Gordon, Reber, Were Federal COVID Relief Funds for Schools Enough? (NBER w29470; Tax Policy and the Economy 36, 2021-11)
Summary
NBER Working Paper No. 29470, dated November 2021, titled Were Federal COVID Relief Funds for Schools Enough?, by Nora E. Gordon and Sarah J. Reber. The paper simulates the net effects of the pandemic and federal aid on the finances of local school districts. The baseline scenario assumes one-time adjustment costs of $500 per pupil plus additional costs of $1,000 per student in poverty and $500 per student not in poverty per year for four years. The authors state that low-poverty districts are projected to face some budgetary shortfalls, while many higher poverty districts are projected to have excess funds. The introduction describes nearly $200 billion of federal aid for schools distributed through the Elementary and Secondary School Emergency Relief (ESSER) funds.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Gordon, Reber, Were Federal COVID Relief Funds for Schools Enough? (NBER w29470; Tax Policy and the Economy 36, 2021-11).”
Original at www.nber.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/gordon-reber-were-federal-covid-relief-funds-for-schools-enough-nber-w-3d2049eada/
, SHA-256 965e2f9a4b96345436ce99af3d7ca8a58d03bd5e466282adfc1cf7f8a7929f7e.Full text
NBER WORKING PAPER SERIES
WERE FEDERAL COVID RELIEF FUNDS FOR SCHOOLS ENOUGH?
Nora E. Gordon
Sarah J. Reber
Working Paper 29470
http://www.nber.org/papers/w29470
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
November 2021
We are grateful to Jeffrey Clemens, Lucy Dadayan, and Kim Reuben for helpful discussion of
state revenue and for sharing their data, and to Nat Malkus for insights on the Return to Learn
data. Becca Lowry and Emily Pelles provided excellent research assistance. The views expressed
herein are those of the authors and do not necessarily reflect the views of the National Bureau of
Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies
official NBER publications.
© 2021 by Nora E. Gordon and Sarah J. Reber. All rights reserved. Short sections of text, not to
exceed two paragraphs, may be quoted without explicit permission provided that full credit,
including © notice, is given to the source.
Were Federal COVID Relief Funds for Schools Enough?
Nora E. Gordon and Sarah J. Reber
NBER Working Paper No. 29470
November 2021
JEL No. H52,H75
ABSTRACT
Congress responded to the COVID pandemic’s disruptions to instruction with unprecedented
federal aid for school districts. While this relief has been widely characterized as a major windfall
for K-12 education, per-pupil amounts vary considerably across districts, as will the costs districts
face for COVID mitigation and recovery. In this paper, we conduct simulations to understand the
potential distribution of net effects of the pandemic and federal aid on the finances of local school
districts in the next several years. In our baseline scenario, we assume one-time adjustment costs
of $500 per pupil plus additional costs of $1,000 per student in poverty and $500 per student not
in poverty per year for four years. Federal aid was distributed proportional to the longstanding
Title I program, which sends more money per pupil to higher poverty districts. Low-poverty
districts are therefore projected to face some budgetary shortfalls, while many higher poverty
districts are projected to have excess funds, which they could direct towards long-standing
challenges. While our findings depend on key assumptions about the COVID-related costs, we
find significant district-level variation in simulated net fiscal impacts, in part but not completely
due to poverty rates, across all the scenarios we consider.
Nora E. Gordon
McCourt School of Public Policy
Georgetown University
306 Old North
37th and O Streets NW
Washington, DC 20057
and NBER
neg24@georgetown.edu
Sarah J. Reber
University of California, Los Angeles
Luskin School of Public Affairs
Department of Public Policy
3250 Public Policy Building
Los Angeles, CA 90095
and NBER
sreber@ucla.edu
I. Introduction
Congress responded to the COVID-19 pandemic and the major disruptions to schooling it caused
with nearly $200 billion of federal aid for schools, distributed through the Elementary and
Secondary School Emergency Relief (ESSER) funds of three relief packages. This is roughly 12
times the typical level of funding distributed through Title I, the US Department of Education’s
single largest program for elementary and secondary schools, and amounts to almost $3,500 per
pupil, compared to average spending per-pupil in the United States of about $13,000. So far,
media attention has focused on a handful of high-profile districts. For example, in Michigan, the
distribution of federal relief funds generated major political controversy (Levin 2020), as Detroit
received more than $6,000 per pupil, while low-poverty suburban districts like Bloomfield Hills
received less than $100 per pupil. And many charter districts that serve students who are
demographically similar to those in Detroit Public Schools received less than $1,500 per pupil
(Malkus 2021).
In this paper, we address the question of whether the federal aid is enough to cover schools’
COVID-related costs systematically and explore how the answer to this question depends on
district characteristics. While almost all districts face considerable costs associated with the
pandemic, some districts have been impacted more than others. And the amount of per-pupil
federal aid that districts received varied considerably across districts: all three ESSER packages
sent funds proportionate to existing Title I allocations, which are based largely, but not entirely,
on child poverty.
We simulate the net fiscal impact of COVID and the federal relief under a range of assumptions
about how much COVID affects costs for schools, how those costs depend on child poverty
rates, and the effect of the pandemic on state aid to school districts. Because so much remains
unknown about district needs, the efficacy of spending to address “learning loss” and other
student needs related to COVID, and the effects of the pandemic on district revenue, this
exercise is by necessity an approximation. In our baseline scenario, we assume districts have
one-time adjustment costs of $500 per pupil that do not depend on student demographics; we
assume districts have to spend an additional $1,000 per student in poverty and $500 per student
not in poverty per year for four years starting in 2020-21 to address learning loss and other
problems created by the disruption to schooling.
In the baseline scenario, we estimate that about 62 percent of all districts, and over 95 percent of
high-poverty districts (those with child poverty rates of 25 percent or more, accounting for about
15 percent of enrollment), have received enough ESSER aid to cover the (assumed) costs
associated with COVID. About 85 percent of low-poverty districts are expected to face
budgetary shortfalls of more than $200, and about 23 percent would experience shortfalls more
than $500 per pupil per year for four years. This is because Title I—and therefore ESSER—sent
1
significantly more funding to higher-poverty districts. Many state governments and local school
districts have the fiscal capacity to address potential funding shortfalls in lower-poverty districts,
though in some cases, they may be constrained by school funding rules. 1 Most high-poverty
districts and many moderate-poverty districts are projected to have excess funds under our
baseline scenario; that is, they received more funding from ESSER than necessary to cover the
assumed COVID-related costs. These districts often have long-standing unaddressed needs and
enroll many students who were not working at grade level even prior to the pandemic; many of
these districts could benefit from capital investments, which the substantial influx of ESSER
funding may provide a unique opportunity to address.
II. Background
Our simulation of the net fiscal impact of the pandemic on school districts requires assumptions
about how COVID changed district costs and revenues. In this section we discuss what guides
the thinking behind these assumptions, which we later make explicit (see Table 3). We begin
with the spending side, then proceed to revenues.
Effect of COVID on district costs
Our goal is to compare the COVID-related changes in revenue to additional costs a school
district might incur to respond to COVID. While it is an intuitively appealing exercise to
estimate the cost of making up for learning that did not occur, and addressing other student needs
wrought by the pandemic, it is not straightforward to make such a calculation. First, we do not
know how much “learning loss” the pandemic caused and for whom (see West and Lake 2021
for a recent overview, suggesting students lost a few months of learning on average), nor do we
know the extent of effects on mental health and other outcomes. Second, there is considerable
uncertainty about what interventions would catch students up academically and address the
mental health and other needs of students.
Moreover, focusing new spending exclusively on “COVID-related” learning loss would make
little practical sense: an effective intervention for a child who is behind in reading is a good
investment even if that child would have been behind in reading absent the pandemic. Well
before COVID many students—especially those living in poverty and Black, Hispanic/Latino,
and American Indian and Alaska Native (AIAN) students—were not scoring at grade level on
standardized tests. Only 35 percent of American fourth graders were assessed as proficient (or
above) in reading, and 41 percent were proficient in math in 2019 on the National Assessment of
1
In addition to state and local laws affecting school finance, including tax and expenditure limitations,
“maintenance of equity” provisions in the American Rescue Plan constrain states in how they distribute reductions
in revenue across school districts, and how school districts distribute reductions in resources across school buildings.
2
Educational Progress, also known as the “Nation’s Report Card” (National Center for Education
Statistics 2021).
Some of this systemic underperformance is related to limited funding, but not all of it. Research
shows additional school funding on average has been associated with improvements in
achievement (Jackson 2020), but these magnitudes do not come close to the impacts found in
experimental evaluations of the most cost-effective programs, such as high-dosage tutoring
(Harris 2009). Leaders face legal, political, and organizational constraints when it comes to
selecting programs, and educators face many challenges replicating successful programs in new
settings and at a larger scale. It is unrealistic to expect schools to suddenly have the tools at their
disposal to resolve these longstanding challenges now, even with the considerable influx of
federal funds, so it would be unrealistic for us to estimate the cost of “undoing the effects of
COVID.” Instead, we consider a range of hypothetical spending scenarios based on the costs of
packages of intervention and mitigation approaches that school districts might plausibly pursue.
To inform the spending levels in our simulations, Table 1 lists some activities districts might
spend on in response to COVID, and their associated costs. Not all districts will be able, or want,
to undertake all of these activities, even with sufficient funding, due to difficulty finding the right
staff, long-term contracts, or other logistical constraints. ESSER funding is temporary (it is a
one-time infusion and districts have four years to spend it), so districts may wish to avoid
entering into new long-term contracts and direct spending away from salaried positions towards
goods or contracted services. While we offer Table 1 as a menu of activities and related prices
rather than transactions, survey data suggest that districts are indeed spending their relief funds
from the American Rescue Plan on many of these activities (The School Superintendents
Association 2021). 2 The COVID school costs appendix explains how we arrive at each of these
estimates in greater depth.
Table 1 lists all costs in per-pupil annualized terms. For some activities, like high-dosage
tutoring, this per-pupil, per-year unit of measurement is intuitive: the cost is incurred based on
the number of pupils served per year. But not all activities involve all students. Cost estimates for
high-dosage tutoring range from about $1,500 to $3,800 per student served; if half of a district’s
students receive tutoring, this would come out to $750 to $1,750 per pupil overall. Given the
disproportionate impact of COVID on economically disadvantaged students, we construct
spending scenarios in which poor students receive more services than students who are not poor.
2
The School Superintendents Association’s survey of school district leaders found that approximately three-quarters
of respondents are using ARP funds for summer programs and additional enrichment activities, and nearly half will
use the funds for high-dosage tutoring. Two-thirds of respondents are using the funding to add support staff,
including counselors, social workers, and reading specialists, to their staff. Additionally, school district leaders
reported using ARP funds for internet connectivity and technology, educator professional development activities,
and the implementation of social-emotional learning practices.
3
For example, we could assume that a district provides high-dosage tutoring each year for half its
students in poverty, and for one-quarter of non-poor students. (In practice, districts might
allocate tutors or funds to schools based on poverty, but would likely assign tutors to individual
students within a school based on academic needs rather than economic disadvantage.)
Early research on how COVID has affected student achievement suggests that impacts were
more negative for students experiencing prolonged periods of distance learning in the 2020-21
school year than for those attending school in person. Some of the costs listed in Table 1 are
relevant for all districts, such as transition costs during initial closures in spring of 2020, and
expenditures that make in-person schooling safer, like HVAC upgrades and school nurse
staffing. Yet the need for other activities, especially those targeting “learning loss,” is likely
greater in districts that operated remotely for longer, while districts who returned to fully in-
person instruction sooner may have less to catch up on. We do not incorporate data about
whether schools operated in-person or remote in the analysis.
Revenues
School districts are financed by a combination of taxes they levy directly (local revenue) and aid
from their state and the federal government. On average, about 46 percent of revenue is local,
and 46 percent is from the state, with the rest coming from federal aid, but these numbers vary
considerably both across and within states. Local, state, and federal funds changed in different
ways in response to the pandemic and the associated recession. We discuss how revenue from
each level of government was affected by the pandemic below.
Federal revenues
Though federal revenue is typically small relative to state and local revenue, the federal
government increased funding substantially as a result of the pandemic. Congress established an
Elementary and Secondary School Emergency Relief (ESSER) Fund with $13.5 billion allocated
proportional to past Title I amounts as part of the Coronavirus Aid, Relief, and Economic
Security (CARES) Act passed in March 2020. Subsequent relief packages authorized two more
ESSER funds with slightly different requirements but the same allocation formula: $54.3 billion
under the Coronavirus Response and Relief Supplemental Appropriations Act (CRRSA) and
$122.7 billion in the American Rescue Plan (ARP). In this paper, we analyze the impact of the
three ESSER funds collectively.3
3
Congress also provided relief to states through a Governor’s Emergency Education Relief (GEER) Fund.
Governors distributed those funds to school districts, as well as child care providers and institutions of higher
education, at their discretion.
4
Estimated ESSER allocation (including all three rounds of ESSER and weighted by district
enrollment) was $3,431 per pupil but varies considerably (we describe how we make these
estimates below): districts in the 10th percentile received an estimated $774 per pupil, while
those in the 90th percentile received $7,192 per pupil (Appendix Table 1). Figure 1 shows the
distribution of per-pupil ESSER funding for all districts and by district poverty rate. If all
COVID-induced costs to districts were incurred uniformly per pupil (that is, without additional
costs for students in poverty), one could simply compare the ESSER levels in Figure 1 with
district costs from COVID to answer the question of whether federal funding is sufficient. For
example, if costs are less than $1,500 per pupil, then 28 percent of districts did not receive
enough federal funds. The simulation exercise described below allows costs to vary based on
student poverty and incorporates potential reductions in state aid to estimate net fiscal impacts of
the pandemic under alternative scenarios.
Title I allocates funding based on the number of poor children in the district, so much of the
variation in per-pupil ESSER funding is due to differences in child poverty rates. At the same
time, Title I funding amounts per eligible child vary even conditional on child poverty rates. For
5
example, consider the poorest districts, with at least a quarter of children in poverty: 8 percent of
students attend districts receiving less than $3,000 per pupil in ESSER funds, while 28 percent
attend districts receiving $7,500 or more per pupil. That is because the formulas rely on lagged
state-level average spending per-pupil, and reward low-population states, as well as states and
districts with historically high poverty levels (relative to newly high-poverty places) (Gordon and
Reber 2020). The formulas also direct more funds to higher enrollment districts conditional on
child poverty rates.
State revenues
State budget shortfalls can lead to cuts in state aid to local school districts, especially since many
states have balanced budget requirements. The instability of state revenue over the business cycle
poses a particular problem to poorer districts, which rely more heavily on state aid, compared to
their lower-poverty counterparts. While local revenue per pupil is negatively correlated with
district poverty rates in nearly all states, the opposite is typically true for per-pupil state and
federal aid. Across the states, state revenue per pupil is either uncorrelated with child poverty
rates (at the school district level), or positively correlated with it—in other words, most states
distribute state aid progressively, providing more funding per pupil to higher-poverty districts.4
Early in the pandemic, many (including us) were concerned that reductions in revenues to state
governments would translate to reductions in state aid for local school districts, as happened
during the Great Recession (Gordon and Reber 2020; Clemens and Veuger 2020). Instead, the
stock market recovery, relative stability of incomes for high-earners, and federal aid to states
meant that the impact of the pandemic on state revenues was short-lived in most (or all) states.
State governments also will receive $195 billion from the ARP.
The impact of the recession on states’ economic activity and their own-source revenue varied
considerably. States that rely more on progressive income taxes and capital gains taxes fared
well as the incomes of higher-income households were steady and the stock market recovered;
states that rely on sales or energy taxes—and those with large leisure and hospitality sectors—
fared worse (Dadayan and Rueben, 2021). Comparing own-source state revenue collected from
April 2020 to March 2021 with the same period a year earlier shows that only a handful of states
experienced major reductions in own-source revenue. 5 But even for those states with the most
drastic reductions—in the most extreme example, revenue fell around 40 to 45 percent over that
time in Alaska—federal relief to states replaced lost own-source revenue multiple times over.
4
This interactive from the Urban Institute shows how redistributive each state is and how the states compare.
5
We thank Lucy Dadayan for sharing unpublished estimates from the Urban Institute’s State and Local Finance
Initiative.
6
For the bulk of our analyses, we assume state revenue to school districts is unaffected by
COVID, since at this point the fiscal position of all states appears to be at least as good as before
the pandemic; many states have surpluses. Some states did experience major shortfalls earlier in
the pandemic, and indeed some did cut funding to schools temporarily, so we consider one
scenario assuming that states cut aid to school districts. We do not incorporate projected state
government surpluses in our analyses, as it is not clear whether or how these surpluses would
translate into state funding for school districts. Though states often earmark a set share of general
revenue or specific funding streams (e.g., lottery or tobacco taxes) for their education budgets,
this would not apply to the federal relief funds for states. However, we note that to the extent that
ESSER funding is not enough to meet the COVID-related, or long standing, needs of some
school districts, many states are in a good fiscal position and could increase their own education
spending to help address these issues.
Local revenues
The majority of local revenue raised by school districts is from property taxes, which are
relatively stable in recessions. During the Great Recession, revenues from property taxes proved
more stable than income or sales tax collections even though property values declined, as local
governments were able to adjust property tax rates to hold revenue constant (Evans, Schwab, and
Wagner 2019). Some school districts are fiscally dependent, meaning they do not levy taxes
themselves but receive local revenue from a “parent government,” usually a city or county
government. These parent governments may also rely on other revenue sources, like sales taxes,
in addition to or instead of property taxes. Still, because most districts depend primarily on
property taxes for local revenue, which were largely unaffected by the pandemic, we do not
incorporate shocks to local revenue into our fiscal impact scenarios.
III. Methods and data
The goal of this paper is to assess whether federal aid to local school districts is enough to cover
COVID-related costs, and how the answer to this question varies across districts with different
characteristics. The three ESSER funds were divided among states and then among districts
within states proportionate to Title I funding in previous years (FY2019 for the CARES Act
funding, and FY2020 for the other two funds), so in theory, one simply needs to compare the
ESSER funding to expected costs. This is not straightforward, however, because the Title I
formulas are quite complex, making it difficult to draw conclusions about how much different
districts will receive by inspecting the formulas. (In fact, the formulas lead to allocations that are
sufficiently confusing that Congress mandated a study to help understand them in 2016.) In
addition, whether or not federal aid is sufficient depends on other factors—notably the costs of
addressing COVID and the impact on state aid to local school districts—that are uncertain. We
therefore simulate changes to school districts’ costs and revenues under a range of assumptions.
7
Parameters
The simulations have four parameters which vary across scenarios. We consider the following
categories of COVID-related spending that school districts might need to undertake: (1) one-time
“all-student” costs; (2) per-pupil “recovery” costs for poor students; and (3) per-pupil “recovery”
costs for non-poor students. We assume the recovery costs are spread over four years and express
them in per-pupil annual terms. The precise division between one-time and recovery costs is not
critical, nor is the time pattern of spending. The key point is that higher-poverty districts are
assumed to need more money to address the effects of COVID. We discuss the assumptions we
make about costs in the different scenarios below.
The fourth parameter is the percentage reduction in state aid. We use estimates of shortfalls in
states’ own-source revenue made earlier in the pandemic (Whitaker 2020). These estimates
overstate the state revenue shortfalls, but reflect the relative impacts across states (for example,
shortfalls are larger in states with large travel and leisure sectors) to provide a sense of the effect
of incorporating impacts on state aid on our understanding of how many districts—and which
districts—received enough federal aid. We assume state aid to school districts is reduced
proportional to each state’s revenue shortfall (as if states made across-the-board cuts to balance
their budget). That is, if a state has a 2 percent revenue shortfall, we assume state aid to all
districts in the state is cut by 2 percent, which will affect districts that rely more on state aid.
Data
To simulate the net fiscal impact of the pandemic under alternative spending and revenue
scenarios, we use school-district level data on enrollment, revenue by source, and Title I funding
from the Annual Survey of School District Finances/F33 for the 2018-19 school year (the most
recent available). We use data on the racial composition of enrollment from the NCES Common
Core of Data and child poverty rates from the Census Small Area Income and Poverty Estimates
(SAIPE). The child poverty rate from SAIPE is based on children who live in the district’s
geographic boundaries, whether or not they attend public schools; this is the measure used to
allocate Title I. Implicitly, our simulations proceed as if the ESSER funding came in 2018-19;
this approach will surely mis-estimate ESSER funding and other aspects of school finance for
some individual districts, but we expect the patterns we identify in the analysis based on 2018-19
data to be similar to what we would see with more up-to-date data.
Table 2 presents summary statistics for the districts in our sample. Due to various data
limitations, we exclude districts in Hawaii, Vermont, and the District of Columbia, districts with
enrollment less than 150, districts that are missing key variables, and charter school districts. Our
working sample of 11,673 districts accounts for about 94 percent of public school enrollment.
The average district enrolled 4,055 students, and the average child poverty rate in the sample is
16.4 percent. (All summary statistics other than enrollment itself are weighted by enrollment.)
8
The average district had enrollment that was 52 percent white, 26 percent Hispanic, 15 percent
Black, 5 percent Asian and 1 percent American Indian and Alaska Native (AIAN) and had about
$13,000 in current expenditure and $15,000 in total revenue per pupil.6 About $1,100 in revenue
was from the federal government, with the remainder roughly split between state and local
sources.
Scenarios
Table 3 shows the parameters (columns 1-4) we use for the four scenarios (rows 1-4) we
consider in the analysis. In the baseline scenario (row 1) we assume $500 per-pupil in one-time
all-student costs, corresponding to the transition costs noted in the last row of Table 1 and
described further in the cost appendix. Because that estimate was based on district reports of
spring 2020 costs and the pandemic is ongoing, we double these costs in the larger all-student
costs scenario (row 2).
We assume districts also have costs associated with COVID recovery that span at least several
years as districts cope with significant disruptions to schooling and COVID-related trauma for
students. These are not estimates of annual total, current, or instructional spending per student,
but of additional spending over the counterfactual, incurred in response to COVID. We assume
districts incur such costs annually over the four-year ESSER spending period. 7 In practice, these
costs might be weighted toward COVID mitigation (personal protective equipment, air filters,
etc.) in the early years and additional instructional support (tutoring, counseling, etc.) in later
years.
Because the impact of COVID has been so much more severe for disadvantaged communities,
we assume higher per-pupil recovery costs for students in poverty in each scenario. Column 2
contains annual per-pupil recovery cost assumptions for poor students, and column 3 for students
who are not poor. Our baseline annual cost estimate is $1,000 per poor student (in rows 1, 2, and
4), and $500 per non-poor student (in each scenario). The activities and corresponding costs
detailed in Table 1 provide a rough way to interpret these magnitudes. For example, a district
spending $500 per non-poor student could provide a quarter of those students with high-dosage
tutoring, or upgrade or replace the HVAC systems for about half the non-poor students. A
district spending $1000 per poor student could provide half of those students with high-dosage
6
Total expenditures include current and capital expenditures. Consistent with the literature, we focus on current
expenditures here.
7
Districts have until September 2024 to obligate funds from the American Rescue Plan (ARP), so the funds could
be spent over more than four years. Similarly, districts mostly did not access CARES funding in the Spring of 2020,
though they knew that funding would become available soon. For the purposes of this exercise, we are less
concerned about the precise timing of spending but want to allow a distinction between one-time costs and ongoing
costs and differential ongoing costs for poor and non-poor students.
9
tutoring, upgrade or replace HVAC systems for all of those students, or send about two-thirds of
poor students to summer school.
Those examples could address some long standing needs that districts have had insufficient
resources to address, but may be insufficient to fully serve students deeply impacted by COVID.
We offer an alternative high costs for poor students scenario, described in row 3, in which
districts spend an additional $3,000 (as opposed to $1,000) per poor student in response to
COVID. $3,000 per pupil is a significant amount of funding compared with average current
expenditure of about $13,000 per pupil (Table 2). It could fund high-dosage tutoring for three-
quarters of students in poverty, along with HVAC upgrades or replacements and more than
doubling socio-emotional supports for all students in poverty.
To explain how the simulation works, we walk through the calculations for Montgomery County
Public Schools (MCPS) in Maryland; see the methods appendix for more detail. In 2018-19,
MCPS enrolled 162,680 students, had a poverty rate of 8.2 percent, received $25.6 million in
Title I funding, $925 million in state aid, and had total revenue of about $3.34 billion. The three
ESSER funds totaled about 12 times a normal year’s Title I allocation, so we estimate MCPS
will receive about $319 million in ESSER funding (12 X $25.6 million), or about $2,000 per
pupil. Under the baseline assumption of $500 per pupil, we estimate $81 million in one-time all-
student costs ($500 X 162,680), leaving $238 million in ESSER funds, which we assume is spent
over four years, so $59.5 million per year is available for recovery. Under the baseline
assumption that recovery costs $500 and $1,000 per-pupil, per-year for non-poor and poor
students, respectively, MCPS would need to spend $88 million per year for four years (there are
roughly 13,322 poor students and 149,357 non-poor students in the district). This means that
under these baseline assumptions, MCPS would face a net shortfall of $28.5 million or $175 per
pupil, equal to less than 1 percent of total revenue. In scenario 4, we assume that state aid
declines by 2.5 percent—or $23 million (2.5 percent of $925 million)—in which case the
shortfall would be $51.9 million or $319 per pupil.
We next do the same calculations for Chicago Public Schools. Chicago has a higher poverty rate
and is estimated to receive about $8,000 per pupil in ESSER funds and to have a surplus of
$1,267 per pupil per year for four years under the baseline assumptions. This does not
necessarily mean that Chicago will be “overfunded;” instead, it means that ESSER funds are
more than enough to cover COVID-related costs under the baseline assumptions. However, as is
common in high-poverty school districts, many of Chicago’s students were not working at grade
level or had unmet needs prior to the pandemic; any “excess” ESSER funding could be put to
good use addressing those needs.
10
IV. Results
Before presenting the results of the simulated net fiscal impacts of COVID under the four
scenarios, we discuss the effects on federal revenue (the estimated ESSER allocations) and state
aid (which applies to scenario 4 only). To understand how the simulated fiscal effects of COVID
vary across districts depending on their poverty rate, racial composition, and enrollment, we
regress the outcomes (ESSER allocations, simulated reductions in state aid, and simulated net
fiscal impacts in the four scenarios) on these district characteristics.
Distribution of effects on federal revenue
Table 4 shows the regression estimates for estimated per-pupil ESSER funding. We assume
ESSER is spent over four years and annualize the effects. This normalization is for ease of
interpretation and not substantive; some districts may choose to spend their funds more quickly
and others spread them out. The average annualized per-pupil ESSER allocation is $858
(corresponding to $3,431 total, Appendix Table 1). All regressions are weighted by enrollment.
Since ESSER was distributed proportionate to Title I and we estimate the ESSER allocation by
simply multiplying a district’s reported Title I funding by 12 (assuming the funding is allocated
as prescribed by law), this analysis is therefore effectively an analysis of how Title I is
distributed.
In the first three columns, we regress per-pupil ESSER on the share of children in poverty, racial
and ethnic composition, and enrollment separately, and we include all three in column 4. In
column 5, we add state fixed effects to understand the correlates of funding within states. For
every 10 percentage point increase in the share of children in poverty, a district’s annualized per-
pupil ESSER allocation is about $610 higher, on average (column 1). The share of children in
poverty explains 64 percent of the variation in ESSER allocations, consistent with the fact that it
is the most important—but not the only—input in the Title I formulas. The Title I formulas also
have some state-specific factors and incorporate non-linearities in the share of children in
poverty (Gordon and Reber 2020).
Districts with high shares of American Indian and Alaska Native (AIAN), Black, or Hispanic
students received more ESSER funding per-pupil, and districts with higher Asian and Pacific
Islander shares received less, compared to districts with more white students (column 2). For
example, for every 10 percentage point increase in the share of students who are Black, annual
per-pupil ESSER is about $200 more, on average. Each 10 percentage point increase in the
Hispanic share is associated with a $100 increase. Racial composition does not enter the Title I
formulas, so any correlation of racial composition with ESSER funding is due to a correlation of
racial composition with variables that are used in the formulas, most importantly, the share of
children in poverty. Larger districts received more per-pupil ESSER funding, on average
(column 3). The distribution of district size is highly skewed: the coefficient implies that moving
11
from the 25th to 75th percentile of the enrollment distribution is associated with an additional
$170 per pupil, and moving from the 10th to the 90th percentile is associated with an additional
$327 per pupil. Enrollment, per se, also does not enter the Title I formulas directly, but two of
the formulas allocate more funding per eligible child to districts with large numbers of poor
students, which benefits larger districts more. Enrollment is virtually uncorrelated with the share
of children in poverty (not reported).
In column 4, we include all three sets of variables at the same time, and in column 5 we add state
fixed effects. Not surprisingly, the coefficients on the racial composition variables and log
enrollment are smaller when the share of children in poverty is also included in the regression
(and the sign is reversed for the coefficient on the share of students who are Asian). Controlling
for the share of children in poverty and district size, school districts serving more AIAN or Black
students received more ESSER funding per pupil. The same is true within states (the coefficients
are slightly larger); and within states, Hispanic-serving school districts also get more, and Asian-
serving districts less ESSER, on average, compared to districts with high white shares of
students. Again, these regressions are simply showing how the way that Title I is allocated
benefits different types of districts. A full explanation of what aspect of the formulas account for
these correlations is beyond the scope of this paper; we refer interested readers to a
Congressionally-mandated study of the formulas (Snyder et al. 2018).
Distribution of effects on state revenue
Table 5 shows the same set of regressions with the simulated reduction in state aid per pupil as
the dependent variable. Recall that this parameter is only relevant for Scenario 4 (State Aid
Shortfall). We simulate the state aid shortfall by multiplying the projected state own-source
revenue shortfall by each district’s state aid. Essentially, we assume that states do across-the-
board cuts proportional to their own revenue shortfalls (which, as discussed above, did not last
nearly as long as initially expected). Districts that depend more on state aid and/or are in states
with bigger revenue shortfalls will have larger simulated state revenue shortfalls. In practice,
projected state-level revenue shortfalls (in percentage terms) are not correlated with district-level
characteristics (not reported); that is, the states where own-source state revenue was more
impacted by the pandemic are similar in terms of the share of students in poverty, racial
composition, and district size, compared to states where state revenue was less affected.
Therefore, the correlations in Table 5 are mostly driven by how much per-pupil state aid
different types of districts receive. With or without controls, districts with higher shares of
children in poverty have higher simulated state revenue shortfalls. This is because most states
allocate more aid to poorer districts, so they would be more affected by (simulated) cuts to state
aid. The coefficients on the racial composition variables are only statistically significant in the
specification with state fixed effects, and in any case, are not substantively large.
12
Distribution of simulated net fiscal impacts
We focus the discussion here on the distribution of net fiscal impacts overall and by district
poverty and racial composition and present summary statistics (Appendix Table 2) and
regression analyses (Appendix Tables 3) for completeness.
Figure 2 shows the distribution of net fiscal impacts, weighted by enrollment, for the four
scenarios. The average net impact across the scenarios ranges from a shortfall of $177 (with the
assumption of higher costs for poor students) to a “windfall” of $151 per-pupil in the baseline
scenario. In each scenario, however, the projected net impact varies a lot across districts. For
ease of exposition, we refer simply to districts throughout the discussion, but recall that these
statistics are weighted by enrollment; we refer to net impacts within $200 per pupil (in either
direction) of the no-COVID counterfactual as minimal fiscal impacts.
For the baseline scenario, about a third of districts would experience minimal fiscal impacts, and
less than 10 percent would face a shortfall of $500 or more (per pupil per year for four years) and
would need to find alternative sources of revenue to cover the costs assumed in this scenario.
About 10 percent of districts would have an extra $500-$1,000 per pupil to spend, and 7 percent
could spend $1,000 or more per pupil beyond the costs assumed in the baseline.
13
In scenario 2, where we assume the one-time all-student costs are $1,000 per pupil instead of
$500, over half of districts would experience a shortfall of more than $200 per pupil (annually
for four years). In scenario 3, which assumes higher recovery costs for poor students, roughly a
third of districts would experience shortfalls more than $500, and another third would have
shortfalls of $200-500; only a small share of districts would experience significant windfalls.
Finally, scenario 4, which incorporates reductions in state aid, has similar impacts to scenario 2
(which raised one-time costs).
In figure 3, we show how the simulated effects vary with district poverty rate. We group districts
with child poverty rates of less than 10 percent (low), 10-25 percent (medium), and more than 25
percent (high) together. The low poverty districts account for 28 percent of districts and
(coincidentally) serve 28 percent of students. The medium poverty districts comprise 55 percent
of districts and 56 percent of students, and 16 percent of districts serving 15 percent of students
are in the high poverty group.
The strong, positive relationship between the share of children in poverty and the per-pupil
ESSER allocation (Figure 1 and Table 4) explains why higher-poverty districts are more likely to
have windfalls and low-poverty districts more likely to have shortfalls across all the scenarios. In
the baseline scenario, almost all high-poverty districts are projected to have enough ESSER
funding to cover their costs, and almost 60 percent would have a windfall of $500 per pupil per
year for four years or more. The results for scenario 3 suggest that most high-poverty districts
received enough ESSER funding to cover $500 per pupil in transition costs plus $500 for every
non-poor child and $3,000 for every poor child each year for four years. Many could afford
substantially more than this.
About 61 percent of low-poverty districts would experience modest annualized shortfalls of
$200-500 per pupil, and 23 percent would face shortfalls greater than $500 per pupil per year
(panel a). In all the other scenarios, large shares of low-poverty districts face substantial
shortfalls. In the case of scenario 2, it is because there are assumed to be substantial all-student
costs, but ESSER effectively only allocates aid for poor students, so the aid does not match the
needs. In scenario 3, ESSER is not enough to cover both the moderate all-student and non-poor
student costs and the high recovery costs for the relatively small number of poor students these
districts serve.
The results for districts with moderate poverty rates, not surprisingly, fall in between those for
low- and high-poverty districts: The typical district in this range would experience minimal net
fiscal impacts in the baseline scenario. Less than a quarter of such districts would experience
large shortfalls (more than $500 per pupil per year) in any scenario, and less than 15 percent
would experience substantial windfalls in any scenario.
14
In Figure 4, we show how the distribution of net fiscal impacts varies depending on the racial
composition of enrollment. We divide districts into three categories based on the share of
enrollment that is non-white: districts that are less than 25 percent non-white (these districts
comprise 30 percent of enrollment nationally), 25 to 75 percent non-white (45 percent of
enrollment), and more than 75 percent non-white (25 percent of enrollment). 8 Recall that the
distribution of ESSER funding does not take race into account explicitly, nor do any of the
parameters assumed in the simulation. Differences in simulated net fiscal impacts by racial
composition of districts arise because of the correlation of racial composition with poverty and
other features of the Title I formulas, as discussed above. Interestingly, even controlling for the
share of children in poverty, districts serving more AIAN, Black, and Hispanic students have
larger (positive) net fiscal impacts across scenarios, though controlling for poverty yields smaller
magnitudes (Appendix Tables 3a-3d). Due to the high correlation (about 0.5) between the non-
white share and the share of children in poverty, the patterns in Figure 4 largely mirror those in
Figure 3.
8
We divide districts according to the non-white share of enrollment for simplicity. Appendix Tables 3a to 3d
present regression analyses that consider all the racial/ethnic categories and enters the enrollment share in the
regression linearly rather than categorically. There are some differences in net fiscal impacts across non-white racial
groups, but the categorical variables in Figure 4 capture the key relationship between net fiscal impacts and racial
composition. (Notably, the sign is flipped for Asians relative to the other non-white groups, but this group is small.)
15
In the baseline scenario, about 7 percent of predominately non-white districts are projected to
face shortfalls more than $200 per-pupil annually for four years, and more than a third are
projected to have a windfall of more than $1,000 per pupil annually for four years. Most high-
poverty districts received enough ESSER funding to afford the higher costs of COVID recovery
for poor students in Scenario 3, and about a third would have a substantial windfall even in that
scenario. Many predominately white districts would need to find other sources of revenue to
cover the COVID costs assumed across scenarios, and the distribution of net fiscal impacts for
districts that are between 25 and 75 percent non-white falls in the middle; most received enough
ESSER funding to cover costs in the baseline scenario, but about a third would face shortfalls.
V. Discussion
The fiscal impact of COVID on states was both less than expected and less than the impact of the
Great Recession, and federal aid to states in response to COVID was substantial. In this context,
ESSER funding is not critical for addressing state revenue shortfalls or for fiscal stimulus. But
school districts have incurred and continue to face new, often substantial, costs because of
COVID. An important limitation of our analysis is that we do not have district-level data on how
16
COVID affected instructional modalities (remote, hybrid, or in-person) and educational
achievement. Without this information, it is hard to know how the revenue windfalls correspond
to the distribution of COVID-induced local needs. The return to in-person instruction and to
universal standardized testing should make it possible for researchers to address these questions.
By design, high poverty districts benefit most from ESSER funding. Even under our scenario
with the highest remediation costs for students in poverty—$3,000 per pupil per year for four
years—we estimate the majority of higher poverty districts will receive more than enough to
cover these costs. The high correlation between child poverty and district racial composition
means that predominately non-white districts received substantially more ESSER funding,
enough to cover considerable COVID-related costs. And while high-poverty districts in states
that operated close to normally during 2020-21 may have fewer needs for interventions targeting
COVID-specific learning loss, they could direct ESSER funding to address long-standing unmet
educational needs of their students.
ESSER funding presents a unique opportunity for many high-poverty districts, many of which
serve predominately non-white student populations, to invest. Because these funds are a one-time
infusion, districts may be reluctant to take on new recurring obligations, including through
hiring, anticipating the “fiscal cliff” when funds run out and are not replaced. But they can enter
into contracts for high-dosage tutoring or teacher coaching, or make improvements to
infrastructure. Ideally, if districts learn that certain programs or investments they make with
ESSER funds are effective, they would look for ways to fund those programs when ESSER
funding runs out—by redirecting funds from less-effective programs or seeking new revenue
sources.
Because of COVID’s disproportionate impact on low income students, the decision to weight
poverty heavily in ESSER allocations was appropriate, but the reliance on Title I allocations
leaves most low poverty districts with moderate to large shortfalls. It also directs substantial
windfalls to many—but not all—high poverty districts, with a relatively quick timeline for
spending. Normally, $16 billion per year is distributed using these formulas; the three rounds of
ESSER funding sent over $190 billion, amplifying concerns about aspects of the Title I formulas
that are not transparent and may appear arbitrary or unfair (Gordon and Reber 2020). Given how
much has already been distributed proportional to Title I, Congress may wish to consider
alternative, simpler ways of incorporating child poverty into federal aid in the future.
States—especially those that are flush with surplus from ARP—may wish to help districts
address shortfalls, which would mean sending money specifically to low-poverty districts.
Congress worried that states would “undo” the progressive allocation of ESSER and in ARP
created “maintenance of equity” requirements for states (governing how they distribute any cuts
to school districts) and districts (governing how they spread any cuts across schools). These
17
requirements could limit the ability of states to direct more funding to lower-poverty districts that
received less ESSER funds. State school finance laws also constrain state’s discretion at
responding to the specific pattern of windfalls and shortfalls created by ESSER. And some state
finance programs limit how much districts can raise locally. Federal funding directed
disproportionate resources to higher poverty districts—with disproportionate needs. However, if
state or local laws, including school finance regimes and tax and expenditure limitations, prevent
these lower-poverty districts from raising revenue to address their own COVID-induced needs,
the progressive nature of federal relief could create unintended problems.
Distributional issues aside, ESSER funding could present an enormous opportunity for many
districts, especially high-poverty districts. Well-chosen interventions and investments could have
larger impacts compared to additional funding induced by changes to state school funding
formulas because the politics of budgeting and staffing can make it challenging to free up
significant funds for new (potentially more cost-effective) programming without a substantial
influx of new funding. District needs vary, and the best use of funds will also vary, and some
districts may choose to use ESSER similarly to how they spend existing revenue.
This analysis was designed to shed light on the magnitudes of federal aid for schools compared
to reasonable assumptions about the costs associated with the COVID pandemic. The analysis
suggests that many low-poverty districts and some moderate-poverty districts could face funding
shortfalls, though most would be modest and state and local governments are mostly in a good
fiscal position to help address these. We estimate that many high-poverty and some moderate-
poverty districts will have moderate to substantial windfalls. This implies that ESSER funding is
more than enough to address the “COVID costs” assumed in our scenarios, but that does not
necessarily mean it is too much, or even enough, to fully address the needs of students, many of
whom were struggling before the pandemic. Though advocates, parents, and policymakers are
interested in how funds can be used to address pandemic-related needs, it is challenging to
disentangle those from long standing unmet needs, particularly in high-poverty districts that
received substantial ESSER funding. In practice, a focus more on which uses of funds are likely
to be cost-effective and less on which of students’ problems were caused by COVID (which is
essentially unknowable) could prove useful in districts with substantial ESSER windfalls.
18
References
Ander, Roseanna, Jonathan Guryan, and Jens Ludwig. 2016. “Improving Academic Outcomes
for Disadvantaged Students: Scaling Up Individualized Tutorials.”
Bureau of Labor Statistics. 2021. “Occupational Outlook Handbook, Registered Nurses.” U.S.
Department of Labor. June 2, 2021. https://www.bls.gov/ooh/healthcare/registered-nurses.htm.
Buttner, Annie. 2021. ”Diagnosing the School Nurse Shortage.” Frontline Education (blog).
January 5, 2021. https://www.frontlineeducation.com/blog/school-nurse-shortage/.
Clemens, Jeffrey and Stan Veuger. 2020. “Implications of the COVID-19 Pandemic for State
Government Tax Revenues.” National Tax Journal, 73 (3): 619-644.
https://doi.org/10.17310/ntj.2020.3.01.
Dadayan, Lucy, and Kim S. Rueben. 2021. “Why States and Localities Need More Federal Aid.”
Tax Policy Center (blog). February 21, 2021. https://www.taxpolicycenter.org/taxvox/why-
states-and-localities-need-more-federal-aid.
Evans, Grace, Heather L. Schwartz, and Benjamin K. Master. 2021. “Lost Learning and the
Costs of COVID-19.” The RAND Blog. February 10, 2021.
https://www.rand.org/blog/2021/02/lost-learning-and-the-costs-of-covid-19.html.
Evans, William N., Robert M. Schwab, and Kathryn L. Wagner. 2019. “The Great Recession and
Public Education.” Education Finance and Policy 14 (2): 298–326.
https://doi.org/10.1162/edfp_a_00245.
Finlaw, Sarah. 2021. “Baker-Polito Administration Announces Pooled Testing Initiative for
Massachusetts Schools, Districts.” Office of Governor Charlie Baker and Lt. Governor Karyn
Polito. January 8, 2021. https://www.mass.gov/news/baker-polito-administration-announces-
pooled-testing-initiative-for-massachusetts-schools.
Gewertz, Catherine. 2021. “Testing for COVID-19 at School: Frequently Asked Questions.”
Education Week. March 16, 2021. https://www.edweek.org/leadership/should-schools-test-
students-and-staff-for-covid-19/2021/03.
Gordon, Nora and Sarah Reber. 2020. “Federal Aid to School Districts during the COVID-19
Recession.” National Tax Journal: 73 (3): 781–804. https://doi.org/10.17310/ntj.2020.3.07.
Government Accountability Office. 2020. “School Districts Frequently Identified Multiple
Building Systems Needing Updates or Replacement.” June 2020.
https://www.gao.gov/assets/gao-20-494.pdf.
Griffith, Michael and Allie Pearce. 2020. “The Air We Breathe: Why Good HVAC Systems Are
an Essential Resource for Our Students and School Staff.” Learning Policy Institute. December
8, 2020. https://learningpolicyinstitute.org/blog/covid-hvac-systems-essential-resource.
19
Harris, Douglas N. 2009. “Toward Policy-Relevant Benchmarks for Interpreting Effect Sizes:
Combining Effects With Costs.” Educational Evaluation and Policy Analysis 31 (1): 3–29.
https://doi.org/10.3102/0162373708327524.
Hamilton, Laura and Betheny Gross, panel co-chairs. “How Has the Pandemic Affected
Students’ Social-Emotional Well-Being? A Review of the Evidence to Date.” Center for
Reinventing Public Education. August 2021.
Jackson, C. Kirabo. 2020. “Does School Spending Matter? The New Literature on an Old
Question.” In Confronting Inequality: How Policies and Practices Shape Children’s
Opportunities, 165–86. APA Bronfenbrenner Series on the Ecology of Human Development.
Washington, DC, US: American Psychological Association. https://doi.org/10.1037/0000187-
008.
Kraft, Matthew A., and Grace Falken. 2020. “A Blueprint for Scaling Tutoring Across Public
Schools.” https://doi.org/10.26300/DKJH-S987.
Levin, Koby. 2020. “‘The wrong formula’: Education advocates call for new approach to how
schools get coronavirus aid.” Chalkbeat Detroit. June 24, 2020.
https://detroit.chalkbeat.org/2020/6/24/21302086/wrong-formula-advocates-call-for-different-
coronavirus-aid
Malkus, Nat. 2021. “Federal COVID Elementary and Secondary School Emergency Relief
Funding District-Level Data Compilation,” Return to Learn Tracker, American Enterprise
Institute, http://www.returntolearntracker.net/esser/.
National Center for Education Statistics. 2020. “School Nurses in U.S. Public Schools.” Institute
of Education Sciences, U.S. Department of Education. April 2020.
https://nces.ed.gov/pubs2020/2020086.pdf.
National Center for Education Statistics. 2021. “Nation’s Report Card. National Assessment of
Educational Progress.” Institute of Education Sciences, U.S. Department of Education. Accessed
August 31, 2021. https://www.nationsreportcard.gov/
Newberry, Laura, Melissa Gomez, & Howard Blume. 2021. “Inside the L.A. School District’s
Massive COVID Testing Effort.” Education Week. https://www.edweek.org/leadership/inside-
the-l-a-school-districts-massive-covid-testing-effort/2021/08
Nickow, Andre, Philip Oreopoulos, and Vincent Quan. 2020. “The Impressive Effects of
Tutoring on PreK-12 Learning: A Systematic Review and Meta-Analysis of the Experimental
Evidence.” w27476. Cambridge, MA: National Bureau of Economic Research.
https://doi.org/10.3386/w27476.
O’Neill, Mark and Kendall Alexander. 2021. “Wolf Administration Directs Vaccine Providers to
Support COVID-19 Vaccination Clinics and Unveils Free COVID-19 Testing Program for
Schools Across Pennsylvania.” August 16, 2021. https://www.media.pa.gov/pages/education-
20
details.aspx?newsid=1177.
Pier, Libby, Michael Christian, Hayley Tymeson, and Robert H Meyer. 2021. “COVID-19
Impacts on Student Learning.”
School Superintendents Association, The. 2021. “School District Spending of American Rescue
Plan Funding.” https://aasa.org/uploadedFiles/ARP-Survey-Findings-090121.pdf.
Schwartz, Heather L., Jennifer Sloan McCombs, Catherine H. Augustine, and Jennifer T.
Leschitz. 2018. “Getting to Work on Summer Learning: Recommended Practices for Success,”
2nd Ed. RAND Corporation, 2018. https://www.rand.org/pubs/research_reports/RR366-1.html.
Renaissance Learning, Inc. 2021. “How Kids Are Performing: Tracking the Mid-Year Impact of
COVID-19 on Reading and Mathematics Achievement.”
https://renaissance.widen.net/s/zvq8rnrp9t/r63370.
Riser-Kositsky, Maya. 2019. “Education Statistics: Facts About American Schools.” Education
Week. January 3, 2019. https://www.edweek.org/leadership/education-statistics-facts-about-
american-schools/2019/01.
Snyder, T., Dinkes, R., Sonnenberg, W., and Cornman, S. 2018. Study of the Title I, Part A
Grant Program Mathematical Formulas (2019-016). U.S. Department of Education. Washington,
DC: National Center for Education Statistics.
West, Martin R. and Robin Lake, panel co-chairs. “How Much Have Students Missed
Academically Because of the Pandemic? A Review of the Evidence to Date.” Center for
Reinventing Public Education. July 2021.
Whitaker, Stephan D. 2020. “How Much Help Do State and Local Governments Need? Updated
Estimates of Revenue Losses from Pandemic Mitigation.” Cfed District Data Briefs, no. cfddb
20200629 (June). https://www.clevelandfed.org/newsroom-and-events/publications/cfed-district-
data-briefs/cfddb-20200629-updated-estimates-of-revenue-losses-from-pandemic-mitigation.
Zhou, Tiffani, Tomas Molfino, and Jonathan Travers. 2021. “The Cost of COVID:
Understanding the Full Financial Impact of COVID-19 on Districts and Schools.”
https://www.erstrategies.org/cms/files/4709-cost-of-covid-updated-2.pdf.
21
Tables
Table 1. Costs of activities schools might undertake in response to COVID
Activity Cost
Match/SAGA high-dosage tutoring as $3,800 per pupil served per year
implemented in Chicago Public Schools (Anders,
Guryan, and Ludwig 2016)
Match/SAGA estimate if operating at scale in $2,500 per pupil served per year
Chicago Public Schools (Anders, Guryan, and
Ludwig 2016)
High-dosage tutoring at scale with AmeriCorps $1,461 per pupil served per year
volunteers (Kraft and Falken 2020)
HVAC replacement or upgrade $949 per pupil per year (spread over 4
years)
COVID testing prior to fall 2021 (universal) $2,430 per pupil per year
COVID testing from fall 2021 forward Cost to school districts: approximately
zero.
Summer school $1,402 per pupil per year
Hiring one full-time nurse per school $172 per pupil per year
Increasing socio-emotional supports $600 per pupil per year
Transition costs $500 per pupil one-time costs
See Cost Appendix for sources and calculations.
22
Table 2. Summary Statistics, District Characteristics
Mean Std Dev Median 10th pctile 90th pctile N
Enrollment 4,055 15,139 1,356 297 7,984 11,683
Child Poverty Rate 16.4% 9.0% 15.7% 5.6% 28.8% 11,673
Percent AIAN 1.1% 5.4% 0.3% 0.1% 1.3% 11,674
Percent Asian 5.4% 7.9% 2.5% 0.4% 14.9% 11,674
Percent Black 15.3% 18.2% 8.2% 0.7% 40.0% 11,674
Percent Hispanic 26.4% 25.3% 16.4% 2.8% 69.1% 11,674
Percent White 51.8% 30.1% 54.3% 9.5% 91.6% 11,674
PP Current Exp 13,029 4,853 11,548 9,005 19,267 11,683
PP Total Revenue 15,314 5,438 13,770 10,473 21,943 11,683
PP Local Revenue 7,105 4,821 5,720 2,631 13,830 11,683
PP State Aid 7,094 3,247 6,479 3,628 11,809 11,683
PP Federal Aid 1,115 738 999 440 1,926 11,683
Authors’ calculations. Summary statistics are weighted by enrollment (except enrollment). Enrollment, expenditure
and revenue data are from Census Annual Survey of School District Finances/F-33 Fiscal Year 2019; racial
composition is from NCES Common Core of Data; child poverty rate is from Census Small Area Income and
Poverty Estimates (SAIPE). Vermont, Hawaii, and Washington, D.C., districts with enrollment less than 150, and
districts with per-pupil total revenue greater than 5,000 or less than 40,000 are excluded.
Table 3. Simulated Cost and Revenue Scenarios
(1) (2) (3) (4)
Per-Pupil Per-Pupil Per-Pupil State Aid
One-Time Annual Annual Impact
Costs Cost for Costs for
Poor Non-Poor
Students Students
1. Baseline $500 $1,000 $500 None
2. Larger All-Student Costs $1,000 $1,000 $500 None
3. High Costs for Poor Students $500 $3,000 $500 None
4. State Aid Shortfall $500 $1,000 $500 Moderate
Per-pupil one-time costs (column 1) are assumed to have occurred once and apply to all students and districts. Per-
pupil costs for poor students (column 2) and for non-poor students (column 3) are average projected annual per-
pupil costs spent over four years for poor and non-poor students, respectively. The Moderate state aid impact
scenario assumes that state aid is reduced by the average of the FY2020 estimated impact and the middle scenario
for FY2021 divided by 3. See text for details.
23
Table 4. Predictors of Projected ESSER Allocation Per Pupil
(1) (2) (3) (4) (5)
Per-Pupil Per-Pupil Per-Pupil Per-Pupil Per-Pupil
ESSER ESSER ESSER ESSER ESSER
Share of Children 6069.2*** 5584.1*** 5731.5***
in Poverty (441.0) (354.4) (51.3)
Share AIAN 2410.9*** 888.6* 1011.7***
(447.7) (435.0) (69.8)
Share Asian -760.7 423.5 -137.7**
(426.3) (303.5) (49.5)
Share Black 1994.0*** 440.4** 624.1***
(205.0) (145.9) (26.0)
Share Hispanic 980.7*** 19.3 105.7***
(124.7) (164.0) (22.5)
ln(Enrollment) 99.3** 46.1* 79.5***
(31.7) (21.8) (2.7)
Constant -137.5*** 307.1*** -88.4 -602.9**
(39.0) (48.2) (280.9) (199.3)
Observations 11,673 11,674 11,683 11,665 11,665
State Fixed Effects No No No No Yes
Mean of the DV 857.9 857.3 857.7 857.4 857.4
R2 0.637 0.391 0.061 0.675 0.763
Robust standard errors in parentheses. Per-pupil ESSER values assume ESSER aid is spent over four years. See
text for details about simulated variables. Vermont, Hawaii, and Washington, D.C., districts with enrollment less
than 150, and districts with per-pupil total revenue greater than 5,000 or less than 40,000 are excluded.
*
p < 0.05, ** p < 0.01, *** p < 0.001
24
Table 5. Predictors of Per-Pupil Shortfall in State Aid, Scenario 4
(1) (2) (3) (4) (5)
PP State PP State PP State PP State PP State
Revenue Revenue Revenue Revenue Revenue
Shortfall Shortfall Shortfall Shortfall Shortfall
Share of Children 204.6** 193.4*** 286.7***
in Poverty (61.1) (47.5) (6.8)
Share AIAN 127.5 64.1 14.6
(124.9) (130.9) (9.2)
Share Asian -48.2 49.2 -96.6***
(35.1) (32.7) (6.5)
Share Black 21.1 0.3 4.0
(24.7) (32.3) (3.4)
Share Hispanic 44.3 35.8 39.1***
(42.3) (44.5) (3.0)
ln(Enrollment) -4.0 -7.7** -1.9***
(3.1) (2.6) (0.4)
Constant 128.4*** 148.1*** 200.5*** 190.2*** 126.8***
(10.7) (9.0) (25.3) (24.8) (3.2)
Observations 11673 11674 11683 11665 11665
State Fixed Effects No No No No Yes
Mean of the DV 162.0 161.9 162.0 161.9 161.9
R2 0.061 0.032 0.009 0.092 0.651
Robust standard errors in parentheses. Dependent variable is simulated per-pupil state aid shortfall; see text for
details. Vermont, Hawaii, and Washington, D.C., districts with enrollment less than 150, and districts with per-pupil
total revenue greater than 5,000 or less than 40,000 are excluded.
*
p < 0.05, ** p < 0.01, *** p < 0.001
25
Appendix Table 1. Summary Statistics, ESSER Funding
Mean Std Dev Median 10th pctile 90th pctile N
Estimated ESSER
Per Pupil 3,431 2,733 2,941 774 7,192 11,683
Estimated ESSER
Per Poor Pupil 19,660 8,487 18,525 11,085 29,774 11,673
ESSER Per Pupil
(AEI Data) 3,249 2,514 2,817 781 6,699 14,147
ESSER Per Poor
Pupil (AEI Data) 19,561 7,472 18,686 11,579 28,228 11,124
Authors’ calculations. Summary statistics are weighted by enrollment. See text for details. Vermont, Hawaii, and
Washington, D.C., districts with enrollment less than 150, and districts with per-pupil total revenue greater than
5,000 or less than 40,000 are excluded. AEI data (Malkus, 2021) include charter school districts, but we do not have
child poverty data for these districts so cannot calculate ESSER per poor child.
Appendix Table 2. Simulated Per-Pupil Funding Net Fiscal Impact, Alternative Scenarios
Mean Std Median 10th 90th N
Dev pctile pctile
1. Baseline 151 648 20 -461 1023 11,673
2. Higher All-Student Costs 26 648 -105 -586 898 11,673
3. Higher Cost for Poor Students -177 522 -302 -606 482 11,673
4. State Aid Shortfall -11 627 -149 -596 828 11,673
Authors’ calculations. Weighted by enrollment; see Methods Appendix for details. Vermont, Hawaii, and
Washington, D.C., districts with enrollment less than 150, and districts with per-pupil total revenue greater than
5,000 or less than 40,000 are excluded.
26
Appendix Table 3a. Predictors of Per-Pupil Net Revenue Impact, Baseline (Scenario 1)
(1) (2) (3) (4) (5)
PP Net PP Net PP Net PP Net PP Net
Impact Impact Impact Impact Impact
Share of Children 5569.2*** 5084.1*** 5231.5***
in Poverty (441.0) (354.4) (51.3)
Share AIAN 2271.3*** 888.6* 1011.7***
(445.8) (435.0) (69.8)
Share Asian -629.6 423.5 -137.7**
(401.1) (303.5) (49.5)
Share Black 1869.0*** 440.4** 624.1***
(197.2) (145.9) (26.0)
Share Hispanic 906.1*** 19.3 105.7***
(127.5) (164.0) (22.5)
ln(Enrollment) 96.2** 46.1* 79.5***
(30.5) (21.8) (2.7)
Constant -762.5*** -366.4*** -765.4** -1227.9*** -1592.0***
(39.0) (44.5) (268.0) (199.3) (23.9)
Observations 11673 11665 11673 11665 11665
State Fixed Effects No No No No Yes
Mean of the DV 150.9 150.5 150.9 150.5 150.5
R2 0.597 0.377 0.064 0.638 0.737
Robust standard errors in parentheses. Dependent variable is simulated annual per-pupil net revenue; see text for
details. Vermont, Hawaii, and Washington, D.C., districts with enrollment less than 150, and districts with per-pupil
total revenue greater than 5,000 or less than 40,000 are excluded.
*
p < 0.05, ** p < 0.01, *** p < 0.001
27
Appendix Table 3b. Predictors of Per-Pupil Net Revenue Impact, Higher All-Student Costs
(Scenario 2)
(1) (2) (3) (4) (5)
PP Net PP Net PP Net PP Net PP Net
Impact Impact Impact Impact Impact
*** ***
Share of Children 5569.2 5084.1 5231.5***
in Poverty (441.0) (354.4) (51.3)
Share AIAN 2271.3*** 888.6* 1011.7***
(445.8) (435.0) (69.8)
Share Asian -629.6 423.5 -137.7**
(401.1) (303.5) (49.5)
Share Black 1869.0*** 440.4** 624.1***
(197.2) (145.9) (26.0)
Share Hispanic 906.1*** 19.3 105.7***
(127.5) (164.0) (22.5)
ln(Enrollment) 96.2** 46.1* 79.5***
(30.5) (21.8) (2.7)
Constant -887.5*** -491.4*** -890.4** -1352.9*** -1717.0***
(39.0) (44.5) (268.0) (199.3) (23.9)
Observations 11673 11665 11673 11665 11665
State Fixed Effects No No No No Yes
Mean of the DV 25.9 25.5 25.9 25.5 25.5
R2 0.597 0.377 0.064 0.638 0.737
Robust standard errors in parentheses. Dependent variable is simulated annual per-pupil net revenue; see text for
details. Vermont, Hawaii, and Washington, D.C., districts with enrollment less than 150, and districts with per-pupil
total revenue greater than 5,000 or less than 40,000 are excluded.
*
p < 0.05, ** p < 0.01, *** p < 0.001
28
Appendix Table 3c. Predictors of Per-Pupil Net Revenue Impact, Higher Costs for Poor Students
(Scenario 3)
(1) (2) (3) (4) (5)
PP Net PP Net PP Net PP Net PP Net
Impact Impact Impact Impact Impact
*** ***
Share of Children 3569.2 3084.1 3231.5***
in Poverty (441.0) (354.4) (51.3)
Share AIAN 1706.6*** 888.6* 1011.7***
(439.3) (435.0) (69.8)
Share Asian -104.5 423.5 -137.7**
(330.4) (303.5) (49.5)
Share Black 1371.8*** 440.4** 624.1***
(169.2) (145.9) (26.0)
Share Hispanic 606.3*** 19.3 105.7***
(143.4) (164.0) (22.5)
ln(Enrollment) 83.7** 46.1* 79.5***
(25.8) (21.8) (2.7)
Constant -762.5*** -561.0*** -974.2*** -1227.9*** -1592.0***
(39.0) (31.2) (220.4) (199.3) (23.9)
Observations 11673 11665 11673 11665 11665
State Fixed Effects No No No No Yes
Mean of the DV -177.1 -177.4 -177.1 -177.4 -177.4
R2 0.378 0.292 0.074 0.442 0.595
Robust standard errors in parentheses. Dependent variable is simulated annual per-pupil net revenue; see text for
details. Vermont, Hawaii, and Washington, D.C., districts with enrollment less than 150, and districts with per-pupil
total revenue greater than 5,000 or less than 40,000 are excluded.
*
p < 0.05, ** p < 0.01, *** p < 0.001
29
Appendix Table 3d. Predictors of Per-Pupil Net Revenue Impact, State Aid Decline (Scenario 4)
(1) (2) (3) (4) (5)
PP Net PP Net PP Net PP Net PP Net
Impact Impact Impact Impact Impact
Share of Children 5364.6*** 4890.7*** 4944.8***
in Poverty (416.9) (340.6) (51.0)
Share AIAN 2143.8*** 824.5* 997.0***
(390.0) (366.6) (69.5)
Share Asian -581.3 374.3 -41.1
(406.3) (292.6) (49.3)
Share Black 1848.0*** 440.1** 620.1***
(188.3) (136.5) (25.8)
Share Hispanic 862.0*** -16.6 66.6**
(94.6) (129.5) (22.3)
ln(Enrollment) 100.2** 53.8* 81.4***
(28.6) (20.5) (2.7)
Constant -890.9*** -514.6*** -966.0*** -1418.1*** -1718.8***
(41.5) (42.4) (254.7) (184.3) (23.8)
Observations 11673 11665 11673 11665 11665
State Fixed Effects No No No No Yes
Mean of the DV -11.1 -11.4 -11.1 -11.4 -11.4
R2 0.592 0.383 0.074 0.642 0.722
Robust standard errors in parentheses. Dependent variable is simulated annual per-pupil net revenue; see text for
details. Vermont, Hawaii, and Washington, D.C., districts with enrollment less than 150, and districts with per-pupil
total revenue greater than 5,000 or less than 40,000 are excluded.
*
p < 0.05, ** p < 0.01, *** p < 0.001
30
Cost Appendix
This appendix explains the estimates in Table 1 and is not intended as a comprehensive summary
of all COVID-related costs schools may incur.
High-dosage tutoring
Anders, Guryan, and Ludwig (2016) conducted a randomized controlled trial of Match/SAGA
tutoring in Chicago Public Schools: on average, ninth and tenth grade students at risk of
dropping out gained one to two additional school years in math in a single year. That program
had one tutor working with two students at a time, teaching on-grade-level material aligned with
current classwork while relying on assessment to identify specific skills missed in previous years
and teaching those as well.
Not all tutoring programs are created equal, but the most effective so-called “high-dosage” ones
are consistently associated with a specific combination of programmatic features. High-dosage
tutoring, optimally implemented, should include: three to five 30-minute tutoring sessions per
week, during the school day at the school site; student:tutor ratios of no more than 4:1 and
ideally 2:1; the same tutor assigned to students for the entire school year; initial and ongoing
tutor training and supervision; and high quality instructional materials (Kraft and Falken 2020).
A meta-analysis drawing on studies conducted in a range of pre-COVID contexts found such
programs produce learning gains corresponding to three to 15 additional months of schooling
(Nickow, Oreopoulos, and Quan 2020).
HVAC replacement or upgrade
According to the GAO, 41 percent of school districts need to replace or update their HVAC
systems; based on this figure, approximately 36,000 schools need updates to or replacements for
their HVAC systems (Government Accountability Office 2020). If, theoretically, half of these
schools needed updates to their ventilation system, and the other half needed completely new
ventilation systems, the total cost for these updates would be approximately $72 billion (Griffith
and Pearce 2020). Dividing this figure by the 36,000 schools in need of new or updated HVAC
systems, this results in a cost of approximately $2 million per school. Based on the average
public school enrollment of 527 students per school, the cost per pupil is approximately $3,795
(Riser-Kositsky 2019). Spreading this over four years, the annual per-pupil cost would be $949.
COVID testing prior to fall 2021 (universal)
RAND reports test prices ranged from $5 (rapid) to $130 (PCR) per test (Evans, Schwartz, and
Master 2021); EdWeek reports a similar range from $10 to $120 (Gewertz 2021). $67.50 is the
midpoint of this range. If each student were tested once a week over a 180-day school year, the
cost per pupil would be $2,430.
COVID testing from fall 2021 forward
As of fall 2021, school districts can turn to a range of funding sources to support COVID testing.
For example, Los Angeles Unified School District, which is held up as a national exemplar with
its weekly universal testing program, is relying mainly on funds from the Federal Emergency
Management Association (Newberry, Gomez, and Blume 2021). The Department of Health and
Human Services is allocating $10 billion of its American Rescue Plan funding to support school
31
testing. In the 2020-21 school year, districts were likely to bear the costs of testing themselves,
though Massachusetts used state funds to support rapid testing from January 2021 (Finlaw 2021).
School districts in Pennsylvania can also access testing funded by the state for the 2021-22
school year (O’Neill and Alexander 2021).
Summer school
RAND reports that a five-week summer program consisting of six hours of academic and
enrichment activities each day would cost between $1,070 and $1,700 per student (Schwartz et
al. 2018). When adjusted to December 2020 dollars, this ranges between $1,340, or $1,464
(Evans, Schwartz, and Master 2021). When the average of these values is taken, the cost per
pupil would be $1,402.
Hiring one full-time nurse per school
Both the American Academy of Pediatrics and the National Association of School Nurses
recommend that each school have at least one registered nurse (Buttner 2021). During the 2015-
16 academic year, only 52 percent of public schools had a full-time nurse (National Center for
Education Statistics 2020). Based on the average school nurse salary of $64,630, and adding 40
percent of this cost for benefits, the average yearly cost associated with adding a full-time school
nurse would be $90,482 (Bureau of Labor Statistics 2021). Given that there is an average of 527
students in each school (Riser-Kositsky 2019), the cost per pupil of hiring one full-time
registered nurse in a school would be about $172.
Increasing socio-emotional supports
The National Association of School Psychologists predicts student needs for socio-emotional
supports have doubled or tripled under COVID. Zhou, Molfino, and Travers (2021) estimate that
increasing spending on these supports commensurately, by 250 percent, would cost
approximately $600 per pupil.
Transition costs
This category could include spending associated with transitioning to remote instruction (such as
purchasing equipment for teachers and students, training teachers, or purchasing curricula
materials) or with opening safely for in-person instruction (such as upgrading ventilation,
reducing group size or operating more busses to increase spacing, or purchasing masks and other
personal protective equipment). We model these costs as a constant per-pupil amount that is
incurred once, though in practice these costs began in the Spring of 2020 and due to the spread of
the delta variant may continue to be relevant in the 2021-22 school year; we assume per-pupil
one-time costs of $500 in the baseline scenario. 9
9
We draw on Zhou, Molfino and Travers (2021), who report on interviews with “leaders of large, urban school
systems.” These leaders reported new costs over the spring of 2020, of up to $750-1,000 per pupil. The largest
component of these new costs was technology for implementing distance learning. They also describe simultaneous
cost savings from cutting some in-person programming, like field trips and sports, and cuts to staff who did not hold
full-time, full-year positions; these savings were limited to $200-300 per pupil. Zhou, Molfino and Travers attribute
these relatively small cost savings to districts’ attempts to remain ready to pivot back to in-person instruction.
Because the additional costs are described as “as high as approximately $750-1,000 per-pupil” and we do not see the
full distribution, we use the lower end of that range for our estimate. If a district incurred $750 in new costs, and
saved $250 (the midpoint of the savings range), the net new costs would be $500.
32
Methods Appendix
We assume that school districts incurred one-time all-student costs. These could be costs related
to the transition to remote learning (like new curriculum or training or devices) or to COVID
mitigation during in-person learning (for example, upgrades to ventilation systems,
reconfiguration of the physical plant, personal protective equipment (PPE), cleaning supplies,
etc.). Separately, we assume that districts face ongoing costs, spread out over four years starting
during the 2020-21 school year; we allow these per-pupil costs to be higher for students in
poverty. This is meant to capture the notion that districts need to spend additional money to
address the educational and other needs of students because of the pandemic. These needs—and
the approaches to addressing them—may vary across districts and change over time. For
example, districts may have spent on additional transportation or reduced group size in 2020-21
to facilitate in-person instruction, but in later years they may shift to funding additional
counselors or tutors to accelerate learning and address mental health issues. As a shorthand, we
refer to these ongoing costs as recovery costs, but they could include a range of costs; the key is
that they vary depending on the child poverty rate of the district. 10
We simulate the net fiscal impact of COVID and federal aid as follows. Parameters in bold are
varied in different scenarios, summarized in Table 3.
1. Simulated Total ESSER is equal to Title I funding (reported in F33) X 12. The total of the
three ESSER funds was about 12 times the total allocated for Title I in the most recent
year, and ESSER funding is proportional to Title I, so scaling Title I provides an estimate
of ESSER funding.
2. Simulated Total One-Time COVID Costs is equal to Per-Pupil One-Time All-Student
COVID Cost (assuming to be $500 in the Baseline Scenario) X enrollment (reported in
F33).
3. Simulated Annualized Net Total ESSER is Simulated Total ESSER less Simulated Total
One-Time COVID Costs, annualized by dividing by four. Dividing by four is an
approximation, as different rounds of ESSER came with different spending timelines.
Simulated Annualized Net Total ESSER is what is available to districts for recovery,
annualized. Note that some districts did not receive enough ESSER funding to cover
these fixed costs. Those districts will have negative Net Total ESSER.
10
Districts have until September 2024 to obligate funds from ARP, so the funds could be spent over more than four
years. Similarly, districts mostly did not access CARES funding in the Spring of 2020, though they knew that
funding would become available soon. For the purposes of this exercise, we are less concerned about the precise
timing of spending but want to allow a distinction between one-time costs and ongoing costs and differential
ongoing costs for poor and non-poor students.
33
4. Simulated Total State Aid is equal to Total State Aid (from F33) X State Revenue Shock
Factor (equal to 1 in the baseline scenario, indicating no change in state aid). In
alternative scenarios, we assume shortfalls in state revenue translate proportionally to
reductions in state aid to school districts, as discussed in the main text.
5. Simulated Total Revenue is equal to Actual Local Revenue (from F33) plus Actual
Federal Aid (from F33) plus Simulated Total State Aid plus Simulated Net Total ESSER.
This assumes no change in local revenue or to federal aid other than ESSER. In the
baseline scenario, this is simply reported total revenue plus estimated annualized ESSER
funding.
6. The Number of Poor Students is equal to Enrollment (from F33) X Child Poverty Rate
(from SAIPE). The Number of Non-Poor Students is equal to Enrollment less the Number
of Poor Students. Note that the child poverty rate in SAIPE is for children living in the
district, so this calculation implicitly assumes that the poverty rate among public school
enrollment is the same as for residents.
7. Simulated Annualized Recovery Cost for Poor Students is equal to the Number of Poor
Students X Per-Pupil Annualized Recovery Cost for Poor Students (equal to $1,000 in
the Baseline Scenario). The Simulated Annualized Recovery Cost for Non-Poor Students
is equal to the Number of Non-Poor Students X Per-Pupil Annualized Recovery Cost for
Non-Poor Students (equal to $500 in the Baseline Scenario).
8. Simulated Total Required Revenue is equal to Actual Total Revenue (from F33) plus
Simulated Annualized Recovery Cost for Poor Students plus Simulated Annualized
Recovery Cost for Non-Poor Students. (Recall that the one-time costs were netted out of
simulated ESSER.)
9. Simulated Per-Pupil Shortfall is equal to Simulated Total Required Revenue less
Simulated Total Revenue, divided by Enrollment (from F33). This is the key output of the
exercise.
34
File and source
- File
- were-federal-covid-relief-funds-for-schools-enough.pdf
- Size
- 599,777 bytes
- SHA-256
- 965e2f9a4b96345436ce99af3d7ca8a58d03bd5e466282adfc1cf7f8a7929f7e
- Original
- www.nber.org