Research Review

Educational Researcher, Vol. 50 No. 6, pp. 392 –400 DOI: 10.3102/0013189X21994488

Article reuse guidelines: sagepub.com/journals-permissions © 2021 AERA. https://journals.sagepub.com/home/edr392 EDUCATIONAL RESEARCHER

In March 2020, the COVID-19 pandemic forced schools around the nation to close physical campuses and shift to distance learning. The pandemic exposed deep cracks in our

education system, with low-poverty schools and students transi- tioning to online participation quickly and students of color in high poverty schools and English learners (ELs) lagging behind (Burke, 2020; Hamilton et al., 2020; Umansky, 2020). A nationally representative survey of teachers conducted by the EdWeek Research Center found that in May 2020, 23% of stu- dents were considered “truant” (i.e., not logging into any online work, not making contact with teacher, etc.) and close to 45% of teachers reported students had “much lower” levels of engage- ment with schoolwork than before the pandemic.1 A report by the Los Angeles Unified School District, the second largest dis- trict in the country, found that participation in online learning of middle and high school students between March 16 and May 22, 2020, never reached 100% and was lower for students in particular subgroups such as low income, ELs, students with dis- abilities, and homeless and foster youth (Besecker et al., 2020).

When schools closed in March 2020, it is safe to assume students were mostly absent at least in the first week or two immediately following the closures (Kuhfeld et al., 2020). This would put most students above average absenteeism levels for regular school years. Students who were consistently absent from

March through June, would have missed 10 weeks of school (50 days) putting them at the far end of the normal absenteeism spectrum. The analysis presented here gives some indication of the negative impact absences could have on both academic and social-emotional learning (SEL) outcomes. These estimates shed further light on the potential learning and social-emotional costs if students continue to be absent from school for long periods in 2020–2021 due to the challenging circumstances posed by the COVID-19 pandemic.

Given the deep inequalities present in our school systems, an increasingly urgent question for schools around the nation is how much learning has been lost due to the COVID-19 pan- demic? And how are different student subgroups affected by this? Are students in the earlier grades losing more ground than students in middle and high school? And, also importantly, how is the social-emotional development of students affected by their absence from school?

Although some data are beginning to emerge to directly answer these questions, testing during the pandemic was cur- tailed in most districts. It is therefore useful to look at past expe- rience with absenteeism to gauge what the potential impact

994488 EDRXXX10.3102/0013189X21994488Educational ResearcherEducational Researcher research-article2021

1University of California, Los Angeles, CA 2University of California, Riverside, CA

The Effects of Absenteeism on Academic and Social-Emotional Outcomes: Lessons for COVID-19 Lucrecia Santibañez1 and Cassandra M. Guarino2

In March 2020, most schools in the United States transitioned to distance learning in an effort to contain COVID-19. A significant number of students did not fully engage in remote learning opportunities due to resource or other constraints. An urgent question for schools around the nation is how much did the pandemic impact student academic and social- emotional development. This paper uses administrative panel data from California to approximate the impact of the pandemic by analyzing how absenteeism affects student outcomes. Our results suggest student outcomes generally suffer more from absenteeism in mathematics than in ELA. Negative effects are larger in middle school. Absences negatively affect social-emotional development, particularly in middle school. Our results suggest districts will face imminent needs for student academic and social-emotional support to make-up for losses due to the COVID-19 pandemic.

Keywords: absenteeism; achievement; achievement gap; COVID-19; descriptive analysis; econometric analysis; regression

analyses; secondary data analysis; social-emotional learning; student behavior/attitude

FEATURE ARTICLES

AUgUST/SEpTEMbER 2021 393

could be. Our study uses administrative panel data from six large school districts in California to analyze the effects of absenteeism in the recent past. We use data from 2014–2015 to 2017–2018 for students in Grades 3 to 12 to understand (1) the average pat- terns of absenteeism occurring during regular school years for all students and by subgroup and (2) the impact on test scores and SEL outcomes of being away from school for all students and by subgroup.

Our article makes several key contributions beyond prior lit- erature on this topic. First, we investigate the potential impact of absenteeism on both academic and social-emotional outcomes. Second, we estimate results for a large span of grades, allowing us to see how patterns related to absenteeism change across K–12 experience. Third, we drill down to results for four vulnerable student subgroups who may experience the effects of the pan- demic in different ways: ELs, students with disabilities, low- income students, and homeless/foster youth. Fourth, we have data from six of the largest, most diverse districts in the most populous U.S. state. Last, our 4-year student-level panel allows us to control for unobserved attributes of students that could bias the relationship between absenteeism and outcomes.

Our findings indicate that absenteeism hurts both academic and social-emotional outcomes with variation by grade and sub- group, as well as in the cumulative effect of different degrees of absence. Students generally suffer more from absenteeism in mathematics than in English language arts (ELA) and experience larger negative potential effects on academic outcomes in middle school than in elementary grades. With regard to social-emotional development, absenteeism is likely to have the greatest negative impact on social awareness and self-efficacy and the likely negative impact is most pronounced in middle school.

Previous Literature

Effects of Absenteeism on Test Scores

It is well established in the literature that absenteeism negatively affects academic outcomes. Using data from elementary students in North Carolina, Aucejo and Romano (2016) find that being absent for 10 days from school would reduce test scores by about 0.03 SD in ELA and 0.06 SD in mathematics. The negative impact is greater in upper elementary grades (fourth and fifth) than in third grade and larger for low-performing versus high- performing students. In mathematics, the detrimental effects of absences in one school year can persist into subsequent grades, suggesting that absences today can have lasting consequences. Other studies have also found that absences affect mathematics more than ELA (Gottfried, 2009, 2011, 2014) and later grades more than earlier grades (Gershenson et al., 2017). Liu et al. (2020) use a high school student panel covering 2002–2003 to 2012–2013 in one large California school district and find that missing 10 mathematics classes in the spring semester reduces mathematics test scores by about 0.07 SD, math course grades by 0.19 SD, and the probability of on-time graduation by 0.08 SD. Results are similar for ELA outcomes. To account for time- varying classroom influences that could bias the relationship between absenteeism and outcomes, they use class period-level absences and control for absences in another subject (i.e., ELA

or mathematics). The coefficient on spring semester absences in their student fixed effects model change only slightly when absences in another subject (class period) are added as a control. Importantly, they find that teacher and subject preferences are relatively stable across years and do not significantly bias the rela- tionship between attendance and student outcomes.

Kuhfeld et al. (2020) estimate typical summer learning loss to project the impact of COVID-19 using a national sample of stu- dents in Grades 3 to 7 who took MAP Growth assessments between 2017–2018 and 2018–2019. Assuming that students lost the 3 months (about 60 instructional days) immediately fol- lowing school closures in March, the authors project students could lose 32% to 27% of the expected yearly learning gains in ELA and 63% to 50% in mathematics when they come back in the fall of 2020. Effects vary across student proficiency catego- ries, with the most significant losses concentrated among stu- dents at low proficiency levels. One advantage of our study over studies like this is that we use absences during the school year to predict academic losses. When students are absent for extended periods during the year, teachers provide homework and supple- mental lesson materials. While these instructional efforts may not be as intensive as those that have been exerted during the pandemic, they do mitigate learning losses due to absence in ways that are more relevant to the current COVID-19 situation than summer learning loss.

Effects of Absenteeism on Social-Emotional Learning Outcomes

SEL skills, such as self-efficacy, self-management, and growth mind-set, have been found in the literature to be correlated with academic outcomes (e.g., Claro et al., 2016; Usher & Pajares, 2009; West, Buckley, et al., 2018). Recent work using Project CORE data suggests that when SEL outcomes improve, so do test scores and behavioral outcomes—this is true across student subgroups and regardless of the baseline level of SEL (Kanopka et al., 2020).

Only a handful of quantitative studies have tried to estimate the effect of being absent from school on SEL outcomes. Gottfried (2014) finds that chronic absenteeism reduces educa- tional and social engagement for kindergartners. West, Pier, et al. (2018), using two years of survey data from CORE districts find that students in Grades 4 to 12 with low ratings on growth management, self-awareness, self-efficacy, and self-management miss more school. They find the strongest negative associations with absences for self-management and self-efficacy.

Data and Method

This study uses rich longitudinal student-level data from the CORE districts—a group of the largest districts in California who formed a collaborative organization in 2010 to cooperate in efforts to implement new academic standards, improve training for teachers and administrators, and pool data.2 We use CORE data from six districts to estimate the potential impact of absen- teeism on academic outcomes and four districts to estimate the impact on social-emotional outcomes.3 The total number of student-year observations in our analyses is over 1.3 million,

394 EDUCATIONAL RESEARCHER

representing close to 600,000 individual students. We use 4 years of data from 2014–2015 through 2017–2018.

Outcome Data

The achievement variables are composed of vertically-scaled test scores on the Smarter Balanced Assessments (SBAC) in ELA and mathematics. These tests are available for Grades 3 to 8 and grade 11. Grades earlier than third grade and Grades 9, 10, and 12 are not tested.

Social-emotional data come from CORE surveys of students. The data provide scale scores generated from survey items using a generalized partial credit model4 measuring the following con- structs: (1) self-management, the ability to regulate one’s emo- tions, thoughts, and behaviors effectively in different situations; (2) growth mind-set, the belief that one’s intelligence is malleable and can grow with effort; (3) self-efficacy, the belief in one’s own ability to succeed in achieving an outcome or reaching a goal; and (4) social awareness, the ability to take the perspective of and empathize with others from diverse backgrounds and cultures, to understand social and ethical norms for behavior, and to recog- nize family, school, and community resources and support (Hough et al., 2017). A validation study found the CORE- generated SEL constructs to have high structural validity, and high reliability in most of the factors (Meyer et al., 2018).5 Scaled SEL scores range from 5.5 to 4.6 depending on the con- struct, but we standardize all scaled scores to a mean of zero and standard deviation of 1 by school year and by construct.6 Table A1 in the online appendix (available on the journal website) con- tains descriptive statistics on all variables used in our analyses.

Student Characteristics/Designations/Behaviors

Data available for each student for every school year include: days attended, days enrolled, grade attended, race/ethnicity, gen- der, EL designation, disability status (SWD), whether student is a homeless or foster youth (HL/FST), income proxied by free and reduced-price lunch (FRPL), behaviors (i.e., suspensions or expulsions during the school year), and enrollment patterns related to school changes.

Method

Studying the impact of absenteeism on student outcomes is chal- lenging because of unobserved factors that could be associated with both absenteeism and student outcomes. If not accounted for, these unobserved factors could bias estimates of the impact of absenteeism on outcomes. To mitigate this potential source of bias, we use a student fixed effect model that essentially uses each student as their own control.7 Such models control for time invariant qualities of individuals—fixed traits or some persistent degree of ability—that could be important sources of bias. Our independent variable of interest—days absent—varies almost yearly within students in the panel, making fixed effects ideally suited to the analysis. A limitation to student fixed effects is that they do not control for time-varying unobserved nonrandom student-specific variation that may contribute to both the out- come variables and absenteeism simultaneously. For example,

students might have family problems during a given year, caus- ing absenteeism to go up and outcomes to go down. If students dislike a particular teacher or subject, this would reveal a time- varying influence on absenteeism that is difficult to observe. This latter potential source of bias seems less of a threat since, as mentioned earlier, Liu et al. (2020) did not find evidence of teacher- or subject-specific unobserved effects on absenteeism.

Our data allow us to control for other time-varying factors that could be potential time-varying sources of bias. We include the number of suspensions and expulsions a student has in a given year, which both controls for some involuntary sources of absenteeism and indicates problems that manifest in disciplinary behaviors. We also include an indicator for whether or not the student experienced a change of schools during the course of the year or in the prior summer, which allows us to control for absences due to adjustment issues or potential underlying factors that caused the school move (e.g., a change in residence, family divorce). We also include program designations such as EL or SWD that vary over time.

Equation (1) presents the basic model we estimate in this article.

Y Abs Abs Enr X git it it it it it t i it= + + + + + + +β β β β τ γ ε1 2

2 3 4  

(1)

In this model, Absit are the number of days student i was absent at time (year) t. We allow a squared term to pick up non- linear relationships with absenteeism at different levels. Enrit are the number of days student i was enrolled at time (year) t. X⃑ it is a vector of time-varying student-level characteristics (i.e., number of suspensions or expulsions, whether the student changed schools that year, and program designations). git is a grade-level indicator. τt is a time (year-level) indicator. The student fixed effect is denoted by γ i . To understand the effects of days absent by grade, we include two-way interactions (e.g., fifth grade * days absent). To estimate the effects of days absent from school by grade and program designation (e.g., EL status) we allow further nonlinearities in certain specifications and use three-way interac- tions (e.g., fifth grade * EL * days absent). Our findings lend themselves to graphical displays, which we provide. All models are estimated with robust standard errors to account for the pos- sibility of arbitrary serial correlation and heteroskedasticity.

To check our estimates against other specifications, we ran models that did not include student fixed effects but did include all observable time-invariant student characteristics, time- varying variables, lagged achievement in one or both subjects,8 and school-level fixed effects. The absenteeism coefficients from these approaches differ from those in the student fixed effects model, suggesting that the student fixed effects go one step fur- ther in eliminating sources of bias (see Table A2 in the online appendix available on the journal website). We are confident that the estimates from the models we present closely approxi- mate the causal parameters.9

Findings

Table 1 displays descriptive patterns of absenteeism. Panel A shows that on average, students in Grades K–12 are absent from

AUgUST/SEpTEMbER 2021 395

school 7.4 days in a regular school year. Absences vary consider- ably by grade: elementary and middle school students spend about 7 days away from school in a regular school year, whereas middle school and high school students are absent 6 and 9 days on average every school year, respectively. Absences are highest for kindergarten and Grades 10 through 12, with 12th graders absent an average of 10.8 days. Absenteeism rates also vary con- siderably by student subgroup. African American students and those classified as SWDs, ELs, and HL/FST youth are much more likely than all students on average to be absent from school.

Panel B shows that 14% of students are absent zero days, 65% are absent 1 to 10 days, 13% are absent 11 to 18 days per year, and 8% are absent 18 days or more—the level at which

absenteeism is considered chronic. Chronic absence is more prevalent in Grades 9 to 12 than in the earlier grades. About 7% of students are absent from school 30 days or more in any given year, indicating that most chronically absent students are absent for longer periods than 18 days.

Effects of Absences on Test Scores

Results from estimating Equation 1 with test scores as outcomes are presented in Table 2 (columns 1 and 2) and Figure 1. We use only Grades 3 to 8 in this analysis.10 Being away from school for 10 days results in a 5% of a standard deviation loss in ELA and an 8% SD loss in mathematics. The squared absence term is very

Table 1 Average Days Absent per Year, by Subgroup

Panel A—Mean Days Absent by Subgroup African

American Asian/P.I.Grade All EL FRPL SWD HL/FST Latinx White

K 9.2 7.9 9.5 11.6 10.5 9.2 8.3 11.8 7.3 1 7.4 6.4 7.7 9.8 9.0 7.3 7.0 10.0 5.5 2 6.8 6.1 7.0 9.0 8.5 6.7 6.7 9.6 4.8 3 6.5 6.0 6.6 8.7 8.1 6.3 6.6 9.3 4.5 4 6.4 6.0 6.5 8.6 7.8 6.2 6.8 9.1 4.3 5 6.2 6.1 6.3 8.6 7.6 6.0 6.9 8.9 4.1 6 5.9 6.2 6.1 8.5 7.6 5.9 6.5 8.2 3.5 7 6.1 6.8 6.3 8.9 8.1 6.1 6.7 8.5 3.5 8 6.3 7.4 6.5 9.2 9.0 6.4 6.9 8.7 3.5 9 7.5 8.8 7.6 10.9 10.0 7.8 7.3 9.5 3.7 10 8.9 10.8 9.0 12.3 11.5 9.4 8.2 10.5 4.5 11 9.5 12.8 9.5 13.1 11.7 10.0 8.9 11.1 5.6 12 10.8 14.4 10.6 14.4 13.2 11.2 10.3 12.7 7.3 Mean 7.4 8.5 7.5 10.3 9.5 7.5 7.5 9.6 4.4 N 572,805 123,589 432,052 69,246 21,023 403,581 54,174 54,378 56,991 % 0.22 0.75 0.12 0.04 0.70 0.09 0.09 0.10

Panel B—Proportion of Students Absent, Various Levels of Absence

Days Absent

Grade 0 1–10 11–18 >18 >30

K 0.07 0.61 0.19 0.12 0.10 1 0.10 0.66 0.16 0.08 0.06 2 0.12 0.67 0.14 0.07 0.05 3 0.13 0.68 0.13 0.06 0.05 4 0.14 0.67 0.13 0.06 0.05 5 0.15 0.67 0.12 0.06 0.04 6 0.17 0.67 0.11 0.06 0.04 7 0.17 0.66 0.11 0.06 0.05 8 0.17 0.66 0.10 0.07 0.06 9 0.17 0.63 0.11 0.09 0.08 10 0.15 0.62 0.12 0.11 0.10 11 0.14 0.62 0.12 0.12 0.10 12 0.10 0.63 0.14 0.14 0.12 Mean 0.14 0.65 0.13 0.08 0.07

Note. Averages over 2014–2015 to 2017–2018 school years. Includes data from six CORE districts. EL = English learner; FRPL = free or reduced-price lunch; SWD = students with disability; HL/FST; homeless or foster youth; P.I. = Pacific Islander.

396 EDUCATIONAL RESEARCHER

small but statistically significant and positive and thus tends to lessen slightly the negative effect of additional days absent on test scores as absenteeism increases.

In the graphical display (Figure 1), what is important to note is the slope of the lines. The intercept for ELA and mathematics differs only because the vertically scaled SBAC scores and their predicted means differ across the subjects—thus the distance between the lines is not due to a difference in absenteeism effects. As the slopes reveal, absences have a clear negative effect on test scores. The rate of predicted academic loss due to absenteeism as it accumulates is steeper for mathematics.

Absences affect test scores differently depending on the stu- dent’s grade level. Predicted effects by grade are found in Figure 2. The slopes on the grade lines are steeper (downward trend) in sixth, seventh and eighth grade, indicating that academic loss due to extended absences is borne more heavily by students in middle school. Predicted effects for elementary students are noticeably flatter—indeed, even slightly positive for third graders, suggesting that the ability of teachers and parents to compensate for lost schooling is greater in the early grades.

The impact of extended absences on academic achievement vary by student subgroup. Overall predicted effects for students classified as EL, FRPL, SWD, and HL/FST can be found in Figure 3. For comparison we add a category of nonvulnerable students (NONVUL) composed of students who do not fall into any of these program designations. The negative effects of absen- teeism are substantial for all students and are the most pro- nounced for students classified as FRPL, SWD, and HL/FST. These findings are concerning, given that in our analytic sample, 77 percent of the student population is classified as FRPL, 13% as SWD, and 4% as HL/FST.11 ELs (18% of sample students) are an exception, as they are less affected even than non-vulnera- ble students (19%). It should be noted that this group includes students that are considered long-term ELs, newcomer ELs, and ELs at various points of English Language Development. More research is needed on variation within this subgroup to better understand these effects. For all subgroups, absences appear to have a more negative impact on test scores in the middle school grades relative to elementary grades (results by subgroup and grade available in the online appendix (Figure A1) available on the journal website.

Effects of Absences on Social-Emotional Learning Outcomes

To estimate the effects of absenteeism on SEL in a normal school year, we estimate Equation 1 using as outcomes the SEL scale scores in each of the four constructs standardized by year—social awareness (SA), self-efficacy (SE), self-management (SM), and growth mind-set (GM). SEL scores are available for Grades 4 to 12, and we include all of those grades in the analysis. As recommended by the construct developers and standard in this literature, we estimate this model on the subsample of stu- dents who answered at least 50% of the items used to generate each of the constructs (Education Analytics, 2018; West, Pier, et al., 2018).

Regression results are shown in Table 2 (columns 3–5). The coefficients on days absent for all four SEL constructs are nega- tive and statistically significant, with very small (close to zero) positive coefficients on the squared term. As shown in Figure 4, being absent from school for 20 or more days, harms all four SEL constructs. Effects for most constructs flatten out after 40 days. However, for SA the decline is more or less linear suggesting

Table 2 Effects of Days Absent on Cognitive and Noncognitive Outcomes

ELA Math GM SA SE SM

(1) (2) (3) (4) (5) (6)

Days absent −0.515*** (0.018) −0.786*** (0.018) −0.003*** (0.000) −0.004*** (0.000) −0.005*** (0.000) −0.004*** (0.000) Days enrolled 0.047*** (0.003) 0.076*** (0.003) 0.000* (0.000) 0.000*** (0.000) 0.000** (0.000) −0.000*** (0.000) Days absent squared 0.002*** (0.000) 0.004*** (0.000) 0.000*** (0.000) 0.000*** (0.000) 0.000*** (0.000) 0.000*** (0.000) N 1,369,902 1,376,758 1,301,220 1,308,211 1,300,709 1,311,810 No. of observations 569,779 572,892 585,949 587,304 585,825 588,066

Note. Robust standard errors in parentheses. Models are estimated as shown in Equation 1. Data cover six CORE districts, for years 2014–2015 to 2017–2018. GM = growth mind-set; SA = social awareness; SE = self-efficacy; SM = self-management.

FIGURE 1. Predicted effects on test scores, by different values of days absent. Note. Graph points represent the predicted outcome (test score, y-axis) for a given value of days absent (0, 10, 20, etc., x-axis) from the models estimated in Equation 1. Confidence intervals of 95% around prediction mean. Data cover six CORE districts, 2014–2015 to 2017–2018.

AUgUST/SEpTEMbER 2021 397

24 00

24 20

24 40

24 60

24 80

25 00

25 20

P re

di ct

ed T

es t S

co re

0 10 20 30 40 50 Days Absent

ELA

24 00

24 20

24 40

24 60

24 80

25 00

25 20

P re

di ct

ed T

es t S

co re

0 10 20 30 40 50 Days Absent

MATH

3 4 5 6 7 8

FIGURE 2. Predicted effects of absences on test scores, by grade level. Note. Graph points represent the predicted outcome (test score, y-axis) for a given value of days absent (0, 10, 20, etc., x-axis) from the models estimated in Equation 1. Confidence intervals of 95% around prediction mean. Data cover six CORE districts, 2014– 2015 to 2017–2018.

FIGURE 3. Predicted effects of absences on test scores, by subgroup. Note. Graph points represent the predicted outcome (test score, y-axis) for a given value of days absent (0, 10, 20, etc., x-axis) from the models estimated in Equation 1. Confidence intervals of 95% around prediction mean. Data cover six CORE districts, 2014– 2015 to 2017–2018.

398 EDUCATIONAL RESEARCHER

a steeper rate of loss on this construct and the greater importance of schools in promoting social awareness.

Figure 5 shows that the SEL constructs are affected differently by absenteeism across grades. To facilitate viewing, we aggregate grades into school levels, but grade-specific results can be found in the online appendix (Figures A2–A5) avaible on the journal website. All constructs are negatively affected by absenteeism. However, middle school is the level at which extended absence from school has the strongest negative impact on social–emo- tional development, with SE and SA having the steepest slopes. At the elementary level, the most affected constructs are SE and SM. At the high school level, the most affected construct is SA.

Absences are detrimental to all subgroups (see online appendix Figure A6 available on the journal website). Absences harm SA and SE more or less equally across groups. Absences harm non- vulnerable students more than others in SM, and they harm non- vulnerable students and SWDs slightly more than others in GM.

Conclusion

After schools shut down in-person instruction in mid-March 2020, districts across the nation scrambled to provide various

− .1

5 −

.1 −

.0 5

0 .0

5 P

re di

ct ed

S co

re

0 10 20 30 40 50 60 Days Absent

SA SE SM GM

FIGURE 4. Predicted effects on social-emotional outcomes, by different values of days absent. Note. Graph points represent the predicted outcome (construct score, y-axis) for a given value of days absent (0, 10, 20, etc., x-axis) from the models estimated in Equation 1. Confidence intervals of 95% around prediction mean. Data cover four CORE districts, 2014–2015 to 2017–2018.

− .2

− .1

0 .1

P re

di ct

ed S

co re

0 10 20 30 40 Days Absent

SA

− .2

− .1

0 .1

P re

di ct

ed S

co re

0 10 20 30 40 Days Absent

SE

− .2

− .1

0 .1

P re

di ct

ed S

co re

0 10 20 30 40 Days Absent

SM

− .2

− .1

0 .1

P re

di ct

ed S

co re

0 10 20 30 40 Days Absent

GM

Elementary Middle High

FIGURE 5. Predicted effects on social-emotional outcomes, by different values of days absent and school level. Note. Graph points represent the predicted outcome (construct score, y-axis) for a given value of days absent (0, 10, 20, etc., x-axis) from the models estimated in Equation 1. Elementary grades include Grades 4 to 5. Middle grades include Grades 6 to 8. High school grades include Grades 9 to 12. Confidence intervals of 95% around prediction mean. Data cover four CORE districts, 2014–2015 to 2017–2018.

AUgUST/SEpTEMbER 2021 399

modes of distance instruction within weeks so students would lose as little learning as possible. In an effort to help assess the possible effects of being away from school during the pandemic, we estimated the impact of absenteeism on academic and social- emotional outcomes from recent pre-COVID experience. This study, using data from six large school districts in California, shows that average absenteeism is low in the regular school year: about 7 days on average, although this is higher in secondary school and for certain subgroups such as homeless/foster youth and students with disabilities.

This article adds important information to the growing evidence on the anticipated negative impact of COVID-19 on student devel- opment and its possible differential impacts by student subgroups. We show that absenteeism negatively affects student achievement, more so for mathematics than ELA and more so for middle school students than elementary students. Although all students experience the negative effects of absenteeism on academic outcomes, certain vulnerable subgroups of students—particularly low-income stu- dents, students with disabilities, and homeless and foster youth— are more subject to learning loss than other students.

Being absent from school harms SEL skills, as well, particu- larly those related to social awareness, self-efficacy, and self- management and, again, more so for middle school students than others. Absences are detrimental to SEL for all subgroups, with some variation across groups.

Taken together with evidence that significant numbers of stu- dents were absent from virtual schooling opportunities for lon- ger periods than normal during the COVID-19 pandemic (EdWeek Research Center, 2020; Hamilton et al., 2020) and that absenteeism was highest among students of color and disad- vantaged groups (Besecker et al., 2020), our results suggest that school disruptions brought on by the pandemic will negatively affect both the academic and social–emotional development of students, particularly for students in certain grades and vulnera- ble subgroups. It is increasingly evident that students will need both academic and social-emotional support to make up for what was lost.

Our study raises some questions as to why these effects occur with different patterns for different groups and grades. For example, heterogeneity within certain groups such as ELs and SWDs across grades and subcategorizations may be driving some of the differences we see. More research is needed with regard to the mechanisms by which absence from school affect students of all types so that the full effects of the pandemic or other such shocks to the school system can be better understood.

This study provides an overview of the effects of absenteeism on both academic and social-emotional development across a large span of grades and for several subgroups of students. As such, it enables districts to gauge the potential effects of absen- teeism to better predict and proactively address the potential effects of COVID-19.

NOTES

We are grateful to the six CORE districts who provided data for this study through the PACE/CORE Research Partnership. Our deep thanks go to Heather Hough and Joe Witte at PACE for facilitating the data needed and supporting our work. Dave Calhoun at CORE facilitated interactions with districts that helped inform our analysis and

disseminate our results. We are grateful to Clemence Darriet at UCLA for research assistance. Thank you to Libby Pier at Education Analytics, Jose Felipe Martinez at UCLA, Anna Bargagliotti at Loyola Marymount University, Alix Gallagher at PACE, Michael Gottfried at University of Pennsylvania, and Susanna Loeb at the Annenberg Institute at Brown University for helpful comments and suggestions. Two anonymous reviewers provided feedback that greatly improved the paper. All errors remain our own.

1EdWeek Research Center Survey. https://www.edweek.org/ew/ articles/2020/04/27/survey-tracker-k-12-coronavirus-response.html

2See http://coredistricts.org/about-us/ 3Two of the six districts used in the academic outcome analysis did

not collect SEL data over the period of our study. 4The construct developers at Education Analytics recommend

using the generalized partial credit model scale scores provided for the type of analyses we conduct (Education Analytics, 2018; Meyer et al., 2018).

5The sole exception was the Growth Mindset factor, which had low reliability in Grade 4.

6Loeb et al. (2019) and West, Pier, et al. (2018) standardize the scale scores for ease of interpretability, thus we follow this procedure.

7This is equivalent to adding an indicator variable for each student. 8It should be noted that there is virtually no difference in the esti-

mated effects of absenteeism when a lagged dependent variable (in one or both subjects) is included in the fixed effects model. This is because the student fixed effects model essentially takes a differencing or aver- aging type of approach, whereas models without student fixed effects necessitate the inclusion of lagged outcome variables. The lagged test scores convey little information beyond what is accounted for in the fixed effects.

9In all studies of the relationship of absenteeism to outcomes there’s a potential threat of bias due to missing outcome data that is correlated with absenteeism. In this study, however, because we have a 4-year panel we have score data for the vast majority of students: 99% have at least 2 or more SBAC scores and 96% have at least 2 or more SEL scores.

10Eleventh grade is excluded from the analysis because in a 4-year panel with no test score data between 8th and 11th grade, there are far fewer students with prior test score observations than in other grades, thus constraining the percentage of 11th-grade students with longitudi- nal data. Moreover, there is evidence that the population of 11th graders taking the test was a more restrictive group, with higher rates of non- testing. According to Warren and Lafortune (2019), 6.2% of enrolled 11th-grade students did not take the 2018 test compared with 2.5% students in Grades 3 to 8.

11These percentages closely mirror the proportions in the overall CORE student population.

REFERENCES

Aucejo, E. M., & Romano, T. F. (2016). Assessing the effect of school days and absences on test score performance. Economics of Education Review, 55(December), 70–87. https://doi.org/ 10.1016/j.econedurev.2016.08.007

Besecker, M., Thomas, A., & Daley, G. (2020). Student engagement online during school facilities closures: An analysis of LA Unified sec- ondary students’ Schoology activity from March 16 to May 22, 2020. http://laschoolboard.org/sites/default/files/IAU%20Report%20 2020%200707%20-%20Student%20Engagement%20 Online%20During%20Closures.pdf

Burke, M. (2020, May 11). A unified plans expanded summer school as “just about every” student is now connected online. EdSource. https://edsource.org/2020/la-unified-plans-expanded-

400 EDUCATIONAL RESEARCHER

summer-school-as-just-about-every-student-is-now-connected- online/631280

Claro, S., Paunesku, D., & Dweck, C. S. (2016). Growth mindset tem- pers the effects of poverty on academic achievement. Proceedings of the National Academy of Sciences of the U S A, 113(31), 8664–8668. https://doi.org/10.1073/pnas.1608207113

Education Analytics. (2018). CORE districts’ SEL survey scale scores (SY2016-17 and SY2017-18). https://www.rand.org/education- and-labor/projects/assessments/tool/2014/core-districts-social- emotional-learning-survey.html

EdWeek Research Center. (2020, April 28). Survey tracker: Monitoring how K-12 educators are responding to coronavirus. EdWeek. https://www.edweek.org/ew/articles/2020/04/27/survey-tracker- k-12-coronavirus-response.html

Gershenson, S., Jacknowitz, A., & Brannegan, A. (2017). Are stu- dent absences worth the worry in US primary schools?. Education Finance and Policy, 12(2), 137–165. https://doi.org/10.1162/ EDFP_a_00207

Gottfried, M. A. (2009). Excused versus unexcused: How student absences in elementary school affect academic achievement. Educational Evaluation and Policy Analysis, 31(4), 392–415. https://doi.org/10.3102/0162373709342467

Gottfried, M. A. (2011). The detrimental effects of missing school: Evidence from urban siblings. American Journal of Education, 117(2), 147–182. https://doi.org/10.1086/657886

Gottfried, M. A. (2014). Chronic absenteeism and its effects on stu- dents’ academic and socioemotional outcomes. Journal of Education for Students Placed at Risk, 19(2), 53–75. https://doi.org/10.1080/ 10824669.2014.962696

Hamilton, L. S., Grant, D., Kaufman, J. H., Diliberti, M., Schwartz, H. L., Hunter, G. P., Setodji, C. M., & Young, C. J. (2020). COVID-19 and the state of K–12 schools: Results and technical docu- mentation from the Spring 2020 American Educator Panels COVID- 19 Surveys. https://doi.org/10.7249/RRA168-1

Hough, H., Kalogrides, D., & Loeb, S. (2017). Using surveys of stu- dents’ social-emotional learning and school climate for account- ability and continuous improvement. Policy Analysis for California Education. https://cepa.stanford.edu/content/using-surveys-students- social-emotional-skills-and-school-climate-accountability-and- continuous-improvement

Kanopka, K., Claro, S., Loeb, S., West, M., & Fricke, H. (2020). What do changes in social-emotional learning tell us about changes in academic and behavioral outcomes? [Policy brief]. Policy Analysis for California Education. https://edpolicyinca.org/publications/ changes-social-emotional-learning

Kuhfeld, M., Soland, J., Tarasawa, B., Johnson, A., Ruzek, E., & Liu, J. (2020). Projecting the potential impacts of COVID-19 school closures on academic achievement (EdWorkingPaper: 20-226). https://doi .org/10.3102/0013189X20965918

Liu, J., Lee, M., & Gershenson, S. (2020). The short- and long-run impacts of secondary school absences (EdWorkingPaper: 20-125). https://doi.org/10.26300/xg6s-z169

Loeb, S., Christian, M. S., Hough, H., Meyer, R. H., Rice, A. B., & West, M. R. (2019). School differences in social–emotional learning gains: Findings from the first large-scale panel survey of students. Journal of Educational and Behavioral Statistics, 44(5), 507–542. https://doi.org/10.3102/1076998619845162

Meyer, R. H., Wang, C., & Rice, A. B. (2018). Measuring students’ social-emotional learning among California’s CORE districts: An IRT modeling approach (Working paper). Policy Analysis for California Education. https://files.eric.ed.gov/fulltext/ED591083 .pdf

Umansky, I. (2020). COVID-19’s impact on English learner students: Possible policy responses. [PACE commentary]. https://edpolicyinca .org/newsroom/covid-19s-impact-english-learner-students

Usher, E. L., & Pajares, F. (2009). Sources of self-efficacy in math- ematics: A validation study. Contemporary Educational Psychology, 34(1), 89–101. https://doi.org/10.1016/j.cedpsych.2008.09.002

Warren, P., & Lafortune, J. (2019). Achievement in California’s pub- lic schools: What do test scores tell us? Public Policy Institute of California. https://www.ppic.org/wp-content/uploads/achievement- in-californias-public-schools-what-do-test-scores-tell-us.pdf

West, M. R., Buckley, K., Krachman, S. B., & Bookman, N. (2018). Development and implementation of student social-emotional surveys in the CORE Districts. Journal of Applied Developmental Psychology, 55(March–April), 119–129. https://doi.org/10.1016/j .appdev.2017.06.001

West, M. R., Pier, L., Fricke, H., Loeb, S., Meyer, R. H., & Rice, A. B. (2018). Trends in student social-emotional learning: Evidence from the core districts (Working paper). Policy Analysis for California Education. https://files.eric.ed.gov/fulltext/ED591084 .pdf

AuTHORS

LUCRECIA SANTIBAÑEZ, PhD, is an associate professor of educa- tion at the University of California, Los Angeles, 405 Hilgard Avenue, Los Angeles, CA 90095; [email protected]. Her research focuses on how to improve teaching and learning for low-income, English learner, and other vulnerable student populations in the United States and Mexico.

CASSANDRA M. GUARINO, PhD, is a professor of education policy at University of California, Riverside, 900 University Avenue, Riverside, CA 92521; [email protected]. Her research focuses on issues of equity in education and the factors that influence student success.

Manuscript received July 13, 2020 Revisions received September 29, 2020,

and November 23, 2020 Accepted November 24, 2020