Projecting the Potential Impact of COVID-19 School Closures on Academic Achievement

Seasonal learning research (including studies to understand the effects of summer learning loss) makes comparisons of student learning patterns when school is in versus out of session. Thus, one way to think about COVID-19 school closures is to consider them extensions of summer break for most students. Research has consistently shown that achievement typically slows or declines over the summer months on average and that the declines tend to be steeper for mathematics than for reading (Quinn & Polikoff, 2017). However, there is much debate about the magnitude of summer loss and the degree to which summer vacation contributes to socioeconomic achievement gaps (von Hippel, 2019).

Prominent early work on summer learning loss found that students lost about a month of learning over the summer, with lower-income students falling behind middle- and high-income students in reading (Alexander et al., 2001; Cooper et al., 1996). Recent summer loss research using the Early Childhood Longitudinal Study, Kindergarten Cohort (ECLS-K) has indicated minimal loss on average during the summers following kindergarten and first grade (von Hippel et al., 2018), while studies using NWEA’s MAP Growth assessment showed fairly sizable drops across grade levels (Atteberry & McEachin, 2020; Kuhfeld et al., 2019). This variability in estimates across studies can be seen in Table 1, where summer drop estimates range from 0.001 to 0.010 SDs per day of school missed across grades/subjects. Despite the inconsistencies in the magnitude of summer loss, research using both recent data sources is consistent in failing to replicate the earlier finding of summer being the primary period in which socioeconomic inequalities widen (e.g., Kuhfeld, 2019; von Hippel, 2019; von Hippel & Hamrock, 2019). Additionally, both datasets indicate that summer is a particularly variable time for learning, with far higher variability in growth rates during the summer than the school year (Atteberry & McEachin, 2020; von Hippel et al., 2018; In fact, some students actually show learning gains during the summer.) However, little is currently known about the underlying sources of this variability. Prior research has found that gender, racial/ethnic, and socioeconomic characteristics only explain between 3% and 13% of the summer learning variability (Burkam et al., 2004; von Hippel et al., 2018).

The literature on school closures due to weather or natural disasters also provides some insight into the potential effect of COVID-19 school closures, especially given such closures occur unexpectedly and disrupt scheduled instruction. Although they occur over a shorter duration, school closures resulting from inclement weather or natural disasters provide an analog to school closures due to COVID-19. Absent the weather event or natural disaster, schools would be in session and learning for most students would occur as normal. Hansen (2011) found that each day of school cancellation due to snow in Colorado reduced eighth grade mathematics achievement by magnitudes ranging from 0.013 to 0.039 SDs, and the effects of snow days on student achievement in Maryland ranged from 0.013 to 0.016 SDs per day. Goodman (2014) studied snow day closures in Massachusetts and found that each day of school closure had null effects on mathematics and reading achievement overall but that students attending schools with a high percentage of students receiving free or reduced-price lunch (FRPL) experienced a decline of 0.014 SDs in mathematics and 0.016 SDs in reading for every day of school closure. Marcotte (2007) and Marcotte and Hemelt (2008) also used data from Maryland to estimate the relations between snowfall and student achievement. However, the measure of achievement was reported as the percentage of students who reached the “proficient” level on the state test, which is not comparable with SDs per day. We include a summary of the results from these studies in Appendix Table A1 (available on the journal website).

A related line of research found that the displacement effect of Hurricane Katrina led to drops in achievement at a magnitude of approximately 0.10 SDs in the following year (e.g., Sacerdote, 2012). Unfortunately, these studies did not investigate effect heterogeneity by student demographics or school poverty. Furthermore, these estimates are not comparable with those provided by the snow day literature due to differences in research design and recorded units of time.

In contrast to the seasonal learning and school closure studies discussed above, an emerging literature on school absenteeism focuses on the impact of instructional time loss due to absences while schools are in session. Unlike the school closure due to the COVID-19 that forced every student to be out of school, not all students are absent during a normal school year. There are numerous reasons why a student might miss school, including lack of access to reliable transportation and need to care for family members. Minority and low-income students tend to have more absences and are more likely to be chronically absent (i.e., missing at least 10% of school days), compared with their more affluent and nonminority peers (Whitney & Liu, 2017).

Research consistently found that absences had negative effects on end-of-year test scores. Several studies that used a value-added type of model found similar effect sizes in both elementary and secondary schools. Specifically, missing 10 school days decreased student mathematics test scores by 0.06 to 0.08 SDs, with slightly smaller effect sizes for ELA (English language arts) scores (Aucejo & Romano, 2016; Gershenson et al., 2017; Liu et al., 2020). Largely due to the specific variation used in estimating the impact of absences, studies that used either flu or snow days as an instrumental variable for absences tended to yield much larger estimates (Aucejo & Romano, 2016; Goodman, 2014). For example, Goodman (2014) found that one moderate snow day–induced absence reduced student mathematics scores by 0.05 SDs.

When considering heterogeneity, existing research suggests that absences are similarly harmful for students in different demographic groups, school types, and grade levels. However, two studies have indicated that low-income students suffer more from school absences than their wealthier peers (Aucejo & Romano, 2016; Gershenson et al., 2017). Another takeaway from the absenteeism literature is that the negative effects of absences were linear, meaning that each additional absence caused similar learning loss no matter how many absences a student had already accrued (Gershenson et al., 2017; Liu et al., 2019).

The literatures on summer vacation, school closures due to weather and natural disasters, and absenteeism indicate that student learning is likely to be negatively affected by being out of school. While there is a fair amount of variability in the effect size estimates by grade and study (Table 1), some clear trends emerge. Students showed bigger losses in mathematics than reading while out of school. Being absent from school is generally associated with larger impacts on learning than being out of school due to summer vacation, particularly in middle school. Finally, our review suggests that studies on summer loss and absenteeism may provide better (if imperfect) models for the impact of COVID-19 than the literature on weather-related school closures, which was sparse (only two studies with effect size estimates), generated inconsistent findings, and tended to rely on small sample sizes from specific geographical settings. Accordingly, we draw on the absenteeism effect sizes reported in Table 1, as well as new summer loss analyses, to produce the projections reported in this study. Before describing our approach, we discuss the unique circumstances of the COVID-19 school closures, including access to virtual instruction and the societal impacts of COVID-19 that go beyond school closures.

The data for this study are from NWEA’s anonymized longitudinal student achievement database. School districts use NWEA’s MAP Growth assessments to monitor elementary and secondary students’ reading and mathematics growth throughout the school year, with assessments typically administered in the fall, winter, and spring. We use the test scores of approximately 5 million third- to seventh-grade students² in 18,958 schools across the United States. In this study, we follow students across two school years (2017–2018 and 2018–2019) and one summer break (summer of 2018). The NWEA data also include demographic information, including student race/ethnicity, gender, and age at assessment, though student-level SES is not available.

Table 2 provides descriptive statistics for the sample by subject and grade. Overall, the sample is 51% male, 47% White, 17% Black, 4% Asian, and 18% Hispanic. School-level FRPL eligibility was obtained from the 2017–2018 Common Core of Data file from the National Center for Education Statistics. The average student in our sample attends a school that is 51% FRPL-eligible. A comparison of the 18,972 schools in our sample relative to the U.S. population of public elementary and middle schools (72,075 schools serving Grades 3–8) is provided in Appendix B of the online supplemental materials (available on the journal website). Overall, the sample closely aligns with the characteristics of U.S. public schools, with a slight overrepresentation of Black students and underrepresentation of Hispanic students.

Student test scores from NWEA’s MAP Growth reading and mathematics assessments were used in this study. MAP Growth is a computer adaptive test that precisely measures achievement even for students above or below grade level and is vertically scaled to allow for the estimation of gains across time. The MAP Growth assessments are typically administered three times a year (fall, winter, and spring) and are aligned to state content standards. Test scores are reported on the Rasch unIT scale, which is a linear transformation of the logit scale units from the Rasch item response theory model.

To answer this question, we made several projections of the effect of COVID-19 school closures on learning, each making a different assumption about what prior out-of-school literature should be the basis for the projection. These scenarios, from best-case to worst-case,³ are as follows:

Learning rates under Typical Growth and COVID Slide were estimated using multilevel growth models that were fit separately by grade/subject with the MAP Growth assessment data. All technical detail behind these projections is provided in Appendix C (available on the journal website). The COVID Slide scenario assumed that typical summer loss patterns would extend through the prolonged school closure. Given the majority of schools in the United States shut down around the week of March 15, 2020 (6.5 months into the school year), we used students’ projected achievement level at 6.5 months as the starting point for the projection and then assumed students lost ground from that point at the monthly rate calculated for each subject/grade. Linear projections were made based on estimated summer learning loss rates but starting from the projected achievement level at 6.5 months and extending to the presumed start of the next school year (12 months, September 1, 2020).

Under the Partial Absenteeism scenario, we assumed that students received remote instruction, but the exposure to instruction during the school closures was approximately half of what students typically receive if schools were open for face-to-face instruction. This assumption is based on teachers’ reports on the challenges of providing instructional remotely during the COVID-19 school closures (Gewertz, 2020). Under this 50% scenario, COVID-19-related learning loss is based on loss due to absenteeism rather than loss due to summer slide. This decision was made because absenteeism can occur intermittently amid days of instruction (as in the scenario described by Gewertz, 2020), whereas summer loss is defined by the complete absence of all schooling for an extended period of time. To make these projections, we drew on existing absenteeism literature. We first calculated an average effect size (in SD units) for each school day missed by subject based on the effect sizes from absenteeism studies reported in Table 1 (e.g., an average of −0.007 SDs per day in mathematics and −0.004 SDs per day in reading). Next we converted these estimates into monthly losses on the Rasch unIT scale using NWEA’s subject- and grade-specific achievement norms (Thum & Kuhfeld, 2020), assuming that there were approximately 20 potential instructional days in a typical month and that students were receiving approximately 50% of normal instruction (e.g., absent half of the time) due to COVID-19 school closures. Because students can only be absent while schools are still in session, we produced absenteeism projections only to the end of the school year (9.5 months) and then assumed typical summer learning loss in the subsequent 2.5 months. Finally, the Full Absenteeism projection was calculated in the same manner as the Partial Absenteeism projection but assuming that students were absent the full 20 instructional days each month.

To display the possible scenarios for learning as a result of the school closures during the 2019–2020 school year, we produced a set of plots to compare these empirical- and literature-based projections with typical learning rates. The plots display students’ estimated learning rates across the 2019–2020 school year and summer of 2020 based on the various projections. In addition to the plots, we also reported the impact of school closures as a percentage of learning gains that students were expected to make relative to a typical school year. These percentages were calculated by estimating the total gains during the school year (subtracting the initial score on September 1, 2019 from the projected score on June 15, 2020) under the three COVID-19 scenarios and dividing those estimates by the total gains expected under typical growth.

We do not expect all students to be affected by COVID-19 school closures equally. Prior summer learning loss research indicated that there is a considerable variability in students’ learning patterns over the summer (e.g., Atteberry & McEachin, 2020; Kuhfeld et al., 2019), most of which cannot be explained by observed student and family characteristics (Borman et al., 2005; Kuhfeld, 2019; von Hippel et al., 2019). In addition to producing average estimates of learning rates during time out of school, we estimated variation across students in (a) out-of-school (summer) learning rates and (b) projected student test scores as schools reopen this fall under both typical and COVID Slide conditions (see Appendix C for more details). The Partial/Full Absenteeism scenarios were based on effect sizes from previous studies, and we do not have student-level data to explore variation in absenteeism rates across students. As a result, the variation across students in projected learning rates in this study is only examined for the COVID Slide projections (based on the MAP Growth summer estimates).

To guide planning to support student learning during this pandemic and school closures, it is important to understand not only the possible impact of school closures on student learning but also whether students with large losses recover at similar or different rates than other students. To investigate this question, we examined the correlation among the learning rates during the summer of 2018 and in the 2018–2019 school year. Though the empirical data are from a typical school year and summer, the results from this analysis can inform decision-making by serving as a proxy for post–COVID-19 student learning recovery during the 2020–2021 school year (assuming that schools are able to operate normally for the majority of the 2020–2021 school year).

Finally, we examined the implications for SES-based achievement gaps using the COVID Slide projections. Since NWEA does not have a student-level indicator of SES, we focused on school-level percentage of students receiving FRPL, which was obtained from the 2017–2018 Common Core of Data. We classified low-SES (high-poverty) schools as schools with >90% FRPL eligibility and high-SES (low-poverty) as those with <10% FRPL eligibility (excluding schools with 10% to 90% FRPL eligibility). We projected school SES achievement gaps in fall 2020 under three scenarios: (a) all students within each school SES level showed typical learning during the 2019–2020 school year, (b) all students within each school SES level showed COVID Slide, and (c) a different percentage of students in low- and high-SES schools showed COVID Slide depending on the likelihood of receiving remote instruction. As described in further detail in Appendix C (available on the journal website), the third scenario draws on the data collected nationally from an Education Week teacher survey that were disaggregated by school SES. Unfortunately, there was no disaggregation by race/ethnicity. Specifically, the survey found that 73% of high-SES schools reported that all students were expected to receive remote learning during the COVID-19 school closures, compared with only 34% of low-SES schools (Herold, 2020). In our third scenario, we assumed that 73% of the students in the high-SES schools showed typical gains, while the remaining 27% of students were assumed to show COVID Slide. Similarly, in the low-SES schools, we assumed 34% of the students showed typical gains while the remaining 66% of students showed COVID Slide. This third scenario is a mixture distribution reflecting different levels of access to remote learning. Importantly, we assumed that access to remote learning can be approximated based on students’ typical growth rate, which is probably an overstatement given the challenges of remote instruction. In the online supplemental materials (available on the journal website), we also consider a fourth scenario that accounts for both differential access to remote instruction and rates of parental unemployment by school SES.

Projected COVID-19 impacts on average academic growth trajectories are presented in Figure 1 for mathematics (Panel A) and reading (Panel B). In a typical year, average academic growth is not constant across the academic year (shown as the curved lines seen in some grades) and generally declines from the last day of school through the summer, with steeper declines in mathematics than in reading. The line directly below Typical Growth shows projected trajectories based on the Partial Absenteeism scenario, the subsequent line shows projected trajectories under the COVID Slide scenario, and the lowest line within each grade displays the Full Absenteeism scenario. Under each projection, standard summer learning rates (shown as dotted lines) were assumed from the end of the 2019–2020 school year to the start of the 2020–2021 school year.

Under all of the COVID-19 projections, students’ learning gains were projected to be lower at the end of the 2019–2020 school year than under typical conditions. Not surprisingly, the COVID Slide projections were consistently lower than both the Typical Growth and Partial Absenteeism projections, while the Full Absenteeism projections were the most extreme, implying steeper drops while students were out of school across all grades and subjects. We also calculated the percentage of learning gains that students would be expected to have made relative to a normal year under each condition. Our results suggest that in the Partial Absenteeism scenario, students would be expected to begin this school year with 60% to 87% of their previous year gains (see Appendix Table C5 in the online supplemental materials available on the journal website). Under the COVID Slide projections, students were projected to end the abbreviated 2019–2020 school year with roughly 63% to 68% of the learning gains in reading but only 37% to 50% of the average gains in mathematics compared with those of a normal school year. Under the Full Absenteeism projections, the story is even more dire, with students in sixth and seventh grade projected to have ended the disrupted 2019–2020 school year with less than 30% of their typical learning gains in both mathematics and reading.

Beyond average achievement, educators may be equally concerned about whether COVID-19 will result in greater variability in the academic skills that students bring with them as school resumes. As seen in Table 3, 8% to 30% of students in math would be expected to show monthly gains during the summer. In reading, approximately the upper third to half of the distribution (35% to 45% of students) showed gains over the summer. That is to say, in a typical summer, we observe a large amount of variation in learning patterns, with a sizable proportion of students showing gains in the absence of formal schooling.

Figure 2 displays the predicted variability in fall achievement under COVID-19 relative to variability under a typical school year. The plot shows the predicted interquartile range (e.g., the 25th, 50th, and 75th percentiles) of projected fall 2020 test scores by grade and subject. Additionally, below each set of lines, the ratio of the SD of the projected scores under COVID Slide relative to a typical fall SD is shown for each grade. The estimated means, SDs, and percentiles scores for each condition and grade/subject are reported in Table C6 in the online supplemental materials (available on the journal website). Across the board, students are projected to begin this fall with lower scores relative to a typical fall. In math, the predicted SDs are projected to be fairly similar to a typical fall, while in reading, the SDs of expected scores are expected to be up to 1.2 times the SDs expected in a typical fall. Thus, students likely are returning not only with lower achievement (on average) but with a wider range of academic skills that may require teachers to further differentiate instruction.

To project whether larger COVID-19 learning losses would be associated with faster growth rates during the 2020–2021 school year, we examined whether students who lost more ground during a typical summer showed slower or faster rates of recovery during the subsequent typical school year. Correlations between students’ summer loss and linear growth during the 2018–2019 school year are presented in Table 3. In mathematics, student-level correlations ranged from −0.41 to −0.43, and in reading, the correlations ranged from −0.45 to −0.46. These correlations imply that students who lost more ground during the summer of 2018 showed steeper growth during the following school year (2018–2019) than students with less summer loss. Accordingly, this suggests that a student who lost ground during the summer does not necessarily continue to lose ground during the next school year; rather, they are likely to gain more ground than students who showed modest summer drops.

Finally, we projected whether SES-based achievement gaps were likely to widen under various COVID Slide scenarios. Figure 3 displays the standardized achievement gap between low- and high-SES schools under three scenarios: (a) typical fall, (b) all students within each school SES level showed COVID Slide, and (c) differential access to remote instruction and thereby different combinations of the first two scenarios by school SES level. Our results showed that under typical conditions there were sizable (approximately 0.75 SDs) achievement gaps between low- and high-SES schools across the grades and subjects examined. Perhaps surprisingly, assuming that all students showed COVID Slide within each school, SES level did not significantly widen the achievement gaps compared with a typical fall. Results from the conditional hierarchical linear models (see Appendix Table C7 available on the journal website) indicated that school SES is not significantly associated with summer learning rates in most subjects/grades, indicating that out-of-school time on its own may not widen achievement gaps. However, our third scenario, which accounted for the fact that students in high-SES schools were more likely to receive remote instruction than students in low-SES schools, resulted in significantly larger achievement gaps in math and slightly larger gaps in reading compared with a typical fall.

While we provide three sets of projections in this study—one based on growth rates calculated from MAP Growth data and the other two based on prior literature on student absenteeism—we acknowledge that it is impossible to accurately weigh the complex range of supports and challenges that students are facing during this period. The school closures caused by COVID-19 have additional aspects of trauma to students, loss of resources, and loss of opportunity to learn that go well beyond a traditional summer break or individual school absence for many families. In other words, families with financial resources, stable employment, and flexible work from home and childcare arrangements will likely weather this storm more easily than families who are renting their housing, working in low-paying fields that are hardest hit by the economic impacts, and experiencing higher rates of food insecurity, family instability, and other shocks from this disruption. Furthermore, the summer of 2020 was very different from typical summers, with many summer camps and programs closed or held online only. As a result, we could be underestimating summer loss between the 2019–2020 and 2020–2021 school years in our projections by relying on data from a prior summer.

A major limitation of our study is that we do not differentially project the impact of COVID-19 on the basis of race. The COVID-19 pandemic simultaneously occurred alongside social unrest and protests about the mistreatment and killings of Black people by police. Moreover, there are higher rates of unemployment among Black households, and COVID-19 infection and death rates are higher in the Black communities than in the White communities (Cerullo, 2020; Krogstad et al., 2020). We decided against providing race-based projections due to a fundamental concern that we might understate just how significant the impact of COVID-19 will be on schooling and learning for Black communities.

There are three primary reasons why we worry that projections might understate the effects of COVID-19 on students of color. First, there is very little information at this stage about how COVID-19 is affecting students of color specifically that we could use as the basis for projections. Projecting SES-based gaps is possible because there are at least emergent data on how COVID-19 might differentially affect instruction on the basis of school SES (Herold, 2020). For example, national teacher and principal surveys have provided insights on the rates of synchronous instruction, virtual attendance, and technology access disaggregated by school SES (see Appendix Table A4 available on the journal website). To our knowledge, there is no such national information disaggregated by student race/ethnicity.

Second, projecting race-based gaps would require knowledge of how factors like trauma and job loss would affect racial minority groups differentially. These heterogenous impacts are likely since there are higher rates of COVID-19 infections and deaths in the African American community (Bouie, 2020). Furthermore, the economic downturn has been particularly damaging for Black and Hispanic parents, who are less likely to be able to work from home during the pandemic (Cerullo, 2020; Krogstad et al., 2020). Additionally, the trauma caused by police shootings of African Americans and the subsequent nationwide protests might themselves interact with any COVID-19-induced trauma to affect students’ achievement in complicated ways. Projecting these vital differences by race/ethnicity and their impact on achievement would have involved unwarranted conjecture.

Third, most research on summer loss found that race-based and SES-based achievement gaps do not typically grow much during the summer (Kuhfeld et al., 2019; von Hippel et al., 2018). Thus, an approach of using gaps in summer learning loss as the basis for any additional race- or SES-based projections could understate the effect of COVID-19 on achievement gaps in profound ways given how extended COVID-19-related school closures may fundamentally differ from typical summer closures. As von Hippel (2020) has argued, parents with resources are not routinely using those resources to build their children’s math and reading advantage during a typical summer vacation, but during the pandemic, those same parents likely were channeling available resources to support student learning through accessing online learning materials, hiring tutors, or forming “learning pods” with other families (Moyer, 2020). This unequal scenario is likely exacerbated by the so-called “digital divide” in technology and internet access by race/ethnicity and SES (Musu, 2018), which contributes to greater inequalities during the COVID-19 pandemic than during a typical summer.

In short, forecasting is inherently speculative, and we feel that the question of how COVID-19 will affect race-based gaps is too important to speculate on given our current moment, especially since projections would have relied on more assumptions than the other projections in our study and summer loss may be a less appropriate model when examining race-based difference in achievement. When data are available for achievement in fall of 2020, estimating the effect of COVID-19 on those gaps should be a foremost priority.

Another limitation directly related to projections on the basis of race is that our SES-based projections may also be conservative. For example, while Figure 3 shows that achievement gaps on the basis of SES may increase by a third for certain subjects and grades, our projections also indicate that gaps may not widen substantively in other subject-grade combinations. Our school SES projections could very well understate the magnitude of COVID-19’s impact on gaps because our forecasts could not adequately model factors unique to this moment like the trauma associated with illness and sudden job loss (though Figure C2 in the online supplemental materials, available on the journal website, attempts to factor in additional impacts of parental job loss). Furthermore, we are unable to account for the likely heterogeneity in the impacts of trauma that may lead to increased variability in student learning patterns, which likely implies that our projections are an underestimation of the impacts of COVID-19 on achievement variability. These limitations should be kept in mind as educators and policymakers take actions to address the potential impacts of COVID-19 on U.S. schools and school systems.

Beyond impacts on race and SES, there are also potential technical limitations that bear mention. In calculating the projected impact of out-of-school time on learning in this study, we assumed that it is appropriate to linearly extrapolate learning loss from research on absenteeism and summer loss across the 3 months of school closure. However, one could plausibly argue that impacts may actually be nonlinear. Campbell and Frey (1970) hypothesize that forgetting learned material may occur nonlinearly, with rapid initial deceleration of knowledge followed by slower drop offs as time passes. Liu et al. (2020) found that additional absences had an approximately linear impact on student learning, though the number of absences assumed in this study (approximately 60 school days) far exceeds the average number of absences observed in their study. We are unaware of any studies that have examined this phenomenon in the context of summer break. If the true effect of being out of school accelerates the longer duration students are out of school, we could be underestimating the impact on learning. But if summer loss simply reflects a process of forgetting and re-remembering that is not directly linked to the amount of time out of school, we could be greatly overestimating the potential impacts on learning. Future research could explore the nonlinearity of summer learning using variation in when schools start and end, though given most U.S. public schools follow mostly standard schedules, it is unclear how much variation in the length of summer break exists across school districts.

Our results indicate that students may be substantially behind, especially in mathematics. Thus, teachers of different grade levels may wish to coordinate in order to determine where to start instruction. Educators will also need to find ways to assess students early, either formally or informally, in order to understand exactly where students are academically.⁴ Second, students are likely to enter school with more variability in their academic skills than under normal circumstances. Prior research suggests that greater heterogeneity in student achievement affects a classroom teacher’s ability to adapt instruction in order to meet the instructional needs of all students (Connor et al., 2009; Evertson et al., 1981). Experts recommend that teaching grade-level content to all students in the fall, while identifying students in need of special support, is likely to help students remain on track (Lynch & Hill, 2020).

Third, under typical schooling conditions, the students who lose the most during the summer tend to gain the most when back in school. Nonetheless, the ground that students have to make up during the 2020–2021 academic year will probably be greater due to COVID-19. Therefore, educators may want to work with students to determine the growth rates needed to catch up and set learning goals for the year that are ambitious but obtainable. These strategies might include establishing out-of-school learning supports during the 2020–2021 school year for the students most affected by school closures.

Finally, the effects of COVID-19 to which our study cannot speak may be the ones most worthy of addressing. Districts are rushing to support educators who are attempting to teach academic content remotely while also caring for their students’ social-emotional well-being. Prior research on students displaced by Hurricane Katrina indicated that students had difficulty concentrating and often manifested symptoms of depression in the months following the hurricane (Picou & Marshall, 2007). Understanding these impacts will be essential to supporting students’ social and emotional needs after this huge disruption of COVID-19. Many students may face greater food insecurity, loss of family income, loss of family members to the coronavirus, and fear of catching the virus themselves (National Association for the Advancement of Colored People, 2020). While the scale of the COVID-19 school closures is novel, the inequalities in our school systems are unfortunately anything but new. Our models cannot account for the reality that the crisis is having an unequal impact on our most underserved communities. Nonetheless, we hope that these analyses, which synthesize what we know from existing bodies of research, will inform tomorrow’s decision making.

Researchers, district leaders, and policy makers are offering research-based program and policy recommendations designed to make up academic ground from COVID-19 school closures (e.g., Allensworth & Schwartz, 2020; Hill & Loeb, 2020). For example, Kraft and Goldstein (2020) point to High-Dosage Tutoring (HDT) as a promising way to aid schools’ efforts facilitate additional instruction as well as an avenue to stimulate the economy. Several Senate bills seek to expand national service programs such as AmeriCorps (Phillips, 2020). At the state level, former Tennessee Governor Bill Haslam’s Foundation is teaming up with Boys and Girls Clubs and other youth-serving organizations in a pilot program called Tennessee Tutoring Corps which aims to recruit more than 1,000 college students to provide essential employment to recent college graduates who otherwise may be unable to enter the job market. These tutors will offer students in grades K-6 critical one-on-one academic support to help them catch up on content they may have missed due to lost instruction time in the classroom during COVID-19 closures (Blad, 2020; Slavin, 2020).

During summer 2020 many school-based summer programs, designed to increase year-round learning opportunities and compensate for typical summer learning loss, pivoted to home-based summer options (Borman, 2020). Such programs have been particularly helpful in boosting reading performance when targeted at low-income (K–8) children (Kim & Quinn, 2013). Home-based summer reading programs (e.g., Kids Read Now, Harvard Graduate School of Education’s READS LAB, etc.) provide students with additional access to reading materials matched to individual interests and students’ reading levels. Moreover, the programs provide parents and teachers guidance to help improve comprehension, and employ gentle nudges (e.g., texts, emails, etc.) to encourage continued reading throughout the summer (Borman, 2020). Additionally, free book distribution programs are also a good way to get quality books to underserved families who have access to fewer literacy supports. Various extended learning time interventions, including week-long “acceleration” academies and summer programs, will likely need to be considered to help struggling students during the 2020–2021 school year and subsequent summer (Allensworth & Schwartz, 2020).

Finally, research and experts have offered strategies to help address students’ emotional needs and build on teacher-student relationships, which will be crucial as districts work through intermittent school closures heading into this academic year. For example, prior research suggests looping, when an educator continues teaching her/his/their current class into the next year, as a promising practice that schools may want to scale in 2020–2021 (Hill & Jones, 2018). Additional recommendations stress the importance of having teachers work together to uncover missed learning and to provide effective responses to trauma in school settings. Using administrative data from online learning platforms, in particular, can equip teachers and support staff with information that can help identify students who have disengaged from instruction and/or who are at risk of dropping out entirely (Hill & Loeb, 2020). For additional recommendations on supporting students’ recovery from the COVID-19 school closures, a series of evidence briefs covering a wide range of topics are available from the EdResearch for Recovery (2020).