Literacy Loss in Kindergarten Children during COVID-19 School Closures

Abstract

By April 15th, 2020, more than 1.5 billion students worldwide experienced school closures in an effort to slow the spread of a novel coronavirus, COVID-19, during a worldwide pandemic (1). These interruptions in formal educational experiences cause irreversible consequences on school-age children’s literacy skills (2). Comparing children’s literacy during the time with or without formal education, we concluded kindergarten children will lose 67% of their literacy abilities during COVID-19 school closures. Reading books daily to children mitigates 10.5% of this loss. Educators and policy makers can promote this simple solution to slow literacy loss during school closures, which may be a common occurrence as nations see the public health benefits of physical distancing for future pandemic outbreaks.

Full text

1
Literacy Loss in Kindergarten Children during COVID-19 School Closures
Authors: Xue Bao1, Hang Qu2, Ruixiong Zhang2, Tiffany Hogan1*.
Affiliations:
1MGH Institute of Health Professions.
2Georgia Institute of Technology.
5
*Correspondence to: thogan@mghihp.edu
Abstract: By April 15th, 2020, more than 1.5 billion students worldwide experienced school
closures in an effort to slow the spread of a novel coronavirus, COVID-19, during a worldwide
pandemic (1). These interruptions in formal educational experiences cause irreversible
10
consequences on school-age children’s literacy skills (2). Comparing children’s literacy during
the time with or without formal education, we concluded that kindergarten children will lose
67% of their literacy abilities during COVID-19 school closures. Our analysis showed that
reading books daily to children mitigates 10.5% of this loss. Educators and policy makers can
promote this simple solution to slow literacy loss during school closures, which may be a
15
common occurrence as nations see the public health benefits of physical distancing for future
pandemic outbreaks.
Introduction
To stop the spread of COVID-19, by April 15th, 91.3% of children enrolled in formal schooling -
1.5 billion children worldwide from 192 countries - were ordered to stay home (1). In the United
20
States, 130,000+ school closings impacted almost 57 million children. All fifty states in the U.S.
had closed their schools for at least 3 weeks and forty-nine states had closed all public and
private schools for the rest of the 2019-20 academic year (3). In these unprecedented times,
children will be out of school half as much as they were in school this academic year (~92 days
out of 180 days) (4).
25
The effect of school closure on academic achievement has been studied in the summer months.
This well-documented, ‘summer slump’ shows that children lose the equivalent of approximately
one month of literacy skills when they are out of school during summer break (5). The slump in
literacy achievement is reduced by attending summer school, visiting libraries, and/or
participating in literacy-rich summer-based activities (5–9). During COVID-19 closures, these
30
activities are not available to children so there is strong motivation to determine alternatives that
will reduce literacy loss. In this study, we use U.S. literacy scores to model the loss of literacy
skills in children during COVID-19 school closures, using data from a large, national
longitudinal study. Furthermore, we examine the impact of daily reading to children for reducing
predicted literacy loss. Our results will inform policy makers, educators, and parents as they
35
make decisions about how best to stimulate literacy skills during school closures, a likely staple
of formal education during future pandemics.

2
Materials and Methods
Data and sampling
The data used in this study were drawn from the Early Childhood Longitudinal Study,
Kindergarten Class of 2010-11 (ECLS-K:2011), sponsored by the National Center for Education
Statistics (NCES). The ECLS-K:2011 was a longitudinal study of 18,170 children who started
5
kindergarten in the 2010-11 school year from both private and public schools. The same children
were followed from kindergarten to fifth grade (2015-16 school year). Data from multiple
sources (e.g., child, family, school and community) were collected through direct assessments,
interviews, and questionnaires. Multi-stage sampling design was used to determine a nationally
representative sample. During data screening and cleaning, we excluded the participants that had
10
incomplete key information (e.g., scaled scores, parent questionnaire answers, etc.) (10, 11).
The sample used in the current study had 3,170 children (Table 1). The scores we used in our
analyses were drawn from kindergarten fall, kindergarten spring, 1st grade fall, and 1st grade
spring reading scaled score, herein referred to as literacy abilities. Data on reading-related
summer activities were collected through the 1st grade fall semester parent questionnaire.
15
n in current sample
Percentage (%)
Gender
Male
1,660
52.4
Female
1,510
47.6
Age entering kindergarten (in years)
≤6
215
6.8
6-7
2,734
86.3
>7
221
7.0
Parent race and ethnicity
Parent 1
White
1,477
50.4
African American
253
8.6
Hispanic
888
30.3
Asian and Pacific Islanders
230
7.8
Others
84
2.9
Parent 2
White
1,220
53.5
African American
129
5.7
Hispanic
686
30.1
Asian and Pacific Islanders
188
8.2
Others
59
2.6
Parent highest education
Less than a high school diploma
287
9.8
High school diploma or equivalent
682
23.2
Some college, no degree
692
23.5
Bachelor’s degree
707
24.0
Graduate degree
577
19.6
Total
3,170
100
Table 1. Demographic information of the current sample.
Reading outcome measures
The ECLS-K:2011 used a screening test and two-stage assessment to measure reading and
20
language skills (10, 11). The purpose of the two-stage design was to optimize measuring

3
accuracy and administration time. The reading assessments were based on the 2019 National
Assessment of Education Progress (NAEP) Reading Framework. Each child completed a 30
minute computer-assisted language screening and reading assessment. The language screener
included two tasks from the Preschool Language Assessment Scale task (preLAS 2003) that
measured receptive and expressive language proficiency. Students who passed the language
5
screening or who spoke English as home language proceeded to the reading assessment. The
reading assessment tested children’s ability in basic reading skills (e.g., phonological awareness,
familiarity with print, letter/word recognition, and sight word), expressive and receptive
vocabulary, and reading comprehension (e.g., recalling, interpretation, and evaluation).
Language and literacy assessments used in the ECLS-K:2011 besides preLAS were Peabody
10
Picture Vocabulary Test – 3rd Edition (PPVT-III), Test of Early Reading Ability – 3rd edition
(TERA-3), and Test of Preschool Early Literacy (TOPEL) (10, 11).
The ECLS-K:2011 used the item analyses to calculate composite scores to quantify literacy
abilities, so that scores were comparable between children.(10, 11). Because this process
comprised a battery of assessments of reading and language, we refer to the composite as
15
‘literacy abilities.’
Data analysis
We used multivariate linear regression with ordinary least square method (12) to obtain the rate
of literacy change using the literacy scale score from four assessments, taken during kindergarten
fall and spring semesters and 1st grade fall and spring semesters (Figure 1). The linear model was
20
interpreted as
!!" # $ %"&"! $%#&#! $ %$&$! $ '!
where
!!
was the literacy scale score,
#
was the reference level of the literacy scale score, i.e. the
literacy scale score at the first assessment,
&"!
was the number of days in the kindergarten after
the first assessment and
%"
was the rate of gain in the kindergarten,
&#!
and
%#
were the number
25
of days and the rate of gain in the summer vacation between kindergarten and 1st grade,
&$!
and
%$
were the number of days and the rate of gain during the 1st grade fall semester,
'!
was a
random error.
To analyze the effects of different factors on the rates of change we categorized and partitioned
the data based for each factor and followed the same approach to obtain the conditional rate of
30
gain from the partitioned data. As some inputs were missing in certain questionnaires, the total
amounts of data for different factors may differ.

4
Figure 1. Increase of literacy abilities based on the multivariate liner model. Differing rates of gain were revealed
during time with and without formal educational instruction. Literacy skills were assessed each fall and spring
semester of kindergarten and 1st grade.
Results
We used the Early Childhood Longitudinal Study, Kindergarten Class of 2010-11 (ECLS-
5
K:2011) data to assess the effect of school closures on rate of literacy gain or loss in kindergarten
children, under the assumption that the literacy changes during COVID-19 school closures is
similar to that of those shown during summer school closures. Using a multivariate linear
regression model, we calculated the rate of literacy change, measured by literacy scale scores,
during kindergarten spring semester as 7.23 points per 100 days across the nation. The rate of
10
gain in literacy scores without formal education decreased to 2.41 points per 100 days,
significantly lower than their rate of gain during the spring semester (p<0.001) (figure 2A and
table 2). Using these projections, COVID-19 school closures will lead to an average of 67% loss
of children’s literacy ability (figure 2A and table 2). Without COVID-19 school closures, literacy
abilities in kindergarten children would improve, on average, 20.9 points in one academic year.
15
Most of the states ordered school closures during the week of March 16th. We predicted that the
average gain of kindergarten literacy score during the 2019-20 academic year will decrease to
16.5 points with schools closed in mid-March until the end of the academic year.
Figure 2. The predicted literacy scale score before, during, and after COVID-19 related school closures. The gray line
20
in Panel A shows the literacy scale score under the business-as-usual scenario. The dotted, green, orange, blue, and
yellow lines represent the literacy scale score (Panel A) and the score (Panel B) change of all children in the sample,
children whose parents never read to them, once or twice a week, three to six times a week, and every day,
respectively. Panel A shows the predicted literacy scale score change over time during 2020 kindergarten spring
semester, COVID-19 school closures, and 2020 1st grade fall semester. Panel B shows the predicted change of
25
literacy scale score during school closures of children whose parents read to them at different frequencies.
Many activities proven to mitigate literacy loss during school closures – summer school, visits to
the library, and summer literacy activities – are not available because of COVID-19 closures (5–
7, 9). Therefore, we determined the effect of reading books to children at home, an activity that
30
can be accomplished during COVID-19 closures. The ECLS-K:2011 data set included parents’
estimates of, how frequently they read to their child during the summer months. They were given
four options: not at all, once or twice, 3-6 times per week, and every day. Results from the
regression model showed that the more frequently children were read to, the more literacy
abilities they gained during the COVID-19 closures (figure 2A). The children whose parents
35
never read to them had the lowest rate of gain (-1.42 points per 100 days of summer) compared
to children who had books read to them every day (2.94 points per 100 days) (figure 2B). Based
on these projections, we predict that children whose parents read to them every day will gain 4.7
points during COVID-19 school closures, whereas children whose parents never read to them

5
will lose 2.3 points during the same period of time (figure 2B). Using the current regression
model, we concluded that reading books at home every day during the COVID-19 school closure
period will result in 40% of the expected literacy increase expected if a child was participating in
business-as-usual spring kindergarten semester formal educational experiences.
5
Factor
n
Baseline
(intercept)
Spring gain
per 100 days
School closures
gain per 100 days
Fall gain per
100 days
Book reading frequency
Less than 1
66
55.3±2.2
8.44±1.57
-1.42±4.83*
8.54±1.59
1-2 per week
627
54.1±0.6
6.78±0.42
1.67±1.22*
8.23±0.42
3-6 per week
1075
55.1±0.4
7.45±0.32
1.90±0.92*
8.84±0.31
Every day
1402
57.2±0.4
7.27±0.29
2.94±0.83*
9.15±0.29
Table 2. Results of the multivariate liner regression model that predicted the change in literacy abilities before, during,
and after the COVID-19 school closures. *p<0.001
Discussion
We used data from a large, U.S. longitudinal study to predict that 67% of kindergarten children’s
10
literacy abilities will be lost during COVID-19 school closures. Past studies show that literacy
loss over the summer can be mitigated by educational efforts such as intensive summer
instruction and frequent library visits (7), however those options are not available during
COVID-19 closures. Reading every day to a child, an option available while staying at home,
reduced the rate of loss by 10.5%. Kindergarten children, who have limited ability to read
15
independently, depend on adults to provide access to books through read alouds. These read
alouds provide, ‘book-language,’ which is higher in sentence complexity and school vocabulary
than conversation and oral storytelling (13, 14). Reading aloud also provides an opportunity to
practice word reading and interaction with text around letters and sounds, and are effective even
if reading the same book multiple times (15). Formal schooling through active online learning
20
will likely yield the greatest maintenance of literacy skills but until schools have robust,
evidence-based systems for delivering high-quality online learning, reading to children is an
effective low-tech alternative, which could also promote a love of books and social-emotional
connections between parents and children.
25
References and Notes:
1. UNESCO, COVID-19 Educational Disruption and Response. UNESCO (2020), (available
at https://en.unesco.org/covid19/educationresponse).
2. UNESCO, Adverse consequences of school closures. UNESCO (2020), (available at
https://en.unesco.org/covid19/educationresponse/consequences).
30
3. Map: Coronavirus and School Closures - Education Week. Educ. Week (2020), (available at
https://www.edweek.org/ew/section/multimedia/map-coronavirus-and-school-
closures.html).

6
4. Office of the Governor, Virginia Governor Ralph Northam - March, (available at
https://www.governor.virginia.gov/newsroom/all-releases/2020/march/headline-855292-
en.html).
5. H. Cooper, B. Nye, K. Charlton, J. Lindsay, S. Greathouse, The Effects of Summer
Vacation on Achievement Test Scores: A Narrative and Meta-Analytic Review. Rev. Educ.
5
Res. 66, 227–268 (1996).
6. D. T. Burkam, D. D. Ready, V. E. Lee, L. F. LoGerfo, Social-Class Differences in Summer
Learning Between Kindergarten and First Grade: Model Specification and Estimation.
Sociol. Educ. 77, 1–31 (2004).
7. J. A. Christodoulou, A. Cyr, J. Murtagh, P. Chang, J. Lin, A. J. Guarino, P. Hook, J. D. E.
10
Gabrieli, Impact of Intensive Summer Reading Intervention for Children With Reading
Disabilities and Difficulties in Early Elementary School. J. Learn. Disabil. 50, 115–127
(2017).
8. J. S. Kim, T. G. White, Scaffolding Voluntary Summer Reading for Children in Grades 3 to
5: An Experimental Study. Sci. Stud. Read. 12, 1–23 (2008).
15
9. J. Johnston, J. Riley, C. Ryan, L. Kelly-Vance, Evaluation of a Summer Reading Program
to Reduce Summer Setback. Read. Writ. Q. 31, 334–350 (2015).
10. M. Najarian, K. Tourangeau, C. Nord, K. Wallner-Allen, Early Childhood Longitudinal
Study, Kindergarten Class of 2010-11 (ECLS-K: 2011), Kindergarten Psychometric Report.
Natl. Cent. Educ. Stat. Inst. Educ. Sci. US Dep. Educ. Wash. DC, 269 (2018).
20
11. M. Mulligan, K. Tourangeau, C. Nord, K. Wallner-Allen, J. Leggitt, Early Childhood
Longitudinal Study, Kindergarten Class of 2010–11 (ECLS-K: 2011), First-Grade and
Second-Grade Psychometric Report. Natl. Cent. Educ. Stat. Inst. Educ. Sci. US Dep. Educ.
Wash. DC, 277 (2018).
12. P. J. Huber, Robust Statistics (John Wiley & Sons, Inc, Hoboken, NJ, USA, 2005).
25
13. A. Biemiller, C. Boote, An Effective Method for Building Meaning Vocabulary in Primary
Grades. J. Educ. Psychol. 98, 44–62 (2006).
14. S. B. Neuman, C. Copple, S. Bredekamp, Learning To Read and Write: Developmentally
Appropriate Practices for Young Children (2000).
15. P. Blewitt, K. M. Rump, S. E. Shealy, S. A. Cook, Shared book reading: When and how
30
questions affect young children’s word learning. J. Educ. Psychol. 101, 294–304 (2009).
Acknowledgments: Funding: This work was funded by an MGH Institute of Health Professions
doctoral fellowship grant awarded to the first author; Author contributions: Xue Bao:
Conceptualization, Formal analysis, Writing – Original Draft, Project administration Hang Qu:
35
Software, Visualization Ruixiong Zhang: Methodology, Data Curation, Tiffany Hogan:

7
Validation, Writing - Review & Editing; Competing interests: Authors declare no competing
interests. Data and materials availability: The ECLS-K:2011 kindergarten–fifth grade public-
use data file used in this study is available at
https://nces.ed.gov/pubsearch/pubsinfo.asp?pubid=2015086 I

8
Supplementary Text
Model description
In this research, we used the literacy assessment data from the kindergarten spring semester
and the 1st grade fall semester, and we removed the data with missing values of the free factors
studied in this research. Then we applied linear regression to obtain the rate of gain in literacy
5
abilities. The linear model can be interpreted as
!!" %%$%"&"! $ %#&#! $ %$&$! $ '!
or
! " (% $ '
Where
!!
is the literacy ability,
%%
is the reference level of the literacy ability, i.e. the
10
literacy ability at the first assessment,
&"!
and
%"
are the days in spring term after the first test
and rate of gain in the spring term,
&#!
and
%#
are the days and the rate of gain in summer
vacation,
&$!
and
%$
are the days and the rate of gain in the fall term,
'!
is a random error. The
corresponding estimator and covariance matrix are:
%
)
"
*
(&(
+
'"(&!
15
,-.
/
%
)0
"
*
(&(
+
'"1#
2
Where
1#
2 are the mean of square of error
"
('"
3
!4 (%
)3
#
.
The comparisons of the rate of gain between summer and spring term (
%#4 %"
) and
between summer and fall term (
%#4 %$
) are then calculated using the corresponding vectors
-""
5
67487876
9 and
-#"
5
67678748
9 such that
-"% " %#4%":
and
-#% " %#4%$
. Let
;""
20
-"%
and
;#" -#%
, so
;"
<
" -"%
)
,-.
*
;"
<+
" -",-.
/
%
)0
-"
&
;#
<
" -#%
)
,-.
*
;#
<+
" -#,-.
/
%
)0
-#
&
25
Further, let
="")!
*
+
,-.
/
)!
*0 and
=#")"
*
+
,-.
/
)"
*0, if there is no difference between the rate of gain
in summer term and spring/fall term,
="
and
=#
following a t-distribution with degrees of freedom
of n-4, and we can then acquire the p-value from the t-distribution.
To assess the effects on the rate of gain from different factors, we partitioned the data by the
factor level, and applied the linear regression in each part of the data, respectively.
30
!!1 " %%1 $ %"1&"!1 $%#1&#!1 $ %$1 &$!1 $'!17 > " 6787?7@
or
!1" (1%1$ '17 > " 6787?7@
For factors with only two levels, the separated model is the same as assigning a new
variable
A
as
35
A "
B
67 CDEFGH:DF:IJKJI:6
87 CDEFGH:DF:IJKJI:8
The separated model can be written as one model of
!!"
*
%%% $ L%A
+
$
*
%"% $ L"A
+
&"! $
*
%#% $ L#A
+
&#! $
*
%$% $ L$A
+
&$! $'!
Where the
L
is the effect of the factor on the intercept as well as the coefficient. By using
this model, we can calculate the p-value for the effect of the factor on the rate of gain.
40