Connecticut Assessment Summary Spring 2021
Connecticut Assessment Summary Spring 2021
Connecticut Assessment Summary Spring 2021
2020-21
August 2021
Table of Contents
Background .............................................................................................................................................................3
Attendance Patterns...............................................................................................................................................4
Test Participation....................................................................................................................................................4
Introduction ................................................................................................................................................................6
Results ..................................................................................................................................................................... 10
Conclusion ............................................................................................................................................................... 30
Executive Summary
Background
From the Smarter Balanced Assessment Consortium (SBAC or Smarter Balanced) exams to the Connecticut SAT
School Day, summative assessments play a key role in the evaluation of student learning and critical thinking.
Moreover, summative assessment results are an important part of Connecticut’s Next Generation
Accountability System. Due to the COVID-19 pandemic, the Connecticut State Department of Education (CSDE)
received approval from the United States Department of Education (USED) to waive both summative testing
and accountability for the 2019–20 school year. In June 2020, the CSDE provided Sensible Assessment
Practices to offer guidance to educators on how to use available data to “assess” their incoming students when
schools re-opened in September without necessarily having to test them.
The 2020–21 school year presented a unique set of challenges, as many Connecticut’s students spent a
significant part of the year learning remotely. In light of these circumstances, the USED approved the CSDE’s
request to waive accountability for a second consecutive year. Still, despite on-going disruptions to learning
due to the pandemic, the CSDE felt it was vital to reaffirm Connecticut’s commitment to equity and administer
all statewide assessments during the 2020–21 school year. Having these scores allows for the monitoring of
long-term trends and the evaluation of the full impact of the pandemic on student achievement and growth. It
also provides accurate data to target support and resources where they are most needed to address and
combat the negative impact of this pandemic on student learning.
To support attendance and engagement of students as they participated in varied school learning models (i.e.,
in-person, hybrid, remote), the CSDE established a new system to collect district learning model data on a
weekly basis and student-level attendance data on a monthly basis in 2020–21. The weekly collection allowed
the CSDE to strategize and support districts locally with local boards. The monthly collecting of attendance and
the separate reporting of in-person and remote days for each student allowed for the CSDE to provide
targeted supports (e.g., roundtables, webinars, guidance, and the Learner Engagement and Attendance
Program [LEAP]), to research and publish attendance-related findings with Attendance Works that informed
policy discussions nation-wide, and to group students based on their predominant learning models. Moreover,
these groupings facilitated the evaluation of 2020–21 summative assessment results by student learning
model.
• Nearly one-quarter of students were fully/mostly in-person (i.e., more than 75 percent of school days
or at least 16 out of an estimated 20 school days in an average month were in-person), nearly half of
students were hybrid learners (i.e., between 25 and 75 percent of school days were in-person), and
nearly one-quarter of students were fully/mostly remote (i.e., below 25 percent or fewer than 5 out of
an estimated 20 school days in an average month were in-person).
• Students with high needs (i.e., those who are English learners, have a disability, and/or are from a low-
income family) tended to be fully remote at greater rates than their peers. Specifically, a greater
proportion of English learners (10.8 percent), students eligible for free meals (12.1 percent), and
students who are experiencing homelessness (16.7 percent) were fully remote for the entire school
year as compared to the state average (7.6 percent).
Test Participation
• A new remote testing approach was implemented with fidelity; 11.5 percent of Grade 3-8 students
tested remotely. Pursuant to CSDE analyses, only results from tests administered in-person are used in
these analyses.
o Nearly 82 percent of Grade 3-8 students tested in-person (which was strongly recommended)
on Smarter Balanced; 11.5 percent of students tested remotely.
o In-person test participation was strongest for students who learned fully/mostly in-person
(97%) or in hybrid (95%) model. Conversely, only 37 percent of fully/mostly remote learners
took the exams in-person, while 45 percent of fully/mostly remote learners took the Smarter
Balanced exams remotely. The CSDE confirmed that at the state-level, the demographics of
those who tested in-person are reasonably similar to those who tested remotely or not at all,
so as to allow for these analyses and related inferences.
Assessment Results
• During the pandemic, in all grades and most student groups, students who learned fully/mostly in-
person lost the least ground academically while those who learned in hybrid or fully/mostly remote
models showed substantially weaker achievement and growth.
o While the academic impacts were seen in all subjects, the observed differences were largest in
math.
• Estimated statewide results from Connecticut’s growth model further indicate the following:
o Growth before the pandemic was much stronger than growth during the pandemic.
o Among low- and high-achieving students, those learning in-person showed greater growth
than those learning in hybrid or remote models.
o During the pandemic, students below proficiency (Levels 1 and 2) grew at lower rates than
those above proficiency; this was not the case before the pandemic.
o Students above proficiency (Levels 3 and 4) who learned in-person neared pre-pandemic
growth in ELA but not in Math.
• Domain score analyses for Smarter Balanced assessments for students in Grades 5 and 8 reveal the
following:
o In Grade 5, the ELA domains of Research (Claim 4) and Evidence/Elaboration (which is part of
Claim 2: Writing) show steeper declines than the other domains among hybrid and
fully/mostly remote learners. Such declines are not observed prior to the pandemic.
o In Grade 8, the ELA domains of Organization/Purpose (part of Claim 2: Writing) shows a slightly
steeper decline than the other domains, especially among fully/mostly remote students. Such
declines are not observed prior to the pandemic
o In Grade 5 Math, Operations and Algebraic Thinking showed slightly lesser decline than the
other domains in all learning models.
o In Grade 8 Math, all domain scores decreased between grades 6 and 8 with the exception of
Statistics and Probability for in-person learners which stayed constant.
Connecticut Alternate Assessment (CTAA) for English Language Arts and Mathematics
The CTAA for ELA and math is available for eligible students in Grades 3-8 and 11. This assessment is presented
to each student individually by a CSDE-trained teacher via an online Test Delivery System. The CTAA, including
the supporting resources, was developed with Connecticut teachers and administrators working closely with
other national state members and experts in the National Center and State Collaborative (NCSC). Only in-
person testing was available for students to take the CTAA in 2020–21.
In the 2020–21 school year, schools across Connecticut used one of three learning models: (1) fully in-person,
where all students attended school in-person on all days; (2) hybrid, where all students attended school in-
With many districts opening in a hybrid format, and around one-third of students statewide learning remotely,
access to a device and internet connectivity quickly became essential school supplies. A donation from the
Partnership for Connecticut brought 60,000 laptops to high school students in the 33 Alliance Districts by July
2020. Additionally, in late July, Governor Lamont launched the Everybody Learns Initiative which brought an
additional 82,102 laptops and Chromebooks, 12,774 hotspots, and broadband cable internet to students.
In 2020–21, the Connecticut State Board of Education resolved to allow school districts to have 177 days of
student instruction versus 180 days, which permitted three days to be used at the beginning of the school year
for the purpose of building capacity to safely transition back to in-person services during the COVID-19
pandemic. In its resolution, the Board affirmed that its authorization for hybrid or remote programming due to
unavoidable emergency is contingent upon school districts providing rigorous learning and engagement
opportunities that are aligned with State standards and Board expectations. The Board also charged the CSDE
with ensuring fidelity to this expectation by collecting whatever data are necessary and making that
information transparent. Therefore, the CSDE established two new data collections: a weekly collection
regarding a district’s learning model and a monthly student-level attendance data collection to collect the
number of days of membership and attendance for each student each month. The data have been
disseminated on the Supporting Student Participation page of the CSDE website.
A new concept of “remote attendance” was introduced for the first time in 2020–21. While the definition of
“in attendance” is unchanged (i.e., presence for at least half a school day), the CSDE’s guidance on how to
track attendance on remote days expects districts to consider synchronous and asynchronous approaches to
determine whether a student is “in attendance.” Specifically, a remote student can be considered as being “in
attendance” on a particular day if the total time spent on one or more of the following activities equals at least
half the school day: synchronous virtual classes; synchronous virtual meetings; time logged in electronic
systems; and/or assignment submission/completion. This has fundamentally changed who determines
attendance, especially in elementary and middle schools. While previously front-office staff may have assumed
Since September is a month of significant change in public education systems – as districts follow up to
determine the status of returning students, while also enrolling new students who may be entering the district
– and data quality for new collections improves over time, CSDE decided to wait until October 2020 to
mandate the separate reporting of in-person and remote days. So, attendance data were collected since the
start of the school year, but data were reported separately for in-person and remote days since the beginning
of October. This allowed for an examination of attendance patterns for different groups of students on in-
person versus remote days.
Each student was assigned a learning model classification for the 2020–21 school year based on their
membership days from the beginning of October through the end of the school year according to the following
criteria:
Learning Models
Table 1 shows the district learning models in 2020–21 based on the most prevalent learning model among the
students in each district. In five of Connecticut’s twelve largest districts, the most prevalent learning model
among their students was fully/mostly remote.
Table 1: District Learning Models in 2020–21 Based on the Most Prevalent Learning Model Among their Students
Figure 1 shows the percentage of Grade 3-8 students in each learning model by high needs status. Nearly a
third (31.2%) of students with high needs (i.e., those who are English learners, have a disability, and/or are
from a low-income family) were fully/mostly remote as opposed to only (15.7%) of students without high
needs. By contrast, only one out of five (20.7%) of students with high needs were fully/mostly in-person as
compared to more than a third (35.6%) of students without high needs.
Figure 2: Test Mode Percentages by Learning Model, High Needs and Non High-Needs, Grades 3-8
Table 2: Proportion of Student Group Among Fully/Mostly Remote Learners by Test Mode (Grades 3-8)
Student Group In-person Testers Remote Testers and Not Tested
High Needs 66.9 70.2
Not High Needs 33.1 29.8
Eligible for FRL 59.9 64.2
Not Eligible for FRL 40.1 35.8
English Learners 14.0 11.7
Not English Learners 86.0 88.3
Students with disabilities 13.3 14.3
Students without disabilities 86.7 85.7
Overall, proficiency is lower in 2020–21 as compared to 2018-19, especially for those who learned in hybrid or
remote models. This is a starting point for interpretation but doesn’t tell the whole story. This chart makes it
appear that students who learned in-person did not lose any ground. That’s not entirely accurate. They too lost
ground but less than the others.
The problem with simply looking at proficiency rates is that it misses any growth/change above/below the
proficiency cut off score. Also, due to grade promotion, attrition, and other reasons, a third of the students in
Grades 3-8 in 2018-19 were not the same as those in the same grades in 2020–21. Additionally, this view does
not account for the fact that students who learned in the three learning models in 2020–21 may have different
levels of achievement in 2018-19; since statewide assessments were not administered in 2019–20, prior
achievement data is only available for students in Grades 5-8 in 2020–21.
Figure 4 shows matched cohort proficiency rates for Grades 5-8 combined by high needs status. So, while this
analysis still looks at proficiency rates, it limits the sample to matched students (i.e., those who tested in 2018-
19 and then two grades higher in 2020–21). The plots in Figure 4 illustrate several key points:
Figures 5 shows the matched cohort proficiency rates in ELA and Math for Grades 5-8 combined by
race/ethnicity. The plots demonstrate similar findings to the high needs group analysis:
• In most race/ethnic groups, students who learned in-person in 2020-21 were higher achieving in 2018-
19. The only exception to this pattern was among Asian students in ELA where the fully/mostly remote
learners were equally high achieving in 2018-19 to their in-person peers.
• In most race/ethnic groups, even students who learned in-person in 2020-21 lost ground as compared
to their 2018-19 achievement, especially in Math, though declines are greater for those who learned in
hybrid or fully/mostly remote models. Asian students who learned in hybrid of fully/mostly remote
models lost the least ground in ELA as compared to their peers from other race/ethnic groups.
• Declines were substantially greater in Math than in ELA.
The three solid lines in Figure 6 represent the matched cohort growth of students in the three learning models
from Grade 3 to Grade 6. The Grade 3 data on the chart represent the average Grade 3 scale scores in 2017-18
for the students who tested in-person in Grade 6 in 2020-21. The Grade 4 data on the chart represent the
average Grade 4 scale scores in 2018-19 for the same cohort.
Figure 6: Matched Cohort Average Scale Scores (ELA Grade 6 in 2020-21) by High Needs Status
Students who learned in-person in 2020-21 had higher scores when they were in Grade 3 and in Grade 4 as
compared to those who learned in hybrid or remote formats. All three learning model groups show a higher
growth rate (i.e., a steeper slope) in the one year from 3rd to 4th grade (i.e., before the pandemic) than they do
from Grade 4 to Grade 6 (i.e., through the pandemic). The declines are greater for those who learned in hybrid
or remote formats. As a result, the gap between the three groups is wider in 2020-21 than it was when they
were in Grade 3 or 4. By comparison, the previous matched cohort – shown in the plot as a dotted line and
labeled in the legend as Class of 2025 (pre-pandemic) – demonstrated relatively steady growth from Grade 3 in
2015-16 to Grade 6 in 2018-19, and this pre-pandemic cohort shows a much steeper growth trajectory from
Figure 7 provides a matched cohort comparison of average scale scores for Grade 6 ELA by race/ethnicity.
Asian students show the smallest gap in growth across student learning models. For all other races, the gap in
average scale scores between the in-person students and the students in the other learning models widened
over time.
Figure 7: Matched Cohort Average Scale Scores (ELA Grade 6 in 2020-21) by Race/Ethnicity
Figure 8 provides a look at growth trajectories in Math using matched cohort average scale scores. A
comparison of average scale scores for Grade 6 Math by High Needs status is used for illustrative purposes; the
CSDE’s analysis showed similar patterns in other grades and for most student groups. Again, we see that
students who learned in-person in 2020-21 had higher scores in 2017-18 when they were in Grade 3 as
compared to those who learned in hybrid or remote formats in 2020-21. All three learning model groups show
a higher growth rate (i.e., a steeper slope) in the one year from 3rd to 4th grade (i.e., before the pandemic) than
they do from Grade 4 to Grade 6 (i.e., through the pandemic). The declines are greater for those who learned
in hybrid or remote formats. As a result, the gap between the three learning model groups is wider in 2020-21
Figure 8: Matched Cohort Average Scale Scores (Math Grade 6 in 2020-21) by High Needs Status
Most notable in comparing Figure 8 (i.e., Grade 6 Math by High Needs Status) to Figure 6 (i.e., Grade 6 ELA by
High Needs Status) is that the impact of pandemic-related factors in Math appears more significant than in
ELA. Though in-person and hybrid students in 2020-21 started higher in Grade 3 than their pre-pandemic
peers, in Grade 6, their achievement lags that of their pre-pandemic peers. The gaps are even greater in Math
between those who learned in-person and those who learned in hybrid or remote formats. Here, this pattern
is shown to hold for students with high needs and students without high needs. The average scale scores in the
hybrid and remote learning models have not increased from Grade 4 to Grade 6 for students with high needs.
Figure 9 provides a matched cohort comparison of average scale scores for Grade 6 Math by race/ethnicity.
Here, the gap in average scale scores between the in-person students and the students in the other learning
models widened over time for all race/ethnicity groups. Most notable in comparing Figure 9 (i.e., Grade 6
Math by Race/Ethnicity) to Figure 7 (i.e., Grade 6 ELA by Race/Ethnicity) is that the impact of pandemic-related
factors in Math appears more significant than in ELA. By grade 6, the gaps are even greater in Math between
those who learned in-person and those who learned in hybrid or remote models. The average scale scores in
Figure 9: Matched Cohort Average Scale Scores (Math Grade 6 in 2020-21) by Race/Ethnicity
Figure 10 provides a look at ELA and Math achievement for students in grades 3 and 4 by high needs status. As
in grades 5-8, students in grades 3 and 4 who learned fully/mostly in-person lost the least ground, while those
who learned in hybrid or fully/mostly remote models in 2020-21 showed substantially lower achievement. The
declines are greater in Math than in ELA. The charts in Figure 10 show that this holds for students with or
without high needs; a similar pattern is seen for most student groups. One important note of caution: There
are no prior achievement scores for these cohorts of students against which to compare this difference, since
summative assessments were not administered in 2019-20 and these students were in Grades 1 and 2 in 2018-
19 and thus did not take these assessments.
• Among low- and high-achieving students, those learning in-person showed greater growth than those
learning in hybrid or remote models.
• During the pandemic, students below proficiency (Levels 1 and 2) grew at lower rates than those
above proficiency; this was not the case before the pandemic (see blue columns).
• Students above proficiency (Levels 3 and 4) who learned in-person neared pre-pandemic growth.
Figure 11: Connecticut Growth Model Matched Cohort ELA (Grades 4-8 Combined)
Figure 12: Connecticut Growth Model Matched Cohort Math (Grades 4-8 Combined)
Figure 13 shows the change in domain scores in ELA from Grade 3 to Grade 5 for students who were in Grade 5
in the 2020-2021 school year; the results are disaggregated by the three learning models. The first panel also
shows the change in domain scores for students who were in Grade 5 in 2018-19 (class of 2026); it serves as a
pre-pandemic reference. For all domains, in-person students start higher than their peers and decrease in
proficiency from 3rd through 5th grade. The differences between 3rd and 5th grade are small (no larger than 0.2)
but consistent. The students in hybrid and fully/mostly remote learning models start lower, but also appear to
decrease more than their in-person peers. Specifically, the domains of Research (Claim 4) and
Evidence/Elaboration (which is part of Claim 2: Writing) show steeper declines than the other domains among
hybrid and fully/mostly remote learners. Such declines are not observed prior to the pandemic among the
class of 2026. Note that the middle grade (i.e., either Grade 4 or Grade 7) in Figures 13-16 represents an actual
data point only for the pre-pandemic class chart and not for the three learning model charts.
The domain scores in mathematics show a similar pattern to the domain scores for ELA. For the students in
Grade 5 (Figure 15), those who learned in-person were the students with the highest domain scores in Grade
3. All three learning models had declines in proficiency from grade 3 to 5. The declines appear to be greatest
for students who learned remotely. Operations and Algebraic Thinking showed slightly lesser decline than the
other domains in all learning models. Again, as with ELA, such declines were not observed prior to the
pandemic among the class of 2026.
For students in Grade 8 (Figure 16), again the students who learned in-person started out at a higher
proficiency level than their peers, with those who learned remotely starting out lower than their peers. All
domain scores decreased between grades 6 and 8 with the exception of Statistics and Probability for in-person
learners which stayed constant. Such declines were not observed prior to the pandemic among the class of
2023.
Figure 17: CT SAT School Day Achievement by High Needs Status (Grade 11)
• As with students in Grade 3 and 4, there may be previous achievement differences between students
in the different learning models.
• Test participation is low among students with high needs (74.5%) as compared to those without high
needs (93.4%).
• Participation among students fully/mostly remote is also low (73.9%) as compared to those in hybrid
or in-person models (91.4%).
• Less than 10 percent of students attended in-person in Grade 11 within both groups (i.e., high needs
and not high needs).
• 49 percent of students with high needs were fully/mostly remote while among students without high
needs, only 28 percent were fully/mostly remote.
Student growth suffered during the pandemic. Students who learned fully/mostly in-person lost the least
ground academically, while those who learned in hybrid or fully/mostly remote models showed substantially
weaker achievement and growth. This pattern held true for students in all grades and most student groups.
While the academic impacts were seen in all subjects, the observed differences were largest in math.
Estimated statewide results from Connecticut’s growth model further indicate that growth before the
pandemic was much stronger than growth during the pandemic. Among low- and high-achieving students,
those learning in-person showed greater growth than those learning in hybrid or remote models. Moreover,
students below proficiency grew at lower rates than those above proficiency which was not the case before
the pandemic. Students above proficiency (Levels 3 and 4) who learned in-person neared pre-pandemic
growth in ELA but not in Math. With regard to the alternate assessments, due to low participation rates, the
aggregate results may not be representative of the population.
The encouraging results among students who learned fully/mostly in-person strengthen the case for offering
full time, in-person instruction during 2021-22. The performance declines, especially in mathematics and
among students who learned in hybrid or fully/mostly remote models, demand the sustained implementation
of evidenced-based solutions.
The CSDE’s American Rescue Plan - Elementary and Secondary School Emergency Relief Fund (ARP-ESSER)
application to the U.S. Department of Education will provide an additional nearly $1 billion to districts across
the state with the majority concentrated in our highest need districts to support a range of evidence-based
activities that are designed to increase student engagement and accelerate learning. The CSDE is also using its
state set-aside of over $100 million to support a wide range of projects including: model curricula; online
curricula and courses; summer enrichment grants; social, emotional, and mental health supports; high dosage
tutoring; specialized initiatives for English learners and students with disabilities; postsecondary access, adult
education, and credit recovery; and boosting engagement of high school students.