The_Lingering_Legacy_of_Redlining_on_School_Funding__Diversity__and_Performance_

1

The Lingering Legacy of Redlining on School Funding, Diversity, and Performance

Christopher Cleveland

Harvard University

Between 1935-1940 the Home Owners' Loan Corporation (HOLC) assigned A (minimal risk) to

D (hazardous) grades to neighborhoods that reflected their lending risk from previously issued

loans and visualized these grades on color-coded maps. These maps arguably influenced

mortgage lenders to provide or deny home loans within residential neighborhoods. In this study,

we leverage a spatial analysis of 144 HOLC-graded core-based statistical areas (CBSAs) to

understand how HOLC maps relate to current patterns of district and school funding, school

racial diversity, and school performance. We find that districts composed of D neighborhoods

have less district per-pupil total revenues, but schools in D neighborhoods have higher per-pupil

2

1 Introduction

The United States has a long history of racially discriminatory policies and practices.

Redlining, where the Home Owners’ Loan Corporation (HOLC) assigned A-D security ratings to

of HOLC A-D security ratings on social and economic outcomes has become increasingly well

documented by social scientists. While much of the literature today shows HOLC maps’ adverse

effects on modern outcomes such as credit scores, family structure, homeownership, home

values, household income, neighborhood segregation, and incarceration (Appel & Nickerson,

2016; Krimmel, 2018; Anders, 2018; Aaronson et al. 2020; Aaronson et al., 2021), very few

century, often resultant from historical discriminatory practices, is predictive of present-day

educational outcomes. With recent work by Aaronson et al. (2020) and Aaronson et al. (2021)

establishing a causal relationship between HOLC maps and neighborhood segregation, housing,

and socioeconomic outcomes, it would be surprising if these results did not spill over to

measures of educational quality given the deep-rooted connection between neighborhoods and

the schools and districts that serve them (Massey & Denton, 1993; Aaronson, 1998; Chetty et al.,

mortgage insurance. While historians agree that the HOLC maps influenced the FHA maps, it is unknown to what degree since all but two FHA

maps (Chicago, IL; Greensboro, NC) have been lost or destroyed (Light, 2010).

3

2016; Bayer et al., 2020; Dalane and Marcotte, 2020). To this end, our paper addresses the

following questions: (1) How do historic HOLC D neighborhoods relate to existing patterns of

district-level and school-level funding, student racial diversity, and student performance? (2) Do

1940 HOLC A-D grades and educational outcomes.

To link historical HOLC maps to contemporary educational institutions, we map 1935-

1940 HOLC A-D neighborhood grades to present-day schools and districts. For schools,

individual HOLC A-D grades are assigned based on overlapping school-level latitude and

longitude geolocations and HOLC A-D geospatial polygons. For districts, we determine HOLC

various plots and statistical tests.

Our analysis identifies important findings. First, for district-level funding, present-day

districts located in D neighborhoods have less district-level per-pupil total revenues but higher

per-pupil federal and state revenues than those mapped to HOLC A, B, and C security ratings.

However, the differences in per-pupil federal and state revenues are not large enough to

overcome the sizeable gaps in per-pupil local funding that favor those districts mapped to higher

HOLC security ratings. These local, state, and federal dynamics drive the differences in per-pupil

total revenue by HOLC A-D grades. Conversely, at the school level, we find that schools

mapped to HOLC D ratings have, on average, more per-pupil total expenditures and per-pupil

4

federal expenditures than schools mapped to HOLC A, B, and C security ratings. Differences in

aggregated per-pupil state and local expenditures between D vs. A, B, and C schools are

statistically insignificant and signal the countervailing force that state funding has on local

largest federal funding program for U.S. public schools. This finding provides a plausible

mechanism for why HOLC D schools, regardless of what HOLC A-D district they fall within,

have the most per-pupil funding of all HOLC A-D mapped schools. Furthermore, our district-

level results remain pertinent as they represent an upper limit on how much a district can allocate

to their most-in-need schools (i.e., D schools) while still maintaining funding levels for their

diverse student populations than their more highly rated HOLC A, B, and C counterparts,

nationally and by region. While schools in higher-rated HOLC A, B, and C neighborhoods have

larger shares of White student bodies, they also have, on average, more diverse student

populations. Even more, school diversity is monotonically increasing in HOLC rating, such that

A schools have the greatest levels of diversity, followed by B, C, and ending with D.

Third, for student performance, we find that schools today located in historic HOLC D

areas have worse school-level average math and reading scores than their more highly rated A,

B, and C peers, nationally and by region. We also observe minor differences in either measure of

average learning rates or educational opportunity changes by A-D HOLC grade. These results

5

are true nationwide and by region. Finally, we document positive time trends for the finance and

diversity outcomes across all HOLC A-D grades from the late 1980s to today, but persistent and

widening gaps between schools in historically redlined D neighborhoods and those in A, B, and

1935-1940 HOLC maps and modern district and school-level outcomes. Overall, our paper

provides some of the first evidence documenting the association between historic HOLC A-D

grades, which captured neighborhood inequality in the 1930s and subsequently contributed to

neighborhood inequality in the following generations, and modern-day educational outcomes.

While we cannot definitively say whether HOLC redlining caused the modern-day educational

neighborhoods when designing modern education interventions focused on improving the life

outcomes of students of color and students from socioeconomically disadvantaged communities.

The remainder of this paper is as follows. Section 2 provides an overview of the history

of HOLC maps, the current set of literature that links HOLC security ratings to social and

economic outcomes, and related literature on neighborhoods and schools. Section 3 describes the

data used for this research. Section 4 and Section 5 outline this paper's methods, including

analysis samples construction, and details the analytic approach. Section 6 provides the results

and a discussion. Section 7 concludes.

2 Literature

6

2.1 HOLC Maps History

In 1932, the Federal Home Loan Bank Board (FHLBB) was created to manage federal

savings and loan associations. In 1933, under the FHLBB, the Home Owners’ Loan Corporation

they created a systematic appraisal process that included neighborhood-level characteristics such

as race, ethnicity, immigration status, household income, homeownership rates, access to public

services, and occupation type (Hillier, 2003; Crossney & Bartelt, 2005; Fishback et al., 2020).

Between 1935 and 1940, HOLC’s department of Research and Statistics used thousands of

realtors, developers, lenders, and appraisers to create neighborhood-by-neighborhood security

neighborhoods. These grades were later solidified as color-coded maps in which D-graded areas

were colored red and prompted the later coinage of the term “redlining.”

Each neighborhood had a standardized assessment sheet used to assign the HOLC grades.

For example, area descriptions used in Los Angeles County in 1939 included eight sections. 1)

Population asked for a record of whether the population was increasing, decreasing, or static. It

also asked for the class and occupation of residents, the percentage of foreign families and their

nationalities, the percentage of Black families, and whether the population trends reflected

shifting or infiltration. 2) Buildings asked for the type and size of the building, construction,

average age, repair status, occupancy rate, owner-occupied, 1935 price bracket, 1937 price

7

bracket, 1939 price bracket, sales demand, predicted price trend, 1935 rent bracket, 1937 rent

bracket, 1939 rent bracket, rental demand, and predicted rent trend. 3) New Construction

recorded the number of new properties built within the past year, the prices for these units, and

Characteristics of Area captured other qualitative detail about the terrain and population.

A current debate exists as to whether HOLC A-D grade assignments were racially biased

or merely a geographic snapshot of an outcome caused by America's long history of racial

discrimination before it (Fishback et al., 2020). There is evidence that race, immigration status,

household income, and ethnicity were often explicit factors in the HOLC grading process, such

standards of the FHA and other private lenders is an ongoing debate. Some authors argue that

access to HOLC maps was limited, while others argue that access to HOLC maps was ubiquitous

and materially influenced public and private lending policies and practices (Jackson, 1980;

Hillier, 2003; Light, 2010; Woods, 2012). Regardless, it is clear that HOLC encouraged the

general rule of using maps to classify the creditworthiness and lending risks of neighborhoods

and broadly are considered to have redirected public and private capital and homeownership for

intergenerational wealth-building to native-born white families and away from African American

and immigrant families (Appel & Nickerson, 2016; Krimmel, 2018; Anders, 2018; Aaronson et

8

al., 2020; Aaronson et al., 2021).

2.2 Modern Outcomes Linked to HOLC Maps

A literature base that leverages both descriptive and causal empirical strategies is

HOLC eight decades ago are low-to-moderate income (~74%) and minority neighborhoods

(~64%) today. Additionally, the authors find greater economic inequality, higher levels of

interaction between Black and White residents, and a stronger positive association between

gentrification and economic change in neighborhoods rated “Hazardous” by HOLC in their

sample. Regionally, the South showed the smallest change in the HOLC-rated “Hazardous”

Three papers have highlighted the impacts of the HOLC maps on housing and

socioeconomic outcomes. Appel and Nickerson (2016) use a spatial regression discontinuity

boundary design to study long-term impacts on home prices. They show that housing

characteristics varied smoothly at the boundaries when the maps were created. Despite this initial

smoothness, they find that HOLC "redlined" neighborhoods have about 4.8% lower home prices,

fewer owner-occupied homes, and more vacant buildings relative to adjacent areas. Similarly,

Aaronson, Hartley, and Mazumder (2020) use a spatial RD boundary propensity score design to

study the effects of the HOLC A-D maps on the long-run trajectories of neighborhoods. They

find that the maps led to reduced homeownership rates, house values, and rents and increased

9

racial segregation in later decades. They conclude that the HOLC maps had sizeable and

persistent impacts on urban-neighborhood development driven by reduced credit access and the

subsequent neighborhood disinvestment (Aaronson et al., 2020). In a follow-up paper, Aaronson,

and an identification strategy like Aaronson et al. (2020), they find sizeable and statistically

significant differences across all outcomes that favor higher-rated HOLC neighborhoods over

lower-rated ones.

Other papers have explored a broader set of outcomes connected to the HOLC maps.

Jacoby, Dong, Beard, Wiebe, and Morrison (2018) use descriptive spatial analysis to evaluate the

of Philadelphia (Beard et al., 2017; Jacoby et al., 2018). Nardone, Casey, Rudolph, Karasek,

Mujahid, and Morello-Frosch (2020) descriptively explore how birth outcomes within California

vary based upon HOLC grade. The authors find that worse HOLC ratings are associated with

adverse birth outcomes. However, their findings are inconsistent when using propensity score

matching and stratifying by metropolitan area (Nardone et al., 2020). Hoffman, Shandas, and

Pendleton (2020) use descriptive spatial analysis to study the association between a

neighborhood's HOLC grade and modern-day surface temperatures. They find that for most areas

in their study, previous HOLC redlined areas have consistently elevated land surface

temperatures compared to non-redlined areas. Cities in the Southeast and Western regions

16

5 Empirical Strategy for Outcomes

Across our outcomes, our primary empirical strategy is inspired by a difference of means

approach taken by Hoffman, Shandas, and Pendleton (2020), which documents the association of

redlined HOLC D neighborhoods differ from those found in historically HOLC A, B, or C

neighborhoods.

This approach allows us to quantify, for a given urban region, how much better

or worse off schools or districts with a particular HOLC grade assignment (e.g., D) are from the

set of all other HOLC schools or districts (e.g., A, B, C) in that urban region. Given the

heterogeneity in state funding mechanisms across the U.S. (which have implications for school

invariant covariates at the city-level are differenced out in our model, which we present below:

schools, 𝐻𝑂𝐿𝐶𝐴, 𝐻𝑂𝐿𝐶𝐵, and 𝐻𝑂𝐿𝐶𝐶 are the estimated average differences in the outcome 𝑌

between schools in what once were HOLC D graded neighborhoods and their contemporaries

and district outcomes to HOLC ratings. In addition, instead of calculating a “delta” variable that consists of demeaned HOLC A-D averages

within a given city by their respective city-wide average, we use CBSA fixed effects.

above section for details. We explored alternative fixed effects specifications including state fixed effects or no fixed effects but are comfortable

17

located in what once were HOLC A, B, or C HOLC graded neighborhoods, respectively. For

districts, the interpretation is similar but with the caveat that HOLC A-D grades represent HOLC

A-D weighted averages, which could but often do not refer to a single HOLC A-D polygon, but

time-variant determinants of our outcomes and (2) unobservable determinants of our outcomes.

yet exist, selection bias remains a threat if past HOLC A-D neighborhood grade assignments

captured pre-existing inequality between neighborhoods variation (e.g., homeownership rates,

home values, access to public services, crime) and is a determinant of our modern-day

educational outcomes. Notably, Aaronson et al. (2020) find evidence that this is indeed the case

while also presenting causal evidence in Aaronson et al. (2021). As indicated in the model

specification, we do not use additional covariates in these models. We explored using covariates

for family income, percent Black, and building age available for a smaller sample of HOLC

polygons as produced by Markley (2021) and did not find the incorporation of these covariates

changed our findings. Therefore, we chose the models without covariates to preserve sample

size.

student performance outcomes s = 3,906 versus 5,124 without covariates

18

their conceptions of neighborhood quality and modern-day educational outcomes.

For the student diversity outcomes, in addition to raw differences in percentages of

different racial groups, we also analyze each school's Simpson's Diversity Index (1-D) for our

diversity (Simpson, 1949; Hirschman, 1964).

6 Results

We present our results by the three overarching outcomes addressed in this paper: (1)

district and school-level finance, (2) school-level racial diversity, and (3) school-level student

performance.

In each section, we provide and discuss the regression results from our CBSA FE

three decades.

6.1 District and School Finance

6.1.1 Current Outcomes – Districts

In Table 2 Panel A, we provide nationwide results of our district-finance outcomes. We

see that, on average, total per-pupil revenues are often lower for districts mapped to HOLC D

grades than those mapped to A, B, and C grades. Nationwide, districts mapped to HOLC D

grades have, on average, $14,402 per-pupil total revenues with differences greatest between A

neither contains data from the 1980s or 1990s.

19

vs. D districts ($1,546, SE = $822) and least between B vs. D districts ($992, SE = $536). All

these differences are statistically significant. At the federal level, districts mapped to HOLC D

grades have, on average, $1,201 in per-pupil federal revenues, with A (-$722, S.E. = $139) and B

per-pupil state funding, with differences largest between A vs. D districts ($3,670, S.E. = $752)

and smallest between C vs. D districts (-$461, S.E. = $495). All but the C vs. D average

comparisons are statistically significant.

Finally, at the local level, we see that across the U.S., districts mapped to HOLC D

grades have, on average, $5,240 in per-pupil local revenues, which is less than those districts

significantly less per-pupil local funding than higher-rated HOLC A, B, and C districts in the

sample. This gap in local funding is abated by redistributive federal and state funds that favor D

mapped districts relative to A, B, and C districts; however, non-local educational funding is not

enough to overcome the sizeable initial gap driven by differences at the local level.

6.1.2 Longitudinal Outcomes - Districts

We expand our analysis to look at how, if at all, the relationship between HOLC A-D

grades and the finance outcomes changed over time. This exercise reduces our original cross-

sectional district-level analysis sample (d = 1,760) by a third (dp = 1,109).

we perform robustness checks across samples. While the distribution of HOLC A-D grades across districts is slightly different between the 2017-

20

In Figure 2, we show A-D averages across time for each educational finance outcome,

weighted by district enrollment. Each outcome is adjusted for inflation and denominated in 2018

USD. Starting with average per-pupil total revenue, one can see near-parallel lines across time

across time is encouraging, it is less so after considering level differences in average per-pupil

total revenue between D districts and their A, B, and C counterparts (Table 2 Panel B). Here we

see an initial and subsequently increasing gap in per-pupil total revenue that favors districts

located in historically non-redlined neighborhoods (i.e., A, B, C). Thus, while growth rates are

similar across HOLC A-D security ratings, per-pupil funding gaps are not. This result is a direct

in the late 1980s and smaller growth rates.

In contrast to per-pupil total and local revenue trends, per-pupil state and federal revenues

favor those districts with lower HOLC security ratings (i.e., C, D) than those with higher (i.e., A,

B). Like our 2017-18 cross-sectional findings, state and federal redistributive policies appear to

benefit those districts most in need and have consistently done so over the three decades we

consider here. For per-pupil state revenues, funding gaps in the late 1980s favor C and D districts

relative to A and B districts. Growth rates lead to C and D district convergence, with B districts

finance outcome regressed on an A-D HOLC indicator variable. This result suggests no statistically significant differences between the weighted

average A-D HOLC grades across these two samples for our F-33 educational finance variables.

21

remaining mostly in parallel; however, there is a distinct increase in the gaps between B, C, and

D districts and their highest-rated A counterparts. These results are further confirmed in Table 3,

where there is limited variation across years for D vs. C and D vs. B comparisons but a clear

funding from the American Recovery and Reinvestment Act of 2009 (ARRA) (U.S. Department

of Education, 2009).

[Insert Figure 2 here]

6.1.3 Current Outcomes – Schools

HOLC A, B, and C schools are monotonically decreasing in HOLC security rating, such that

gaps in per-pupil total expenditures are widest between schools mapped to A and D HOLC

grades (-$1,539, S.E. = $852) and smallest between schools mapped to C and D HOLC grades (-

$882, S.E. = $334). These differences are all statistically significant.

While the 2018-19 NERDS data affords us school-level finance data, it is more coarsely

disaggregated relative to the NCES F-33 district finance data, limiting our ability to deconstruct

the underlying variation per-pupil total expenditures into cleanly delineated federal, state, and

local buckets. That said, we can isolate per-pupil federal expenditures and investigate the

combined sum of state and local expenditures. Like the district-level finance results, we find that,

22

on average, schools mapped to D schools have more per-pupil federal expenditures than A, B,

and C schools. These differences are all statistically significant and once again are largest

between A and D schools (-$602, S.E. = $39) and smallest between A and C schools ($141, S.E.

patterns in the average differences between D vs. A, B, C schools as we do for total and federal

per-pupil expenditures; however, not all differences are statistically significant, namely, A vs. D

differences (-$1,131, S.E. = $860). That said, average differences between D vs. B (-$1,010, SE

= $590) and C (-$745, SE = $323) schools are statistically significant. Negative per-pupil state

and local averages mean one of two things – per-pupil local expenditures outpace per-pupil state

higher HOLC A, B, or C grades relative to those mapped to redlined HOLC D grades, we argue

the latter is most realistic.

6.1.4 Reconciliation of District and School Finance Results

In this section, we attempt to reconcile the differences in the district and school-level

results by proposing hypotheses and discussing their likelihood. Starting with the first, the

differences in the district and school-level results may simply be an artifact of the outcomes

chosen since the district-level analysis uses per-pupil revenues and the school-level analysis uses

per-pupil expenditures.

If notable variation exists between funding allocations versus reported

it. NCES F-33 expenditures do not provide this level of detail.

23

spending in non-uniform ways by HOLC A-D security rating, one would expect to see these

differences play out in our regression results. When we substitute district per-pupil total

expenditures for district per-pupil total revenues, our district-level results are indistinguishable

strategies we use to assign HOLC grades. Recall, for districts, HOLC A-D mapping is based on

the weighted average of the portions of HOLC A-D polygons contained within that district;

however, for schools, HOLC A-D mapping is based on a one-to-one match of an individual

school to a HOLC grade. By construction, a district with a given HOLC grade assignment will

contain the largest share of that HOLC polygon type, in square miles, relative to other HOLC

schools cluster in HOLC polygons different from the district HOLC A-D mapping, the

distribution of HOLC school mappings will not accurately reflect the HOLC polygons that make

up that district.

[Insert Figure 3 here]

To check this, we construct the underlying HOLC A-D school distributions of each

HOLC A-D district. In Figure 2, notable patterns emerge. First, higher (lower) rated HOLC

districts have larger shares of higher (lower) rated HOLC schools, with the greatest being that of

our district and school mapping strategies, the district will be assigned a HOLC A grade, whereas schools will be assigned B-D grades. A

scenario such as this could explain our district and school-level finance results.

24

the HOLC A-D school rating congruent with the HOLC A-D district rating. For example,

districts mapped to HOLC A security ratings have the greatest share of HOLC A schools in that

district. Alternatively, districts mapped to HOLC D security ratings have the greatest share of

A. Conversely, HOLC A districts have the largest share of A mapped schools, second of B, third

of C, and the least of D. This pattern holds for B and C districts. These results validate our

mapping strategies, and suggest schools are not clustering in polygons discordant from the

mapped district HOLC A-D grade.

A final likely explanation is that districts are simply allocating resources to those most in

before. We find that nationally, schools mapped to HOLC D grades have, on average, 79.3

percent of students qualifying for free and reduced-price lunch, where percentage point

differences are greatest between A vs. D schools (-0.37, S.E. = 0.02) and smallest between A vs.

C schools (-0.06, S.E. = 0.01). We observe a similar pattern across all regions. All differences

between D vs. A, B, and C schools are statistically significant nationwide and by region.

Combined with the district and school finance results, these findings suggest that districts,

regardless of their A-D assigned grade, systematically target and allocate more funding to

schools located in historically HOLC D neighborhoods because those schools today serve the

largest shares of students from low-income households. These results can be seen as an extension

25

of recent research that finds redlined HOLC D neighborhoods are worse off, both in terms of

homeownership rates and home values, relative to their higher-rated peers (Aaronson et al.,

2020) and provide some of the first evidence of secondary ripple effects that stem from the

need schools (i.e., D schools) while still maintaining adequate funding levels for all other schools

they serve. Given that we find that districts’ HOLC A-D mapping is predictive of their total per-

pupil funding, such that districts mapped to higher-rated HOLC grades have, on average, more

per-pupil total funding than those mapped to lower-rated HOLC grades, one might expect lower-

rated HOLC schools in higher-rated HOLC districts to receive and spend more money relative to

support this theory, the CBSA FE regression results do not, which is likely an artifact of our

smaller subsample and less within-CBSA variation in HOLC A-D districts. Zooming out to the

state and regional levels, our results from the state F.E. and region F.E. support this narrative and

align with the distributional patterns in Figure 3.

This evidence suggests that districts with more resources in our sample (i.e., A, B, C),

relative to those with less (i.e., D), distribute more to their most-in-need schools (i.e., D schools).

when clustering at the regional level but remain statistically significant for D vs. C differences.

32

and shrink to 0.54 SD units (S.E. = 0.10) or around 1.6-grade levels in the Northeast (Table

S0.4). Comparing B and D schools paints a similar picture to above, albeit muted, with schools

separated by 0.34 SD units (S.E. = 0.03) or just above one grade level. These gaps favoring B

comparisons, with gaps shrinking to single digits nationwide and across regions. Overall, C and

D schools are separated by 0.13 SD units (S.E. = 0.02) or just under one-half grade level.

Differences in educational opportunity are largest for the South region with a gap of 0.15 SD

units (S.E. = 0.05), while the Northeast gap is smallest at 0.10 SD units (S.E. = 0.02) (Table

S0.4). Both nationwide and by region, these differences in D vs. A, B, and C educational

cannot speak to whether gaps have risen or fallen over the past several decades. Without a

complete accounting of these trends over time, we cannot measure progress nor bring historical

context to bear. For example, suppose trends in educational opportunity favored D vs. A, B, and

C schools over the past half-century. In that case, one might view the current gaps in this

outcome as a historical lower bound and vice versa for a historical upper bound. Unfortunately,

apart from the current 2009-2018 SEDA panel, we lack historical data on educational

performance measures. Even so, we can make prognostications on what might be if the status

quo remains. Given the large gaps in educational opportunity by HOLC A-D security rating and

the near-zero S.D. unit changes in it for each HOLC A-D grade, the educational opportunity gap

33

is expected to remain unabated into the future. The equality exhibited in average learning rates

and average educational opportunity changes by HOLC A-D grade, which standalone might be a

positive finding, will lead to a continued inequality in average educational opportunity across

in educational opportunity would remain constant over time, a positive average change in

educational opportunity uniform across A-D HOLC grades could, in time, raise all schools to and

above the minimum educational opportunity threshold. Unfortunately, our findings do little to

support this claim, as changes in educational opportunity, while positive, are small and often fall

below 0.01 SD units.

banks, and mortgage lenders evaluated the creditworthiness and risk of home loans and mortgage

insurance within residential neighborhoods over the next several decades. With the release of

newly digitized HOLC A-D maps from the University of Richmond lead “Mapping Inequality

Project,” there has been a recent surge in research quantifying the negative impacts of redlining

on long-term social and economic outcomes (Appel & Nickerson, 2016; Krimmel, 2018; Anders,

2018; Aaronson et al., 2020; Aaronson et al., 2021). However, this effort has yet to extend to K-

12 public school educational outcomes to the best of our knowledge.

This paper examines the relationship between historic HOLC A-D maps and modern-day

district and school funding patterns, racial diversity, and student performance. We employ a

34

novel mapping strategy that links 1935-1940 HOLC A-D neighborhood grades to present-day

districts and schools. At the district level, we find those mapped to historic HOLC D grades have

the least favorable overall and local district-finance outcomes relative to those mapped to higher-

differences. For example, we find those districts mapped to historic HOLC D grades have the

most favorable state and federal district-finance outcomes today. These findings show a

redistributive system targeting districts with higher percentages of students eligible for free or

reduced-price lunch. However, the results also suggest that past neighborhood inequality lingers

well into the future. For example, those districts that receive redistributive funding to equalize

HOLC A-D grades and funding at the school level. Schools mapped to the worst HOLC grades

(i.e., D) have the most favorable school-finance outcomes than their higher-rated counterparts

(i.e., A, B, and C). In reconciling the district and school-level finance results, we find that D

schools have the greatest share of students eligible for free or reduced-price lunch relative to A,

B, and C schools. Together, the district and school-level findings suggest that districts, regardless

of their HOLC A-D grade, systematically target and allocate more money to schools in historic

HOLC D neighborhoods and do so because these schools serve the largest shares of students

from low-income households. Finally, these results persist across time, with overall positive time

grade polygon areas.

35

trends in outcome measures regardless of HOLC A-D grade but widening gaps between D vs. A,

B, and C districts.

Overall, these results align with recent research that finds HOLC negatively impacted the

HOLC D neighborhoods (and therefore mapped to HOLC D grades in our paper) should have

lower assessed property values, lower property tax bases, and less local funding than those

districts composed primarily of A, B, and C HOLC neighborhoods. In general, any local funding

shortfalls are addressed through targeted state and federal redistributive funding programs (e.g.,

Title I, IDEA) that allocate dollars to low-income districts to reduce funding gaps and better

hypothesized mechanism underlying them hints at a lingering historical legacy of redlining,

where HOLC neighborhood grades assigned in 1935-1940 predict local funding gaps today and

where federal and state funds are needed to equalize funding for districts composed of primarily

D neighborhoods.

We also find those schools located today in historically redlined residential

neighborhoods to have, on average, larger shares of Black and non-White student bodies and less

diverse student populations. These differences are persistent and growing over time for the racial

diversity outcomes, albeit for a smaller, less representative sample. However, while A assigned

public schools have the highest percent White student populations, they also exhibit the highest

36

student racial diversity levels via Simpson’s Diversity Index. That said, A schools have the

lowest Exposure Index values across all HOLC A-D grades and race-group pairings (i.e., White-

Asian, White-Black, White-Hispanic, White-Non-White).

percent, whereas shares of Hispanic 5-17-year-olds have increased to 25 percent from 16 percent

(NCES, 2019). These public-school demographic shifts mirror the increasing racial diversity of

the U.S. population, driven in part by diversifying urban demographics resulting from

nationwide migration patterns that brought White families into cities from the suburbs and Black,

Hispanic, and Asian families out to them (Wells et al., 2020). Even so, more diverse populations

between Whites and Non-White students has decreased since the 1990s, with Non-White

students attending schools with fewer shares of White students (Fiel, 2013).

countervailing forces could overwhelm, or at a minimum, limit the benefits that a more diverse

U.S. population has on school racial diversity.

Finally, we also find that schools located today in what once were historically redlined

areas have worse average ELA and math scores. However, there is no difference in both average

learning rates and trends in test scores across A-D schools. Notably, these findings only provide

group segregation. This is reflected in the negative trends in percent White and positive trends in percent Hispanic and percent Non-White we

present in this paper.

37

a snapshot of present-day differences in educational performance measures by HOLC A-D

grade. They do not lend insight into whether gaps by HOLC A-D grade have abated over time.

That said, looking forward, if the status quo continues, we will predict the gap in average test

maps with modern educational outcomes and highlights the transmission of past neighborhood

inequality to the present. In addition, these results suggest that education policymakers need to

consider the historical implications of past neighborhood inequality on present-day

neighborhoods when designing and implementing complex modern interventions that target

inequitable outcomes between students of different socioeconomic and racial groups.

38

References

Aaronson, D. (1998). Using Sibling Data to Estimate the Impact of Neighborhoods on Children’s

Educational Outcomes. The Journal of Human Resources, 33(4), 915–946.

Aaronson, D., Hartley, D., Mazumder, B. (2020). The Effects of the 1930s HOLC “Redlining”

Maps. Federal Reserve Bank of Chicago. Working Paper.

Aaronson, D., Faber, J., Hartley, D., Mazumder, B., & Sharkey, P. (2021). The long-run effects

39

Clotfelter, C.T., Hemelt, S.W., Ladd, H.F., & Turaeva, M. (2019). School segregation in the era

of immigration, school choice and color-blind jurisprudence – the case of North Carolina.

EdWorkingPaper No. 19-101. Providence, RI: Annenberg Institute at Brown University.

Retrieved March 1, 2021, from https://www.edworkingpapers.com/ai19-101.

Clotfelter, C.T., Ladd, H.F., Clifton, C.R., & Turaeva, M.R. (2021). School Segregation at the

Classroom Level in a Southern ‘New Destination’ State. Race and Social Problems,

EdWorkingPaper No. 20-338. Providence, RI: Annenberg Institute at Brown University.

Retrieved March 1, 2021, from https://www.edworkingpapers.com/ai20-338.

Edunomics Lab (2021). NERD$: National Education Resource Database on Schools (Version

1.0). Georgetown University. Retrieved 06/2021 from https://edunomicslab.org/nerds/

Fiel, J.E. (2013). Decomposing School Resegregation: Social Closure, Racial Imbalance, and

Racial Isolation. American Sociological Review, 78(5), 828-848.

Filardo, M., Allen, M., Huvendick, N., Sung, P., Garrison, D., Turner, M. A., Comey, J.,

Williams, B., & Guernsey, E. (2008). Quality Schools, Healthy Neighborhoods, and the

Future of DC: Policy Report. In 21st Century School Fund. 21st Century School Fund.

Fishback, P., LaVoice, J., Shertzer, A., Walsh, R. (2020). Race, Risk, and the Emergence of

Hillier, A. E. (2003). Redlining and the Home Owners’ Loan Corporation. Journal of Urban

Hirschman, A. O. (1964) The paternity of an index. The American Economic Review, 54(5), 761-

Hoffman, J. S., Shandas, V., & Pendleton, N. (2020). The Effects of Historical Housing Policies

8(1), 12.

6(4), 419–452.

impact of historical and structural racism on urban violence in Philadelphia. Social

Katz, B. & Turner, M.A. (2009). Rethinking U.S. Rental Housing Policy. In Opportunity 08 (2nd

Krimmel, J. (2018). Persistence of Prejudice: Estimating the Long Term Effects of Redlining.

Light, J. (2010). Nationality and Neighborhood Risk at the Origins of FHA Underwriting.

40

Lutz, B. (2011). The end of court-ordered desegregation. American Economic Journal. Economic

Policy, 3(2), 130-168.

Markley, S. (2021, September 25). Tabulating HOLC Area Description Sheet Data.

https://doi.org/10.31235/osf.io/dktah

Massey, D., & Denton, N. (1988). The Dimensions of Residential Segregation. Social Forces,

67(2), 281-315.

content/uploads/dlm_uploads/2018/02/NCRC-Research-HOLC10.pdf.

Monarrez, T., Kisida, B., & Chingos, M. (2020). The Effect of Charter Schools on School

Segregation. EdWorkingPaper No. 20-308. Providence RI: Annenberg Institute at Brown

University. Retrieved March 1, 2021, from https://www.edworkingpapers.com/ai20-308.

Murnane, R.J. (2021). Can policy interventions reduce inequality? Looking beyond test scores

for evidence. William T. Grant Foundation.

Nardone, A. L., Casey, J. A., Rudolph, K. E., Karasek, D., Mujahid, M., & Morello-Frosch, R.

(2020). Associations between historical redlining and birth outcomes from 2006 through

2015 in California. PloS One, 15(8), E0237241.

Nelson, R.K., Winling, L., Marciano, R., Connolly, N., et al. (2021). Mapping Inequality.

Court-Ordered School Desegregation and the Resegregation of American Public Schools.

Reardon, S. F., Ho, A. D., Shear, B. R., Fahle, E. M., Kalogrides, D., Jang, H., & Chavez, B.

from http://purl.stanford.edu/db586ns4974.

Unequal? New Evidence on School Segregation and Racial Academic Achievement Gaps

Analysis. Retrieved March 1, 2021, from http://cepa.stanford.edu/wp19-06.

Weighted Student Funding Implementation. Public Budgeting and Finance. Retrieved

Simpson, E. (1949). Measurement of diversity. Nature, 163, 688.

Federal Government Can Do to Foster Both. In Urban Institute. Urban Institute.

Institute. Retrieved March 1, 2021, from https://apps.urban.org/features/funding-

41

U.S. Census Bureau. (2020). 2019 TIGER/Line Shapefiles (machine readable data files).

Retrieved November 8, 2020, from https://www.census.gov/geographies/mapping-

files/time-series/geo/carto-boundary-file.html.

U.S. Census Bureau. (2021). Public School Spending Per Pupil Increases by Largest Amount in

11 Years. Retrieved June 8, 2021, from https://www.census.gov/newsroom/press-

releases/2021/public-school-spending-per-pupil.html.

Statistics. Retrieved February 25, 2020, from https://nces.ed.gov/pubs2019/2019038.pdf.

U.S. Department of Education, Institute of Education Sciences, National Center for Education

Statistics. (2020). Retrieved October 2, 2020, from

https://nces.ed.gov/programs/edge/Home.

U.S. Department of Education, Institute of Education Sciences, National Center for Education

Statistics. (2021). Retrieved December 11, 2021, from https://nces.ed.gov/ccd/elsi.

U.S. Government Accountability Office. (2016). K-12 Education: Better Use of Information

Could Help Agencies Identify Disparities and Address Racial Discrimination. Retrieved

October 2, 2020, from https://www.gao.gov/products/gao-16-345.

Wells, A.S., Fox, L., Cordova-Cobo, D., & Kahlenberg, R.D. (2016). How racially diverse

48

regress the outcome on HOLC grade indicators with the “D” security rating as the reference category. No controls are

included. 2017-18 total students weight all regressions. All models use city-level fixed effects to account for any

differences that are fixed at the local level and that differ between CBSAs. All dollars denominated in USD 2018. ***

49

Table 4: School Racial Diversity and 1935-1940 HOLC A-D Grades, 2018-2019

Panel A: 2018-2019

Panel B: Margins & Differences Over Time, Racial Diversity Outcomes, 1988-2019

on HOLC grade indicators with the “D” security rating as the reference category. No controls are included. 2018-19

total students weight all regressions. All models use city-level fixed effects to account for any differences that are

50

regress the outcome on HOLC grade indicators with the “D” security rating as the reference category. No controls

are included. All regressions are weighted by 2009-2018 total number of math and ELA tests for pooled SEDA

estimates. All models use city-level fixed effects to account for any differences that are fixed at the local level and

that differ between CBSAs. *** p< 0.01, ** p< 0.05, * p< 0.1.

51

Online Appendix A Table A.1: Diversity Sample Overview

Panel A: All 2018-19 U.S. Public Schools vs. HOLC Sample

52

in this sample if they were matched to 1935-1940 HOLC A-D maps. Shares represent match rates for “Schools” and

53

Panel A: Full vs. Panel, Finance (USD 2018)

54

Panel B: Full vs. Panel, Racial Diversity Outcomes

different racial diversity outcomes. Cluster-robust standard errors are in parentheses with clustering done at the city-

level. No controls are included. All models use city-level fixed effects to account for any differences that are fixed at

the local level and that differ between CBSAs. *** p< 0.01, ** p< 0.05, * p< 0.1.

55

Figure A.2: Los Angeles and Surrounding Districts & 1935 – 1940 HOLC Maps

Figure A.3: School Total PPE by HOLC A-D District Grades – Only “D” Schools (USD 2018)

Revenue; [bottom right] Per-Pupil Local Revenue. All values represented in USD 2018. Coefficients (solid dots) are

HOLC A-D weighted averages from regressions of respective finance outcomes on HOLC A-D indicators without a

constant term. The 95% confidence intervals (lines) are calculated using Huber-White heteroskedasticity-consistent

standard errors. Regressions are run using both the 2017-18 full sample and the 2017-18 data from the panel sample.

The full sample is the original 2017-18 cross-sectional sample and includes all d = 1,760 districts. The panel sample

is the time series sample spanning 1989-90 through 2017-18 school years and includes dp = 1,109 districts.

57

Diversity (1-D). Comments on full and panel samples here. Coefficients (solid dots) are HOLC A-D weighted averages