore the relationship between D graded areas with different
population compositions at the time of grading and two outcome variables: homeownership rates
and the percentage of black residents in both 1960 and 2000. We find an inconsistent
relationship between receiving a D versus a C grade when breaking D graded areas down by
population as well as when stratifying by region. These results imply two things. The first is that
it is possible that the FHA made lending decisions in C and D HOLC-graded areas that diverged
from what the HOLC grade would imply. In other words, in some regions, the FHA may have
restricted lending across both C and D HOLC-graded areas, so that the possibility of
homeownership attainment was relatively similar in both.
The second is that, contrary to our hypothesis, we find that D graded areas with any black
individuals at the time of grading have very few differences in homeownership rates from C

41
graded areas across regions. We do see, however, that the patterns of racial concentration over
time are highly persistent and, in many cases, D graded areas with any black residents at the time
of grading still have significantly higher percentages of black residents for decades to come.
While these results suggest less deleterious impacts of D grading than past studies, it
does not mean that mortgage discrimination and racial steering was not occurring. Rather, it is
possible that these processes may have been operating across both C and D neighborhoods and
were focused acutely on the presence of black residents Similar to studies of the spatial
mismatch hypothesis (Hellerstein et al., 2008), which find that race, not space, is the salient
predictive factor to understand future job growth, we find that persistent patterns of black
resident concentration across time are predicted by initial black household residence, not an
initial D grade. This suggests that future studies of the impact of racial segregation and racial
steering should focus on these initial conditions, rather than use HOLC maps as a proxy.

42
References
Aaronson, Hartley, D., & Mazumder, B. (2021). The Effects of the 1930s HOLC “Redlining”
Maps. American Economic Journal. Economic Policy, 13(4), 355–392.
https://doi.org/10.1257/pol.20190414
Aaronson, D., Hartley, D., Mazumder, B., & Stinson, M. (2023). The Long-Run Effects of the
1930s Redlining Maps on Children. Journal of Economic Literature, 61(3), 846–862.
https://doi.org/10.1257/jel.20221702
Akbar, P. A., Hickly, S. L., Shertzer, A., & Walsh, R. P. (2022). Racial Segregation in Housing
Markets and the Erosion of Black Wealth. The Review of Economics and Statistics, 1–45.
https://doi.org/10.1162/rest_a_01276
Appel, I., & Nickerson, J. (2016). Pockets of poverty: The long-term effects of redlining.
Available at SSRN 2852856.
Chetty, R., Friedman, J. N., Hendren, N., Jones, M. R., & Porter, S. R. (2018). The Opportunity
Atlas: Mapping the Childhood Roots of Social Mobility. IDEAS Working Paper Series from
RePEc.
Cutler, D. M., Glaeser, E. L., & Vigdor, J. L. (1999). The Rise and Decline of the American
Ghetto. The Journal of Political Economy, 107(3), 455–506. https://doi.org/10.1086/250069
Brodkin, K., (1998). How Jews became white folks and what that says about race in America.
Rutgers University Press.
Eriksson, & Ward, Z. (2019). The Residential Segregation of Immigrants in the United States
from 1850 to 1940. The Journal of Economic History, 79(4), 989–1026.
https://doi.org/10.1017/S0022050719000536
Faber, J. W. (2020). We Built This: Consequences of New Deal Era Intervention in America’s
Racial Geography. American Sociological Review, 85(5), 739–775.
https://doi.org/10.1177/0003122420948464
Fishback, P. V, LaVoice, J., Shertzer, A., & Walsh, R. (2020). Race, Risk, and the Emergence of
Federal Redlining. In National Bureau of Economic Research Working Paper Series (Vol.
28146).
Fishback, P., Rose, J., Snowden, K. A., & Storrs, T. (2022). New Evidence on Redlining by
Federal Housing Programs in the 1930s. Journal of Urban Economics, 103462.
https://doi.org/10.1016/j.jue.2022.103462

43
Gotham, K. F. (2002). Beyond Invasion and Succession: School Segregation, Real Estate
Blockbusting, and the Political Economy of Neighborhood Racial Transition. City &
Community, 1(1), 83–111. https://doi.org/10.1111/1540-6040.00009
Greer, J. (2013). The Home Owners’ Loan Corporation and the Development of the Residential
Security Maps. Journal of Urban History, 39(2), 275–296.
https://doi.org/10.1177/0096144212436724
Hershberg, Burstein, A. N., Ericksen, E. P., Greenberg, S., & Yancey, W. L. (1979). A Tale of
Three Cities: Blacks and Immigrants in Philadelphia: 1850-1880, 1930 and 1970. The
Annals of the American Academy of Political and Social Science, 441(1), 55–81.
https://doi.org/10.1177/000271627944100106
Hellerstein, J. K., Neumark, D., & McInerney, M. (2008). Spatial mismatch or racial
mismatch?. Journal of Urban Economics, 64(2), 464-479.
Hillier, A. E. (2005). Residential Security Maps and Neighborhood Appraisals: The Home
Owners’ Loan Corporation and the Case of Philadelphia. Social Science History, 29(2),
207–233. https://doi.org/10.1215/01455532-29-2-207
Krimmel, J. (2018). Persistence of prejudice: Estimating the long term effects of redlining.
10.31235/osf.io/jdmq9
Light, J. S. (2010). Nationality and neighborhood risk at the origins of FHA underwriting.
Journal of Urban History, 36(5). https://doi.org/10.1177/0096144210365677
Markley, S. (2022). Tabulating Home Owners' Loan Corporation area description sheet data.
Environment and Planning B: Urban Analytics and City Science, in press.
https://doi.org/10.1177/23998083221133112
Massey, D. S., & Denton, N. A. (1993). American apartheid : segregation and the making of the
underclass (pp. 170–181). Harvard University Press.
Michney, T. M., & Winling, L. (2020). New Perspectives on New Deal Housing Policy:
Explicating and Mapping HOLC Loans to African Americans. Journal of Urban History,
46(1), 150–180. https://doi.org/10.1177/0096144218819429
Michney, T. M. (2022). How the City Survey’s Redlining Maps Were Made: A Closer Look at
HOLC’s Mortgagee Rehabilitation Division. Journal of Planning History, 21(4), 316–
344. https://doi.org/10.1177/15385132211013361
Robert K. Nelson, LaDale Winling, Richard Marciano, Nathan Connolly, et al., “Mapping
Inequality,” American Panorama, ed. Robert K. Nelson and Edward L. Ayers,
accessed May 9, 2021,
https://dsl.richmond.edu/panorama/redlining/#loc=5/39.1/94.58&text=downloads

44
Rothstein, R. (2017). The color of law : a forgotten history of how our government segregated
America (First edition.). Liveright Publishing Corporation, a division of W.W. Norton
& Company.
Shertzer, & Walsh, R. P. (2019). Racial Sorting and the Emergence of Segregation in American
Cities. The Review of Economics and Statistics, 101(3), 415–427.
https://doi.org/10.1162/rest_a_00786
Shertzer, Twinam, T., & Walsh, R. P. (2016). Race, Ethnicity, and Discriminatory
Zoning. American Economic Journal. Applied Economics, 8(3), 217–246.
https://doi.org/10.1257/app.20140430
Steven Manson, Jonathan Schroeder, David Van Riper, Tracy Kugler, and Steven Ruggles.
IPUMS National Historical Geographic Information System: Version 15.0 [dataset].
Minneapolis, MN: IPUMS. 2020. http://doi.org/10.18128/D050.V15.0
Taylor, K. Y. (2019). Race for profit: How banks and the real estate industry undermined black
homeownership. UNC Press Books.
Xu, W. (2022). Legacies of Institutionalized Redlining: A Comparison Between Speculative and
Implemented Mortgage Risk Maps in Chicago, Illinois. Housing Policy Debate, 32(2), 249–
274. https://doi.org/10.1080/10511482.2020.1858924

45
Appendix
Table 1: Nativity and ethnic breakdown of foreign-born Chicago population
Grades: A B C D
Chicago
Asia, n (%) 0 (0%) 0 (0%) 0 (0%) 0 (0%)
Eastern Europe, n (%) 0 (0%) 14 (10%) 103 (28%) 77 (36%)
Jewish, n (%) 1 (3%) 6 (4%) 15 (4%) 7 (3%)
Mixed, n (%) 0 (0%) 3 (2%) 46 (13%) 25 (12%)
Nordic, n (%) 0 (0%) 3 (2%) 12 (3%) 3 (1%)
South America, n (%) 0 (0%) 0 (0%) 0 (0%) 5 (2%)
Southern Europe, n (%) 0 (0%) 1 (1%) 39 (11%) 42 (20%)
Western Europe, n (%) 0 (0%) 1 (1%) 33 (9%) 24 (11%)

46
Chapter 3. Historical Neighborhood Conditions and Modern-Day Disparities in Asthma
Prevalence and Air Quality: An Exploration of Redlining and Industrial Impact
1. Introduction
Asthma is a chronic condition with far reaching impacts across the population. In the United
States in 2023, approximately 27 million people, about 1 in 12, indicated that they had asthma
(National Center for Health Statistics, 2023); however, the burden of this disease is unequally
distributed demographically. Not only is asthma more prevalent among minorities, particularly
black, Puerto Rican, and indigenous persons, but asthma severity as measured by Emergency
Department visits is also higher among minority groups (Pate, 2021). Indeed, mortality among
black individuals is more severe, with this group being about three times more likely to die from
asthma than white individuals (National Center for Health Statistics, Wonder Data).
One reason for these health disparities is that minorities are disproportionately exposed to
lower-quality housing as well as neighborhoods with higher levels of air pollution that can
exacerbate asthma symptoms (Bryant-Stephens et al., 2021; Liu et al., 2021; Mohai et al., 2009;
Perez & Coutinho, 2021). The sorting of families into houses and neighborhoods is the result of
non-random processes. Minority families have historically been pushed into neighborhoods that
have higher exposure to poverty (Logan, 2011) and decreased quality of city services
(Trounstine, 2018) due to practices such as white flight (Shertzer & Walsh, 2019) and
blockbusting (Gotham, 2002; Rothstein, 2017). The result is that minority families have
historically had fewer housing options at a lower quality.
This kind of structural racism, which refers to using systems to both create and reinforce
racial discrimination as well as to manipulate the distribution of resources to favor certain groups
over others, is increasingly becoming a variable of interest in understanding modern day health

47
disparities (Bailey et al., 2017). One federally sanctioned practice of structural racism is known
as redlining, which refers to policies that denied specific neighborhoods, largely those located in
the inner city and those with minority populations, access to favorable housing finance. Scholars
have begun using maps created by the Home Owner’s Loan Corporation (HOLC) in the late
1930s/early 1940s to proxy structural racism in the form of redlining and to understand its
impacts on health disparities (Friedman et al., 2022; Kraus et al., 2023; Swope et al., 2022).
These maps were created by appraisers who assigned grades to neighborhoods across the United
States according to their perceived credit risk, leaning on the teachings of the real estate industry
that associated the presence of minority and foreign families with decreased value (Greer, 2013;
Light, 2010; Rothstein, 2017; Taylor, 2019).
However, the HOLC as an institution largely did not practice redlining. In fact, in many
cities, the HOLC itself issued loans to black homeowners in a greater than proportionate share
(Michney & Winling, 2020). Additionally, the HOLC did not start making these maps until it had
already closed about 90% of its own loans (Michney, 2022). The Federal Housing Administration
(FHA), however, also created maps grading neighborhoods. These maps, which have since been
destroyed, directly influenced FHA lending behavior, with the FHA systematically denying any
loans in D-graded neighborhoods (Xu, 2022). The extent to which the HOLC maps can be
indicative of redlining therefore relies on them sharing common borders and grades with the
maps of institutions with a history of discriminatory lending behavior such as the FHA
(Aaronson et al., 2023). Directly measuring this overlap nationwide is impossible given the
destruction of the FHA maps; however, for the FHA map of Chicago still in existence, the
overlap of grades is substantial but not exact (Xu, 2022).

48
Despite this, the HOLC maps still provide rich information about neighborhood
conditions in the mid 20th century that go beyond the credit-risk grade. Each graded
neighborhood is accompanied by an area description file that gives a breakdown of
neighborhood characteristics such as sales trends, population composition, and average sales
value. These files also include a remarks section in which the HOLC appraiser can write a
general impression of the neighborhood. The information within these paragraphs often includes
detailed descriptions of the overall character of the neighborhood and any influences deemed
favorable or detrimental to its future.
One benefit of this data, therefore, is that it provides a snapshot of historical
neighborhood conditions, which can be predictive of modern-day outcomes. While certain data
such as housing values, demographics, and income characteristics can be found in the 1940
census, understanding the more nuanced characteristics of place, such as whether the
neighborhoods suffered from smoke or noise pollution, is more difficult to find. In this way, the
HOLC area description files provide a comprehensive, if imperfect, historical description of
neighborhoods across the United States.
This paper utilizes the information provided in the area description files to target a
neighborhood characteristic linked to disparate asthma outcomes – industrial activity. In this
sense, it offers a way beyond the HOLC risk grades given to categorize neighborhoods nd
examines a potential mechanism beyond mortgage discrimination linking historic neighborhood
conditions to future health disparities. For comparability to past work linking HOLC maps to
health outcomes, this paper also analyzes neighborhood outcomes based on having received a D
versus a C grade. Inverse propensity score weighting is used in both analyses to construct a
treatment and control group that is characteristically similar.

49
The outcome variables of interest are modern-day levels of asthma prevalence as well as
modern-day measures of air quality proxied by fine particulate matter (PM2.5) and nitrogen
dioxide (NO2) concentrations. Results of these weighted regressions show that neighborhoods
that received a D-grade versus a C-grade have significantly higher levels of asthma, PM2.5
concentrations, and NO2 concentrations. When exposure is measured based on industrial impact
as identified through area description file remarks, there are also significantly higher rates of all
outcome variables of interest in areas described as being impacted by industrial activity
compared to those that are not. However, the discrepancy in asthma prevalence as well as NO2
concentrations is greater in C- versus D-graded neighborhoods than in neighborhoods
historically impacted by industry versus those that were not.
2. Literature Review
Neighborhood conditions and health disparities
Neighborhood conditions, particularly housing conditions and environmental exposures, have
been shown to impact asthma outcomes. Low-quality housing increases the risk of exposure to
indoor environmental hazards such as mouse and cockroach allergens, mold, and dust which can
both trigger and worsen asthma symptoms (Simoneau & Gaffin, 2023). The number of in-home
maintenance deficiencies is positively associated with asthma risk to the point that living in a
more deteriorated unit approximately doubles the odds of asthma (Rosenbaum, 2008). A 2021
study of the impacts of home repairs that specifically targeted dust, mold, and pests found that
within a year of completed repairs, hospitalizations and Emergency Department visits for
children with asthma living in the home decreased (Bryant-Stephens et al., 2021).
Certain neighborhood level characteristics are also associated with asthma. Living in a
neighborhood with little or no social cohesion is associated with increased odds of being

50
diagnosed with asthma (Rosenbaum, 2008). Also, moving to a lower poverty neighborhood is
associated with significant improvements in children’s asthma symptoms, improvements not
found to be mediated by exposure to pest allergen levels or smoke (Pollack et al., 2023).
Proximity to traffic-related air pollution and energy-related drilling activities are positively
associated with the count of asthma-related hospital visits (Li et al., 2023). Evidence linking air
pollution to asthma exacerbation is longstanding (Guarnieri & Balmes, 2014). Given the
increased likelihood of racial/ethnic minority communities to be exposed to higher levels of air
pollution (Liu et al., 2021) and to live within a mile of a polluting facility (Mohai et al., 2009)
than non-Hispanic white populations, environmental exposures have been cited as drivers of
health disparities (Morello-Frosch & Lopez, 2006).
Structural racism and health disparities
Research has focused on the ways that differential exposure to housing and neighborhood
conditions among racial and ethnic groups has contributed to disparities in asthma prevalence
and severity (Perez & Coutinho, 2021). Certain population groups being systematically exposed
to neighborhoods with worse health outcomes and air quality is the product of decades of public
and private decision making that has essentially created two housing markets: one for white
buyers and one for minority (and particularly black) buyers (Taylor, 2019). Black prospective
homeowners have been historically discriminated against from buying houses in white areas,
through tactics such as intimidation, violence, and racially restrictive covenants (Rothstein,
2017). Additionally, neighborhoods with black homeowners have faced economic devaluation
(Akbar et al., 2022) due to discriminatory beliefs linking race to housing quality (Light, 2010),
white flight (Boustan, 2010), and reductions in the quality and frequency of city services
(Trounstine, 2018). Redlining, or the federally backed practice of hindering housing

51
development in certain neighborhoods through systematic finance restriction, has also
particularly disadvantaged inner-city neighborhoods with minority communities.
Studies have utilized grades given by the Home Owner’s Loan Corporation as a proxy to
measure the impact of redlining specifically and structural racism more broadly, finding links
between HOLC grades and both modern-day asthma outcomes and adverse environmental
exposure. Emergency department visits due to asthma (Nardone et al., 2020) as well as
childhood rates of acute asthma (Friedman et al., 2022) are significantly higher in formerly Dgraded. Schuyler and Wenzel (2022) find a monotonic increase in severe asthma from grades A
to D as well as a decrease in ambient and filterable PM2.5 emissions as the distance from Dgraded neighborhoods increased. In fact, NO2 and PM2.5 disparities are found to be larger by
grade than by race and ethnicity (Lane et al., 2022). Additionally, median levels of air pollution
are significantly higher in redlined neighborhoods (Jung et al., 2022).
Contribution
Scholars have pushed back on the idea that HOLC maps should serve as a proxy of
mortgage discrimination (Fishback et al., 2022; Michney & Winling, 2020), citing that the
HOLC itself did not limit lending in neighborhoods that it later gave a C or D grade and
questioning the extent to which the HOLC maps influenced the choices of future lenders.
Additionally, studies exploring the relationship between HOLC grades and health disparities are
often cross-sectional in nature, directly comparing the outcomes of D-graded neighborhoods to
the highest A-graded neighborhoods. Such comparisons ignore the myriad of ways in which
these neighborhoods differed and were on fundamentally different development paths before
being graded. This, combined, with the uncertainty regarding the extent to which the HOLC

52
grades represent a lack of mortgage access, call into question the causal mechanisms behind the
observed disparities in the development trajectories of HOLC graded neighborhoods.
This work utilizes data from the HOLC maps to explore one potential mechanism of
differing health and air quality outcomes over time in graded neighborhoods by categorizing
neighborhoods according to industrial activity in addition to by grade. This work also contributes
to the literature by analyzing disparate outcomes using a more plausibly comparable treatment
and control group. This is accomplished through restricting the sample to C and D graded
neighborhoods as well as by weighting areas according to their propensity of having received
treatment.
3. Data
Independent variables of interest
Home Owner’s Loan Corporation “Redlining” Maps
This analysis uses Home Owner’s Loan Corporation data pulled from the University of
Richmond’s Mapping Inequality Project (Nelson & Winling, 2023). The Mapping Inequality
Project digitized the original HOLC maps that assigned risk grades to cities across the United
States and includes the boundaries of the original graded neighborhoods, the grades given, and
the area description files associated with each neighborhood that provide additional information
on conditions and characteristics. The current work focuses specifically on areas graded within
the Midwest and Northeast as these two regions have a particularly rich history of industrial
activity. Areas were further filtered so that only those with at least 100 graded neighborhoods
were included. The final sample includes observations from the following: Chicago, IL; Detroit,
MI; Cleveland, OH; Essex County, NJ; Union County, NJ; Queens County, NY; and Pittsburgh,

53
PA. Finally, only C and D graded areas are included in the analysis as these are most likely to be
observationally similar, leading to a sample of 517 neighborhoods.
Coding of neighborhoods by industrial activity
In order to determine which neighborhoods were subject to industrial activity, the area
description files that accompany the Home Owner’s Loan Corporation maps were used. In
addition to information such as population and building characteristics, these files contain a
section where neighborhood appraisers were able to describe the general conditions of the
neighborhood being graded. This section of text was coded using NVivo to identify areas
influenced by industrial activity.
Appraisers largely describe industrial influences in three ways. The first is proximity, i.e.
whether industry is physically located in the neighborhood being graded. An example of this type
of language from a Chicago neighborhood is: “there is some industrial concentration at the east
and south along 5th Ave”. In some cases, descriptions of industry are a bit less specific,
describing industrial activity that is broadly “to the west” for example. The second is whether the
neighborhood is impacted by the by-products of industrial activity, i.e. whether the area
experiences odors, smoke, etc. from nearby industrial activity not necessarily located within the
neighborhood. An example of this type of language, from Cleveland, is: “the area also suffers
from obnoxious odors emanating from the glue factory.” It is not certain if the industry is located
within the particular neighborhood; however, it is apparent that the neighborhood is being
impacted. Finally, an appraiser might note whether a neighborhood is specifically zoned for
industry.
In this analysis, if any of this language appears either alone or in combination within an
area description file, that neighborhood is flagged as being impacted by industry. Out of the 517

54
C- and D-graded neighborhoods in the sample, 124 are coded as industrial - of these, 57% also
have a D grade. While a majority of D-graded areas in the sample are impacted by industry, there
are still many C-graded areas also impacted.
This coding strategy has several limitations. The first is that the HOLC appraisal process
was not standardized. This is particularly relevant as data on industrial impact, if it does appear,
is found within the open remarks section of the area description file. While it is reasonable to
assume that an appraiser would mention industrial activity when reporting neighborhood
characteristics, such a report was not mandated. It could be the case, therefore, that the analysis
undercounts areas impacted by industry, which would bias results downwards. Similarly, the files
do not have data on where within a neighborhood the industrial activity is precisely located.
When measuring air quality, this is particularly relevant information as it would aid in the
estimation of a more precise radius of impact.
Outcome variables of interest
This paper uses three variables to explore variations in individual and environmental health:
asthma prevalence, fine particulate matter concentrations, and nitrogen dioxide concentrations.
Asthma
The primary outcome variable of interest is asthma prevalence among adults (age ≥ 18)
which is taken from the CDC’s 500 Cities Project data (PLACES, 2024). This project occurred
between 2016 and 2019 and was a collaborative effort to produce small area health estimates in
the 500 largest cities within the Unites States. Three data sources were used to create this
measure: the Behavioral Risk Factor Surveillance System data (2017), Census Bureau population
data (2010), and the American Community Survey (2013-2017) estimates. The measure is based
on affirmative responses to the questions: “Have you ever been told by a doctor, nurse, or other

55
health professional that you have asthma?” and “Do you still have asthma?”. This analysis uses
responses aggregated to the census tract level.
Air quality
In addition to asthma prevalence, fine particulate matter (PM2.5) and nitrogen dioxide
(NO2) outdoor concentrations are also outcomes of interest. Both measures are 2010 estimates at
the census tract level developed by the Center for Air, Climate, and Energy Solutions (CACES)
(Saha et al., 2021). Sources of fine particulate matter include traffic related factors such as onroad diesel and gasoline (Venecek et al., 2019) as well as industrial related processes (Kundu &
Stone, 2014). Similarly, NO2 come from sources such as cars and power plants (EPA, 2024).
Both are of interest due to their links to asthma (Guarnieri & Balmes, 2014). According to one
study, about 13% of new cases of pediatric asthma globally could be attributed to NO2 pollution
(Achakulwisut et al., 2019), while PM2.5 concentrations has been associated with asthma
emergency department visits (Fan et al., 2016).
These air quality measures allow for the comparison of whether predictors of areas with
the highest asthma prevalence align with predictors of areas with the worst air quality.
Additionally, they allow for the assessment of the extent to which past industrial activity is
related to future air quality.
Alignment of Boundaries
Asthma prevalence, PM2.5 concentrations, and NO2 concentrations are all aggregated to
the census tract level. HOLC maps, on the other hand, are drawn according to neighborhood
boundaries that overlap with but do not align with administratively defined census tracts. To
assess outcomes at the HOLC-neighborhood level, a weighting strategy similar to that found in

56
(Aaronson et al., 2021) is used. First, the percent area overlap of census tracts and HOLCdefined neighborhoods is calculated. Then, outcome variables within each tract are weighted
according to percent overlap. These weighted variables are aggregated to the HOLCneighborhood level. Tracts that have a less than 15% area overlap with a HOLC neighborhood
are excluded.
4. Statistical Analysis
One challenge when analyzing this data is that treatment is not randomly assigned, meaning that
treatment status is related to the characteristics of the neighborhood being studied. Nonrandomized treatment status has the potential to bias regression results as any treatment effect
may be capturing the impact of an omitted variable rather than of the treatment itself.
In the current analysis, treatment is defined in two different ways. The first is based on
whether the neighborhood was given a D grade (where the control group is C-graded areas) and
the second is based on whether the neighborhood is described as being impacted by industry
(where the control group is both C and D-graded areas that were not described as being impacted
by industry). In both cases, treatment is non-randomized as both grades and industrial siting
decisions depend on existing neighborhood characteristics.
One method of mitigating the impacts of this type of bias is by constructing a treatment
and control group that are characteristically as similar as possible. For this reason, the sample is
restricted to only C and D-graded neighborhoods, dropping any neighborhood that received an A
or a B grade. Having the lowest HOLC scores, C and D-graded areas typically are closer to the
city center, have more diverse populations with less socioeconomic power, and older, less
valuable housing than A and B-graded areas.

57
Inverse Propensity Score Weighting (IPW) is then used to make the treatment and control
groups even more plausibly comparable. This method uses observable characteristics to calculate
the likelihood that any given neighborhood was treated and then assigns weights so that
observations in the treatment group characteristically most similar to observations in the control
group receive the most weight and vis-versa. Such a strategy is suitable for this analysis for a
variety of reasons. When comparing D-graded areas to C-graded areas, propensity score
weighting relies on the fact that grades were subjectively given, so that the assignment of a C
versus a D grade may have come down to the personal preferences of the appraiser rather than
due to some objective difference. It is possible that a different appraiser, with different tastes and
preferences, might have assigned an alternate grade. Regarding the siting of industry, this
strategy relies on the fact that not all neighborhoods deemed suitable for industrial siting would
have seen such activity. Multiple neighborhoods may have been candidates for the siting of a
particular industry, but the ultimate neighborhood chosen may have been due to facts external to
existing characteristics.
Given that this analysis considers two different definitions of treatment, two different
regressions were used to create the weights for the different analyses. The first predicts the
likelihood of receiving a D-grade and the second predicts the likelihood of having been impacted
by industry. The variables used to predict these weights are median rent, percent black, percent
owner, percent vacant, percent of units needing repair, and the neighborhood’s distance from the
center of the city. All variables, except for the measure of distance from the city center, which
was calculated using data from Open Street Maps, are taken from the 1940 census.
The impacts of this weighting strategy can be seen in Tables 1 and 2. Table 1 shows a
comparison of weighted and unweighted mean neighborhood characteristics when treatment

58
status is based on having received a D-grade while Table 2 shows the same mean comparison
when treatment status is based on whether the neighborhood is impacted by industrial activity.
Table 1: Comparison of weighted and unweighted neighborhood characteristics by HOLC grade
Unweighted Weighted
C-Grade
(N = 346)
D-Grade
(N = 171) p-value
C-Grade
(N =
498)
D-Grade
(N = 452) p-value
Median Value
1940
4,007
(1,608)
2,419
(1,590) <0.001 3,643
(1,720)
3,031
(1,617) 0.002
Median Rent
1940
32
(11)
20
(11) <0.001 29
(12)
26
(11) 0.008
Percent Black
1940
1.1%
(3.6%)
7.5%
(16.9%) <0.001 1.9%
(5.9%)
4.4%
(11.4%) 0.005
Percent
Immigrant
1940
19.6%
(5.2%)
21.3%
(7.3%) 0.001 19.7%
(5.6%)
21.7%
(6.9%) 0.003
Percent
Owner 1940
40%
(19%)
34%
(19%) 0.002 39%
(19%)
38%
(20%) 0.6
Percent
Vacant 1940
4.7%
(8.2%)
4.8%
(8.2%) 0.4 4.53%
(8.23%)
4.67%
(8.09%) 0.9
Percent
Repairs 1940
4.7%
(4.0%)
9.8% (
8.4%) <0.001 5.9%
(5.5%)
7.1%
(6.5%) 0.028
Distance from
City Center
12.1
(5.4)
10.0
(6.2) <0.001 11.5
(5.5)
11.0
(5.7) 0.4
Industry 15% 42% <0.001 32% 68% 0.002
Table 2: Comparison of weighted and unweighted neighborhood characteristics by industry
status
Unweighted Weighted
NonIndustrial
(N =
393)
Industrial
(N = 124) p-value
NonIndustrial
(N =
522)
Industrial
(N = 443) p-value
Median Value
1940
3,782
(1,737) 2,529 (1,508) <0.001 3,520
(1,796)
3,052
(1,487) 0.014
Median Rent
1940 30 (12) 21 (10) <0.001 28 (13) 25 (10) 0.001
Percent Black
1940
3.5%
(11.1%) 2.4% (8.7%) 0.012 3.5%
(11.0)
2.8%
(10.3) 0.002

59
Percent
Immigrant
1940
19.8%
(6.1) 21.3% (5.7) 0.004 20.0%
(6.2)
20.8%
(5.6) 0.13
Percent Owner
1940 39% (20) 36% (16) 0.2 38% (20) 39% (16) 0.6
Percent Vacant
1940
5.19%
(9.09) 3.32% (3.91) <0.001 4.92%
(8.51)
3.01%
(3.18) <0.001
Percent Repairs
1940
5.8%
(5.6) 8.4% (7.9) <0.001 6.2% (6.2) 6.8% (6.2) 0.2
Distance from
City Center
12.2
(5.7) 8.9 (5.2) <0.001 11.4 (5.7) 10.4 (5.0) 0.2
D-Grade 25% 57% <0.001 41% 59% <0.001
As a result of the weighting strategy, treated and untreated neighborhoods are
observationally more similar across both definitions of treatment. In comparison to C-graded
neighborhoods, D-graded neighborhoods are closer in median value, median rent, percent
ownership, and distance from the city center (among others) after weighting. The same is true for
industrial versus non-industrial neighborhoods after weighting.
This study conducts two different sets of analyses. To replicate the strategy of other
works looking at the relationship between redlining and health, this paper first defines treatment
as having received a D-grade. This analysis then diverges from past work by using the area
description files to more precisely categorize neighborhoods according to whether appraisers
described the impact of industrial activity when grading. While inverse propensity score
weighted regressions are the preferred method of analysis, ordinary least squares regressions are
also included for comparison.
5. Results
Exploration of neighborhood characteristics by treatment status
Before examining the weighted regression analyses, I first compare the characteristics of
neighborhoods by treatment status to understand how categorizing neighborhoods by industrial

60
impact compares to categorizing them by the more commonly used HOLC grade. All variables
used besides distance from the center of the city come from 1940 census data. Table 3 shows the
result of this comparison. Within this sample, D-graded neighborhoods and neighborhoods that
are described as having any industrial activity look fairly similar. There is no statistically
significant difference between the two in terms of mean house value, mean rent, percent
ownership, or percent repairs.
Industrial neighborhoods are slightly closer to the city center than D-graded
neighborhoods on average, while D-graded neighborhoods have a slightly higher percentage of
vacant units. Demographically, D-graded neighborhoods and industrial neighborhoods have
roughly the same percentage of immigrant residents. The largest difference between the two
appears in the overall percentage of black residents, with D-graded neighborhoods having a
significantly higher percentage of black residents.
It is also important to note the overlap between the two neighborhood types. Roughly
42.5% of D-graded neighborhoods are also coded as having industry while a majority of
industrially coded areas (57.2%) also have a D grade.

61
Table 3: Characteristics by treatment status
Weighted Regression Analysis
In the following regression analyses, all variables are standardized so that the magnitude of
impact of each estimated coefficient is directly comparable within and across each regression.
All coefficients should be interpreted in terms of standard deviation increases.
D-graded vs C-graded areas
Table 4 shows the results of the analysis when treatment is defined by having received a D grade.
Here, the reference group is all C-graded neighborhoods within the sample. Models 1-3 show the
results of unweighted OLS while models 4-6 show the results of inverse propensity score
weighted linear regression. Across all models, receiving a D-grade is significantly associated
with all outcome variables of interest. In comparison to C-graded areas, D-graded areas have
higher levels of asthma as well as higher levels of PM2.5 and levels of NO2. From the OLS to
IPW models, the magnitude of the D-grade coefficient for the three outcome variables of interest
all increase.
D-Graded Industrial P-Value
Mean House Value 1940 2,458.51 (1,607.0) 2,529.14 (1,508.0) 0.699
Mean Rent 1940 20.4 (11.1) 20.64 (10.0) 0.876
Percent black 1940 7.4% (16.8) 2.4% (8.6) <0.05
Percent immigrant 1940 21.2% (7.3) 21.3% (5.7) 0.856
Percent owner 1940 34.9% (19.6) 36.1% (16.5) 0.584
Percent vacant 1940 4.79 (8.18) 3.32 (3.91) <0.05
Percent repairs 1940 9.6 (8.4) 8.4 (7.9) 0.185
Distance from city center 10.1 (6.2) 8.9 (5.2) 0.079
Percent Industrial 42.5% - -
Percent D graded - 57.2% -
N 174 124 -

62
In the preferred IPW models, the association between having received a D grade and
asthma prevalence is highest, with D-graded areas having a 0.47 standard deviation greater
prevalence of asthma than comparable C-graded areas. D-graded vs C-graded areas also have a
higher concentration of PM2.5 particles (0.296 standard deviations higher) and NO2 particles
(0.367 standard deviations higher).
Table 4 reveals other interesting relationships between neighborhood conditions and
asthma outcomes. Regarding 1940 characteristics, there is a positive relationship between the
percentage of black residents living in the neighborhood and asthma (0.159 sd) but no
relationship with air quality. Conversely, there is a positive relationship between percent
immigrant and NO2 concentrations (0.126 sd) but not with asthma. There are additionally
potentially counterintuitive associations between median rent, percent owner, and percent vacant.
There is a positive relationship between median rent and asthma, PM2.5, and NO2; a positive
relationship between percent ownership and asthma and PM2.5; and a positive relationship
between percent vacancy and asthma - albeit a negative relationship between this variable and
NO2.
Finally, there is a statistically significant relationship between a neighborhood’s distance
from the center of the city and asthma prevalence, with a greater distance being associated with
lower asthma prevalence.
Table 4: OLS and IPW Regression results for D-grade treatment status
OLS IPW OLS
Asthma PM2.5 NO2 Asthma PM2.5 NO2
(1) (2) (3) (4) (5) (6)
D-grade 0.369***
(0.111)
0.191*
(0.108)
0.288***
(0.103)
0.470***
(0.086)
0.296***
(0.086)
0.367***
(0.083)
Median Rent
1940
0.272***
(0.058)
0.151***
(0.056)
0.443***
(0.054)
0.268***
(0.054)
0.176***
(0.054)
0.455***
(0.052)

63
Percent Black
1940
0.190***
(0.053)
0.020
(0.052)
0.056
(0.049)
0.159***
(0.057)
-0.037
(0.057)
-0.021
(0.055)
Percent
Immigrant 1940
0.044
(0.049)
-0.004
(0.048)
0.169***
(0.046)
0.063
(0.045)
-0.030
(0.045)
0.126***
(0.043)
Percent Owner
1940
0.482***
(0.060)
0.097*
(0.059)
-0.079
(0.056)
0.547***
(0.059)
0.160***
(0.059)
-0.085
(0.057)
Percent Vacant
1940
0.130**
(0.054)
-0.108**
(0.053)
-0.179***
(0.051)
0.155***
(0.053)
-0.064
(0.053)
-0.154***
(0.051)
Percent Repairs
1940
-0.083*
(0.049)
0.023
(0.048)
0.131***
(0.046)
-0.128***
(0.048)
0.021
(0.048)
0.137***
(0.046)
Distance -0.311***
(0.061)
-0.072
(0.060)
0.059
(0.057)
-0.323***
(0.061)
-0.109*
(0.061)
0.076
(0.059)
Constant -0.096*
(0.056)
-0.181***
(0.055)
-0.081
(0.053)
0.470***
(0.086)
0.296***
(0.086)
0.367***
(0.083)
Note: *p<0.1; **p<0.05; ***p<0.01
Industrial vs non-industrial areas
In Table 5, a neighborhood is considered treated if there are any mentions of industrial activity or
impacts of industry within the HOLC area description files. Once again, we see a significant
positive relationship between treatment status and asthma, PM2.5, and NO2 that largely
increases in magnitude between the OLS and IPW models. The relationship between industrial
activity in the 1940s and modern-day asthma prevalence is more tampered here than the
relationship between having received a D grade and asthma.
In the IPW model, neighborhoods that were described as having industrial activity have a
prevalence of asthma that is .283 standard deviations higher than neighborhoods where no
industrial activity was noted. While still a positive and statistically significant association, this
estimation is about 60% of the estimate for D-graded neighborhoods seen in Table 4. While the
relationship between industry and concentration of NO2 is also a bit lower in magnitude here,
neighborhoods described as having industrial activity are significantly associated with PM2.5 at
a magnitude that is slightly higher than the association seen when the treatment status is defined
by receiving a D grade (0.323 sd compared to 0.296 sd).

64
There is also a similar relationship between neighborhood characteristics and the
outcomes variables here as was seen when treatment was defined by grade. A positive
relationship exists between the percentage of black households and asthma but once again, there
is no relationship between this metric and air quality. Neighborhoods with higher percentages of
immigrant residents, however, have higher levels of NO2. Median rent and percent owner are
positively related to asthma prevalence (1940 median rent is also positively related to both
PM2.5 and NO2). Additionally, areas with greater vacancies are associated with higher asthma
prevalence but lower concentrations of both PM2.5 and NO2. Finally, distance from city center
is negatively associated with asthma prevalence but not air quality.
Table 5: OLS and IPW Regression results for industrial impact treatment status
OLS IPW OLS
Asthma PM2.5 NO2 Asthma PM2.5 NO2
(1) (2) (3) (1) (2) (3)
Industry 0.194*
(0.111)
0.226**
(0.107)
0.359***
(0.101)
0.283***
(0.088)
0.323***
(0.086)
0.347***
(0.082)
Median Rent
1940
0.235***
(0.057)
0.146***
(0.055)
0.438***
(0.052)
0.256***
(0.055)
0.154***
(0.054)
0.437***
(0.051)
Percent Black
1940
0.242***
(0.052)
0.055
(0.051)
0.108**
(0.048)
0.277***
(0.052)
0.040
(0.051)
0.077
(0.048)
Percent
Immigrant 1940
0.071
(0.049)
0.008
(0.047)
0.188***
(0.045)
0.026
(0.049)
-0.034
(0.048)
0.151***
(0.045)
Percent Owner
1940
0.461***
(0.060)
0.086
(0.058)
-0.095*
(0.055)
0.473***
(0.062)
0.100
(0.061)
-0.087
(0.058)
Percent Vacant
1940
0.134**
(0.055)
-0.105**
(0.053)
-0.174***
(0.050)
0.118*
(0.063)
-0.138**
(0.062)
-0.149**
(0.059)
Percent Repairs
1940
-0.055
(0.049)
0.032
(0.047)
0.144***
(0.045)
-0.045
(0.048)
0.062
(0.047)
0.154***
(0.045)
Distance -0.286***
(0.062)
-0.051
(0.060)
0.091
(0.057)
-0.240***
(0.061)
0.045
(0.060)
0.086
(0.057)
Constant -0.020
(0.051)
-0.171***
(0.049)
-0.070
(0.046)
-0.033
(0.058)
-0.179***
(0.057)
-0.075
(0.054)
Note: *p<0.1; **p<0.05; ***p<0.01

65
6. Discussion
This analysis reveals several key insights between historic neighborhood conditions and modernday disparities in asthma prevalence and air quality exposure. Using inverse propensity score
weighting, receiving a D-grade versus a C-grade is associated with a 0.470 sd increase in asthma
prevalence, a 0.296 sd increase in PM2.5 concentrations, and a 0.367 sd increase in NO2
concentrations. When treatment status is assigned based on appraiser descriptions of industrial
impact within HOLC area description files, there remains a statistically significant relationship
between treatment and asthma (0.283 sd), PM2.5 concentrations (0.323 sd), and NO2
concentrations (0.347 sd), albeit with decreases in magnitude for asthma prevalence and NO2
concentrations in comparison to when treatment is defined by receiving a D grade.
Interestingly, despite the connection between asthma, air quality, and industrial activity
being well documented (Guarnieri & Balmes, 2014), this study shows that having received a D
versus a C grade is a stronger predictor of future asthma prevalence than an area that was
described as having industrial activity versus one that was not. This is also true for NO2
concentrations. This implies that observed health and air quality disparities in C and D-graded
areas are not solely due to historical industrial siting decisions.
This work therefore confirms what past research has found - that having received a D
grade is positively related to an area’s future rate of asthma prevalence and air quality (Nardone
et al., 2020; Schuyler & Wenzel, 2022; Lane et al., 2022). In directly assessing neighborhoods
described as being impacted by industry, this paper additionally provides insight into the
mechanisms that may be driving such findings. This analysis is particularly useful in light of
work questioning the extent to which HOLC grades are indicative of systematic neighborhoodlevel mortgage denial, calling into question which mechanisms that HOLC grades are proxying

66
(Fischback et al., 2022). In other words, if observed differences in outcomes across grades are
not due to lack of credit access, then what is causing the disparate development pathways?
In regards to asthma, there could be several drivers of the finding that asthma prevalence
is higher in HOLC D-graded areas than in areas historically described as being impacted by
industry regardless of grade. Industry encompasses only one potential asthma trigger. While
asthma outcomes are related to air quality, they are also related to a myriad of other
environmental triggers and socioeconomic factors associated with asthma such as poor housing
quality (Simoneau et al., 2023, Bryant-Stephens et al., 2021), high crime rates (Beck et al.,
2016), and high neighborhood poverty rates (Pollack et al., 2023). It is also possible that Dgraded areas, which were some of the most neglected and under-developed neighborhoods, are
highly correlated with planning decisions detrimental to individual health such as the building of
highways and lack of greenspace. Industrial areas, on the other hand, are often described in the
HOLC area description files in more favorable terms since workers found convenience to
industry, and therefore to a place of work, to be a positive. One C-graded industrial area is
described as “a heterogenous development…in constant demand because of its proximity to local
industrial employment.” Some industrial areas are even described as having deed restrictions.
This difference in relative opinion of D-graded versus industrial areas might have impacted
future development.
Structural racism
D-graded areas may have been more likely to have developed such characteristics in part
due to the consequences of structural racism. D-graded neighborhoods in this analysis show
statistically higher concentrations of black households than industrial neighborhoods. Indeed,
black households were almost exclusively located in D-graded neighborhoods initially (Greer,

67
2013) and their relative concentration in these areas grew over time as black households had
limited housing options outside of the most disadvantaged areas (Taylor, 2019; Akbar et al.,
2022). The elevated concentration of black households within HOLC D-graded neighborhoods is
quite persistent. D-graded neighborhoods with any black residents at the time of grading had
significantly higher percentages of black residents than C-graded areas even decades after the
HOLC grading (Smith & Painter, 2023).
Black residents looking for housing faced particular challenges as they not only
experienced targeted violence when purchasing housing (Massey & Denton, 1993) but were
exploited by the real estate industry through practices such as blockbusting, which
simultaneously caused black households to pay relatively more for housing while eroding black
wealth (Akbar et al., 2022), and being systematically sold dilapidated housing (Taylor, 2019).
Additionally, neighborhoods with minority households not only saw a decline in the quality of
public services and amenities over time (Trounstine, 2018), but at the time of the HOLC grading,
minority households were already being filtered into neighborhoods with hazardous conditions.
Indeed, one D-graded neighborhood in Detroit, MI is described as “a sparsely built area
developing as a negro colony. Section is inconvenient to shopping centers – sewers are lacking.”
The consequence of such policies and actions was the decline in the economic value and
structural integrity of urban, predominately black neighborhoods, declines which could likely
contribute to higher levels of asthma-exacerbating characteristics in these areas. Today, there is a
clear relationship between a neighborhood’s demographic composition and the prevalence of inhome asthma triggers, with a negative correlation existing between the percentage of nonHispanic white households and the number of reported in-home asthma triggers (Lemire et al.,

68
2022). The statistically significant relationship seen between the percentage of black residents in
the 1940s and modern-day asthma prevalence highlights this point.
Limitations
When interpreting results, it is important to recognize the limitations of the measure of
historical industrial impact. The measure encompasses any neighborhood mentioned as having
industry within it, any neighborhood mentioned as having some product of industrial activity
such as smoke or odors, and/or any neighborhood specifically zoned for industry at the time of
the grading of the maps. The measure is therefore broad and does not pinpoint the exact location
of industrial facilities. Also, because the measure is based on a description, it relies on the
accuracy of the appraisers, which is not a guarantee as the HOLC appraisal process was not
highly standardized with reporting norms varying by city. It could be the case that appraisers did
not describe industry or industrial zoning even when it was present or conversely, that they
described industrial activity in an area even if the neighborhood was actually out of the range of
any harmful by-products.
Finally, the measure of industrial impact is necessarily a historical one and says nothing
about the relationship between current industrial activity and asthma outcomes. While it is
reasonable to assume that areas with industrial activity in the 1940s may remain industrial sites
for decades to come, this is not a guarantee. Furthermore, areas without industrial activity at the
time of the HOLC grades could have become industrial sites in subsequent decades. This is
especially likely for neighborhoods with shifts to higher concentrations of minority populations
over time, as research shows that toxic facilities are disproportionately sited in areas where
higher concentrations of minority populations live (Pastor et al., 2001).

69
7. Conclusion
The current work contributes to the literature by restricting analysis from the entire set of graded
neighborhoods to the more plausibly comparable C and D-graded neighborhoods, which are
further weighted to increase comparability using inverse propensity scores. In doing so, this
paper proposes a new way of utilizing the data associated with the HOLC maps to understand the
impacts of historical neighborhood characteristics on modern day outcomes. This paper therefore
explores a mechanism besides mortgage discrimination that might be contributing to the
disparate health outcomes observed in D-graded neighborhoods that have been found in past
work. By doing so, this paper contributes to the understanding of how historic neighborhood
characteristics, and in this case industrial siting decisions, impact future conditions.
Perhaps contrary to expectation, this paper finds that differences in asthma prevalence are
more pronounced in D-graded areas, when the comparison is C-graded areas, than they are in
areas described as being impacted by industry per the HOLC area description files, when the
comparison is those that were not. This additionally holds for concentrations of NO2.
Concentrations of PM2.5 is the only measure for which areas impacted by industry are a stronger
predictor.
These results imply that differences in industrial activity are not driving the disparate
outcomes observed in C versus D-graded areas, but rather, that they are one factor among others
associated with D-graded areas leading to divergent outcomes in asthma and air quality over
time. Given these findings, it is important to take into consideration not only the entirety of
characteristics distinguishing D-graded areas at the time of grading but also the forces driving
certain populations into D-graded neighborhoods over time to better understand their
development trajectories and relationship with modern-day health disparities.

70
References
Aaronson, D., Hartley, D., & Mazumder, B. (2021). The effects of the 1930s HOLC “redlining”
maps. American Economic Journal: Economic Policy, 13(4), 355-392.
Aaronson, D., Hartley, D., Mazumder, B., & Stinson, M. (2023). The long-run effects of the
1930s redlining maps on children. Journal of Economic Literature, 61(3), 846-862.
Achakulwisut, P., Brauer, M., Hystad, P., & Anenberg, S. C. (2019). Global, national, and urban
burdens of paediatric asthma incidence attributable to ambient NO2 pollution: estimates
from global datasets. The Lancet Planetary Health, 3(4), e166-e178.
Akbar, P. A., Hickly, S. L., Shertzer, A., & Walsh, R. P. (2022). Racial segregation in housing
markets and the erosion of black wealth. Review of Economics and Statistics, 1-45.
Bailey, Z. D., Krieger, N., Agénor, M., Graves, J., Linos, N., & Bassett, M. T. (2017). Structural
racism and health inequities in the USA: evidence and interventions. The lancet,
389(10077), 1453-1463.
Beck, A. F., Huang, B., Ryan, P. H., Sandel, M. T., Chen, C., & Kahn, R. S. (2016). Areas with
high rates of police-reported violent crime have higher rates of childhood asthma
morbidity. The Journal of pediatrics, 173, 175-182. e171.
Boustan, L. P. (2010). Was postwar suburbanization “white flight”? Evidence from the black
migration. The Quarterly Journal of Economics, 125(1), 417-443.
Bryant-Stephens, T. C., Strane, D., Robinson, E. K., Bhambhani, S., & Kenyon, C. C. (2021).
Housing and asthma disparities. Journal of Allergy and Clinical Immunology, 148(5),
1121-1129.
Fan, J., Li, S., Fan, C., Bai, Z., & Yang, K. (2016). The impact of PM2. 5 on asthma emergency
department visits: a systematic review and meta-analysis. Environmental Science and
Pollution Research, 23, 843-850.
Fishback, P., Rose, J., Snowden, K. A., & Storrs, T. (2022). New evidence on redlining by federal
housing programs in the 1930s. Journal of Urban Economics, 103462.
Friedman, E., Lee, B., Kalman, C., & Wilson, N. (2022). Historic racism in Kansas City affects
Today's pediatric asthma burden. Health & place, 78, 102927.
Gotham, K. F. (2002). Beyond invasion and succession: School segregation, real estate
blockbusting, and the political economy of neighborhood racial transition. City &
Community, 1(1), 83-111.
Greer, J. (2013). The home owners’ loan corporation and the development of the residential
security maps. Journal of Urban History, 39(2), 275-296.

71
Guarnieri, M., & Balmes, J. R. (2014). Outdoor air pollution and asthma. The Lancet, 383(9928),
1581-1592.
Jung, K. H., Pitkowsky, Z., Argenio, K., Quinn, J. W., Bruzzese, J.-M., Miller, R. L., . . .
Lovinsky-Desir, S. (2022). The effects of the historical practice of residential redlining in
the United States on recent temporal trends of air pollution near New York City schools.
Environment international, 169, 107551.
Kraus, N. T., Connor, S., Shoda, K., Moore, S. E., & Irani, E. (2023). Historic redlining and
health outcomes: A systematic review. Public Health Nursing.
Kundu, S., & Stone, E. A. (2014). Composition and sources of fine particulate matter across
urban and rural sites in the Midwestern United States. Environmental science: Processes
& impacts, 16(6), 1360-1370.
Lane, H. M., Morello-Frosch, R., Marshall, J. D., & Apte, J. S. (2022). Historical redlining is
associated with present-day air pollution disparities in US cities. Environmental science
& technology letters, 9(4), 345-350.
Lemire, E., Samuels, E. A., Wang, W., & Haber, A. (2022). Unequal Housing Conditions And
Code Enforcement Contribute To Asthma Disparities In Boston, Massachusetts: Study
examines housing conditions, code enforcement, and asthma disparities in Boston,
Massachusetts. Health Affairs, 41(4), 563-572.
Li, J., Jahan, J., & Newcomb, P. (2023). Environmental characteristics and disparities in adult
asthma in north central Texas urban counties. Public Health, 217, 164-172.
Light, J. S. (2010). Nationality and neighborhood risk at the origins of FHA underwriting.
Journal of Urban History, 36(5), 634-671.
Liu, J., Clark, L. P., Bechle, M. J., Hajat, A., Kim, S.-Y., Robinson, A. L., . . . Marshall, J. D.
(2021). Disparities in air pollution exposure in the United States by race/ethnicity and
income, 1990–2010. Environmental health perspectives, 129(12), 127005.
Logan, J. R. (2011). Separate and unequal: The neighborhood gap for blacks, Hispanics and
Asians in Metropolitan America. Project US2010 report, 1-22.
Michney, T. M. (2022). How the city survey’s redlining maps were made: a closer look at
HOLC’s mortgagee rehabilitation division. Journal of Planning History, 21(4), 316-344.
Michney, T. M., & Winling, L. (2020). New perspectives on new deal housing policy: explicating
and mapping HOLC loans to African Americans. Journal of Urban History, 46(1), 150-
180.

72
Mohai, P., Lantz, P. M., Morenoff, J., House, J. S., & Mero, R. P. (2009). Racial and
socioeconomic disparities in residential proximity to polluting industrial facilities:
evidence from the Americans' Changing Lives Study. American journal of public health,
99(S3), S649-S656.
Morello-Frosch, R., & Lopez, R. (2006). The riskscape and the color line: examining the role of
segregation in environmental health disparities. Environmental research, 102(2), 181-
196.
Nardone, A., Casey, J. A., Morello-Frosch, R., Mujahid, M., Balmes, J. R., & Thakur, N. (2020).
Associations between historical residential redlining and current age-adjusted rates of
emergency department visits due to asthma across eight cities in California: an ecological
study. The Lancet Planetary Health, 4(1), e24-e31.
Pastor, M., Sadd, J., & Hipp, J. (2001). Which came first? Toxic facilities, minority move‐in, and
environmental justice. Journal of urban affairs, 23(1), 1-21.
Pate, C. A. (2021). Asthma surveillance—United States, 2006–2018. MMWR. Surveillance
Summaries, 70.
Perez, M. F., & Coutinho, M. T. (2021). Focus: Health Equity: An Overview of Health
Disparities in Asthma. The Yale journal of biology and medicine, 94(3), 497.
Pollack, C. E., Roberts, L. C., Peng, R. D., Cimbolic, P., Judy, D., Balcer-Whaley, S., . . . Davis,
M. F. (2023). Association of a housing mobility program with childhood asthma
symptoms and exacerbations. Jama, 329(19), 1671-1681.
Rosenbaum, E. (2008). Racial/ethnic differences in asthma prevalence: the role of housing and
neighborhood environments. Journal of health and social behavior, 49(2), 131-145.
Rothstein, R. (2017). The color of law: A forgotten history of how our government segregated
America. Liveright Publishing.
Saha, P. K., Hankey, S., Marshall, J. D., Robinson, A. L., & Presto, A. A. (2021). High-spatialresolution estimates of ultrafine particle concentrations across the continental United
States. Environmental Science & Technology, 55(15), 10320-10331.
Schuyler, A. J., & Wenzel, S. E. (2022). Historical redlining impacts contemporary
environmental and asthma-related outcomes in Black adults. American Journal of
Respiratory and Critical Care Medicine, 206(7), 824-837.
Shertzer, A., & Walsh, R. P. (2019). Racial sorting and the emergence of segregation in American
cities. Review of Economics and Statistics, 101(3), 415-427.
Simoneau, T., & Gaffin, J. M. (2023). Socioeconomic determinants of asthma health. Current
Opinion in Pediatrics, 35(3), 337-343.

73
Smith, R. B., & Painter, G. (2023). People and place: Does the reason for redlining impact the
long-term trajectory of neighborhoods. Journal of Urban Affairs, 1-19.
Swope, C. B., Hernández, D., & Cushing, L. J. (2022). The relationship of historical redlining
with present-day neighborhood environmental and health outcomes: a scoping review and
conceptual model. Journal of Urban Health, 99(6), 959-983.
Taylor, K.-Y. (2019). Race for profit: How banks and the real estate industry undermined black
homeownership. UNC Press Books.
Tessum, C. W., Paolella, D. A., Chambliss, S. E., Apte, J. S., Hill, J. D., & Marshall, J. D. (2021).
PM2. 5 polluters disproportionately and systemically affect people of color in the United
States. Science advances, 7(18), eabf4491.
Trounstine, J. (2018). Segregation by design: Local politics and inequality in American cities.
Cambridge University Press.
Venecek, M. A., Yu, X., & Kleeman, M. J. (2019). Predicted ultrafine particulate matter source
contribution across the continental United States during summertime air pollution events.
Atmospheric Chemistry and Physics, 19(14), 9399-9412.
Xu, W. (2022). Legacies of institutionalized redlining: A comparison between speculative and
implemented mortgage risk maps in Chicago, Illinois. Housing Policy Debate, 32(2),
249-274.

74
Chapter 4. Do publicly funded neighborhood investments impact individual level healthrelated outcomes? A longitudinal study of two neighborhoods in Pittsburgh, PA from 2011
to 20189
1. Introduction
Scholars have long been interested in the role that the neighborhood has on health-related
outcomes (Neumark & Simpson, 2015). Such interest stems from the recognition that the impact
of where one lives goes beyond the physical housing unit to include a bundle of neighborhoodlevel characteristics (Arcaya et al., 2016). These characteristics may be significantly associated
with a variety of health and health-related outcomes, even when controlling for socioeconomic
status and individual-level factors (Schüle & Bolte, 2015). And yet, disentangling the bundle of
characteristics that make up the neighborhood to understand which specifically impact individual
well-being is difficult. This lack of understanding is particularly relevant to research on the built
environment, as metrics are inconsistent and largely rely on cross-sectional studies of static
neighborhood features (Schüle & Bolte, 2015).
This study contributes to the literature by utilizing a longitudinal research design that
focuses on a feature of neighborhood development that is malleable at the policy level: fully or
partially funded public investments in the form of residential, commercial, business, or
recreational construction or renovation. We assess the effect of these neighborhood-level
investments on five health-related outcomes: food insecurity, stress, perceived neighborhood
safety, neighborhood satisfaction, and dietary quality. We choose these outcome measures as they
touch on several aspects of individual health that may be related to neighborhood investments.
9 Co-authored with Mathew D Baird, Gerald P. Hunter, Bonnie Ghosh-Dastidar, Andrea S. Richardson, Jonathan H.
Cantor, Tamara Dubowitz

75
For example, safety has theoretically been linked to environmental improvements such as better
street lighting and aesthetic enhancements while decreased stress has been connected to aesthetic
improvements, particularly greening efforts. The study utilizes five separate waves of data
collected between 2011 and 2018 from a cohort of individuals residing in one of two historically
disadvantaged neighborhoods in the City of Pittsburgh, the Hill District and Homewood.
Throughout the waves of data collection, these two neighborhoods were experiencing a variety
of structural changes such as the addition of a full-service supermarket, renovation of green
spaces, and rebuilding and renovation of housing, among others (Holliday et al., 2020).
This study design has several benefits. The first is that our survey participants are longterm residents of the study neighborhoods. This means they have not selected into the
neighborhood in anticipation of changes to the built environment, an endogeneity concern that
might bias results. Additionally, since participants do not enter the study over time, our results
are not influenced by new residents moving in with a potentially different set of characteristics.
Using a longitudinal study design also allows us to control for unobservable and unchanging
individual-level characteristics, such as an individual propensity towards healthy behaviors.
Finally, this longitudinal design allows us to pick up on associations that may only become
apparent over a longer exposure timeline.
In this study, we respond to two research questions. The first is whether changes to the
built environment in the form of new or renovated residential, commercial, business, and
recreational spaces impacts individual-level health-related outcomes. The second is to understand
how sensitive the results are to changing geographic levels of exposure. In other words, we are
investigating whether living closer to such investments magnifies, diminishes, or has no impact
on the association between investment and our health-related outcomes of interest. To respond to

76
this second question, we define an individual’s exposure to investments at three different
geographic levels. We first look at cumulative investment costs that an individual is exposed to
within their neighborhood across each survey year. We then restrict this sum to include only
those investments occurring within ½ mile of an individual’s place of residence and those
occurring between a ½ mile and 1-mile of an individual’s place of residence within each survey
year. For all measures, we only include investments occurring in either the Hill District or
Homewood. We additionally stratify investments by type.
2. Literature Review
The term neighborhood encompasses a wide variety of both social and physical characteristics.
As a result, there is no consensus on which neighborhood characteristics should be of focus when
analyzing their impact on individual-level outcomes (Ndjila et al., 1998). The majority of studies
estimating the relationship between neighborhood effects and health have utilized census-based,
aspatial variables to describe neighborhood conditions, with a minority taking an approach
similar to ours and defining neighborhood conditions based on non-aggregated contextual
variables (i.e., developing a proximity variable) (Arcaya et al., 2016). Here, we summarize some
of the relevant literature linking neighborhoods to health and health-related outcomes. Given the
amount of work on this subject, this review is by no means exhaustive. While we review work
that conceptualizes neighborhoods as a whole, we also focus more specifically on papers that
study neighborhood conditions through the lens of physical characteristics as these types of
changes are the focus of our study.
Neighborhood Context and Health
When analyzing the relationship between neighborhoods and individual level outcomes,
scholars have chosen to exploit individual movement in order to study neighborhood effects. One

77
such study is the Moving to Opportunities (MTO) project (Katz et al., 2001). The identification
strategy relies on people being randomized to receive relocation support and vouchers to more to
higher income areas versus a control group. Initial studies showed that the strongest and most
consistent associations were found between moving and improvements in adult mental health,
with no relationship being found between moving and adult socio-economic outcomes (Kling et
al., 2007). Subsequent analysis of this cohort has shown that the length of exposure to lowerpoverty areas mattered, with children who moved at a younger age being more likely to attend
college and having substantially higher incomes as adults (Chetty et al., 2016). Additionally,
(Pollack et al., 2019) find that for participants of this study, having received a housing voucher as
a child is associated with lower hospitalization rates and yearly inpatient spending. These studies,
however, ultimately operate at the individual-level, meaning that individuals who moved were
randomized but the neighborhoods that they moved to were not (Clampet-Lundquist & Massey,
2008), calling into question the extent to which these results can speak to the evidence of
potential neighborhood effects. Additionally, because this study bundles across neighborhoods, it
can not assess the causal mechanisms connecting neighborhood characteristics and outcomes
(Sampson, 2008).
(Santiago et al., 2014) exploits the randomized assignment of households to public
housing units located in a variety of neighborhood types to study the relationship between
neighborhoods and a variety of individual-level outcomes. In regard to physical and behavioral
health, this work concludes that Latino and African-American children living in neighborhoods
with a lower social problems index and crime rates along with high walkability, land use mix,
and occupational prestige exhibit comparatively better health outcomes.

78
One common way to conceptualize neighborhood conditions is by assessing the degree to
which physical disorder is present. Physical disorder is often measured using characteristics such
as the prevalence of graffiti, vacant lots, vandalism, cleanliness, etc. When estimating the
relationship between neighborhood disorder and self-reported health, variables related to the
social characteristics of a neighborhood (i.e., social cohesion and feelings of safety) are often
found to be greater predictors of self-reported health than are variables related to physical
disorder (Kim, 2010; Ross & Mirowsky, 2001; Ruijsbroek et al., 2016). However, when
physiological responses to disorder are measured directly, pedestrians walking past greened
vacant lots have lower heart rates than when walking by non-greened vacant lots, suggesting that
one’s physical surroundings directly impact levels of acute stress (South et al., 2015). This
positive relationship between physical disorder and stress levels has been found by others
(Henderson et al., 2016; Martz et al., 2021; Matthews & Yang, 2010; Yang & Matthews, 2010).
Another common way to assess a neighborhood’s physical characteristics is by measuring
walkability, which has varied in its conceptualization but the physical measurement has generally
been represented by whether environment promote walking through the availability of features
including sidewalks and traffic calming measures. When analyzing the connection between a
neighborhood’s walking environment and self-reported physical and mental health, results have
been mixed. Using survey data and multilevel structural equation modeling, (Barile et al., 2017)
find that a neighborhood’s walkability has no association with either perceived physical or
mental health. However, when assessing levels of physical activity directly, neighborhood
walkability was associated with higher levels of activity (Sallis et al., 2009). These studies utilize
demographic controls but not randomization to control for selection bias.

79
Few papers have assessed the relationship between structural developments, rather than
existing conditions, on health or health-related outcomes. A notable exception to this includes
(Moyer et al., 2019). This paper analyzes the relationship between a greening intervention and
firearm shootings via a natural experiment involving randomly assigning 110 geographically
contiguous clusters to either a control group or one of two treatment groups: a greening
intervention or a less-intensive mowing and trash cleanup. A difference-in-difference analysis
revealed that both intervention clusters saw significantly less shootings than the control clusters.
(Agbai, 2021) utilizes gentrification as a metric to capture exposure to increasing levels of
investment and amenities. Estimating the relationship between living in a gentrifying census tract
and self-reported health, this paper finds that each additional year of living within a gentrifying
tract is associated with an increase in health. Finally, structural improvements to owner-occupied
units were found to be associated with a decrease in total crime rates (South et al., 2021). This is
significant as overall feelings of safety are associated with better self-rated health, lower stress,
and lower levels of depression (Henderson et al., 2016; Kim, 2010; Ruijsbroek et al., 2016).
Contribution
The studies utilizing randomized control trials that involve individuals moving to neighborhoods
with different characteristics is a particularly strong analytic approach to study individual level
outcomes as it largely removes the issue of selection bias. There are several features of such
studies, however, that call into question whether this type of research design is an adequate way
to understand neighborhood effects. The first is that ultimately, the randomization process occurs
at the individual, rather than the neighborhood, level so that individuals are constrained in their
movement by unit availability as well as culturally and economically entrenched sorting
processes. Indeed, the MTO experiment has been criticized for the fact that even among

80
households in the treatment group receiving residential moving assistance, 72% ended up in an
area that was objectively segregated (Clamptet-Lundquist and Massey, 2008). Additionally, such
experiments are unable to say much about the specific features of neighborhoods that are
causally related to outcomes. On the other hand, studies that do attempt to analyze the
relationship between specific neighborhood features and individual-level outcomes do not
typically address issues of selection bias.
This paper contributes to the literature in that it seeks to expand on the call to better
understand specific mechanisms through which neighborhoods and individual-level health
outcomes might be connected (Harding et al., 2010). Our data allows us to not only examine the
potential associations between changes to the built environment and individual-level health
outcomes, but we are able to explore these relationships depending on different types of
investment as well as different levels of physical proximity to the investments. Additionally, our
paper is positioned to control for time invariant, personal characteristics that may otherwise bias
results through using a fixed effects design.
While this paper can not establish causality, our method of applying individual and year fixed
effects to a sample of non-movers has the advantage that any “observed inter-temporal
variations” in our outcome variables of interest “are coming purely from exogenous sources and
are not being selected by the individual resident on the basis of either income or unobservables”
(Galster & Hedman, 2013). We can also correct for one aspect of selection bias since we restrict
our sample to those initially surveyed, i.e. no one is moving into the sample throughout our study
period. Our study is prone to other forms of selection bias, however, as individuals do leave the
sample, either because they have left the neighborhood or are unable to respond to additional
survey waves. This attrition could push our results in either direction.

81
3. Methods
Study Population and Analytic Sample
Data are from the Pittsburgh Research on Neighborhood Change and Health (PHRESH)
studies (Baird et al., 2022; Baird et al., 2020; Dubowitz et al., 2015; Holliday et al., 2020;
Richardson et al., 2017) which are a suite of studies analyzing the impact of neighborhood
investments on resident’s health in two Pittsburgh neighborhoods: the Hill District and
Homewood. The two neighborhoods were matched on geography, socio-economic and racial
composition, and are about four miles apart in Pittsburgh, Pa. In 2011, participants from a
probability sample of households were enrolled from both neighborhoods with subsequent data
collection waves in 2013, 2014, 2016, and 2018. To be eligible to participate, individuals had to
be above 18. Individuals were enrolled in person by recruiters going to each randomly sampled
address. This resulted in less mobile residents, i.e. those who were older or without children,
being more likely to respond and subsequently participate. More detail regarding recruitment of
participants can be found in Dubowitz (2015). This analysis is limited to neighborhood changes
occurring between 2011 and 2018 and examines five outcomes among residents: food insecurity,
stress, perceived neighborhood safety, neighborhood satisfaction, and dietary quality (measured
by the Healthy Eating Index (HEI-2015)). Food insecurity, stress, and diet were collected only in
2011, 2014, and 2018. Neighborhood safety and satisfaction were collected in 2011, 2013, 2014,
2016, and 2018.
Table 1 shows the sample characteristics of individuals who completed 1 through 5
waves of data. Out of the 1,372 individuals who took the initial survey in 2011, 478 did not
complete any subsequent surveys while 903 completed at least 2 waves of surveys and are
therefore included in our regression analysis. Comparing the characteristics of individuals who

82
only completed 1 wave to those who completed more than 1, there are several differences
between the groups. In addition to expected time variant characteristics such as the mean age of
the sample and the mean years in the neighborhood being larger for those who stayed in the
sample longer, other differences stand out. The sample that responded to at least 2 waves of data
collection have larger mean incomes and have a slightly higher percentage who received a high
school diploma or less. Additionally, the sample containing individuals who responded to at least
2 waves of data collection have a lower percentage of renters (65% compared to 79%), a lower
percentage of males (23% compared to 31%), and a higher percentage of employed individuals
(35% to 33%).

83
Table 1: Characteristics of individuals by number of survey waves completed
1-wave
(N=478)
2-waves
(N=208)
3-waves
(N=204)
4-waves
(N=147)
5-waves
(N=344)
1-wave
(N=478)
>1-wave
(N=903) p-value
Mean Age 50
(19)
53
(18)
54
(18)
59
(17)
61
(13)
50
(19)
59
(16) <0.001
Mean Yrs in
Neigh
22
(23)
25
(23)
28
(23)
34
(23)
36
(23)
22
(23)
33
(23) <0.001
Mean Income
(thousands of
USD)
20
(16)
22
(19)
21
(19)
20
(19)
23
(20)
20
(16)
22
(20) <0.001
Educ: High
School or Less 50% 52% 54% 55% 51% 50% 52% 0.3
Educ: BA
graduate or
higher
17% 13% 15% 11% 15% 17% 14% 0.2
Percent with
Children at
Home
28% 32% 34% 19% 16% 28% 22% 0.002
Percent Male 31% 25% 27% 22% 20% 31% 23% <0.001
Percent
Employed 33% 31% 34% 32% 38% 33% 35% 0.4
Percent Renter 79% 75% 75% 63% 59% 79% 65% <0.001
Percent Black 91% 93% 94% 95% 95% 91% 94% 0.013
Total Number
of Participants
Per Wave
1,372 855 680 519 388 - - -

84
Our analysis includes those individuals who remained within the study area for at least
two survey waves. Because new participants were not added into the survey after the initial
selection, our results are not biased by individuals moving into the neighborhood over time. Our
analysis is impacted, however, by selection bias issues that arise due to individuals leaving the
study. For example, when comparing those who left the sample after 1 survey wave to those who
stayed for additional waves of data collection, we see that a higher percentage of those who left
were renters. Such selective attrition has implications for the generalizability of our results, i.e.
neighborhood investment may drive more vulnerable renters out.
Measuring Neighborhood Investments
We compiled a list of all investments that were partially or fully funded by public dollars
in the study neighborhoods that occurred between 2011 and 2018. These data were gathered
through interviews and data requests to the following four Pittsburgh public agencies: Housing
Authority of the City of Pittsburgh, the Urban Redevelopment Authority, the Pennsylvania
Housing Finance Agency, and the City of Pittsburgh. All investments included in this analysis
can be found in Table 2 of the appendix. For a more complete description of how the
revitalization efforts that have been occurring in these two neighborhoods, please see Baird et al.
(2020).
Hill District had significantly more investments throughout the seven-year time period,
with a cumulative total of 19 investments that were at least partially funded by public funds
completed between 2011 and 2018, compared with Homewood, which had a cumulative total of
seven investments over the same time period. The geographic distribution of these investments is
shown in the figures below. Figure 1 shows that the majority of investments in the Hill District

85
were concentrated towards the southern end of the neighborhood. Homewood’s investments are
more geographically dispersed as seen in figure 2.
Figure 1: Hill District Investment Location 2011-2018
Note: neighborhood is delineated at the parcel level

86
Figure 2: Homewood Investment Location 2011-2018
Note: Neighborhood is delineated at the parcel level
Figure 3 shows a comparison of total development costs per study area. While Hill
District had an overall greater level of investments, we do see that Homewood is also
experiencing changes within these study years. This figure also depicts that between 2011 and
2015, the Hill District neighborhood experienced much more fiscal investment, and between
2015 and 2018, investment in Homewood increased.

87
Figure 3: Total Development Costs (TDC) by year and neighborhood
Individual investments across each neighborhood vary in terms of cost, ranging from
$426,000 to $120,000,000, as well as in type. Of the 24 developments included in this study, just
under half were housing related, with commercial and business developments comprising the
next most common type. We also observe investments in the form of an animal shelter, park
renovations, a playground, and a recreational facility. These different development types have
different theoretical implications. For example, we might expect business and commercial
investments to spur the local economy while recreation-related developments may have a larger
impact on lifestyle. Because of this, we separate by investment type in our analysis in order to
more precisely explore causal pathways.
Measuring Exposure to Investments
Using Geographic Information System mapping, we were able to calculate the network
distance between each participant’s residence and every fully constructed investment between
2011 and 2018. With this measure in hand, we constructed three primary measures of exposure to

88
investment: (1) individual exposure at the neighborhood level, (2) individual exposure to
investments within ½ a mile and a 1-mile radius of place of residence, and 3) individual exposure
to investments within a ½ mile radius of place of residence. For all measures, we use a
cumulative measure of total development costs to capture exposure to investments over time.
From 2011 to 2018, the mean individual distance from investment is 0.74 miles with a
standard deviation of 0.43 miles. We used these values as a guide in choosing to focus on
investments that occur within a 1-mile and a ½ mile radius of an individual’s place of residence.
As shown in Table 2, we see that when measuring exposure to investments within ½ mile of an
individual’s home, there is still roughly 28% of the sample with no exposure by 2018. Measuring
distance at any value below ½ mile (for example, at the level of a neighborhood block) would
therefore limit our sample of exposed individuals substantially.
Table 2 also shows that while exposure to investments is increasing across time across all
distance measures, however, individuals are exposed to a more drastic increase at the
neighborhood level than at any other distance measure.
Table 2: Count of individuals with zero exposure to investments by year and distance
Count of Individuals
year neighborhood ½ to 1 mile ½ mile
2011 101 (29.36%) 590 (43.00%) 182 (52.91%)
2013 0 (0.00%) 103 (12.04%) 134 (38.95%)
2014 0 (0.00%) 91 (13.34 %) 134 (38.95%)
2016 0 (0.00%) 22 (4.24%) 103 (29.94%)
2018 0 (0.00%) 7 (1.80%) 96 (27.91%)
Note: percent of total yearly sample in parentheses
Outcomes of Interest
Food Insecurity

89
Food Insecurity was measured using the U.S. Adult Food Security Survey Module. The survey
assesses conditions and behaviors over the past 12 months that assess the extent to which
households are able to financially meet basic food needs (Coleman-Jensen et al., 2014) (e.g.,
how often the respondent’s household had to cut or skip meals due to financial concerns). Each
item is categorized as 0 (indicating no food insecurity) or 1 (indicating food insecurity). The food
insecurity scale is formed by summing the 10 items (range 0–10), with a score of 0 indicating the
highest level of food security and a score from 6-10 indicating low or very low food security
(USDA).
Perceived Stress
To assess perceived stress, respondents answered the following questions which are a part
of the Perceived Stress Score-4 (Cohen, 1988; Cohen et al., 1983): “1) how often have you felt
that you were unable to control the important things in your life?; 2) how often have you felt
confident about your ability to handle your personal problems?; 3) how often have you felt that
things were going your way?; and 4) how often have you felt difficulties were piling up so high
that you could not overcome them?” Potential responses include never, almost never, sometimes,
fairly often, and very often, with these categories being assigned a value of 0 through 4. The
maximum score is 16. Questions 2 and 3 were reverse coded so that higher values indicate more
stress.
Perceived Neighborhood Safety
Perceived neighborhood safety was measured using 4 items rated on a 5-point Likert
scales ranging from 1 (strongly agree) to 5 (strongly disagree) The questions included: “1) you
feel safe walking in your neighborhood during the day; 2) you feel safe walking in your
neighborhood during the evening; 3) your neighborhood is safe from crime; and 4) violence is a

90
problem in your neighborhood.” Responses were coded so that higher scores reflected a greater
perception of safety (Sampson et al., 1997).
Neighborhood Satisfaction
Individual-level neighborhood satisfaction was measured using the following survey
question (Peterson et al., 2004): “All things considered, would you say you are very satisfied,
satisfied, dissatisfied, very dissatisfied, or neutral – neither satisfied nor dissatisfied with your
neighborhood as a place to live?” Responses were assigned values 1 to 5, so that higher values
indicate greater satisfaction.
Healthy Eating Index (HEI)
We assessed resident diet using the Automated Self-Administered 24-hour dietary
assessment tool, which was administered by interviewers during the in-person surveys and
repeated 7–12 days later by phone (Subar et al., 2012). We derived Healthy Eating Index (HEI)-
2010 scores to measure overall dietary quality based on compliance with the US Dietary
Guidelines for Americans, calculating a single per-person score for each wave on the basis of the
average of the 2 recalls (Guenther et al., 2013). Scores range from 0 to 100; a score greater than
80 indicates good dietary quality, a score of 51–80 reflects a need for improvement, and a score
less than 51 indicates poor diet.
4. Analysis
To examine the relationship between exposure to investments and health-related
outcomes, we utilize several individual-level fixed effects models. This allows us to control for
all time-invariant individual level characteristics. Since individuals remain within the study areas
throughout the analysis, this model also controls for time-invariant characteristics of place. We
additionally include year fixed effects to control for yearly fluctuations that may impact all

91
individuals within our sample. The base model we use to estimate this association is represented
by the following equation:
= + + +
+ +
Where is the outcome of interest for individual at time , and estimates the
association between our primary variables of interest, , and the outcome
variable measured for individual at time . captures total cumulative
investment up to time t and therefore measures continuous treatment intensity. Total development
costs are accumulated at three different levels: the neighborhood level, between a ½ mile and 1
mile of an individual’s place of residence, and within ½ mile of the individual’s place of
residence. We first run our analysis at the neighborhood level and then run a separate set of
regressions to examine potential spillover effects. Finally, we include income and employment as
control variables to explore their potential mediating impacts. In a separate series of regressions,
we categorize investments based on whether they are classified as residential, commercial, or
business in order to observe their marginal impact. In all models, we include individual level
fixed effects
, which control for time-invariant individual characteristics, as well as year fixed
effects,
, which control for yearly trends impacting all individuals. Our analysis therefore uses
two-way fixed effects to measure within person variation in outcomes controlling for yearly
trends.
The Hill District and Homewood are divided into nine different subneighborhoods.
captures subneighborhood conditions experienced by individual i at time t. Since these
conditions change over time but do not change consistently across individuals, they are not
controlled for by the included fixed effects. We pick these subneighborhood-level controls based

92
on work exploring the relationship between neighborhood deprivation and public health by
Andrews et al. (2020), Slotman et al. (2022), and Roux and Mair (2010)10. These variables were
pulled using the ndi package from R (Buller ID, 2022).
5. Results
Sample characteristics
Table 3 shows summary statistics for our five outcome variables of interest stratified by
the Hill District and Homewood neighborhoods. We compare outcomes from the base survey
wave in 2011 to the final survey wave included in this study completed in 2018. We also
compare overall means for both neighborhoods across these two years. Looking at food
insecurity in 2011, 11.9% report having low to very low food security across all neighborhoods,
with this number decreasing to 7.6% by 2018. The difference between the change in food
insecurity is statistically significant across both neighborhoods over time. Looking at how food
insecurity breaks down across neighborhoods, we see that the Hill District initially starts with a
slightly higher mean food insecurity score than Homewood, but that by 2018, the Hill District is
showing a lower mean food insecurity. Neither difference in score between neighborhoods across
years is statistically significant.
10 Neighborhood level controls are: median household income; the percent of households receiving public
assistance; the percent of households who work in a management, business, science, or arts occupation; the percent
of households whose incomes fall below the poverty line; the unemployment rate, and the percent of households
who have received a high school education or higher

93
Table 3: Outcome variables per survey wave
2011 Diff 2018 Diff ALL Diff
Hill
District
Homewood p-value
Hill
District
Homewood p-value 2011 2018 p-value
Mean (SD)
Food Insecurity
1.85
(2.50)
1.70 (2.53) 0.6202
0.93
(1.96)
1.29 (2.46) 0.2014
1.81
(2.50)
1.04
(2.12)
p<0.00
1
Min 0 0 0 0 0 0
Max 10 9 10 10 10 10
Low to very low
security
11.5% 12.8% 6.2% 10.9% 11.9% 7.6%

Stress
Mean (SD)
4.41
(3.28)
4.42 (3.35) 0.9849
4.15
(3.07)
3.91 (3.09) 0.5224
4.41
(3.30)
4.08
(3.07)
0.1659
Min 0 0 0 0 0 0
Max 15 14 14 12 15 14

Safety
Mean (SD)
8.96
(2.81)
7.52 (2.54) p<0.001
9.22
(2.49)
7.49 (2.30) p<0.001
8.54
(2.81)
8.72
(2.56)
0.3741
Min 3 3 3 3 3 3
Max 15 13 15 13 15 15

Satisfaction
Neighborhood
Mean (SD)
3.84
(1.11)
3.42 (0.96) p<0.001
3.95
(1.04)
3.37 (1.09) p<0.001
3.72
(1.08)
3.78
(1.04)
0.4298
Min 1 1 1 1 1 1
Max 5 5 5 5 5 5
HEI
Mean (SD)
50.88
(11.38)
52.67
(12.81)
0.2299
50.87
(11.49)
50.81
(11.85)
0.9666
51.40
(11.82) (11.5)
50.85
0.5386
Min 26.02 25.25 24.99 15.37 25.25 15.37
Max 83.26 85.04 80.83 87.471 85.04 87.47
Poor diet 53.3% 43.4% 49% 49.5% 50% 49%

94
Perceived stress scores in this sample range from 0 to 15, with a higher number indicating
more stress. Across both neighborhoods, the mean of this stress score in each year is lower than
the latest population norms found for an age group between 55 and 64 (5.76), with about 34% of
the sample having higher than this average in 2011 and therefore experiencing higher perceived
stress levels (Wartiig et al., 2013). We see that in both years, there is no statistically significant
difference between stress levels of residents in the Hill District when compared to residents in
Homewood. Turning to perceived neighborhood safety, the sample max for this measure is 15
with a mean range between 8.54 and 8.76. Higher scores represent a greater sense of safety. In
both 2011 and in 2018, residents of the Hill District report statistically higher safety scores than
Homewood residents.
We also see that Hill District residents are statistically more satisfied with their
neighborhood than Homewood residents although the difference in magnitude of the scores is
very small. In all, the mean scores measuring neighborhood satisfaction indicate that on average,
residents are neutral to satisfied living in their neighborhood. The dietary quality of the sample is
low, with 50% having poor dietary quality as measured by the healthy eating index (HEI) in
2011. The change in HEI between 2011 and 2018 across both neighborhoods is insignificant.
Comparing between neighborhoods, Hill District has a slightly lower mean HEI score than
Homewood in 2011, but by 2018, the two scores are almost identical.
Full Sample of Investment Types
Tables 4 and 5 shows the results of our fixed effects model where the cumulative
investments variable is defined by the total development costs of all investments that an
individual is exposed to up to time t. We first look at cumulative investments occurring at the
neighborhood level. Table 4 shows the analysis at the neighborhood level both with and without

95
controlling for income and employment status. Without these additional controls, when
measuring the direct impact of investments on our outcome variables of interest, we see a
statistically significant relationship between a standard deviation increase in invesments and food
insecurity, stress, safety, and neighborhood satisfaction. More specifically, food insecurity and
stress decrease while safety and neighborhood satisfaction both increase. We see no relationship
between cumulative investments and HEI.
When we control for income and employment, the magnitude of the relationship between
investments and food insecurity, safety, and neighborhood satisfaction decreases but remains
significant. The relationship between investments and stress is no longer significant when
including this controls.

96
Table 4: Regression output for full investment sample at neighborhood level of exposure
Note:
*p<0.1;
**p<0.05;
***p<0.01. All continuous variables are standardized. X indicates that the model includes fixed effects
public assistance; the percen
(individual and year) and the subneighborhood level controls of: median household income; the percent of households receiving
t of households who work in a management, business, science, or arts occupation; the percent of
households whose incomes fall below the poverty line; the unemployment rate, and the percent of households who have received a
high school education or higher
Food Insecurity Stress Safety
Neighborhood
Satisfaction
HEI
(1) (2) (1) (2) (1) (2) (1) (2) (1) (2)
Level
Neighborhood
Invest
Cumulative
-
0.294
***
(0.074)
-0.270
***
(0.074)
-0.166
*
(0.092)
-0.133
(0.092)
0.231
***
(0.064 )
0.237
***
(0.064)
0.201
***
(0.068)
0.201
***
(0.068)
0.006
(0.093)
0.006
(-0.093)
Income -
-0.097
**
(0.038)
-
-0.086
*
(
0.047)
-
-0.010
(0.027)
-
-0.022
( 0.029)
-
0.018
(0.048)
Employed -
-0.204
***
(
0.070)
-
-0.318***
(0.087)
-
-0.011
(0.052)
-
-0.013
(0.056)
-
-0.112
(0.088)
Fixed Effects X X X X X X X X X X
NeighControls X X X X X X X X X X

97
Table 5 replicates the above analysis with a slight variation. Here, we are examining
whether the associations found at the neighborhood level change depending on an individual’s
promixity to investments. As Table 5 shows, once we start measuring investments closer to an
individual’s place of residence, the relationship between investments and most of our health
outcomes of interest are no longer statistically significant. An exception to this is food insecurity,
which continues to be negatively associated with increasing investments when individuals are
exposed within a ½ mile of their place of residence but not when exposed between a ½ mile and
1 mile.

98
Table 5: Regression output for full investment sample at varying distances of exposure
Note:
*p<0.1;
**p<0.05;
***p<0.01. All continuous variables are standardized. X indicates that the model includes fixed effects
public assistance; the percen
(individual and year) and the subneighborhood level controls of: median household income; the percent of households receiving
t of households who work in a management, business, science, or arts occupation; the percent of
households whose incomes fall below the poverty line; the unemployment rate, and the percent of households who have received a
high school education or higher
Food Insecurity Stress Safety
Neighborhood
Satisfaction
HEI
(1) (2) (1) (2) (1) (2) (1) (2) (1) (2)
Invest 1/2
Cumulative
-
mile radius
-0.064
*
(0.037)
-0.060
*
(0.036)
-0.021
(0.045)
-0.019
(0.045)
0.028
(0.028)
0.021
(0.028)
0.015
(0.030)
0.014
(0.030)
0.038
(0.046)
0.035
(0.046)
Invest ½ to 1
Cumulative
-
mile radius
-0.033
(0.047)
-0.019
(0.047)
-0.045
(0.058)
-0.026
(0.058)
0.027
(0.035)
0.022
(0.035)
0.003
(0.037)
0.003
(0.037)
0.015
(0.059)
0.020
(0.059)
Income -
-0.100
***
(0.038)
-
-0.088
*
(0.048)
-
-0.007
(0.027)
-
-0.020
(0.029)
-
0.015
(0.048)
Employed -
-0.222
***
(0.071)
-
-0.324
***
(0.088)
-
-0.011
(0.053)
-
-0.012
(0.056)
-
-0.110
(0.089)
Fixed Effects X X X X X X X X X X
NeighControls X X X X X X X X X X

99
Investments by Type
In order to more precisely explore the relationship between investments and our health
outcomes of interest, we now define cumulative exposure according to investment type. More
specifically, we divide investments according to whether they are defined as commercial,
business, or residential. We then calculate an individual’s exposure to the cumulative investment
total of each type measuring proximity at the neighborhood level, between a ½ mile and 1 mile,
and within a 1/2 mile of place of residence.
Table 6 shows the result of these regressions. We see that the relationship between
increasing investments and decreasing food insecurity is largely led by commercial investments
and residential investments. The association between food insecurity and commercial
investments specifically is significant at all geographic measures of exposure. The absence of a
relationship between food insecurity and business-type invesments is particularly suprising as
included in business investments is the grocery store Shop N’ Save. While the addition of a
grocery store is intuitively related to food insecurity (a grocery store may provide cheaper food
options), this lack of statistical significance is consistent with other studies exploring this cohort.
Dubowitz et al. (2015) find that improvements in diet seen in the Hill District, the neighborhood
where the grocery store is located, were not significantly linked with frequency of shopping at
the new supermarket, bur rather that both frequent and infrequent shoppers were seeing benefits.
In Table 6, perceived stress responds to both residential investments (decreases at the
neighborhood level) as well as business invesments (decreases when exposure is within a ½ mile
of place of residence). Interestingly, while neighborhood safety is significantly positively related
to increasing invesments across all investment types, Table 5 shows more nuance when looking
at spillover effects. When residents are within a ½ mile of commercial investments, the

100
association is negative; however, when exposure is at distances greater than a 1/2 mile, the
association flips to being positive. We see a similar kind of pattern with neighborhood
satisfaction. When commercial investments occur over a ½ mile away from an individual’s place
of residence, the relationship between investments and neighborhood satisfaction is positive
while the reverse is true within closer proximity to development.

101
Table 6: Regression output for investments by investment type
Food
Insecurity
Stress Safety
Neighborhood
Satisfaction
HEI
Comm. Bus. Resid. Comm. Bus. Resid. Comm. Bus. Resid. Comm. Bus. Resid. Comm. Bus. Resid.
Level
Neigh.
Invest
Cumulative
Level
Neigh
-0.565
***
(0.154)
0.026
(0.062)
-
0.500
***
(0.128)
-0.219
(0.191)
-0.117
(0.079)
-0.284
*
(0.159)
0.057
(0.125)
0.018
(0.027)
0.020
(0.103)
0.231
*
(0.132)
-0.029
(0.029)
0.136
(0.109)
0.187
(0.193)
-0.043
(0.083)
0.047
(0.161)
effects
Fixed
X X X X X X X X X X X X X X X
Controls X X X X X X X X X X X X X X X
Invest 1/2
Cumulative
Spillover
-
mile radius
-0.264
*
(0.140)
-0.032
(0.084)
-0.047
(0.034)
-0.087
(0.172)
-0.194
*
(0.107)
-0.005
(0.042)
-
0.346
***
(0.121)
-0.012
(0.053)
-0.011
(0.026)
-0.295
**
(0.128)
0.077
(0.056)
-0.010
(0.027)
-0.038
(0.174)
0.048
(0.112)
0.034
(0.042)
1
Invest ½ to
Cumulative
-mile
radius
-0.386
**
(0.153)
-0.004
(0.058)
-0.004
(0.040)
-0.135
(0.191)
-0.070
(0.073)
-0.010
(0.050)
0.272
**
(0.124)
-0.002
(0.035)
-0.014
(0.029)
0.258
*
(0.132)
-0.049
(0.037)
-0.020
(0.031)
-0.103
(0.280)
-0.046
(0.077)
0.022
(0.050)
effects
Fixed
X X X X X X X X X X X X X X X
Controls X X X X X X X X X X X X X X X
Note:
*p<0.1;
**p<0.05;
***p<0.01. All continuous variables are standardized. X indicates that the model includes fixed effects
public assistance; the percen
(individual and year) and the subneighborhood level controls of: median household income; the percent of households receiving
t of households who work in a management, business, science, or arts occupation; the percent of
households whose incomes fall below the poverty line; the unemployment rate, and the percent of households who have received a
high school education or higher

102
6. Discussion
Neighborhoods not only define a resident’s physical and sociocultural environment, but they
encompass a spatial distribution network that influences the relative accessibility of healthrelated resources (Bernard et al., 2007). An ongoing task of the literature therefore is to better
understand the causal mechanisms that exist between these spatial networks and health outcomes
of interest. This task poses several challenges. This first is that it is not straightforward to
disentangle neighborhood characteristics to understand which specific features are impacting
health. The built environment is just one component of a larger system of theorized connections
between neighborhood and health (Northridge et al., 2003). The second is a question of
geographic scope. Different geographic areas, such as an administratively defined census tract
versus the block on which a person resides, represent a bundle of characteristics that may be
differentially relevant to the health outcome of interest (Diez Roux, 2001). Finally, it is difficult
to overcome selection bias, which results from individuals selecting into or out of neighborhoods
in response to structural changes.
This paper contributes to the literature by anaylzing the impact of a specific
neighborhood-level change. Using longitudinal data following individuals living in two lowincome neighborhoods, we are able to assess how investments impact health-related outcomes of
residents within our two focus neighborhoods. In this way, we analyze the relevance of a specific
neighborhood-level intervention, investments to the built environment, to individual-level health.
Additionally, since our data is collected at the residential level, we are able to test associations
between investments and health-related outcomes at different geographic distances. While this
longitudinal design following a specific cohort ensures that our results are not biased by

103
individuals who have recently moved into the neighborhood, our results may still be biased by
those who have left the study, a sample that skews towards being renters and male.
We find that when measured at the neighborhood level across all investment types, an
additional million dollars in investments is associated with decreased food insecurity, decreased
stress, increased safety, and increased neighborhood satisfaction. At this level, there is no
relationship between individual stress or the healthiness of food consumption. When including
income and employment status as controls, the magnitude of the relationship decreases slightly
for food insecurity and neighborhood satisfaction but actually increases for safety. When
exposure is measured closer to an individual’s place of residence, we see that only the negative
relationship between increasing investments and food insecurity remains for those living within a
½ mile of investments.
When looking at exposure to investments by type, we see that commercial investments
and residential investments are largely driving the decrease in food insecurity. Commercial
investments are the only investment type correlated with safety and neighborhood satisfaction.
While neighborhood safety was positively associated with increases in aggregated investment
types, it is negatively related to increases in commercial investments, although this relationship
only exists when the individual lives within a ½ mile of investments. For individuals living
between a ½ mile and a mile of commercial investments, this relationship once again becomes
positive. This same pattern arises with neighborhood satisfaction, implying that investments may
have deleterious impacts for residents living closer to the sites of these developments.
These results provide several insights. While we see that food insecurity is significantly
associated with increasing commercial investments across several models, we see no relationship
between any investment type and overall dietary quality. A potential explanation for this is that

104
all physical changes are interacting with local context and individual habits. Residents have
enduring collective identities and properties that impact how they interact with space, meaning
that habits such as what one typically eats are less likely to quickly respond to changes in the
built environment (Macintyre et al., 2002). Food insecurity, on the other hand, is largely
connected to socio-economic status. This also explains why we do not see a statistically
significant relationship between business investments, which include a a grocery store in the Hill
District, and food insecurity. As past work has found (Dubowitz et al., 2015), the opening of this
grocery store did not largely change eating habits of residents and any improvemenents in diet
seen were observed across both frequent and infrequent shoppers. It may therefore be the case
that commercial investments are more directly related to economic activity.
This leads to the question of the mechanisms through which these investments may be
acting. There are several possible pathways. To test the hypothesis that these investments are
positively impacting economic status, we ran two regressions (Appendix Table 2 and Appendix
Table 3) which test the association between investments and both income and employment
status. We find a positive relationship between investments and income as well as a positive
relationship between investments and the likelihood of being employed. It is plausible for all
investment types to impact this aspect of an individual’s life. Commerical and business
invesments may provide new economic opportunity, while residential investments could offer
individuals with a stable, affordable housing option. In addition to economic change, there are
other channels through which these investments may be acting as well. All investment types will
change the physical appearance of the neighborhood such as greened spaces due to landscaping
efforts and new buildings. These spaces might encourage walking or other stress relieving

105
activities. Similarly, aesthetic changes such as revitalized storefronts and housing complexes
might encourage greater overall neighborhood satisfaction.
This work also point to the potentially deleterious outcomes for certain groups. We see
negative impacts on safety and neighborhood satisfaction for individuals living within a ½ mile
of commercial investments. Additionally, we are unable to ascertain the impacts of investments
on the individuals who left the study sample over time. While we cannot test this, it could be the
case that the investments actually caused worse economic or health outcomes for those
individuals who left the study. This impacts the generalizability of this study, which is skewed
towards older individuals who are female and have lived in the neighborhood for several
decades.
Limitations
One of the most important limitations of this study is that of selection bias. While our
methods allow us to control for the potential bias of individuals entering the sample, we are
unable to control for those leaving the sample. This limitation has ambiguous impacts on our
estimations. It could be the case that as individuals became more healthy, they moved to a
different neighborhood entirely and biased our results downwards. Conversely, individuals may
have become worse off and moved, biasing our results upward. We are unable to test for this
since we do not have data for individuals who did not complete survey waves for any reason,
including moving. This selection bias issue has implications for the generalizability of our study
as well as for the interpretation of our estimates. In comparison to the sample that left the study
after 1 wave of data collection, the sample that stayed in the study for at least two waves of data
collection tend to have more females, fewer renterss, older individuals, and individuals who have
lived in the neighborhood longer. This could imply that older, female individuals more

106
entrenched in the neighborhood are better positioned to receive the potential health benefits of
investments. On the other hand, investments may actually be pushing out more vulnerable or
transient populations, though testing this hypothesis is outside of the scope of this paper.
Additionally, while we were able to estimate the associations between distance to
investments and our health-related outcomes of interest, we are limited in that we are only able
to measure exposure from an individual’s residence. Thus our study ignores how people move
throughout space. As (Kwan, 2018) highlights, focusing on an individual’s place of residence
when studying environmental exposures “ignores people’s daily mobility and exposures to
nonresidential contexts.” In the case of this analysis, while an individual’s place of residence
may be close to an investment, that person’s average exposure as they move throughout the day
may be quite low. This is especially probable if this person spends very little time at home or in
their neighborhood of residence. While we do not hypothesize that this is a particular problem
impacting the results of our study, it would be a problem if, for example, the impacts of the
neighborhood characteristic being measured were highly dependent on lengths of physical
exposure.
This analysis also does not examine how individuals are interacting with investments.
This limits our understanding of the potential mechanisms through which neighborhood
investments are working. For example, we cannot say whether individuals were actively utilizing
these new facilities, or simply benefitting from improved landscaping. Additionally, we do not
weight an investment's potential area of impact according to its relative development costs. It is
plausible that a larger investment as measured by total development costs would impact a
relatively larger geographic area than a smaller investment. Along these lines, we utilize a linear
model which assumes a constant relationship between investments and outcomes; however, this

107
may not be the case. This model does not account for the potential decreasing marginal returns of
neighborhood level investments.
7. Conclusion
By providing evidence to better understand the specific types of health-related outcomes
that are associated with neighborhood economic development, our results can assist
policymakers in targeting interventions accordingly. Within the timeframe of our analysis, we
observe different responses to economic developments depending on the type of investments as
well as the distance between the investment and an individual’s place of residence. When looking
at all investment types at the neighborhood level, we see a statistically significant relationship
between food insecurity, stress, safety, and neighborhood satisfaction. When disaggregating by
investment type, we see slightly more nuanced results, including that living closer to commercial
investments is negatively associated with both safety and neighborhood satisfaction.
These findings suggest several things. The first is that economic investments have the
potential to impact health-related variables that are tied to economic security, such as food
insecurity. Because this study does not investigate the ways in which individuals are interacting
with investments, understanding the specific mechanisms of action are beyond the scope of this
analysis. However, there are several plausible pathways through which these investments might
impact the economic security of residents such as through increased income and a higher
likelihood of being employed. It could also be the case that the developments are improving
health through aesthetic improvements to the neighborhood.

108
References
Agbai, C. O. (2021). Shifting neighborhoods, shifting health: A longitudinal analysis of
gentrification and health in Los Angeles County. Social science research, 100, 102603.
Arcaya, M. C., Tucker-Seeley, R. D., Kim, R., Schnake-Mahl, A., So, M., & Subramanian, S.
(2016). Research on neighborhood effects on health in the United States: a systematic
review of study characteristics. Social Science & Medicine, 168, 16-29.
Baird, M. D., Cantor, J., Troxel, W. M., & Dubowitz, T. (2022). Job loss and psychological
distress during the COVID‐19 pandemic: Longitudinal Analysis from residents in nine
predominantly African American low‐income neighborhoods. Health Economics, 31(9),
1844-1861.
Baird, M. D., Schwartz, H., Hunter, G. P., Gary-Webb, T. L., Ghosh-Dastidar, B., Dubowitz, T.,
& Troxel, W. M. (2020). Does large-scale neighborhood reinvestment work? Effects of
public–private real estate investment on local sales prices, rental prices, and crime rates.
Housing policy debate, 30(2), 164-190.
Barile, J. P., Kuperminc, G. P., & Thompson, W. W. (2017). Resident characteristics and
neighborhood environments on health‐related quality of life and stress. Journal of
Community Psychology, 45(8), 1011-1025.
Bernard, P., Charafeddine, R., Frohlich, K. L., Daniel, M., Kestens, Y., & Potvin, L. (2007).
Health inequalities and place: a theoretical conception of neighbourhood. Social science
& medicine, 65(9), 1839-1852.
Chetty, R., Hendren, N., & Katz, L. F. (2016). The effects of exposure to better neighborhoods on
children: New evidence from the moving to opportunity experiment. American Economic
Review, 106(4), 855-902.
Clampet-Lundquist, S., & Massey, D. S. (2008). Neighborhood effects on economic selfsufficiency: A reconsideration of the Moving to Opportunity experiment. American
Journal of Sociology, 114(1), 107-143.
Cohen, S. (1988). Perceived stress in a probability sample of the United States.
Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A global measure of perceived stress.
Journal of health and social behavior, 385-396.
Coleman-Jensen, A., Gregory, C., & Singh, A. (2014). Household food security in the United
States in 2013. USDA-ERS Economic Research Report(173).
Diez Roux, A. V. (2001). Investigating neighborhood and area effects on health. American
journal of public health, 91(11), 1783-1789.

109
Dubowitz, T., Ghosh-Dastidar, M., Cohen, D. A., Beckman, R., Steiner, E. D., Hunter, G. P., . . .
Sloan, J. C. (2015). Diet and perceptions change with supermarket introduction in a food
desert, but not because of supermarket use. Health Affairs, 34(11), 1858-1868.
Galster, G., & Hedman, L. (2013). Measuring neighbourhood effects non-experimentally: how
much do alternative methods matter? Housing Studies, 28(3), 473-498.
Guenther, P. M., Casavale, K. O., Reedy, J., Kirkpatrick, S. I., Hiza, H. A., Kuczynski, K. J., . . .
Krebs-Smith, S. M. (2013). Update of the healthy eating index: HEI-2010. Journal of the
Academy of Nutrition and Dietetics, 113(4), 569-580.
Harding, D. J., Gennetian, L., Winship, C., Sanbonmatsu, L., & Kling, J. R. (2010). Unpacking
neighborhood influences on education outcomes: Setting the stage for future research.
Henderson, H., Child, S., Moore, S., Moore, J. B., & Kaczynski, A. T. (2016). The influence of
neighborhood aesthetics, safety, and social cohesion on perceived stress in disadvantaged
communities. American Journal of Community Psychology, 58(1-2), 80-88.
Holliday, S. B., Troxel, W., Haas, A., Ghosh-Dastidar, M. B., Gary-Webb, T. L., Collins, R., . . .
Dubowitz, T. (2020). Do investments in low-income neighborhoods produce objective
change in health-related neighborhood conditions? Health & place, 64, 102361.
Katz, L. F., Kling, J. R., & Liebman, J. B. (2001). Moving to opportunity in Boston: Early results
of a randomized mobility experiment. The quarterly journal of economics, 116(2), 607-
654.
Kim, J. (2010). Neighborhood disadvantage and mental health: The role of neighborhood
disorder and social relationships. Social science research, 39(2), 260-271.
Kling, J. R., Liebman, J. B., & Katz, L. F. (2007). Experimental analysis of neighborhood
effects. Econometrica, 75(1), 83-119.
Kwan, M.-P. (2018). The limits of the neighborhood effect: Contextual uncertainties in
geographic, environmental health, and social science research. Annals of the American
Association of Geographers, 108(6), 1482-1490.
Macintyre, S., Ellaway, A., & Cummins, S. (2002). Place effects on health: how can we
conceptualise, operationalise and measure them? Social science & medicine, 55(1), 125-
139.
Martz, C. D., Hunter, E. A., Kramer, M. R., Wang, Y., Chung, K., Brown, M., . . . Chae, D. H.
(2021). Pathways linking census tract typologies with subjective neighborhood disorder
and depressive symptoms in the Black Women's Experiences Living with Lupus
(BeWELL) Study. Health & place, 70, 102587.

110
Matthews, S. A., & Yang, T.-C. (2010). Exploring the role of the built and social neighborhood
environment in moderating stress and health. Annals of Behavioral Medicine, 39(2), 170-
183.
Moyer, R., MacDonald, J. M., Ridgeway, G., & Branas, C. C. (2019). Effect of remediating
blighted vacant land on shootings: a citywide cluster randomized trial. American journal
of public health, 109(1), 140-144.
Ndjila, S., Lovasi, G. S., Fry, D., & Friche, A. A. (1998). Measuring Neighborhood Order and
Disorder: a Rapid.
Neumark, D., & Simpson, H. (2015). Place-based policies. In Handbook of regional and urban
economics (Vol. 5, pp. 1197-1287). Elsevier.
Northridge, M. E., Sclar, E. D., & Biswas, P. (2003). Sorting out the connections between the
built environment and health: a conceptual framework for navigating pathways and
planning healthy cities. Journal of urban health, 80, 556-568.
Peterson, C. E., Sastry, N., Pebley, A. R., Ghosh-Dastidar, B., Williamson, S., & Lara-Cinisomo,
S. (2004). The Los Angeles family and neighborhood survey: codebook.
Pollack, C. E., Blackford, A. L., Du, S., Deluca, S., Thornton, R. L., & Herring, B. (2019).
Association of receipt of a housing voucher with subsequent hospital utilization and
spending. JAMA, 322(21), 2115-2124.
Richardson, A. S., Ghosh-Dastidar, M., Beckman, R., Flórez, K. R., DeSantis, A., Collins, R. L.,
& Dubowitz, T. (2017). Can the introduction of a full-service supermarket in a food
desert improve residents' economic status and health? Annals of epidemiology, 27(12),
771-776.
Ross, C. E., & Mirowsky, J. (2001). Neighborhood disadvantage, disorder, and health. Journal of
health and social behavior, 258-276.
Ruijsbroek, A., Droomers, M., Hardyns, W., Groenewegen, P. P., & Stronks, K. (2016). The
interplay between neighbourhood characteristics: The health impact of changes in social
cohesion, disorder and unsafety feelings. Health & place, 39, 1-8.
Sallis, J. F., Saelens, B. E., Frank, L. D., Conway, T. L., Slymen, D. J., Cain, K. L., . . . Kerr, J.
(2009). Neighborhood built environment and income: examining multiple health
outcomes. Social science & medicine, 68(7), 1285-1293.
Sampson, R. J. (2008). Moving to inequality: Neighborhood effects and experiments meet social
structure. American journal of sociology, 114(1), 189-231.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A
multilevel study of collective efficacy. science, 277(5328), 918-924.

111
Santiago, A. M., Galster, G., Lucero, J., Ishler, K., Lee, E. L., Kypriotakis, G., & Stack, L.
(2014). Opportunity neighborhoods for Latino and African American children. Available
at SSRN 2563141.
Schüle, S. A., & Bolte, G. (2015). Interactive and independent associations between the
socioeconomic and objective built environment on the neighbourhood level and
individual health: a systematic review of multilevel studies. PloS one, 10(4), e0123456.
South, E. C., Kondo, M. C., Cheney, R. A., & Branas, C. C. (2015). Neighborhood blight, stress,
and health: a walking trial of urban greening and ambulatory heart rate. American journal
of public health, 105(5), 909-913.
Subar, A. F., Kirkpatrick, S. I., Mittl, B., Zimmerman, T. P., Thompson, F. E., Bingley, C., . . .
McNutt, S. (2012). The automated self-administered 24-hour dietary recall (ASA24): a
resource for researchers, clinicians and educators from the National Cancer Institute.
Journal of the Academy of Nutrition and Dietetics, 112(8), 1134.
Yang, T.-C., & Matthews, S. A. (2010). The role of social and built environments in predicting
self-rated stress: A multilevel analysis in Philadelphia. Health & place, 16(5), 803-810.

112
Appendix
Table 1: List of Neighborhood Investments/Developments
Development
Name
Development
Type
Completed
Year
Total
Development
Cost
Neighborhood
ID
Dinwiddie Phase 1 Housing 2011 $8,272,390.00 Hill District
Kaufman Center Commercial 2011 $6,000,000.00 Hill District
New Granada Theater Commercial 2011 $1,200,000.00 Hill District
Petersen Sports
Complex Commercial 2011 $29,000,000.00 Hill District
Oak Hill Phase II Housing 2011 $37,225,828.00 Hill District
Thelma Lovette YMCA
Recreational
Facility 2012 $13,000,000.00 Hill District
Dinwiddie Phase 2 Housing 2012 $7,575,639.00 Hill District
Oak Hill Commons Commercial 2012 $8,000,000.00 Hill District
Shop N' Save Business 2013 $11,000,000.00 Hill District
Wheel Mill Business 2013 $500,000.00 Homewood
Kaboom, Susquehanna Playground 2013 $37,000.00 Homewood
Dinwiddie Phase 3 Housing 2013 $9,295,662.00 Hill District
Energy Innovation
Center Education 2014 $47,100,000.00 Hill District
Grayson Center Housing 2014 $1,500,000.00 Hill District
Skyline Terrace I/
Addison Housing 2015 $56,718,523.00 Hill District
Homewood Station
Senior Apartments Housing 2015 $11,516,268.00 Homewood
Larimer Pointe Housing 2015 $21,000,000.00 Homewood
Bridgeway Capital
Building Business 2015 $15,000,000.00 Homewood
August Wilson Park Park Renovations 2016 $1,300,000.00 Hill District
Skyline Terrace II
Bentley Drive Housing 2016 $56,718,523.00 Hill District
Cornerstore Village Housing 2016 $120,000,000.00 Homewood
Dinwiddie Phase 4 Housing 2016 $8,176,002.00 Hill District
Animal Rescue League Animal Shelter 2017 $15,000,000.00 Homewood
MOKA Art Gallery Commercial 2018 $426,000.00 Hill District

113
Chapter 5. Conclusion
Understanding the impacts of neighborhoods and housing on individual well-being is
integral to not only better understand the origins of modern-day population-level disparities, but
to develop interventions and techniques to rectify them. This dissertation has analyzed both
modern-day and historical neighborhood level processes with the goal of peeling back the
myriad of characteristics that make up a neighborhood and to pinpoint specific mechanisms
linked to outcomes. To do this, chapters 2 and 3 dive into an increasingly popular data source,
the HOLC redlining maps, to both test whether these maps are indicative of systematic housing
finance restrictions and to suggest ulterior ways that this data may be useful to use past
neighborhood characteristics to predict modern-day outcomes. Chapter 4 departs from a
historical lens by analyzing the impacts that economic development within two historically
disadvantaged, low-income neighborhoods has on residents’ individual-level health-related
outcomes.
Chapter 2 looks at two commonly analyzed outcome variables within the redlining
literature, homeownership rates and black population concentrations, and tests whether the initial
population composition of D-graded areas or the D grade itself is a more salient predictor of
these outcomes in subsequent decades. This paper finds that when categorized by initial
population composition, D-graded areas do not have outcomes with a similar direction of impact
over time, calling into question whether observed outcomes in other papers can be attributed to
systematic mortgage discrimination. Nationally, while neighborhoods categorized as “Any
Black” have lower homeownership rates than C-graded areas in both 1960 and 2000,
neighborhoods categorized as “Immigrant” have no statistically significant difference from Cgraded areas in terms of homeownership rates by 2000. The most robust findings are related to

114
future racial composition, with “Any Black” D-graded areas having consistently higher
percentages of black populations than C-graded in both 1960 and 2000, while “Immigrant”
neighborhoods have significantly lower percentages than or no difference from C-graded areas.
Chapter 3 uses the HOLC area description files to investigate the connection between
past neighborhood characteristics and modern-day asthma prevalence as well as exposure to
concentrations of fine particulate matter (PM2.5) and nitrogen dioxide (NO2). Using the remarks
section of the area description files, I classify neighborhoods according to described industrial
impact. Predicting future asthma prevalence, PM2.5 concentrations, and NO2 concentrations
based on past industrial impact as well as by HOLC grade, I find that both having received a D
grade and having been impacted by industry are statistically significant predictors of all
outcomes. However, the estimated coefficient of a D versus C-graded neighborhood is greater in
magnitude for both asthma prevalence and NO2 concentrations than the estimated coefficient for
an industrial versus non-industrial area. These results suggest that while industrial activity may
be one variable impacting future asthma prevalence and disparities in poor air-quality exposure,
it does not entirely describe the observed differences that have been seen in D versus C-graded
areas.
Chapter 4 uses longitudinal survey data to assess individual-level health-related outcomes
in two low-income, high-minority Pittsburgh neighborhoods. Specifically, it analyzes the
association between changes to the built environment and food insecurity, perceived stress,
perceived neighborhood safety, neighborhood satisfaction, and dietary quality. Findings reveal
several important insights. The first is that residential distance from developments impacts
outcomes. While food insecurity decreases as the total dollar amount of investments increase
regardless of an individual’s residential distance to a new or renovated development, other

115
outcomes are dependent on distance. Both safety and neighborhood satisfaction are positively
related to investments at the neighborhood level, but negatively related when individuals are
living within a ½ mile of commercial developments specifically. Additionally, which food
insecurity responds to investments, dietary quality does not, suggesting that investments may
impact ability to purchase food but does not necessarily impact what is consumed.
1. Implications of findings
Chapters 2 and 3 explore the utility of the HOLC maps and accompanying area
description files, contributing to the debate as to whether the HOLC maps are indicative of
mortgage discrimination. Chapter 2 is conclusive in the finding that the initial racial make-up of
D-graded areas is a more salient predictor of future homeownership rates and concentrations of
black populations than the D grade itself. Since homeownership rates are likely to be indicative
of ability to secure a mortgage, we should see a similar dampening of rates across D-graded
neighborhoods regardless of population composition, particularly in 1960 before the passage of
fair housing laws, if credit was denied across all D-graded areas but available to C-graded areas.
Since this is not the case, it is implausible that D-graded neighborhoods universally had
restricted credit access in comparison to C-graded areas. Rather, it is more likely that racial
filtering into some of the worst-off neighborhoods in combination with other types of
discriminatory practices, such as blockbusting and the movement of investments away from the
city, have contributed to the starker outcomes seen between graded neighborhoods.
Interestingly, chapter 3 shows that a D grade is a more salient predictor of asthma
prevalence and future air quality than whether the neighborhood had industrial activity. This
finding seems to be contradictory to chapter 2 – in a broad sense chapter 2 concludes that grades
do not matter while chapter 3 implies that they do. However, these two analyses are looking at

116
different angles of our understanding of the utility of HOLC maps. Chapter 2 is testing an
outcome variable, homeownership, which is theoretically and practically linked to relative credit
access to determine whether a D grade is suggestive of universal credit restriction. Chapter 3, on
the other hand, is testing whether past industrial activity is driving disparities in asthma and air
quality outcomes seen within D-graded areas. The fact that industrial activity does not entirely
describe the association between D-graded neighborhoods and future asthma prevalence and air
quality suggests that more research needs to be done to better determine what factor or
combination of factors were occurring in D-graded areas that correlates with these two outcome
variables.
While chapter 3 does not directly test potential omitted variables beyond past industrial
activity, there are several factors beyond restricted credit access that may have been occurring
within D-graded areas. Asthma prevalence is not only related to industrial activity but is also
impacted by factors such as high exposure to car exhaust and low-quality housing. As affluent
homeowners left the inner city, the individuals who stayed or who subsequently moved to the
inner city saw a decreased quality of public services (Trounstine, 2018) and a decline in housing
value, both of which sapped the expected wealth gains of homeownership (Akbar et al., 2022).
Additionally, low-income, particularly black households were sold dilapidated housing under
false pretenses and were faced with often unmanageable maintenance costs (Taylor, 2019). Both
declining housing values and predatory lending would impact a household’s ability to fix
structural issues, leading to housing units with characteristics negatively associated with health
outcomes. Additionally, since D-graded neighborhoods were some of the lowest valued
communities in the city, both socially and economically, it is also likely that over time, they
would be the recipient of planning decisions such as being cut by highways. They might have

117
also been overlooked for aesthetic improvements such as greenspaces and economically friendly
developments. Such decisions would both negatively impact rates of asthma and air quality.
Industrially impacted neighborhoods may have avoided such decisions over time as according to
the area description files, they had a level of respect as working class communities that many Dgraded areas lacked.
Both chapters 2 and 3 also point to the overlap of the HOLC-graded maps with minority
move in. Chapter 2 shows that areas with any black households at the time had significantly
higher rates of black populations decades later, even after fair housing laws had been
implemented. Chapter 3 suggests that one mechanism potentially linking D-graded areas to
worse asthma and air quality outcomes at rates higher than industrial areas is the higher
prevalence of black and minority individuals within D-graded areas. As mentioned,
neighborhoods with minority households were subject to discriminatory practices beyond just
credit restriction. Rather than leaning on the HOLC maps to proxy redlining, the impact of these
other practices, which include but are not limited to economic and social discrimination, need to
be given greater weight in studies of how structural racism impacts modern disparities.
An additional contribution of both chapters 2 and 3 is using data beyond just the grade
provided to better understand how past neighborhood conditions are related to future outcomes.
Regardless of whether the HOLC maps signify mortgage discrimination, they still provide a
snapshot of neighborhood conditions in the late 1930s and early 1940s, which can be useful to
understand modern-day conditions. Future work should take advantage of the language used to
describe these different neighborhoods as it gives an insight into how the real-estate industry
perceived these neighborhoods and subsequently acted within them.

118
Chapter 4 works to uncover specific neighborhood characteristics that are linked to
health-related outcomes. Neighborhood effects literature often draws from individual movement
into new neighborhoods to assess changes in outcomes. Such an approach cannot identify
impacts of specific neighborhood features, as a bundle of characteristics change when one
moves. Additionally, such research tends to ignore what may be happening in the original
neighborhood, providing no real insight into changes that could be implemented to improve
conditions of those unable or unwilling to move. Instead, this work traces a naturally occurring
intervention within a historically disadvantaged, low-income neighborhood and finds that there
are significant impacts on individuals. Namely, there is lower food insecurity, lower stress, a
greater sense of safety, and greater levels of neighborhood satisfaction.
In this way, chapter 4 shows that investments impacting the built environment of a
historically disadvantaged neighborhood can have positive outcomes for residents. One of the
biggest limitations of this paper, however, is that while it pinpoints a specific neighborhood level
intervention that can impact individual well-being, it does not analyze how individuals are
interacting with these developments. It is therefore limited in its discussion of the mechanisms
connecting the intervention to the studied outcomes. Chapter 4 does offer one testable way that
investments are impacting individuals which is by improving socio-economic status. Investments
are positively associated with both income and the likelihood of employment. This could be due
to the fact that investments themselves are increasing employment opportunities or, in the case of
residential investments, that they are changing the housing landscape in a way that improves
housing security, allowing individuals to reap the socioeconomic benefits that such security
brings. This final paper, as a kind of juxtaposition to the first two, therefore exemplifies what can
happen in an area and its residents when intentional development occurs.

119
2. Policy Recommendations
This dissertation fundamentally explores the role that neighborhood-based investment and
development plays in economic and health outcomes and has important implications for placebased versus person-based policies and interventions. Policy interventions stemming from
neighborhood effects literature often focus on housing vouchers which are designed to move
individuals into lower poverty neighborhoods (Chetty et al., 2016; Katz et al., 2001; Kling et al.,
2007). Similarly, housing programs are changing to incentivize the siting of low-income housing
developments in low-poverty, high-opportunity neighborhoods (Ellen & Horn, 2018), which also
works at the individual level by moving people out of certain neighborhoods and into others.
These program interventions often ignore what is going on in the neighborhoods being “left” and
does not consider the outcomes of individuals who cannot or do not want to move. This
dissertation focuses on place-based interventions, examining the consequences in the short and
long-term of decisions made within some of the most economically and social disadvantaged
areas.
Investment and development do not just refer to credit access, but rather reference a wide
array of spatial decision-making from city planning and public services to new construction.
Chapters 2 and 3 explore the ways in which race, ethnicity, and industrial siting impacted the
long-term trajectory of neighborhood-level characteristics. While both are rooted in historic
processes, they do have policy implications relevant to today’s landscape. Both papers suggest
that how neighborhoods and the people living within them are treated in the present can impact
these areas and their residents for decades to come. Even though more research needs to be done
to understand how specific mechanisms, specifically those occurring within D-graded
neighborhoods, are linked to future outcomes, certain lessons can still be learned. Policy today

120
should strive towards more equitable resource distribution across neighborhoods and avoid overconcentrating health averse developments such as highways and industry within certain areas.
Policy should also be particularly cognizant of potentially discriminatory real-estate and
appraisal practices which can serve to significantly disadvantage certain areas and households.
While chapters 2 and 3 analyze potentially lacking development, chapter 4 focuses on
how a historically disadvantaged area might be impacted by intentional, structural investments.
Findings here show that re-directing city funds towards renovation and/or new construction in
low-income neighborhoods could serve to benefit residents. While this dissertation does not
directly test whether such investments are the best use of funds to improve residential health and
well-being, it does provide evidence that such investments are beneficial to residents. This is
promising as development projects are perhaps easier to implement than larger structural changes
such as access to better educational or job opportunities.
Ultimately, this dissertation adds to the literature exploring connections between where a
person lives and a variety of outcomes. It suggests more nuanced ways to understand what might
be going on beneath the bundle of neighborhood effects and suggests routes that policy makers
and planners can take to address spatially engrained disparities. This work cements the idea that
where one lives matters and that neighborhoods and housing should continue to be studied to
understand and address patterns of inequality.

121
References
Akbar, P. A., Hickly, S. L., Shertzer, A., & Walsh, R. P. (2022). Racial segregation in housing
markets and the erosion of black wealth. Review of Economics and Statistics, 1-45.
Chetty, R., Hendren, N., & Katz, L. F. (2016). The effects of exposure to better neighborhoods on
children: New evidence from the moving to opportunity experiment. American Economic
Review, 106(4), 855-902.
Ellen, I. G., & Horn, K. M. (2018). Points for place: Can state governments shape siting patterns
of low-income housing tax credit developments? Housing Policy Debate, 28(5), 727-745.
Katz, L. F., Kling, J. R., & Liebman, J. B. (2001). Moving to opportunity in Boston: Early results
of a randomized mobility experiment. The quarterly journal of economics, 116(2), 607-
654.
Kling, J. R., Liebman, J. B., & Katz, L. F. (2007). Experimental analysis of neighborhood
effects. Econometrica, 75(1), 83-119.
Taylor, K.-Y. (2019). Race for profit: How banks and the real estate industry undermined black
homeownership. UNC Press Books.
Trounstine, J. (2018). Segregation by design: Local politics and inequality in American cities.
Cambridge University Press.

122
References
Aaronson, D., Faber, J., Hartley, D., Mazumder, B., & Sharkey, P. (2021). The long-run effects of
the 1930s HOLC “redlining” maps on place-based measures of economic opportunity and
socioeconomic success. Regional science and urban economics, 86, 103622.
Aaronson, D., Hartley, D., Mazumder, B., & Stinson, M. (2023). The Long-Run Effects of the
1930s Redlining Maps on Children. Journal of Economic Literature, 61(3), 846–862.
https://doi.org/10.1257/jel.20221702
Aaronson, Hartley, D., & Mazumder, B. (2021). The Effects of the 1930s HOLC “Redlining”
Maps. American Economic Journal. Economic Policy, 13(4), 355–392.
https://doi.org/10.1257/pol.20190414
Achakulwisut, P., Brauer, M., Hystad, P., & Anenberg, S. C. (2019). Global, national, and urban
burdens of paediatric asthma incidence attributable to ambient NO2 pollution: estimates
from global datasets. The Lancet Planetary Health, 3(4), e166-e178.
Agbai, C. O. (2021). Shifting neighborhoods, shifting health: A longitudinal analysis of
gentrification and health in Los Angeles County. Social science research, 100, 102603.
Akbar, P. A., Hickly, S. L., Shertzer, A., & Walsh, R. P. (2022). Racial Segregation in Housing
Markets and the Erosion of Black Wealth. The Review of Economics and Statistics, 1–45.
https://doi.org/10.1162/rest_a_01276
Appel, I., & Nickerson, J. (2016). Pockets of poverty: The long-term effects of redlining.
Available at SSRN 2852856.
Arcaya, M. C., Tucker-Seeley, R. D., Kim, R., Schnake-Mahl, A., So, M., & Subramanian, S.
(2016). Research on neighborhood effects on health in the United States: a systematic
review of study characteristics. Social Science & Medicine, 168, 16-29.
Bailey, Z. D., Krieger, N., Agénor, M., Graves, J., Linos, N., & Bassett, M. T. (2017). Structural
racism and health inequities in the USA: evidence and interventions. The lancet,
389(10077), 1453-1463.
Baird, M. D., Cantor, J., Troxel, W. M., & Dubowitz, T. (2022). Job loss and psychological
distress during the COVID‐19 pandemic: Longitudinal Analysis from residents in nine
predominantly African American low‐income neighborhoods. Health Economics, 31(9),
1844-1861.
Baird, M. D., Schwartz, H., Hunter, G. P., Gary-Webb, T. L., Ghosh-Dastidar, B., Dubowitz, T.,
& Troxel, W. M. (2020). Does large-scale neighborhood reinvestment work? Effects of
public–private real estate investment on local sales prices, rental prices, and crime rates.
Housing policy debate, 30(2), 164-190.

123
Barile, J. P., Kuperminc, G. P., & Thompson, W. W. (2017). Resident characteristics and
neighborhood environments on health‐related quality of life and stress. Journal of
Community Psychology, 45(8), 1011-1025.
Beck, A. F., Huang, B., Ryan, P. H., Sandel, M. T., Chen, C., & Kahn, R. S. (2016). Areas with
high rates of police-reported violent crime have higher rates of childhood asthma
morbidity. The Journal of pediatrics, 173, 175-182. e171.
Bernard, P., Charafeddine, R., Frohlich, K. L., Daniel, M., Kestens, Y., & Potvin, L. (2007).
Health inequalities and place: a theoretical conception of neighbourhood. Social science
& medicine, 65(9), 1839-1852.
Boustan, L. P. (2010). Was postwar suburbanization “white flight”? Evidence from the black
migration. The Quarterly Journal of Economics, 125(1), 417-443.
Brodkin, K., (1998). How Jews became white folks and what that says about race in America.
Rutgers University Press.
Bryant-Stephens, T. C., Strane, D., Robinson, E. K., Bhambhani, S., & Kenyon, C. C. (2021).
Housing and asthma disparities. Journal of Allergy and Clinical Immunology, 148(5),
1121-1129.
Chetty, R., Friedman, J. N., Hendren, N., Jones, M. R., & Porter, S. R. (2018). The Opportunity
Atlas: Mapping the Childhood Roots of Social Mobility. IDEAS Working Paper Series from
RePEc.
Chetty, R., Hendren, N., & Katz, L. F. (2016). The effects of exposure to better neighborhoods on
children: New evidence from the moving to opportunity experiment. American Economic
Review, 106(4), 855-902.
Clampet-Lundquist, S., & Massey, D. S. (2008). Neighborhood effects on economic selfsufficiency: A reconsideration of the Moving to Opportunity experiment. American
Journal of Sociology, 114(1), 107-143.
Cohen, S. (1988). Perceived stress in a probability sample of the United States.
Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A global measure of perceived stress.
Journal of health and social behavior, 385-396.
Coleman-Jensen, A., Gregory, C., & Singh, A. (2014). Household food security in the United
States in 2013. USDA-ERS Economic Research Report(173).
Cutler, D. M., Glaeser, E. L., & Vigdor, J. L. (1999). The Rise and Decline of the American
Ghetto. The Journal of Political Economy, 107(3), 455–506. https://doi.org/10.1086/250069

124
D'Alessandro, D., & Appolloni, L. (2020). Housing and health. An overview. ANNALI DI
IGIENE MEDICINA PREVENTIVA E DI COMUNITÀ, 32(5 Supple 1), 17-26.
Diez Roux, A. V. (2001). Investigating neighborhood and area effects on health. American
journal of public health, 91(11), 1783-1789.
Dubowitz, T., Ghosh-Dastidar, M., Cohen, D. A., Beckman, R., Steiner, E. D., Hunter, G. P., . . .
Sloan, J. C. (2015). Diet and perceptions change with supermarket introduction in a food
desert, but not because of supermarket use. Health Affairs, 34(11), 1858-1868.
Ellen, I. G., & Horn, K. M. (2018). Points for place: Can state governments shape siting patterns
of low-income housing tax credit developments? Housing Policy Debate, 28(5), 727-745.
Eriksson, & Ward, Z. (2019). The Residential Segregation of Immigrants in the United States
from 1850 to 1940. The Journal of Economic History, 79(4), 989–1026.
https://doi.org/10.1017/S0022050719000536
Faber, J. W. (2020). We Built This: Consequences of New Deal Era Intervention in America’s
Racial Geography. American Sociological Review, 85(5), 739–775.
https://doi.org/10.1177/0003122420948464
Fan, J., Li, S., Fan, C., Bai, Z., & Yang, K. (2016). The impact of PM2. 5 on asthma emergency
department visits: a systematic review and meta-analysis. Environmental Science and
Pollution Research, 23, 843-850.
Fischer, M. J. (2003). The relative importance of income and race in determining residential
outcomes in US urban areas, 1970-2000. Urban affairs review, 38(5), 669-696.
Fishback, P. V, LaVoice, J., Shertzer, A., & Walsh, R. (2020). Race, Risk, and the Emergence of
Federal Redlining. In National Bureau of Economic Research Working Paper Series (Vol.
28146).
Fishback, P., Rose, J., Snowden, K. A., & Storrs, T. (2022). New Evidence on Redlining by
Federal Housing Programs in the 1930s. Journal of Urban Economics, 103462.
https://doi.org/10.1016/j.jue.2022.103462
Friedman, E., Lee, B., Kalman, C., & Wilson, N. (2022). Historic racism in Kansas City affects
Today's pediatric asthma burden. Health & place, 78, 102927.
Galster, G., & Hedman, L. (2013). Measuring neighbourhood effects non-experimentally: how
much do alternative methods matter? Housing Studies, 28(3), 473-498.
Gotham, K. F. (2002). Beyond Invasion and Succession: School Segregation, Real Estate
Blockbusting, and the Political Economy of Neighborhood Racial Transition. City &
Community, 1(1), 83–111. https://doi.org/10.1111/1540-6040.00009

125
Greer, J. (2013). The Home Owners’ Loan Corporation and the Development of the Residential
Security Maps. Journal of Urban History, 39(2), 275–296.
https://doi.org/10.1177/0096144212436724
Guarnieri, M., & Balmes, J. R. (2014). Outdoor air pollution and asthma. The Lancet, 383(9928),
1581-1592.
Guenther, P. M., Casavale, K. O., Reedy, J., Kirkpatrick, S. I., Hiza, H. A., Kuczynski, K. J., . . .
Krebs-Smith, S. M. (2013). Update of the healthy eating index: HEI-2010. Journal of the
Academy of Nutrition and Dietetics, 113(4), 569-580.
Harding, D. J., Gennetian, L., Winship, C., Sanbonmatsu, L., & Kling, J. R. (2010). Unpacking
neighborhood influences on education outcomes: Setting the stage for future research.
Hellerstein, J. K., Neumark, D., & McInerney, M. (2008). Spatial mismatch or racial
mismatch?. Journal of Urban Economics, 64(2), 464-479.
Henderson, H., Child, S., Moore, S., Moore, J. B., & Kaczynski, A. T. (2016). The influence of
neighborhood aesthetics, safety, and social cohesion on perceived stress in disadvantaged
communities. American Journal of Community Psychology, 58(1-2), 80-88.
Hershberg, Burstein, A. N., Ericksen, E. P., Greenberg, S., & Yancey, W. L. (1979). A Tale of
Three Cities: Blacks and Immigrants in Philadelphia: 1850-1880, 1930 and 1970. The
Annals of the American Academy of Political and Social Science, 441(1), 55–81.
https://doi.org/10.1177/000271627944100106
Hillier, A. E. (2005). Residential Security Maps and Neighborhood Appraisals: The Home
Owners’ Loan Corporation and the Case of Philadelphia. Social Science History, 29(2),
207–233. https://doi.org/10.1215/01455532-29-2-207
Holliday, S. B., Troxel, W., Haas, A., Ghosh-Dastidar, M. B., Gary-Webb, T. L., Collins, R., . . .
Dubowitz, T. (2020). Do investments in low-income neighborhoods produce objective
change in health-related neighborhood conditions? Health & place, 64, 102361.
Jung, K. H., Pitkowsky, Z., Argenio, K., Quinn, J. W., Bruzzese, J.-M., Miller, R. L., . . .
Lovinsky-Desir, S. (2022). The effects of the historical practice of residential redlining in
the United States on recent temporal trends of air pollution near New York City schools.
Environment international, 169, 107551.
Katz, L. F., Kling, J. R., & Liebman, J. B. (2001). Moving to opportunity in Boston: Early results
of a randomized mobility experiment. The quarterly journal of economics, 116(2), 607-
654.
Kim, J. (2010). Neighborhood disadvantage and mental health: The role of neighborhood
disorder and social relationships. Social science research, 39(2), 260-271.

126
Kling, J. R., Liebman, J. B., & Katz, L. F. (2007). Experimental analysis of neighborhood
effects. Econometrica, 75(1), 83-119.
Kraus, N. T., Connor, S., Shoda, K., Moore, S. E., & Irani, E. (2023). Historic redlining and
health outcomes: A systematic review. Public Health Nursing.
Krimmel, J. (2018). Persistence of prejudice: Estimating the long term effects of redlining.
10.31235/osf.io/jdmq9
Kundu, S., & Stone, E. A. (2014). Composition and sources of fine particulate matter across
urban and rural sites in the Midwestern United States. Environmental science: Processes
& impacts, 16(6), 1360-1370.
Kwan, M.-P. (2018). The limits of the neighborhood effect: Contextual uncertainties in
geographic, environmental health, and social science research. Annals of the American
Association of Geographers, 108(6), 1482-1490.
Lane, H. M., Morello-Frosch, R., Marshall, J. D., & Apte, J. S. (2022). Historical redlining is
associated with present-day air pollution disparities in US cities. Environmental science
& technology letters, 9(4), 345-350.
Lemire, E., Samuels, E. A., Wang, W., & Haber, A. (2022). Unequal Housing Conditions And
Code Enforcement Contribute To Asthma Disparities In Boston, Massachusetts: Study
examines housing conditions, code enforcement, and asthma disparities in Boston,
Massachusetts. Health Affairs, 41(4), 563-572.
Li, J., Jahan, J., & Newcomb, P. (2023). Environmental characteristics and disparities in adult
asthma in north central Texas urban counties. Public Health, 217, 164-172.
Light, J. S. (2010). Nationality and neighborhood risk at the origins of FHA underwriting.
Journal of Urban History, 36(5). https://doi.org/10.1177/0096144210365677
Liu, J., Clark, L. P., Bechle, M. J., Hajat, A., Kim, S.-Y., Robinson, A. L., . . . Marshall, J. D.
(2021). Disparities in air pollution exposure in the United States by race/ethnicity and
income, 1990–2010. Environmental health perspectives, 129(12), 127005.
Logan, J. R. (2011). Separate and unequal: The neighborhood gap for blacks, Hispanics and
Asians in Metropolitan America. Project US2010 report, 1-22.
Macintyre, S., Ellaway, A., & Cummins, S. (2002). Place effects on health: how can we
conceptualise, operationalise and measure them? Social science & medicine, 55(1), 125-
139.
Markley, S. (2022). Tabulating Home Owners' Loan Corporation area description sheet data.
Environment and Planning B: Urban Analytics and City Science, in press.
https://doi.org/10.1177/23998083221133112

127
Martz, C. D., Hunter, E. A., Kramer, M. R., Wang, Y., Chung, K., Brown, M., . . . Chae, D. H.
(2021). Pathways linking census tract typologies with subjective neighborhood disorder
and depressive symptoms in the Black Women's Experiences Living with Lupus
(BeWELL) Study. Health & place, 70, 102587.
Massey, D. S., & Denton, N. A. (1988). The dimensions of residential segregation. Social forces,
67(2), 281-315.
Massey, D. S., & Denton, N. A. (1993). American apartheid : segregation and the making of the
underclass (pp. 170–181). Harvard University Press.
Matthews, S. A., & Yang, T.-C. (2010). Exploring the role of the built and social neighborhood
environment in moderating stress and health. Annals of Behavioral Medicine, 39(2), 170-
183.
Michney, T. M. (2022). How the City Survey’s Redlining Maps Were Made: A Closer Look at
HOLC’s Mortgagee Rehabilitation Division. Journal of Planning History, 21(4), 316–
344. https://doi.org/10.1177/15385132211013361
Michney, T. M., & Winling, L. (2020). New Perspectives on New Deal Housing Policy:
Explicating and Mapping HOLC Loans to African Americans. Journal of Urban History,
46(1), 150–180. https://doi.org/10.1177/0096144218819429
Mohai, P., Lantz, P. M., Morenoff, J., House, J. S., & Mero, R. P. (2009). Racial and
socioeconomic disparities in residential proximity to polluting industrial facilities:
evidence from the Americans' Changing Lives Study. American journal of public health,
99(S3), S649-S656.
Morello-Frosch, R., & Lopez, R. (2006). The riskscape and the color line: examining the role of
segregation in environmental health disparities. Environmental research, 102(2), 181-
196.
Moyer, R., MacDonald, J. M., Ridgeway, G., & Branas, C. C. (2019). Effect of remediating
blighted vacant land on shootings: a citywide cluster randomized trial. American journal
of public health, 109(1), 140-144.
Nardone, A., Casey, J. A., Morello-Frosch, R., Mujahid, M., Balmes, J. R., & Thakur, N. (2020).
Associations between historical residential redlining and current age-adjusted rates of
emergency department visits due to asthma across eight cities in California: an ecological
study. The Lancet Planetary Health, 4(1), e24-e31.
Ndjila, S., Lovasi, G. S., Fry, D., & Friche, A. A. (1998). Measuring Neighborhood Order and
Disorder: a Rapid.
Neumark, D., & Simpson, H. (2015). Place-based policies. In Handbook of regional and urban
economics (Vol. 5, pp. 1197-1287). Elsevier.

128
Northridge, M. E., Sclar, E. D., & Biswas, P. (2003). Sorting out the connections between the
built environment and health: a conceptual framework for navigating pathways and
planning healthy cities. Journal of urban health, 80, 556-568.
Pastor, M., Sadd, J., & Hipp, J. (2001). Which came first? Toxic facilities, minority move‐in, and
environmental justice. Journal of urban affairs, 23(1), 1-21.
Pate, C. A. (2021). Asthma surveillance—United States, 2006–2018. MMWR. Surveillance
Summaries, 70.
Perez, M. F., & Coutinho, M. T. (2021). Focus: Health Equity: An Overview of Health
Disparities in Asthma. The Yale journal of biology and medicine, 94(3), 497.
Peterson, C. E., Sastry, N., Pebley, A. R., Ghosh-Dastidar, B., Williamson, S., & Lara-Cinisomo,
S. (2004). The Los Angeles family and neighborhood survey: codebook.
Pollack, C. E., Blackford, A. L., Du, S., Deluca, S., Thornton, R. L., & Herring, B. (2019).
Association of receipt of a housing voucher with subsequent hospital utilization and
spending. JAMA, 322(21), 2115-2124.
Pollack, C. E., Roberts, L. C., Peng, R. D., Cimbolic, P., Judy, D., Balcer-Whaley, S., . . . Davis,
M. F. (2023). Association of a housing mobility program with childhood asthma
symptoms and exacerbations. Jama, 329(19), 1671-1681.
Richardson, A. S., Ghosh-Dastidar, M., Beckman, R., Flórez, K. R., DeSantis, A., Collins, R. L.,
& Dubowitz, T. (2017). Can the introduction of a full-service supermarket in a food
desert improve residents' economic status and health? Annals of epidemiology, 27(12),
771-776.
Robert K. Nelson, LaDale Winling, Richard Marciano, Nathan Connolly, et al., “Mapping
Inequality,” American Panorama, ed. Robert K. Nelson and Edward L. Ayers,
accessed May 9, 2021,
https://dsl.richmond.edu/panorama/redlining/#loc=5/39.1/94.58&text=downloads
Rosenbaum, E. (2008). Racial/ethnic differences in asthma prevalence: the role of housing and
neighborhood environments. Journal of health and social behavior, 49(2), 131-145.
Ross, C. E., & Mirowsky, J. (2001). Neighborhood disadvantage, disorder, and health. Journal of
health and social behavior, 258-276.
Rothstein, R. (2017). The color of law : a forgotten history of how our government segregated
America (First edition.). Liveright Publishing Corporation, a division of W.W. Norton
& Company.

129
Ruijsbroek, A., Droomers, M., Hardyns, W., Groenewegen, P. P., & Stronks, K. (2016). The
interplay between neighbourhood characteristics: The health impact of changes in social
cohesion, disorder and unsafety feelings. Health & place, 39, 1-8.
Saha, P. K., Hankey, S., Marshall, J. D., Robinson, A. L., & Presto, A. A. (2021). High-spatialresolution estimates of ultrafine particle concentrations across the continental United
States. Environmental Science & Technology, 55(15), 10320-10331.
Sallis, J. F., Saelens, B. E., Frank, L. D., Conway, T. L., Slymen, D. J., Cain, K. L., . . . Kerr, J.
(2009). Neighborhood built environment and income: examining multiple health
outcomes. Social science & medicine, 68(7), 1285-1293.
Sampson, R. J. (2008). Moving to inequality: Neighborhood effects and experiments meet social
structure. American journal of sociology, 114(1), 189-231.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A
multilevel study of collective efficacy. science, 277(5328), 918-924.
Santiago, A. M., Galster, G., Lucero, J., Ishler, K., Lee, E. L., Kypriotakis, G., & Stack, L.
(2014). Opportunity neighborhoods for Latino and African American children. Available
at SSRN 2563141.
Schüle, S. A., & Bolte, G. (2015). Interactive and independent associations between the
socioeconomic and objective built environment on the neighbourhood level and
individual health: a systematic review of multilevel studies. PloS one, 10(4), e0123456.
Schuyler, A. J., & Wenzel, S. E. (2022). Historical redlining impacts contemporary
environmental and asthma-related outcomes in Black adults. American Journal of
Respiratory and Critical Care Medicine, 206(7), 824-837.
Shertzer, & Walsh, R. P. (2019). Racial Sorting and the Emergence of Segregation in American
Cities. The Review of Economics and Statistics, 101(3), 415–427.
https://doi.org/10.1162/rest_a_00786
Shertzer, Twinam, T., & Walsh, R. P. (2016). Race, Ethnicity, and Discriminatory
Zoning. American Economic Journal. Applied Economics, 8(3), 217–246.
https://doi.org/10.1257/app.20140430
Simoneau, T., & Gaffin, J. M. (2023). Socioeconomic determinants of asthma health. Current
Opinion in Pediatrics, 35(3), 337-343.
Smith, R. B., & Painter, G. (2023). People and place: Does the reason for redlining impact the
long-term trajectory of neighborhoods. Journal of Urban Affairs, 1-19.

130
South, E. C., Kondo, M. C., Cheney, R. A., & Branas, C. C. (2015). Neighborhood blight, stress,
and health: a walking trial of urban greening and ambulatory heart rate. American journal
of public health, 105(5), 909-913.
Steven Manson, Jonathan Schroeder, David Van Riper, Tracy Kugler, and Steven Ruggles.
IPUMS National Historical Geographic Information System: Version 15.0 [dataset].
Minneapolis, MN: IPUMS. 2020. http://doi.org/10.18128/D050.V15.0
Subar, A. F., Kirkpatrick, S. I., Mittl, B., Zimmerman, T. P., Thompson, F. E., Bingley, C., . . .
McNutt, S. (2012). The automated self-administered 24-hour dietary recall (ASA24): a
resource for researchers, clinicians and educators from the National Cancer Institute.
Journal of the Academy of Nutrition and Dietetics, 112(8), 1134.
Swope, C. B., & Hernández, D. (2019). Housing as a determinant of health equity: A conceptual
model. Social Science & Medicine, 243, 112571.
Swope, C. B., Hernández, D., & Cushing, L. J. (2022). The relationship of historical redlining
with present-day neighborhood environmental and health outcomes: a scoping review and
conceptual model. Journal of Urban Health, 99(6), 959-983.
Taylor, K. Y. (2019). Race for profit: How banks and the real estate industry undermined black
homeownership. UNC Press Books.
Tessum, C. W., Paolella, D. A., Chambliss, S. E., Apte, J. S., Hill, J. D., & Marshall, J. D. (2021).
PM2. 5 polluters disproportionately and systemically affect people of color in the United
States. Science advances, 7(18), eabf4491.
Trounstine, J. (2018). Segregation by design: Local politics and inequality in American cities.
Cambridge University Press.
Venecek, M. A., Yu, X., & Kleeman, M. J. (2019). Predicted ultrafine particulate matter source
contribution across the continental United States during summertime air pollution events.
Atmospheric Chemistry and Physics, 19(14), 9399-9412.
Wilson, W. J. (2012). The truly disadvantaged: The inner city, the underclass, and public policy.
University of Chicago Press.
Xu, W. (2022). Legacies of Institutionalized Redlining: A Comparison Between Speculative and
Implemented Mortgage Risk Maps in Chicago, Illinois. Housing Policy Debate, 32(2), 249–
274. https://doi.org/10.1080/10511482.2020.1858924
Yang, T.-C., & Matthews, S. A. (2010). The role of social and built environments in predicting
self-rated stress: A multilevel analysis in Philadelphia. Health & place, 16(5), 803-810.

Abstract (if available)

Abstract This dissertation is comprised of three essays that analyze both modern-day and historical neighborhood level processes with the goal of peeling back the myriad of characteristics that make up a neighborhood and pinpointing how specific mechanisms link to outcomes. Chapters 2 and 3 dive into redlining maps created by the Home Owner's Loan Corporation (HOLC) to both test whether these maps are indicative of systematic mortgage discrimination and to suggest ulterior ways this data may be useful to link past neighborhood characteristics with modern-day outcomes. Specifically, chapter 2 explores heterogeneity in homeownership rates and racial composition according to a D-graded area's initial population composition. Chapter 3 focuses on how HOLC grades compare to past industrial activity as predictors of future asthma prevalence and air quality. Chapter 4 departs from a historical lens by analyzing the impacts of economic development on individual-level health-related outcomes within two historically disadvantaged neighborhoods. These papers suggest nuanced ways of understanding the relationship between neighborhood-level characteristics and individual outcomes, while reaffirming that such characteristics are integral to our understanding of current population-level disparities.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Spatial dimensions of stratification: neighborhood change, urban inequality, and the neighborhood-school link in the U.S.

PDF

The role of public policy in the decisions of parents and caregivers: an examination of work, fertility, and informal caregiving

PDF

The economy-environment trade-off in China's air pollution policies

PDF

More knowledge, better plans: a study of heterogeneity in dementia prevalence and mid-life cognitive changes

PDF

Outcomes-based contracting through impact bonds: ties to social innovation, systems change, and international development

PDF

Affordable south Los Angeles: survival, support, and different futures

PDF

A better method for measuring housing affordability and the role that affordability played in the mobility outcomes of Latino-immigrants following the Great Recession

PDF

The risks and rewards of city management: how city managers evaluate the nature of the job and compensation

PDF

Redlining revisited: spatial dependence and neighborhood effects in mortgage lending

PDF

Changing workplace and inequality in work-family policies

PDF

Three essays on the political economy of federal spending

PDF

The interactions between housing and business

PDF

The impact of social capital: a case study on the role of social capital in the restoration and recovery of communities after disasters

PDF

Innovation and good intentions: evaluations of three cross-sectoral programs for at-risk populations in Southern California

PDF

The economic and political impacts of U.S. federal carbon emissions trading policy across households, sectors and states

PDF

The dollars and sense of caregiving: exploring the financial financial realities of providing informal care to an older adult

PDF

Porridge people: the heirloom grains of the past are fueling the porridge revolution of the present

PDF

Essays on fiscal outcomes of cities in California

PDF

Household mobility and neighborhood impacts

PDF

Housing market crisis and underwater homeowners

Asset Metadata

Creator Smith, Rebecca Brooks (author)

Core Title Redlining, neighborhood change, and individual outcomes: an exploration of how space shapes the landscape of inequality from past to present

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Public Policy and Management

Degree Conferral Date 2024-05

Publication Date 05/17/2024

Defense Date 05/17/2024

Publisher Los Angeles, California (original), University of Southern California (original), University of Southern California. Libraries (digital)

Tag Health policy,Housing policy,neighborhoods,OAI-PMH Harvest,racial disparities,redlining

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Painter, Gary (committee chair), Faber, Jacob (committee member), Owens, Ann (committee member)

Creator Email rbsmith@usc.edu,rbsmith8@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113939973

Unique identifier UC113939973

Identifier etd-SmithRebec-12938.pdf (filename)

Legacy Identifier etd-SmithRebec-12938

Document Type Dissertation

Format theses (aat)

Rights Smith, Rebecca Brooks

Internet Media Type application/pdf

Type texts

Source 20240517-usctheses-batch-1153 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu