ESSER Funding and School System Jobs: Evidence from Job Posting Data — CALDER (No. 297-0225-2, February 2025)
Summary
A February 2025 working paper, ESSER Funding and School System Jobs: Evidence from Job Posting Data, by Dan Goldhaber, Grace Falken and Roddy Theobald (Working Paper No. 297-0225-2). The abstract states that the Elementary and Secondary School Emergency Relief Fund was the largest one-time federal investment in K-12 schools in history, funneling almost $200 billion to states and school districts. It states that the authors use web-scraped job posting data from Washington State to investigate the impact of ESSER spending on district hiring between January 2022 and September 2024. It states that they find strong evidence that ESSER funding significantly increased teacher hiring, with the greatest impact in the early ESSER years.
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
- Issuer
- Source documents
- Document type
- Esser Funding And School System Jobs Evidence From Job Posting Data No
- Case
- Esser Funding And School System Jobs Evidence From Job Posting Data No
Cite this document
“ESSER Funding and School System Jobs: Evidence from Job Posting Data — CALDER (No. 297-0225-2, February 2025).”
Original at caldercenter.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/esser-funding-and-school-system-jobs-evidence-from-job-posting-data-no/
, SHA-256 06a7405e319356daa31881c88de5e6e3ad8806d89c83550efc6243e9ef1f5c06.Full text
ESSER Funding
and School
System Jobs:
Evidence from
Job Posting Data
Dan Goldhaber
Grace Falken
Roddy Theobald
February 2025
WORKING PAPER No. 297-0225-2
ESSER Funding and School
System Jobs: Evidence from
Job Posting Data
Dan Goldhaber
American Institutes for Research / CALDER
University of Washington
Grace Falken
University of Washington
Roddy Theobald
American Institutes for Research / CALDER
Contents
Contents ..................................................................................................................................................................... i
Acknowledgments .................................................................................................................................................... ii
Abstract ................................................................................................................................................................... iii
1. Introduction ...................................................................................................................................................... 1
2. Background on ESSER Allocations and the Washington State Context .......................................................... 4
3. Data and Measures ........................................................................................................................................... 8
4. Empirical Methods ......................................................................................................................................... 15
5. Results ............................................................................................................................................................ 18
6. Discussion and Conclusions ........................................................................................................................... 23
References .............................................................................................................................................................. 28
Figures and Tables.................................................................................................................................................. 33
Appendix A. Aligning S-275 Activity Codes with Job Posting Categories ........................................................... 41
Appendix B. Lower-Bound Position Cost Estimation Method and Results ........................................................... 43
i
Acknowledgments
We would like to thank the Washington Office of Superintendent of Public Instruction (OSPI) for
funding that allowed us collect job postings from May 5, 2022, to June 30, 2023. This research was
also supported by the Center for Analysis of Longitudinal Data in Education Research (CALDER),
which is funded by a consortium of foundations. Additional thanks to Nate Brown, Luke Caldwell,
Trevor Gratz, and Nathaniel Marcuson for research assistance, and to Chad Aldeman, Matt
Barnum, Michael DeArmond, Nora Gordon, Kaori Strunk, and Gema Zamarro for comments that
improved the manuscript.
Disclosure Statements. The Washington Office of Superintendent of Public Instruction (OSPI)
provided funding that allowed us collect job postings from May 5, 2022, to June 30, 2023. This
research was also supported by the Center for Analysis of Longitudinal Data in Education
Research (CALDER), which is funded by a consortium of foundations; for more information on
CALDER’s funding sources, see here: https://caldercenter.org/about-calder. Dr. Theobald’s time
has been funded by the following organizations: William T. Grant Foundation, Institute for
Education Sciences, Spencer Foundation, Bill and Melinda Gates Foundation, Smith Richardson
Foundation, and the National Science Foundation.
CALDER working papers have not undergone final formal review and should be cited as working
papers. They are intended to encourage discussion and suggestions for revision before final
publication. Any opinions, findings, and conclusions expressed in these papers are those of the
authors and do not necessarily reflect the views of our funders or the institutions to which the
authors are affiliated. All errors and opinions are our own.
CALDER • American Institutes for Research
1400 Crystal Drive 10th Floor, Arlington, VA 22202
202-403-5796 • www.caldercenter.org
ii
ESSER Funding and School System Jobs: Evidence from Job Posting Data
Dan Goldhaber, Grace Falken, & Roddy Theobald
CALDER Working Paper No. 297-0225-2
February 2025
Abstract
The Elementary and Secondary School Emergency Relief Fund (ESSER) was the largest
onetime federal investment in K–12 schools in history, funneling almost $200 billion to states
and school districts. We use web-scraped job posting data from Washington State to investigate
the causal impact of ESSER spending on district hiring between January 2022 and September
2024. We employ an instrumental variables strategy, exploiting idiosyncrasies in the Title I
funding formula to isolate plausibly exogenous variation in ESSER. We find strong evidence
that ESSER funding significantly increased teacher hiring, and that this impact was greatest
during the early ESSER years.
iii
1. Introduction
In response to the COVID-19 pandemic, the federal government provided $190 billion in
funding to states and school districts as part of the Elementary and Secondary School Emergency
Relief Fund (ESSER)—the largest one-time federal investment in K–12 schools in history.
Lawmakers intended this program to help schools safely restart in-person instruction and recover
from significant pandemic-related learning losses (Jack et al., 2023; Lewis et al., 2021; National
Center for Education Statistics, 2022a, 2022b). While two recent studies find positive impacts of
ESSER dollars on student achievement (Dewey et al., 2024; Goldhaber & Falken, 2024),
surprisingly little is known about the specific ways that Local Education Agencies (LEAs) spent
these funds.
There is evidence from district spending proposals—structured plans required for districts
to receive their allocated funds—that a major area of intended ESSER investment was personnel
(DiMarco & Jordan, 2022), but assessing ESSER’s effects on staffing is complicated by both
data and methodological issues. ESSER claims data might appear to provide the most direct
measure of staffing spending, but there are important limitations to the use of these data. First,
the claims data categories are quite broad, such that it is not possible to discern whether funding
is going towards new or existing staff salaries. Second, school spending is fungible (Brunner et
al., 2022; Gordon, 2004; Lauth & Robbins, 2002), hence itemized ESSER claims may not
accurately reflect positions that were created as a consequence of ESSER funding. Put another
way, if districts use ESSER funds to pay for things they would have spent money on regardless
of receiving ESSER (and used other funding to pay for), studying claims data would lead to
erroneous conclusions about the causal effects of ESSER funding on school resource use. As a
concrete example, LEAs in Washington claimed $288,000 of ESSER funds to pay for
1
superintendent salaries, but it is unlikely that any of these LEAs would have operated without a
superintendent in the absence of ESSER funding.
The core question in this paper is the extent to which ESSER caused hiring by LEAs that
would not have happened without this funding. Understanding the impact of ESSER spending on
hiring sheds light on policymaker choices when provided with significant, time-limited funding.
The number and jobs created by ESSER also provides an estimate of the number of jobs, in the
absence of additional funding, that may be at risk with ESSER’s expiration. We address this
issue by collecting job postings directly from school district websites in Washington State and
use these scraped job postings data as our preferred measure of LEA hiring to address the
following research questions:
1. What is the relationship between ESSER spending and school district hiring
across different job categories?
2. What is the causal impact of ESSER spending on school district hiring, both
overall and specifically for teachers?
3. How did the impact of ESSER spending on hiring vary over time?
Our analysis of RQ1 is purely descriptive, but the positive association between ESSER
and district hiring supports findings from earlier analyses that staffing was a major focus of
spending plans, with a particular focus on teachers (DiMarco & Jordan, 2022; Schwartz &
Bolves, 2022). Our interest in the causal impact of ESSER for RQ2 and RQ3 necessitates a more
rigorous methodological approach than prior descriptive analyses that can address two sources of
bias. First, naïve estimates of the relationship between ESSER spending and staffing decisions
may be confounded by other unobserved differences between high-poverty and low-poverty
districts that are strongly correlated with ESSER spending. High-poverty districts, for example,
2
face exacerbated issues such as chronic absenteeism and achievement gaps following the
pandemic which could have driven them to hire more staff in the absence of ESSER funds. This
would result in an upward bias in the relationship between remaining ESSER funds and hires. At
the same time, those districts may also have limited administrative bandwidth to expand hiring
operations in the short term, leading to a negative association between ESSER funds and hires.
Second, the pace of ESSER spending in a district may be correlated with the number or type of
hires. For example, districts may have more remaining ESSER funds because of limited
administrative capacity to execute hiring processes or due to greater hiring challenges such as
sparse labor supply. In this case, having more unspent ESSER dollars would be negatively
correlated with hiring, resulting in a downward bias in our estimator of interest.
In our preferred specification, we account for these sources of bias by using counts of
formula-eligible children (FEC) for the 2020–21 school year—the primary determinant of Title I
funding—to instrument for unspent ESSER funds. The identifying assumption in these models is
that the number of FEC living in a district (the majority of which come from census estimates of
children in poverty from four years earlier) does not affect district hiring plans once we control
for district enrollment, free or reduced-price lunch (FRPL) qualification, demographic
representation, historical district revenue, and prior teacher attrition. That is, the identifying
variation in remaining ESSER funds comes from the time lag in poverty data reporting for FEC,
differences in populations within a district’s geography versus those enrolled in its public
schools, and measurement error in the construction of FEC estimates.
Using this approach, we find positive but statistically insignificant impacts of ESSER
spending on district hiring overall, but we do find significant and meaningful impacts
specifically for teacher hiring. Our estimates show that each $10,000 in unspent ESSER funds
3
caused districts to hire about $1,500 in additional teaching staff who they would not have
otherwise, on average across the three years of our data (2022-2024); these estimated effects are
strongest and only statistically significant in 2022 and 2023.
This paper contributes to our understanding of the ways school systems responded to
student needs in the wake of the COVID pandemic, a topic over which there is significant
speculation but little quantitative evidence. More broadly, we contribute to the literature on how
school systems allocate resources when provided with a large increase in revenue (Lauth &
Robbins, 2002; Sun et al., 2022). Because we examine hiring across a range of position types,
our findings reflect what school systems value in the absence of constraints that might link
spending to any specific areas or student types (Fisher & Papke, 2000).
Our results also have practical implications. Understanding the ways school systems
responded to ESSER funding can shed light on what we might expect now that ESSER funding
has ended. Our analysis suggests that districts in Washington hired roughly 8,400 new teachers
because of ESSER. Although it is likely that, just as in the aftermath of the Great Recession,
some of the necessary downsizing of the state’s teacher workforce in the absence of these funds
can be managed through attrition, we argue that the end of ESSER will once again lead to
teacher layoffs, especially in the districts that received larger ESSER allocations.
2. Background on ESSER Allocations and the Washington State Context
2.1 Background on ESSER Allocations
ESSER temporarily increased federal funding for public schools to deal with the
consequences of the COVID-19 pandemic. Across three waves of grants (ESSER I, II, and III),
the ESSER program allocated a total of $190 billion to K-12 school systems. ESSER represents
more than triple the $60 billion of federal funding allocated to districts in the 2019–20 school
year (Cornman et al., 2022) and nearly four times the extra funding districts received through the
4
American Recovery and Reinvestment Act of 2009 (Shores & Steinberg, 2022). In Washington
State, ESSER boosted district budgets by an average of about $2,300 per pupil ($760 per pupil
per year), or about 20% of state-provided funding for districts in 2019–20.
The U.S. Department of Education allocated ESSER in proportion to Title I funding for
2019–20 (ESSER I) and 2020–21 (ESSER II & III). Title I comprises four distinct grant
formulas, all of which are based on the number and percentage of FEC in a district area. The
majority of FEC consist of the number of children in poverty ages 5 to 17 in a school district
area, regardless of their public school enrollment status (Snyder et al., 2019). To be eligible for
each of the four Title I grants, LEAs must be above threshold numbers and percentages of FEC.
While some evidence documents schools manipulating poverty measures within LEAs
(Matsudaira et al., 2012), allocations to LEAs use FEC and other time-lagged, non-manipulable
measures. Allocations are scaled by the number of FEC and adjusted by state per-pupil
expenditures, hold harmless provisions, and state minimum provisions (Gordon & Reber, 2023;
Snyder et al., 2019).
The American Rescue Plan (ESSER III) distributed the bulk of this funding. ESSER III
required LEAs to earmark 20% of their allocation for recovering learning loss, but the remaining
80% was not restricted. Federal guidance explicitly notes that districts could use the funds to
support activities such as “continuing to employ existing staff of the LEA” (U.S. Department of
Education, 2022, p. 12), and even funds designated for supporting learning loss could be used for
this purpose. As a result, LEAs likely used some ESSER funds to maintain staffing levels in the
face of contracting enrollments (Schwartz et al., 2023) as well as to hire new staff.
Analyses of ESSER spending priorities according to LEA proposals (as distinct from
actual expenditures) suggest that this was the case. DiMarco and Jordan (2022) identify the three
5
largest budget priorities by dollar amount across a sample of 5,004 school districts nationwide as
staffing, academic recovery, and facilities and operations, with staffing accounting for 27% of
the total $64 billion budgeted by these districts. Spending priorities also appear to differ across
LEA poverty level (Jordan & DiMarco, 2022b) and geography (Jordan & DiMarco, 2022a). For
example, districts in the highest poverty quartile were the only group whose first priority was
HVAC investment—all other quartiles prioritized staffing (Jordan & DiMarco, 2022b).
Two reports from Rhode Island shed light on differences between proposed and realized
ESSER spending. Schwartz and Bolves (2022) use data capturing districts’ proposed ESSER
spending to identify priorities, similar to DiMarco and Jordan (2022). They project that if ESSER
budgets for staff were entirely directed toward new hires, the grant would fund about 1,100 full-
time equivalent (FTE) instructors throughout Rhode Island, approximately 10% of the
instructional workforce. In a follow-up report looking at ESSER expenses, Schwartz et al. (2023)
find that ESSER spending supported some new positions but predominantly (49%) paid for
existing staff with overall staffing levels remaining unchanged. Similarly, DiMarco and Kelleher
(2023) find that much of the ESSER spending on staff in North Carolina funded onetime bonuses
(e.g., retention incentives) rather than supporting new positions.
Although these findings may not be generalizable to other contexts, it is possible that the
high volume of ESSER dollars originally budgeted for staffing provided a smaller boost to hiring
than initially expected in other contexts, too. There is limited national evidence about how
ESSER funding influenced district spending, but data from the Bureau of Labor Statistics and
Census Bureau shows that nationwide K–12 employment has grown over the last three years.
Public schools now employ more full-time instructional staff than before the pandemic
(Aldeman, 2023a, 2023b). But, as we emphasize above, the fungibility of funding across budget
6
items means that we cannot draw conclusions about the extent to which ESSER caused hiring:
districts may have chosen to hire more or different personnel after the pandemic even in the
absence of ESSER funds.
2.2 ESSER Allocations and Spending in Washington State
Districts in Washington began claiming reimbursements for ESSER-funded expenses in
July 2020, March 2021, and August 2021 for ESSER I, II, and III, respectively. The three waves’
deadlines to obligate funds were September 30th, 2022, 2023, and 2024. By January 2022,
districts had claimed (i.e., spent and had reimbursed) 26% of total ESSER allocations—98% of
ESSER I, 50% of ESSER II, and 7% of ESSER III (Washington Office of Superintendent of
Public Instruction, 2023a). An additional 34%, 28%, and 12% of all ESSER funds were claimed
in the 2022, 2023, and 2024 calendar years that followed (Washington Office of Superintendent
of Public Instruction, 2023a). It is difficult to pin down exactly when districts incur expenses
because funds flow through a chain of reporting in order to appear on state summaries
(Silberstein & Roza, 2023), but these high-level totals suggest that spending volume peaked in
the middle of the ESSER grant period.
Districts varied tremendously in their spending in part because the amount of funding
districts received and funds as a share of total revenue differed widely across the state. Figure 1
illustrates how spending varies widely across districts due to meaningful differences in initial
allocated funding and differential spending rates. For instance, some (wealthy) districts in
Washington had small unspent allocations despite having spent relatively little of their total grant
funding; Snoqualmie School District, for instance, had $517 per pupil unspent in July 2022
despite having spent only 36% of their total ESSER allocation. Others received far more funding
and therefore had larger per pupil unspent amounts despite, in some cases, having spent out a
7
larger share of total ESSER allocations. Granger School District, for instance, still had $1,800
remaining per pupil as of July 2022 despite having spent their allocation at twice the rate (73%)
of Snoqualmie. Much of the variation in unspent funds was due to the link to the Title I
allocation formula. ESSER funding is allocated per FEC so relative district poverty and the share
of the local population that FEC represent lead to large differences in funds per enrolled pupil.
One aspect of ESSER funding not captured in our analysis is the potential impact of
state-reserved ESSER funds. ESSER funds flowed first to SEAs which withheld 10% of the
grants for administrative costs and state initiatives. SEAs funneled the remaining 90% to LEAs
(National Conference of State Legislatures, 2022). Washington SEA funds went toward staffing
21st Century Learning Centers, afterschool programs managed by LEAs, and community groups;
the state also supported health services for regional clusters of districts (Washington Office of
Superintendent of Public Instruction, 2023b). Arrangements like these pose an issue for
identifying the impacts of ESSER on overall employment that may benefit districts.
In Washington, most district funding is derived from state revenue (about 75% of funding
on average) with a strong equalization formula and limited caps of local levy revenue (Knight et
al., 2022). This focus on funding equity is also reflected in the state use of SEA funding:
Washington funneled SEA ESSER funds to districts that were getting little to no ESSER dollars
according to the grant formula. All districts in Washington received at least $75,000 across the
three grants due to this supplemental provision of SEA ESSER funding to districts.
3. Data and Measures
3.1 Web-Scraped Job Postings Data
We leverage novel, web-scraped job postings to measure impacts on the number of new
positions and the implied costs of those positions. We scraped district websites and job boards
twice weekly (Mondays and Fridays) starting in the fall of 2021 to collect information on
8
advertised public school jobs; we exclude posts that were already on district websites prior to
2022 to isolate those that we can observe appearing online. Thus, we have information about the
district-by-calendar-year-level of job postings for the full years of 2022, 2023, and through
September 30th of 2024.
Because district websites were scraped continuously, we also observe when postings are
removed. We assume posts that are removed from district websites represent positions that were
filled and refer to them as “filled posts” (as we note below, our results are robust to using all
posts instead of filled posts). We focus on this measure of hiring throughout our analysis because
we are interested in the impacts of ESSER on LEA hiring (and relatedly the budgetary burden
implied by ESSER-induced hiring). Using job titles listed in postings, consistent with Goldhaber,
Falken, et al. (2023), we categorize each post into one of 11 categories: administration, athletics,
facilities, food services, health, paraeducator, principal, superintendent, teaching, transportation,
and a catch-all, other.
In attributing jobs to district budgets, we must consider measurement issues associated
with assigning filled posts to a school year or calendar year. Specifically, for the initial data of
job postings in fall 2021, we cannot know when the jobs in this period were initially posted,
meaning we attribute all posts that were open in the fall to that initial season, even if they were
posted in the prior summer. If we construct school-year totals, we would thus be overcounting
posts for the 2021-22 school year. The tail end of our data in 2024 introduces a different
measurement issue: because we limit analyses to filled posts, and these posts have had less time
to be taken down than those of prior years, we may undercount positions at the end of our period.
We therefore assign posts to calendar years to minimize the potential for bias associated with
each of these measurement issues (though as we note below, the findings are similar if we
9
instead use the school year and if we use all posts instead of filled posts). Specifically, we know
from data in 2022 and 2023 that about 30% of posts that are new in the fall are filled in the
following calendar year, whereas 40% of filled posts in the fall were initially posted in the
preceding summer (a different school year). This suggests that the potential risk of overcounting
posts by using school years (and including fall 2021) is a greater concern than that of
undercounting posts by using calendar years.
Job postings data have two additional shortcomings worth mentioning. First, these data
fail to cover the entirety of the ESSER funding period. School systems had access to ESSER
funds from March 2020 through September 2024. Because the web scraping only began in the
fall of 2021, the data we use is missing more than a full school year of the grant period. That
said, we suspect ESSER-induced hiring was low in this early period for the following reasons:
school systems did not yet have the lion’s share of ESSER funding (ESSER III, approved in
March 2021); the implementation of any large-scale budget initiatives is often lagged from the
receipt of funding, as evidenced by the fact that the average district in Washington had only
spent 17% of their total ESSER allocation as of July 2021; and, a majority of districts were
operating remotely during this period (Goldhaber, Kane, et al., 2023). Additionally, spending is
delayed by the school district budget adoption process. Funds cannot typically be spent before
the district budgeting process, meaning that although ESSER II was passed in December 2020, it
would not impact district hiring until after budget deliberations in the spring and summer of 2021
(Seattle Public Schools Board of Directors, 2021).
A second limitation is that job postings (including those we classify as filled posts) are
only an approximation of hires. These records aren’t linkable to observed hires, making them
difficult to validate; earlier research has illustrated that filled postings at the district-level are
10
highly correlated with observed teacher hires (Goldhaber, Falken, et al., 2023), but does not
assess validity for non-teaching positions. Postings do have the appeal, however, of capturing
temporary hires for categories such as athletics which may not appear in administrative data.
3.2 Salary and Staffing Data in Washington
The Washington Office of Superintendent of Public Instruction (OSPI) maintains several
publicly available datasets we also use for this analysis. We use OSPI’s personnel dataset that
tracks all staff positions in Washington public schools over time, the S-275, to construct the
typical pay across different job functions in each district, which we use as a scaling factor for job
posts, and to calculate the number of all new hires and new teachers hired in each district in each
year, which we use as a lagged control in our models. Because this dataset is created for
accounting purposes, it disaggregates the compensation of each staff member across each distinct
job function they hold. Job functions are identified using program, activity, and duty root codes,
as well as certification status; we provide a detailed breakdown of which activity codes and
certification categories we align with each of our job posting categories in Appendix A.
We assign each position in the S-275 to the same 11 job categories specified above to
calculate the average pay for positions within each category for all Washington districts in each
year. Because newly-hired teachers are likely disproportionately novice, we also estimate a
“lower bound” post cost using average position salaries among new staff. By construction,
results with this outcome are attenuated in magnitude but qualitatively similar to those we
present below. We describe the construction of this measure and present results using this
outcome in Appendix B.
Because individuals may hold multiple position types—for example, a classroom teacher
may also coach a sports team—we include the portion of annual pay attributed to each position
11
category in the calculations for average pay in each separate category. This means that we
calculate the cost per position, rather than per FTE. For many job types, this is a more
appropriate cost estimate because nonteaching positions are often coded as part time (see
Theobald et al., 2025), extracurricular positions often do not have an associated FTE estimate,
and it is common for staff to span multiple positions in a school. Because not all job posts reflect
equal implied cost, our primary outcome scales each filled post by the average salary for that
position type in that district in that year. We multiply our counts of filled posts in each category
and district by the associated salary to approximate the future staffing costs implied by the filled
job posts. These projected post costs are our preferred outcome for ease of interpretation.
We present the average number of job postings in each category for each year, average
pay for each job category, and average projected post costs in Table 1. The first three columns
here break down the average number of posts in a district for each year of our panel. In each
year, the largest category by average posted positions is paraeducators, followed by teachers.
Across years, 2022 has the highest volume of job posts. The total number of posts across all
categories drops by 20% in 2023 and again in 2024, although, as we note above, we cut off 2024
posts as of September 30th, the ESSER funding obligation deadline. The rightmost three columns
present average measures across all districts in all three years of data. The projected post costs
show that while paraeducator positions had the largest number of posts in the typical district, the
post costs of teaching position far outweighed investments in other types of staff, totaling an
average $10 million in a typical district-year. By contrast, while post demand for athletic staff is
generally the fourth-highest category by posts, the relatively low cost of the typical athletics
position means that the projected position costs for athletics are only 5% of overall post costs
across all areas.
12
3.3 Other Administrative Data Sources
As we note above, district context may influence district hiring needs; hence, in the
models described below, we use district-by-year characteristics from Washington State Report
Card data as controls. These measures include district enrollment, counts of students qualifying
for FRPL, student demographics, urbanicity, and average scores on the Washington State
standardized assessment (Smarter Balance Assessment Consortium assessment, SBAC). We link
all covariates, which are on a school year timeline, to the calendar year corresponding to the
spring of school year. In practice this means that student enrollment for 2021-22 (which is
captured in fall 2021) is the control for predicting posts in the 2022 calendar year. We also
calculate and include distances to the nearest Teacher Education Program (TEP) using data from
the Education Demographic and Geographic Estimates (EDGE) given evidence that this
proximity influences teacher vacancies (Edwards et al., 2024). To account for labor market
dynamics (Rucinski, 2023), we use county-level unemployment data from the Local Area
Unemployment Statistics dataset maintained by the Bureau of Labor Statistics. As a control for
pre-pandemic funding, we use 2019–20 district revenue from the Common Core of Data.
Our IV analysis depends on data that the U.S. Department of Education (ED) uses to
calculate Title I allocations. The ED allocates Title I funds across districts according to the count
of FEC in the district’s geographic area. Formula-eligible children (FEC) counts include children
in poverty between the ages of 5 and 17, children receiving Temporary Assistance for Needy
Families (TANF), neglected and delinquent children, and foster children. The lion’s share of
FEC children (97%) fall into the first category (Stephenson & Kaiser, 2018), calculated each
year by the Census’ Small Area Income and Poverty Estimates (SAIPE) program. SAIPE data
are published with a slight delay relative to when Title I allocations are calculated: SAIPE data
from 2018 determined Title I allocations in the 2020–21 school year which, in turn, impacted
13
ESSER II and III allocations. This is important for our instruments as we use district-level counts
of FEC in Washington to instrument for a time-varying measure of unspent ESSER funding. We
expand on the identifying variation isolated by this instrument below.
Finally, we use data from OSPI documenting district ESSER allocations and spending in
each fiscal year; fiscal years in Washington run from July 1st through June 30th. We construct a
time-varying measure we call “unspent ESSER funds” as our preferred measure of ESSER.
Unspent ESSER funds are the total allocations minus cumulative spent funds as of the prior
summer. For 2022 calendar-year job postings, this means we use unspent ESSER funds as of
July 1st, 2021 as our predictor of interest. The benefit of this construction is that it approximates
a remaining grant budget that administrators faced as they planned their hiring for the following
year. While budget decisions are typically contained to a single-year cycle, the multi-year nature
of this grant likely means that districts planned spending considerate of their total remaining
funds and longer-term plans for the grant period.
Our sample includes 277 districts (of 295 in Washington) in 2022, 290 districts in 2023,
and 289 districts in 2024 for a total of 856 district-year observations. Differences in our sample
are due to an expansion of our job-scraping infrastructure in late 2022 and enrollment dropping
to zero in one district in 2024. That said, in each year we observe posts for districts serving at
least 99.8% of students in Washington public schools and 99.7% of staff. Omitted districts are
extremely small, generally rural, and often are omitted for lacking an active, public-facing hiring
website to scrape. In Figure 1, we visually present differences in per-pupil ESSER allocations
across Washington State districts; in Table 2, we describe the characteristics of these districts,
weighted by enrollment. Because there are vast differences in ESSER allocations per pupil
across districts, Table 2 also presents differences across the subsamples of district-years in the
14
bottom quartile of total ESSER allocations per pupil (Column 2) and the top quartile of total
ESSER allocations per pupil (Column 3). The average ESSER allocation per pupil among
districts in our sample was $3,075. Districts in the top quartile by ESSER allocations per pupil
received $6,105 per student, whereas those in the bottom quartile received $1,009 on average.
Not surprisingly given that ESSER allocations are tied to Title I, there are significant
differences in district characteristics associated with allocations. Districts that received the
highest ESSER allocations per pupil tended to be larger than the typical Washington district,
were more commonly in urban settings or towns, and served higher rates of students of color,
English-language learners, and students receiving FRPL. Generally, districts receiving the lowest
ESSER allocations per pupil exhibit the inverse of these patterns, with a notable
overrepresentation of suburban districts relative to the state average. These differences across
districts are driven partly by differences in the relative poverty of the district—that is, what share
of the school-age population are FEC. High-ESSER districts posted about twice as many
positions across all years as their low-ESSER counterparts.
4. Empirical Methods
We use three approaches to understand the relationship between ESSER funding and
LEA hiring in Washington. First, we estimate a seemingly unrelated regression (SUR) to
estimate the naïve association between unspent ESSER funds and hiring for each job category
we observe. This model involves a system of 11 equations, each of which takes the following
generalized form:
𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 = 𝛽𝛽0𝑐𝑐 + 𝛽𝛽1𝑐𝑐 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑑𝑑𝑑𝑑 + 𝛽𝛽2𝑐𝑐 𝑋𝑋𝑑𝑑𝑑𝑑 + 𝜀𝜀𝑐𝑐𝑐𝑐𝑐𝑐 (1)
15
The outcome of Equation 1, 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 , is the projected position costs of filled posts
𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢
in a category (c) in a district (d) in a year (t). Our main predictor of interest, 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑑𝑑𝑑𝑑 ,
captures district-by-year remaining ESSER funding as of the prior July.
The entire system includes 11 equations predicting this outcome for each observed
category in a district. We also include a vector, 𝑋𝑋𝑑𝑑𝑑𝑑 , of time-varying covariates likely to be
correlated with ESSER funding and district staffing needs, including a cubic of district
enrollment, district-level counts of students receiving FRPL, the number of underrepresented
minority (URM) students, the 2019–20 total district revenue, average SBAC scores, the change
in enrollment between the prior and current school year (standardized), the county
unemployment rate, indicators for district urbanicity, the number of new hires in a district as of
the prior October, and a log of the distance to the nearest TEP. We intend for these covariates to
capture some district-level variation in both labor market environment (e.g., unemployment and
proximity to a supply of teachers) and school environments associated with variation in staffing
patterns (e.g., the household income and demographic composition of enrolled students). We
control for pre-pandemic district revenue to capture potential associations between sustained
differences in funding and staffing, the drivers of which would be funding from federal and local
sources, given the strong equalization embedded in the state funding structure. We weight all
models by district enrollment and interpret estimates as those for the average student.
These SUR estimates disaggregate naïve ordinary least squares (OLS) estimates of the
association between unspent ESSER and total hiring across categories, which we present
alongside these results. For the coefficient on unspent ESSER to be interpreted as causal, we
would need to account for all aspects of districts related to the amount of unspent ESSER funds
16
and to hiring. As noted above, there are good reasons to worry that “observably similar” districts
according to the variables in 𝑋𝑋𝑑𝑑𝑑𝑑 may differ in ways that impact both ESSER allocations and
hiring, suggesting that these estimates could suffer from omitted variable bias. Our specific
concern is that if districts that receive more funds because of variation in other unobservable (to
us) factors also would have more job postings in the absence of ESSER funds, then these
omissions could bias our estimates of the relationship between unspent ESSER and posts.
Our solution is to instrument for unspent ESSER funds with the number of FEC within
district boundaries. Using a two-stage least squares model (2SLS), our first stage of this
estimation takes the following form:
𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢
𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑑𝑑𝑑𝑑 = 𝛽𝛽0 + 𝛽𝛽1 𝐹𝐹𝐹𝐹𝐹𝐹𝑑𝑑 + 𝛽𝛽2 𝑋𝑋𝑑𝑑𝑑𝑑 + 𝜀𝜀𝑑𝑑𝑑𝑑 (2)
All variables in equation 2 are defined as above and we also include the specified
instrument, 𝐹𝐹𝐹𝐹𝐹𝐹𝑑𝑑 , to predict district unspent funds. We present our first stage results in Table 3,
which we discuss further below. Instrumenting for unspent ESSER funds using time-invariant
FEC removes differences in unspent funds attributable to rates of spending, essentially assuming
the rates of spending are constant across districts. This proves a strong instrument for spending,
with an individual F-statistic of 78.45 (Andrews et al., 2019; Lee et al., 2022).
Using FEC as an instrument also accounts for covariances between student needs and
hiring demand in addition to those between hiring and district spending rates. The identifying
variation when we use FEC to instrument for unspent ESSER comes from three main sources.
First, the number of FEC in a district’s geographic area, predominantly the number of children
living in poverty in 2018, differs from 2021–22 district poverty because of the time lag between
when the data are available and the year of funding they determine. Second, differences between
17
the local population of children in a district area and those enrolled in public schools may differ
if there are high rates of private school enrollment or homeschooling. Third, while FEC data are
based on estimates anchored to the most recent American Community Survey, our control for
students qualifying for free- or reduced-price lunch are based on observed students in the
district—a difference that introduces measurement error as a source of variation.
The assumption underlying FEC as an instrument is that, after accounting for factors like
the number of students qualifying for FRPL and lagged district revenue (including Title I
revenue), FEC in a district only affects staffing decisions through its impact on ESSER funding.
If this exclusion restriction holds—and if this variable is sufficiently predictive of ESSER—then
we can estimate 2SLS models that isolate the causal effect of unspent ESSER funds on district
postings. A threat to this exclusion restriction comes from the first source of identifying variation
described above: if districts becoming relatively more or less economically disadvantaged in
recent years (e.g., with more students qualifying for FRPL than we’d expect given prior FEC)
also have different staffing trajectories and needs, then lagged FEC might not be exogenous to
staffing decisions. We assume that this is not the case, but we acknowledge this as a limitation of
our identification strategy when we estimate variants of equation 2.
5. Results
5.1 Associations Between Unspent ESSER Funding and Hiring Outcomes Across
Categories
In Table 4 we present naïve associations between unspent ESSER and hiring outcomes
for each job category. We first show a simple OLS model (Column 1) predicting total filled
postings (Panel A) and post costs (in $10,000s, Panel B). These estimates suggest that a $10,000
18
increase in unspent ESSER is associated with a statistically significant 0.042 increase in filled
posts and a $1,390 increase in total position costs across all categories.
The SUR results (Columns 2 through 11) disaggregate this overall association across job
categories, shedding light on which areas are driving this association. Predictably, we find
mixed impacts of ESSER allocations on category-specific filled posts and total post costs,
suggesting that the impact of ESSER varies across job types. For example, we find the largest
magnitude of effect for teaching positions, relative to all other position types, with a suggestive
impact of $910 increased posted position costs for a $10,000 increase in ESSER allocations. We
also find significant positive associations between ESSER allocations and projected position
costs for several job categories, the largest of which is administrative staff ($430 increase per
$10,000 of ESSER).
We take the above findings as evidence that the association between ESSER and hiring
was greatest for teaching positions. The relative costliness of teaching positions is reflected in
the divergence between the estimated associations for posts and post costs for teaching staff. Our
estimates in Table 4 suggest that 33% of the total job posting effect (column 1) was teacher job
postings (column 10) while teachers made up 65% of the overall impact for post costs. So, while
other areas such as administrative staff exhibit meaningful associations in terms of number of
filled positions, teaching positions appear to make up most of ESSER investments in staff.
There is one significant, negative association between ESSER allocations and projected
posting costs; athletic staff position costs are estimated to be $430 lower with a marginal
increase of $10,000 in unspent ESSER funds. That ESSER would reduce hiring in any area is
surprising, however, recall that these results are descriptive. It may also be possible that high-
need districts which received more ESSER funds are also less likely to invest in sports programs
19
in the wake of the pandemic, for instance because the school systems are more likely to be online
(Goldhaber et al., 2022).
5.2 Causal Impacts of ESSER on Hiring Outcomes Overall and for Teachers
Our first stage results, presented in Table 3, suggest that FEC is an effective instrument
for unspent ESSER funding. Specifically, in the average year a marginal increase in FEC
predicts a $18,700 increase in unspent ESSER, with annual estimates ranging from $22,100 in
2022 to $14,900 in 2024. This average across years is notably similar to the descriptive weighted
average funding per FEC in our sample, about $17,000 ($18,000 unweighted).
In Table 5 we present results aiming to remove the two potential sources of endogeneity
we discussed above. Columns 1 and 3 present OLS estimates as a point of comparison,
reiterating the estimates in Column 1 of Table 4. Columns 2 and 4 instrument for unspent
ESSER funding using FEC. Considering the difference between results in column 1 and 2, we
find that instrumenting diminishes our estimated impact on total filled posts (Panel A) by 52%
and decreases the filled post costs estimate (in $10,000s, comparing columns 4 and 5) by 59%.
Our estimates suggest that a $10,000 increase in unspent ESSER funds for districts leads to an
insignificant increase of 0.020 posts and an increase in costs of $570.
When we narrow our perspective to teacher-specific outcomes in Panel B, however, we
find our results are more consistent across OLS and 2SLS specifications. The 2SLS estimates
using FEC to instrument for unspent ESSER are slightly larger, though statistically
indistinguishable from, those in our OLS framework. We find that a $10,000 increase in unspent
ESSER translates to a 0.019 increase in filled teacher postings. In terms of costs, this marginal
increase in ESSER drives a $1,500 increase in posted teacher position costs. We interpret this
20
difference as suggestive evidence that non-teaching positions may be subject to more
endogenous variation than teaching positions.
We conduct several robustness checks (results available upon request) to assess the
specificity of these estimates to our modeling decisions. If we instead estimate unweighted
regressions—that is, estimated relationships for the average district in Washington rather than
the average student—the findings are similar, though slightly larger in magnitude for
(aggregated) total posts and total post costs. Using all posts instead of limiting to filled posts
returns estimates almost identical to those presented in Table 5, suggesting that right-censoring
of posts does not materially impact our findings. Finally, when we assign posts to school years
instead of calendar years, which offers a more intuitive time unit at the cost of introducing left-
censoring of posts in 2021-22, we again find results qualitatively similar to those depicted here.
Taken together, our causal estimates suggest that 15% of ESSER funding went
specifically towards teaching positions that would not have existed in the absence of this
funding. These estimates are far more conservative than prior estimates from one analysis of 22
states that about 50% of ESSER was spent on personnel (Silberstein & Roza, 2023), but that
figure includes spending both on new hires and existing staff, the latter of which are generally
paid more and thus make up a greater expense (DiMarco & Kelleher, 2023; Schwartz et al.,
2023; Silberstein & Roza, 2023).
5.3 Heterogeneous Impacts of ESSER on Teacher Job Postings
It seems likely that ESSER funding would have heterogeneous impacts on hiring. For
instance, it is highly unlikely that ESSER funds would equally impact hiring across all grant
years. Thus, to get a better sense of the evolution of ESSER across years, we estimate equation
21
(2) for each calendar year separately, allowing us to observe a flexible relationship between
unspent ESSER funds and hiring outcomes over time.
We present the magnitude of these teacher-specific estimates in Figure 2 (posts) and
Figure 3 (post costs). Each figure displays the predicted values of the indicated outcome in the
typical district for a specific level of unspent ESSER funding. That is, we visualize the marginal
differences in hiring predicted by our ESSER point estimate, with the leftmost predicted value
corresponding to districts with zero predicted unspent ESSER and the rightmost predicted value
corresponding to the 95th percentile of predicted unspent ESSER in each year.
Figure 2 shows how districts at the 95th percentile of predicted remaining ESSER hired
about 90 and 30 more teachers they would not have hired without the ESSER funding in 2022
and 2023, and about 20 fewer teachers in 2024. In Figure 3 this same comparison illustrates a
$7.2 million impact on teacher post costs of going from no predicted unspent ESSER to the 95th
percentile in 2022. This drops to $2.4 million in 2023 and a decline of $2.5 million in 2024. In
sum, these figures show a slowdown of the impact of ESSER funding on teacher hiring as the
slope presented in each year shifts from meaningfully positive to a slight negative.
There are arguments for expecting heterogeneity across other district characteristics too.
For instance, while there is relatively limited variation in per pupil spending in Washington,
ESSER funding as a share of district revenues varies across districts and districts vary in terms of
their student characteristics. We do not, however, find evidence that there are differential effects
of ESSER on hiring across district characteristics. We assess this by again estimating our models
for subsamples of our data. We find no consistent patterns in our estimated effects when we
group districts into the share of students eligible for free or reduced-price lunch, the share of
students in under-represented minority groups, or by pre-pandemic revenue. Our assessment is
22
that there is no meaningful evidence of heterogeneous effects along these dimensions of district
financial situation and student characteristics, but we also have limited power to detect
heterogeneity across districts because there are only 290 in the sample.
6. Discussion and Conclusions
We find strong, arguably causal, evidence that public school districts hired more teaching
staff in response to the availability of ESSER funding. To put the financial magnitude of these
hires into perspective, our estimates suggest that districts in Washington hired $665 million
worth of teaching staff over the course of the three years that would not have been hired in the
absence of the ESSER funding. Consistent with a recent review of ESSER proposed spending
data (Brooks & Springer, 2024), it appears that much of this hiring activity focused on teaching
positions.
Given these results, what might the end of ESSER funding mean, both for school budgets
and school staffing? Nationwide, the one-year drop in per pupil funding implied by the end of
ESSER is comparable to funding drops during the Great Recession that were seen over three
years—about $1,400 per student from the 2011–12 to 2013–14 school years (Roza & Silberstein,
2023; Shores & Steinberg, 2022). Because personnel costs make up 76% of education
expenditures (Executive Office of the President & Council of Economic Advisors, 2025),
personnel are highly likely to be affected by these meaningful budget changes. Estimates of how
many personnel will be affected can vary depending on the method.
Schwartz and Bolves’ (2022) approach estimates the number of positions that would not
have been funded if ESSER had not happened by dividing budgeted ESSER funds by the typical
salary in a position type. If we apply this approach to Washington, dividing teacher salary claims
by our lower bound teacher salary measure, it suggests that roughly 8,300 teaching positions
were created by ESSER. However, this may overstate the number of positions likely at risk
23
because it assumes (a) that all funds went to support new hires rather than existing staff, and (b)
that budgets are not fungible.
Aldeman (2023b) takes a different approach, exploiting changes in student–teacher ratios
between the 2018–19 and 2021–22 school years. By applying his method to changes in ratios
between 2018-19 and 2023-24 in Washington, we find that the state’s teacher workforce needs to
be reduced by about 2,400 teaching positions to return to pre-pandemic staffing levels. But this
estimate likely understates the number of teachers at risk post-ESSER. The argument here is
specific to Washington: student-teacher ratios in Washington likely would have risen in the
absence of ESSER funding because, in recent years, teacher salaries have risen faster than what
was guaranteed by increases in state funding in the wake of the state’s McCleary reforms
(Knight & Fujioka, 2023).
Our alternative is to estimate the impact of unspent ESSER funding on filled job postings
to estimate the number of teaching positions that were created by ESSER, which will potentially
be at risk when ESSER ends. We use the coefficient for unspent ESSER on teacher job postings
from our 2SLS model (Table 5, Column 3) and the average unspent ESSER funding across our
three panel years in Washington (about $1.5 billion) to estimate the likely statewide impact of
ESSER funds on the employment over the three years that districts could use the funds. We find
that unspent ESSER funding increased the number of filled teacher posts in Washington districts
by 8,428 positions between 2022 and 2024, with a 90% confidence interval of 5,633 to 11,222.
This point estimate is higher than our estimate using the method of Aldeman (2023b) but similar
to that using the method of Schwartz and Bolves (2022).
For some additional perspective, our estimate of the number of teaching jobs created by
ESSER (8,428) represents 13% of the current Washington public teaching workforce and 24% of
24
all new hires over this three-year period. Importantly, this estimate of ESSER-created jobs likely
overstates the number of current staff whose jobs are at risk. That is because districts can address
some staffing reductions through attrition (roughly 6,000 teachers leave the state teaching
workforce in a typical year). As an example for how this can play out in practice, in the wake of
the Great Recession (between 2008–09 and 2011–12), the number of teachers in Washington
declined by 3,030. The state’s school districts managed much of this decline by hiring fewer new
teachers. Of the 15,080 teachers who left the workforce during this period, districts hired only
12,050 new teachers to replace them. As a result, only 561 teachers (i.e., about 18% of the 3,030
reduced positions) were actually laid off statewide (Goldhaber et al., 2016).
An important caveat to these back-of-the-envelope estimates is that, while we construct
these numbers from our average effects, the distributions of hiring and potential layoffs are not
equal across districts. As we discussed above (and is reported in Table 2), the students in districts
in the top quartile by ESSER allocations serve different student populations than those in the
bottom quartile. 55% of students in the average top-quartile district are URM, 20% are ELL, and
67% receive FRPL. In the bottom quartile, by contrast, the average district serves only 29%, 8%,
and 30% of these student groups, respectively. If we apply the same back-of-the-envelope
method based on the average unspent ESSER in these quartiles, we find that the typical district
in the bottom quartile by ESSER allocations hired an additional 12 teachers across three years
that they would not have hired in the absence of this funding. In the top quartile, however, the
average district hired 38 additional teachers—more than triple that of the bottom quartile
average. The point is that districts that received large allocations of ESSER funding did the bulk
of the additional hiring; these same districts are also the most likely to face budget challenges
25
and the risk of layoffs now that the grant period is over. The students most likely to be affected
by these layoffs are relatively disadvantaged.
Of course, in practice, layoffs occur within districts and not statewide. This complicates
the extent to which policymakers can use attrition to address possible ESSER-related staffing
problems. For example, as shown in Figure 1, many districts received small ESSER allocations;
attrition in these districts will not address the need to reduce staff in districts where large
allocations were associated with ESSER-induced hires. Similarly, within districts, attrition may
not occur in the right subjects. If a district has hired elementary teachers with its ESSER funds, it
cannot manage reductions with retirements in high school math departments. If districts are
unable to manage reductions via attrition, it is likely that the end of ESSER funding will be
followed by significant layoffs, as was the case with the Great Recession. This is concerning
because typical approaches to layoffs (e.g., seniority-based, last-in-first-out layoffs and
overissuing “pink slips”) can have direct and indirect negative effects on student achievement
(Goldhaber et al., 2016; Goldhaber & Theobald, 2013; Kraft, 2015).
To the degree that the end of ESSER leads to teacher layoffs, school districts should be
exploring ways to mitigate their harm. For example, they might consider protecting teachers (and
their students) from layoffs in specific shortage areas, like special education and STEM subjects;
or they might protect teachers in hard-to-staff schools that can least afford to lose teachers
(Bleiberg & Kraft, 2023). District and union leaders also need to be concerned about the process
that governs layoff notices. Following the Great Recession, school districts in Washington issued
about five times as many layoff notices as the number of actual layoffs. This was due to
collective bargaining agreements, which often require districts to notify all teachers who might
be at risk of losing their jobs before the next school year (Goldhaber et al., 2016). Yet, research
26
suggests that the receipt of a layoff notice induces greater mobility among teachers (Goldhaber et
al., 2016), which can harm student achievement (Ronfeldt et al., 2013) and teacher performance
(Strunk et al., 2018). Admittedly, given the way teacher shortages dominated the news during the
pandemic layoffs seem unthinkable. But if the end of ESSER, budget constraints, and enrollment
shifts create the need for layoffs, district leaders will have to find better ways to deal with them if
they want to avoid unintentionally harming teachers and students.
27
References
Aldeman, C. (2023a, July 31). New employment data: 5 things to know about the state of the
education workforce. The 74. https://www.the74million.org/article/new-employment-
data-5-things-to-know-about-the-state-of-the-education-workforce/
Aldeman, C. (2023b, October 25). Schools could lose 136,000 teaching jobs when federal
COVID funds run out. The74. https://www.the74million.org/article/schools-could-lose-
136000-teaching-jobs-when-federal-covid-funds-run-out/
Andrews, I., Stock, J. H., & Sun, L. (2019). Weak Instruments in Instrumental Variables
Regression: Theory and Practice. Annual Review of Economics, 11(Volume 11, 2019),
727–753. https://doi.org/10.1146/annurev-economics-080218-025643
Bleiberg, J. F., & Kraft, M. A. (2023). What Happened to the K–12 Education Labor Market
During COVID? The Acute Need for Better Data Systems. Education Finance and
Policy, 18(1), 156–172. https://doi.org/10.1162/edfp_a_00391
Brooks, C. D., & Springer, M. G. (2024). ESSER-ting preferences: Examining school district
preferences for using federal pandemic relief fundings (No. 24–913; EdWorkingPaper).
Annenberg Institute at Brown University. https://edworkingpapers.com/ai24-913
Brunner, E., Hoen, B., & Hyman, J. (2022). School district revenue shocks, resource allocations,
and student achievement: Evidence from the universe of U.S. wind energy installations.
Journal of Public Economics, 206, 104586.
https://doi.org/10.1016/j.jpubeco.2021.104586
Cornman, S. Q., Phillips, J. J., Howell, M. R., & Zhou, L. (2022). Revenues and expenditures for
public elementary and secondary education: FY 20 (NCES 2022-301). U.S. Department
of Education, National Center for Education Statistics.
https://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=2022301
Dewey, D., Fahle, E. M., Kane, T. J., Reardon, S. F., & Staiger, D. O. (2024). Federal panemic
relief and academic recovery. Education Recovery Scorecard.
https://educationrecoveryscorecard.org/wp-content/uploads/2024/06/June2024ERS-
Report.pdf
DiMarco, B., & Jordan, P. W. (2022, July 7). Financial trends in local schools’ Covid-aid
spending. FutureEd. https://www.future-ed.org/financial-trends-in-local-schools-covid-
aid-spending/
DiMarco, B., & Kelleher, M. (2023). How North Carolina’s schools are staffing their pandemic
recovery. FutureEd. https://www.future-ed.org/how-north-carolinas-schools-are-staffing-
their-pandemic-recovery/
Edwards, D. S., Kraft, M. A., Christian, A., & Candelaria, C. A. (2024). Teacher Shortages: A
Framework for Understanding and Predicting Vacancies. Educational Evaluation and
Policy Analysis, 01623737241235224. https://doi.org/10.3102/01623737241235224
28
Executive Office of the President, & Council of Economic Advisors. (2025). Economic Report of
the President. U.S. Government Publishing Office. https://www.whitehouse.gov/wp-
content/uploads/2025/01/ERP-2025.pdf
Fisher, R. C., & Papke, L. E. (2000). Local government responses to education grants. National
Tax Journal, 53(1), 153–153.
Goldhaber, D., & Falken, G. T. (2024). ESSER and student achievement: Assessing the impacts
of the largest one-time federal investment in K12 schools (CALDER Working Paper No.
301–0624). Center for Analysis of Longitudinal Data in Education Research at the
American Institutes for Research. https://caldercenter.org/publications/esser-and-student-
achievement-assessing-impacts-largest-one-time-federal-investment-k12
Goldhaber, D., Falken, G. T., & Theobald, R. (2023). What do teacher job postings tell us about
school hiring needs and equity? (CALDER Working Paper No. 282–0323). American
Institutes for Research.
Goldhaber, D., Imberman, S. A., Strunk, K. O., Hopkins, B. G., Brown, N., Harbatkin, E., &
Kilbride, T. (2022). To what extent does in-person schooling contribute to the spread of
Covid-19? Evidence from Michigan and Washington. Journal of Policy Analysis and
Management, 41(1), 1–384. https://doi.org/10.1002/pam.22354
Goldhaber, D., Kane, T. J., McEachin, A., Morton, E., Patterson, T., & Staiger, D. O. (2023).
The educational consequences of remote and hybrid instruction during the pandemic.
American Economic Review: Insights, 5(3), 377–392.
https://doi.org/10.1257/aeri.20220180
Goldhaber, D., Strunk, K. O., Brown, N., & Knight, D. S. (2016). Lessons learned from the
Great Recession: layoffs and the RIF-induced teacher shuffle. Educational Evaluation
and Policy Analysis, 38(3), 517–548.
Goldhaber, D., & Theobald, R. (2013). Managing the teacher workforce in austere times: The
determinants and implications of teacher layoffs. Education Finance and Policy, 8(4),
494–527. https://doi.org/10.1162/EDFP_a_00111
Gordon, N. (2004). Do federal grants boost school spending? Evidence from Title I. Journal of
Public Economics, 88(9), 1771–1792. https://doi.org/10.1016/j.jpubeco.2003.09.002
Gordon, N., & Reber, S. (2023). Title I of ESEA: How the formulas work (Understanding and
Improving Title I of ESEA). All4Ed. https://all4ed.org/publication/title-i-of-esea-how-
the-formulas-work/
Jack, R., Halloran, C., Okun, J., & Oster, E. (2023). Pandemic schooling mode and student test
scores: Evidence from US school districts. American Economic Review: Insights, 5(2),
173–190. https://doi.org/10.1257/aeri.20210748
29
Jordan, P. W., & DiMarco, B. (2022a, March 1). National, Regional Trends in Educators’ Covid-
Relief Spending - FutureEd. FutureEd. https://www.future-ed.org/national-reading-
trends-covid-relief-spending/
Jordan, P. W., & DiMarco, B. (2022b, June 8). How district poverty levels influence Covid-relief
spending. FutureEd. https://www.future-ed.org/how-district-poverty-levels-influence-
covid-relief-spending/
Knight, D. S., & Fujioka, K. (2023, April 14). Post McCleary, WA school funding doesn’t add
up. The Seattle Times. https://www.seattletimes.com/opinion/post-mccleary-wa-school-
funding-doesnt-add-up/
Knight, D. S., Pooya, A., & Berge, J. (2022). Washington school finance: Exploring the history
and present-day challenges for fiscal equity (No. 23–702; EdWorkingPaper). Annenberg
Institute at Brown University. https://www.edworkingpapers.com/ai23-702
Kraft, M. A. (2015). Teacher layoffs, teacher quality, and student achievement: evidence from a
discretionary layoff policy. Education Finance and Policy, 10(4), 467–507.
https://doi.org/10.1162/EDFP_a_00171
Lauth, T. P., & Robbins, M. D. (2002). The Georgia Lottery and state appropriations for
education: Substitution or additional funding? Public Budgeting & Finance, 22(3), 89–
100. https://doi.org/10.1111/1540-5850.00082
Lee, D. S., McCrary, J., Moreira, M. J., & Porter, J. (2022). Valid t-Ratio Inference for IV.
American Economic Review, 112(10), 3260–3290. https://doi.org/10.1257/aer.20211063
Lewis, K., Kuhfeld, M., Ruzek, E., & McEachin, A. (2021). Learning during COVID-19:
reading and math achievement in the 2020-21 school year. NWEA, Center for School
and Student Progress.
Matsudaira, J. D., Hosek, A., & Walsh, E. (2012). An integrated assessment of the effects of
Title I on school behavior, resources, and student achievement. Economics of Education
Review, 31(3), 1–14. https://doi.org/10.1016/j.econedurev.2012.01.002
National Center for Education Statistics. (2022a). 2022 NAEP mathematics assessments:
Highlighted results at grades 4 and 8 for the nation, states, and districts (NCES
2022124). U.S. Department of Education.
https://www.nationsreportcard.gov/highlights/mathematics/2022/
National Center for Education Statistics. (2022b). 2022 NAEP reading assessment: Highlighted
results at grades 4 and 8 for the nation, states, and districts (NCES 2022126). U.S.
Department of Education. https://www.nationsreportcard.gov/highlights/reading/2022/
National Conference of State Legislatures. (2022, January 25). Elementary and Secondary
School Emergency Relief Fund tracker. NCSL. https://www.ncsl.org/in-dc/standing-
committees/education/elementary-and-secondary-school-emergency-relief-fund-tracker
30
Ronfeldt, M., Loeb, S., & Wyckoff, J. (2013). How teacher turnover harms student achievement.
American Educational Research Journal, 50(1), 4–36.
https://doi.org/10.3102/0002831212463813
Roza, M., & Silberstein, K. (2023). The ESSER fiscal cliff will have serious implications for
student equity. Brookings Institution. https://www.brookings.edu/articles/the-esser-fiscal-
cliff-will-have-serious-implications-for-student-equity/
Rucinski, M. (2023). The effects of economic conditions on the labor market for teachers. In
Annenberg Institute for School Reform at Brown University (No. 23–884;
EdWorkingPaper). Annenberg Institute for School Reform at Brown University.
https://eric.ed.gov/?id=ED638891
Schwartz, N., & Bolves, A. J. (2022). Investing in the Future: How Rhode Island Districts are
Spending Federal COVID-19 Funding and Why It Matters | Annenberg Institute at
Brown (EdResearch for Recovery). Annenberg Institute at Brown University.
https://annenberginstitute.org/ri-research/briefs/esser
Schwartz, N., Donohue, K., Bolves, A. J., & Rodriguez Benitez, G. (2023). Tracking Rhode
Island’s COVID-19 recovery: A first look at districts’ strategic staffing and personnel
investments (No. 2; Rhode Island COVID-19 Recovery Spending Series). Annenberg
Institute at Brown University.
Seattle Public Schools Board of Directors. (2021). July 7, 2021 – Regular Board Meeting
[Regular Legislative Session]. https://www.seattleschools.org/board-meetings/july-7-
regular-board-meeting/
Shores, K., & Steinberg, M. P. (2022). Fiscal federalism and K–12 education funding: Policy
lessons from two educational crises. Educational Researcher, 51(8), 551–558.
https://doi.org/10.3102/0013189X221125764
Silberstein, K., & Roza, M. (2023, April 4). The massive ESSER experiment: Here’s what we’re
learning. Education Next. https://www.educationnext.org/the-massive-esser-experiment-
heres-what-were-learning/
Snyder, T. D., Dinkes, R., Sonnenberg, W., & Cornman, S. Q. (2019). Study of the Title I, Part A
Grant Program mathematical formulas (NCES 2019-016). Institute for Education
Sciences, National Center for Education Statistics.
Stephenson, T., & Kaiser, P. (2018, February 9). Title I allocation formulas. National Title I
Conference, Philadelphia, Pennsylvania.
https://www2.ed.gov/about/offices/list/oese/oss/technicalassistance/titleiallocationformul
astitleiconfppt22018.pdf
Strunk, K. O., Goldhaber, D., Knight, D. S., & Brown, N. (2018). Are there hidden costs
associated with conducting layoffs? The impact of reduction-in-force and layoff notices
on teacher effectiveness. Journal of Policy Analysis and Management, 37(4), 755–782.
https://doi.org/10.1002/pam.22074
31
Sun, M., Candelaria, C. A., Knight, D., LeClair, Z., Kabourek, S. E., & Chang, K. (2022). The
effects and local implementation of school finance reforms on teacher salary, hiring and
turnover. In EdWorkingPapers.com. Annenberg Institute at Brown University.
https://www.edworkingpapers.com/ai22-585
Theobald, R., Kaler, L., Bettini, E., & Jones, N. (2025). A descriptive portrait of the
paraeducator workforce in Washington State. Exceptional Children, 91(2), 166-186.
U.S. Department of Education. (2022). Frequently asked questions: Elementary and Secondary
School Emergency Relief Programs Governor’s Emergency Education Relief Programs.
U.S. Department of Education.
Washington Office of Superintendent of Public Instruction. (2023a). ALL ESSER dashboard.
ESSER Funding Dashboard.
https://app.smartsheet.com/b/publish?EQBCT=73a2ba95e92b45d0b89eb8dc29d25f92
Washington Office of Superintendent of Public Instruction. (2023b). Washington State plan for
the American Rescue Plan Elementary and Secondary School Emergency Relief Fund.
U.S. Office of Elementary & Secondary Education.
https://ospi.k12.wa.us/sites/default/files/2023-08/washington-state-arp-esser-state-plan-
approved-11-24-2021.pdf
32
Figures and Tables
Figure 1. Total ESSER Allocations and Unspent ESSER per Pupil as of July 1st, 2022
Note. Each bar represents a single district in the state. Allocations are time-invariant and scaled
per pupil in 2022-23. Unspent ESSER allocations are funds remaining as of July 1, 2022 and are
scaled by enrollment in 2022-23. We label the bars of two school districts (SDs) discussed in the
text, Snoqualmie and Granger.
33
Figure 2. Predicted Filled Posts Across Unspent ESSER
Notes. Markers represent predicted outcome values for the level of unspent ESSER from our
2SLS model in which we instrument for unspent ESSER with district FEC. Each panel presents
results from a single-year estimation. Brackets around each marker represent 95% confidence
intervals for those predicted values. The y-axis here represents the predicted number of filled
teachers posts for a typical district in the indicated year across different amounts of unspent
ESSER funding. Unspent ESSER represents the range from zero to the 95th percentile of
predicted values for that variable from our first stage estimation. ESSER=Elementary and
Secondary School Education Relief; FEC=formula-eligible children.
34
Figure 3. Predicted Filled Post Costs Across Unspent ESSER
Note. Markers represent predicted outcome values for the level of unspent ESSER from our
2SLS model in which we instrument for unspent ESSER with district FEC. Each panel presents
results from a single-year estimation. Brackets around each marker represent 95% confidence
intervals for those predicted values. The y-axis here represents the predicted total cost of filled
teachers posts for a typical district in the indicated year across different amounts of unspent
ESSER funding. Unspent ESSER represents the range zero to the 95th percentile of predicted
values for that variable from our first stage estimation. ESSER=Elementary and Secondary
School Education Relief; FEC=formula-eligible children.
35
Table 1. District Average Position Demand, Pay, and Projected Costs
Year-specific averages Panel averages
2022 2023 2024 Salary
Projected
Category average average average Posts per
post costs
posts posts posts position
Administration 49.7 34.1 23.5 35.8 59,098 2,124,649
Athletics 68.9 62.1 46.3 59.1 18,420 1,318,383
Facilities 26.6 20.7 15.3 20.9 58,173 1,205,427
Food 31.0 25.5 20.7 25.7 25,798 682,048
Health 32.2 24.6 20.4 25.7 72,434 1,889,677
Other 89.0 60.5 53.7 67.7 25,704 2,320,797
Paraeducator 148.8 139.0 103.9 130.5 30,521 4,136,981
Principal 8.5 5.5 5.1 6.4 134,570 825,547
Superintendent 0.7 0.2 0.2 0.4 241,229 99,152
Teaching 137.3 103.7 95.3 112.0 89,108 10,179,742
Transportation 9.4 8.7 5.8 8.0 40,083 298,063
Total 602.1 484.6 390.3 492.0 25,080,468
Notes. The first three columns present the average number of filled posts in the
typical district, by category, for each year of the panel. The final three columns
present the average filled posts in a district, average salary for a position type, and
projected post costs (i.e., the product of the prior two columns) averaged across
the three years in our panel. Each cell is weighted by year-specific district
enrollment.
36
Table 2. District-Year Characteristics, Weighted by Enrollment
Bottom Quartile Top Quartile ESSER
All Districts
ESSER districts districts
City district (%) 39.40 13.24 52.07
Rural district (%) 7.53 3.38 7.97
Suburban district (%) 39.27 72.62 21.31
Town district (%) 13.80 10.76 18.65
District enrollment (1,000s) 14.99 13.25 17.54
(12.00) (8.94) (10.76)
Underrepresented minority (%) 41.32 28.87 56.10
(18.57) (8.69) (21.37)
English language learners (%) 12.48 7.22 19.24
(9.49) (4.29) (11.90)
Special education (%) 14.51 13.21 15.28
(2.27) (2.38) (2.09)
Free or reduced-price lunch (%) 46.31 27.92 66.98
(20.53) (13.08) (11.91)
ESSER II & III allocations ($1,000,000s) 34.19 10.50 68.23
(37.37) (8.85) (39.79)
ESSER II & III claims ($1,000,000s) 21.21 7.90 35.17
(25.17) (6.21) (20.93)
District per pupil spending ($1,000s) 15.90 15.38 16.42
(1.58) (1.36) (1.39)
Total filled job posts 543.90 373.85 777.01
(401.04) (226.70) (484.39)
Administration filled posts 47.00 34.10 72.58
(42.21) (23.44) (63.50)
Athletics filled posts 57.89 45.94 104.03
(48.94) (28.68) (75.44)
Facilities filled posts 24.56 20.73 36.80
(25.25) (15.63) (39.42)
Food services filled posts 29.65 18.91 46.62
(31.25) (22.61) (34.28)
Health filled posts 26.34 19.21 29.10
(20.26) (11.07) (16.96)
Other filled posts 75.39 35.76 127.60
(110.11) (21.60) (160.43)
Paraeducator filled posts 132.07 94.58 155.55
(106.36) (80.81) (115.12)
Principal filled posts 7.94 7.38 9.91
(7.65) (7.87) (9.68)
Superintendent filled posts 0.67 0.44 0.97
(0.87) (0.68) (1.07)
Teacher filled posts 135.36 91.84 186.02
(125.49) (71.77) (113.69)
Transportation filled posts 7.04 4.95 7.82
(10.80) (5.32) (10.16)
N 276 69 69
Note. All averages are weighted by district enrollment. The first column presents average characteristics across all observed
districts. The second column reports averages for districts in the bottom quartile of total ESSER allocations per pupil; the
final column reports averages for districts in the top quartile of total ESSER allocations per pupil. Underrepresented
minority category includes American Native, Black, Hispanic, and Multiracial students.
37
Table 3. First Stage Results Predicting District-By-Year Unspent ESSER
Funds ($10,000s)
(1)
Formula-eligible children 1.87***
(0.21)
N 856
2
Adjusted R 0.85
Instrument F-stat 78.45
Notes. *p<.10; **p<.05; ***p<.01. We report standard errors, clustered by
district, in parentheses. All models are weighted by enrollment and include the
following district-year-level controls: district enrollment (cubic), 2019-20
district revenue, the number of free/reduced-price eligible students, the number
of URM students, average SBAC scores, standardized difference in student
enrollment from the prior year, average unemployment in the surrounding
county, attrition of teaching staff relative to the prior year, attrition of
nonteaching staff relative to the prior year, indicators for district urbanicity,
total new hires as of the prior October, and log distance to the nearest teacher
education program. All models include year fixed effects.
38
Table 4. Naïve Associations Between Unspent ESSER Funds and Hiring Outcomes
OLS Seemingly unrelated regression
Food Para- Super-
Total Admin Athletic Facilities Health Other Principal Teachers Transport
services educator intendent
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Panel A. Filled posts
Unspent ESSER 0.042*** 0.008*** 0.004*** 0.003*** 0.002*** -0.000 0.007*** 0.005*** 0.000*** 0.000*** 0.014*** -0.000**
funds ($10,000s) (0.011) (0.000) (0.001) (0.000) (0.000) (0.000) (0.001) (0.001) (0.000) (0.000) (0.001) (0.000)
n 856 856
Panel B. Filled post costs ($10,000s)
Unspent ESSER 0.015*** 0.043*** -0.043*** 0.016*** 0.004*** -0.003* 0.020*** 0.009** 0.003*** 0.001*** 0.091*** -0.001**
funds ($10,000s) (0.003) (0.001) (0.004) (0.001) (0.001) (0.002) (0.005) (0.004) (0.001) (0.000) (0.008) (0.000)
n 856 856
Notes. *p<.10; **p<.05; ***p<.01. We report standard errors, clustered by district, in parentheses. All models are weighted by enrollment and include the following district-
year-level controls: district enrollment (cubic), 2019-20 district revenue, the number of free/reduced-price eligible students, the number of URM students, average SBAC
scores, standardized difference in student enrollment from the prior year, average unemployment in the surrounding county, attrition of teaching staff relative to the prior
year, attrition of nonteaching staff relative to the prior year, total new hires as of the prior October, indicators for district urbanicity, log distance to the nearest teacher
education program, a control for ESSER funds expended as of the prior July, and year fixed effects.
39
Table 5. Estimated Impacts of ESSER on Hiring Outcomes
Filled post costs
Outcome Filled posts
($10,000s)
Specification OLS 2SLS OLS 2SLS
(1) (2) (3) (4)
Panel A. All job categories
Unspent ESSER funds ($10,000s) 0.042*** 0.020 0.139** 0.057
(0.011) (0.024) (0.068) (0.116)
n 856 856 856 856
Panel B. Teaching positions
Unspent ESSER funds ($10,000s) 0.015*** 0.019** 0.110*** 0.150**
(0.003) (0.008) (0.032) (0.068)
n 856 856 856 856
Notes. *p<.10; **p<.05; ***p<.01. We report standard errors, clustered by
district, in parentheses. All models are weighted by enrollment and include the
following district-year-level controls: district enrollment (cubic), 2019-20
district revenue, the number of free/reduced-price eligible students, the number
of underrepresented minority students, average SBAC scores, standardized
difference in student enrollment from the prior year, average unemployment in
the surrounding county, attrition of teaching staff relative to the prior year,
attrition of nonteaching staff relative to the prior year, total new hires as of the
prior October (total teacher hires in Panel B estimates), indicators for district
urbanicity, and log distance to the nearest teacher education program. All models
include year fixed effects. Unspent ESSER funds are grant balances as of the
prior July. The 2SLS specification in columns 2 and 4 use district formula-
eligible children to instrument for unspent ESSER funding. 2SLS=two-stage
least squares; FEC=formula-eligible children; OLS=ordinary least squares;
SBAC=Smarter Balance Assessment Consortium (test scores).
40
Appendix A. Aligning S-275 Activity Codes with Job Posting Categories
To estimate typical costs per position type at the district-level, we have assigned activity
codes in the S-275 to the 11 position categories we created for our job posting data. We also
assign individuals within each activity code an indicator for certificated staff or classified staff
because some activity codes (e.g., 27 Teaching) include multiple job posting categories that can
be distinguished by their certification status. In the example of teaching, classroom teachers are
identified as certificated staff whereas paraprofessionals are classified as teaching staff.
Table A.1 Assignment of activity codes to job posting categories
Classified /
Activity Code Position Category
Certificated
Classified Administration
11 – Board of Directors
Certificated Administration
Classified Administration
12 – Superintendent’s Office
Certificated Superintendent
Classified Administration
13 – Business Office
Certificated Administration
Classified Administration
14 – Human Resources
Certificated Administration
Classified Administration
15 – Public Relations
Certificated Administration
Classified Teaching
21 – Teaching & Learning Supervision
Certificated Teaching
Classified Other (nonteaching)
22 – Learning Resources
Certificated Other (nonteaching)
Classified Administration
23 – Principal’s Office
Certificated Principal
Classified Health
24 – Guidance & Counseling
Certificated Health
Certificated Other (nonteaching)
25 – Pupil Management & Safety
Classified Other (nonteaching)
Classified Health
26 – Health & Related Services
Certificated Health
Classified Paraeducator
27 –Teaching
Certificated Teaching
Classified Athletics
28 – Extracurricular
Certificated Athletics
31 – Instructional Professional Development Classified Other (nonteaching)
41
Certificated Other (nonteaching)
Classified Other (nonteaching)
32 – Instructional Technology
Certificated Other (nonteaching)
Classified Other (nonteaching)
33 – Curriculum
Certificated Other (nonteaching)
Classified Food Service
41 – Food Service Supervision
Certificated Food Service
Classified Food Service
44 – Food Service Operations
Certificated Food Service
Classified Transportation
51 – Transportation Supervision
Certificated Transportation
Classified Transportation
52 – Transportation Operations
Certificated Transportation
Classified Transportation
53 – Transportation Maintenance
Certificated Transportation
Classified Other (nonteaching)
58 – Remote Learning
Certificated Other (nonteaching)
Classified Facilities
61 – Building Supervision
Certificated Facilities
Classified Facilities
62 – Grounds Maintenance
Certificated Facilities
Classified Facilities
63 – Operation of Buildings
Certificated Facilities
Classified Facilities
64 – Building Maintenance
Certificated Facilities
Classified Other (nonteaching)
67 – Building and Property Security
Certificated Other (nonteaching)
Classified Other (nonteaching)
72 – Information Systems
Certificated Other (nonteaching)
Classified Other (nonteaching)
73 – Printing
Certificated Other (nonteaching)
Classified Facilities
74 – Warehousing & Distribution
Certificated Facilities
Classified Other (nonteaching)
75 – Motor Pool
Certificated Other (nonteaching)
Classified Other (nonteaching)
91 – Public Activities
Certificated Other (nonteaching)
42
Appendix B. Lower-Bound Position Cost Estimation Method and Results
Our main results use the average salary for each position category in Washington state in
a year—where we bin all staff into categories according to the activity code and
classified/certified status described in Appendix A. Averages are calculated at the district-level
and only account for the portion of the salary that is attributable to the indicated position,
allowing us to disentangle the portion of individual pay coming from their work as an athletic
coach from the portion coming from their work as a teacher. It is possible, however, that when
we scale counts of filled posts by these salary estimates we might overstate the costs of these
positions because newly hired staff may be entering the salary schedule at a lower pay grade than
the average staff member.
In this appendix, we present results using an alternative measure of salaries to estimate
posted position costs which we call “lower-bound post costs.” We construct the salary estimates
using similar methods to those described above, with two exceptions. First, instead of using the
average positions salary of all staff in a district in a category, we here use average salaries among
employees who were not working in their observed district and category in the year prior.
Second, we calculate the average salary amongst these newly hired staff within an Education
Service District (ESD), category, and year. Washington state is divided into nine ESDs which
were the 295 school districts in the state. We use this larger grouping because it captures regional
differences in cost of living and pay while also avoiding missingness in the data if we calculated
these averages within each district, category, and year. Table B1 repeats the structure of Table 1,
summarizing the lower-bound position costs for each category in the average district. Table B2
43
presents results for the SUR estimation for lower-bound post costs, and Table B3 replicates our
OLS and IV models for this outcome.
Table B1. District-Year Average Position Demand, Pay, and Projected Costs
Lower-bound Lower-bound
Average number
Category salary per projected post
of filled posts
position ($) costs ($)
Administration 35.8 46,287 1,656,682
Athletics 59.1 4,496 259,003
Facilities 20.9 48,728 987,382
Food 25.7 17,376 437,608
Health 25.7 54,393 1,371,522
Other 67.7 11,877 880,697
Paraeducator 130.5 27,097 3,542,651
Principal 6.4 85,552 512,890
Superintendent 0.4 160,647 60,139
Teaching 112.0 67,845 7,517,068
Transportation 8.0 29,929 224,967
Total 492.0 17,450,612
Notes. Each cell presents the average outcome measure for each job category,
weighted by district enrollment for all district-year observations in our sample
(n=856). Projected post costs are the district-by-year-by-category product of
the number of posts an estimate of position salary.
44
Table B2. Seemingly Unrelated Regression Associations Between Unspent ESSER Funds and Lower-Bound Post Costs ($10,000s)
Food Para- Super- Transport-
Admin Athletic Facilities Health Other Principal Teachers
services educator intendent ation
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Unspent ESSER 0.037*** 0.001*** 0.011*** 0.001*** -0.005*** 0.006*** 0.004 0.003*** 0.001*** 0.056*** -0.001**
funds ($10,000s) (0.002) (0.000) (0.001) (0.000) (0.001) (0.001) (0.004) (0.001) (0.000) (0.006) (0.000)
n 856
Notes. *p<.10; **p<.05; ***p<.01. We report standard errors, clustered by district, in parentheses. All models are weighted by enrollment and include the
following district-year-level controls: district enrollment (cubic), 2019-20 district revenue, the number of free/reduced-price eligible students, the number of
URM students, average SBAC scores, standardized difference in student enrollment from the prior year, average unemployment in the surrounding county,
attrition of teaching staff relative to the prior year, attrition of nonteaching staff relative to the prior year, indicators for district urbanicity, log distance to the
nearest teacher education program, a control for ESSER funds expended as of the prior July, and year fixed effects.
45
Table B3. Estimated Impacts of ESSER Lower-Bound Filled Post Costs ($10,000s)
Specification OLS 2SLS
(1) (2)
Panel A. All job categories
Unspent ESSER funds ($10,000s) 0.114** 0.042
(0.047) (0.092)
n 856 856
Panel B. Teaching positions
Unspent ESSER funds ($10,000s) 0.062** 0.093*
(0.028) (0.051)
n 856 856
Notes. *p<.10; **p<.05; ***p<.01. We report standard errors, clustered by district, in
parentheses. All models are weighted by enrollment and include the following district-
year-level controls: district enrollment (cubic), 2019-20 district revenue, the number
of free/reduced-price eligible students, the number of underrepresented minority
students, average SBAC scores, standardized difference in student enrollment from
the prior year, average unemployment in the surrounding county, attrition of teaching
staff relative to the prior year, attrition of nonteaching staff relative to the prior year,
indicators for district urbanicity, and log distance to the nearest teacher education
program. All models include year fixed effects. Unspent ESSER funds are grant
balances as of the prior July. The 2SLS specification in column 2 uses district
formula-eligible children to instrument for unspent ESSER funding. 2SLS=two-stage
least squares; FEC=formula-eligible children; OLS=ordinary least squares;
SBAC=Smarter Balance Assessment Consortium (test scores).
46
File and source
- File
- esser-funding-and-school-system-jobs-evidence-from-job-posting-data-no.pdf
- Size
- 896,976 bytes
- SHA-256
- 06a7405e319356daa31881c88de5e6e3ad8806d89c83550efc6243e9ef1f5c06
- Original
- caldercenter.org