Berkeley Unified School District
Exploring where Berkeley's public schools fall short in their performance
Timeline
5 months
Project Type
Internship
Role
UX Researcher
Tools
Deepnote, G Suite
Skills
Data Science, Data Analysis, Data Visualization, Visual Design
Summary
I analyzed and visualized the outcomes for students enrolled in Berkeley's public elementary and middle schools.
Through Berkeley's Data Science Discovery Program, I had the opportunity to work on a local project for the Berkeley Unified School District. This boosted my coding confidence and deepened my passion for using data to drive positive social impact. I also designed the 2 final deliverables.
Code can be found here.
Jump to visualizations
Background
Berkeley Unified School District (BUSD) strives for academic excellence but continues to grapple with longstanding achievement gaps, highlighting the need for more effective strategies and community-driven solutions.
BUSD is the local public school system in Berkeley. It has a history of being one of the highest achieving schools in Alameda County and the state of California, however, for decades there has also been an underlying story of inequity. Despite significant investments in educational programs, the achievement gaps persist, indicating that current strategies may not be yielding the desired results.
A key effort to address this is the Local Control Accountability Plan (LCAP), where the Parent Advisory Committee (PAC) reviews student performance data and provides feedback to the district. However, interpreting this data remains a challenge, making it difficult to drive meaningful improvements.
Problem
How can we effectively interpret the data to understand its real-world implications for schools, and then clearly present these insights to the PAC?
Objectives
This project aimed to use evidence-based storytelling to make student performance data more accessible and actionable for stakeholders through 3 key goals.
Analyze data from the California Assessment of Student Performance and Progress (CAASPP) - Smarter Balanced Assessment (SBA) to identify patterns in achievement gaps.
Create visualizations to convey insights related to student performance in English Language Arts (ELA) and Mathematics, focusing on socioeconomic disparities.
Develop a reproducible workflow for ongoing analysis and future iterations.
Students of Interest
Rather than focus on race or ethnicity, we wanted to look at an intersectional population through an economic lens.
Since the data was collected by each school and could not be reduced to each student, we followed five cohorts.
1
Fourth grade cohort each year
2
Eighth grade cohort each year
3
Psuedo-cohort* each year
4
All 11 elementary schools
5
All 3 middle schools
*We would not be able to track each student as they advanced to each new grade, so we followed the same class of students starting as 3rd graders in 2015 to 11th graders in 2023.
Methodology
CAASPP testing data spanned across 7 years, all of which was cleaned before visualizing.
All data for this project was collected from the California Assessment of Student Performance and Progress (CAASPP) - Smarter Balanced Assessment (SBA). Data from all seven available testing years were collected: 2015, 2016, 2017, 2018, 2019, 2022, 2023. No testing was administered in 2020 and 2021 because of COVID-19.
Additional datasets containing numerical keys were also collected, ie. datasets containing codes for all represented subgroups.
While this analysis highlights the achievement gap between SED and non-SED students, the dataset does not include the proportion of each group at each school. As a result, the findings reflect overall performance trends but do not account for how student distribution may impact the gap at the school level.
If you would like to know how I cleaned the data before visualizing it, please email me at monicacortes@berkeley.edu.
Visualizations
Utilizing Python's visualization libraries, I visualized the percentage of students meeting or exceeding state standards.
SED = Socio-economically disadvantaged students; Non-SED = Socio-economically disadvantaged students
4th graders each year (2015 - 2023)
Amongst fourth graders, there is no significant decrease in the achievement gap in the past 8 years.
The highest achievement gaps in ELA and Math topics were both recorded in 2015: 52% (ELA) and 51% (Math).
The achievement gap had already been decreasing before 2020, but its increase in 2022-2023 suggests that new BUSD programs, ie. 2020 Vision have not provided a long-term, tangible benefit.
However, because state testing was not done in 2020-2021, it is also difficult to confirm this just looking at CAASPP data alone.
8th graders each year (2015 - 2023)
Amongst eighth graders, there is ultimately no significant decrease in the achievement gap as students leave the eighth grade and enter middle school.
The highest achievement gaps in ELA and Math topics were both recorded in 2022: 46.12% (ELA) and 43.8% (Math).
This could be explained by students returning to schools and this exam after Covid-19 procedures.
While at least 70% of non-SED students consistently meet or exceed state standards, SED students doing the same remains below 50%.
Pseudo-cohort each year (followed the same, one class from 2015 - 2022)
Achievement gaps show a significant decrease as students progress to each next level, but both SED and non-SED students fall in terms of progress.
This indicates that SED students are able to catch up their non-SED counterparts as they each progress in school.
Unfortunately, both sets of students were affected by the Covid-19 crisis, indicated by a sharp decline in both ELA and Math results in 11th grade (2022).
Since newer programs have only recently been introduced (beyond 2018/6th grade) it's difficult to say whether they were contributing to the decreasing achievement gap.
All 11 elementary schools (2022 - 2023)
Within 2022 to 2023, state test results did not indicate a significant decrease achievement gap.
Non-SED students meeting or exceeding the state standard averaged 78.6% (ELA) and 76.4% (Math), while SED students results averaged at 35.2% (ELA) and 33.7% (Math).
Berkeley Arts Magnet at Whittier and Rosa Parks continued having 2 of the lowest achievement gaps, while Washington and Cragmont remained having 2 of the higher achievement gaps.
All 3 middle schools (2022 - 2023)
Within 2022 to 2023, state test results did not indicate a significant decrease achievement gap, with students performing in a similar range as the elementary schools.
Non-SED students meeting or exceeding the state standard averaged 78.6% (ELA) and 65.1% (Math), while SED students results averaged at 35.2% (ELA) and 28.4% (Math).
Conclusion
Persistent achievement gaps highlight the need for further research.
Persistent achievement gaps in BUSD highlight the need for further research. While data suggests potential trends, disruptions from COVID-19 limit definitive conclusions. Future work could explore:
Enrollment Data: Understanding how student demographics influence achievement gaps.
Qualitative Insights: Interviews with students, parents, and teachers for deeper context.
Best Practices: Examining schools with smaller gaps for scalable strategies.
Elementary-to-Middle School Transitions: Identifying key intervention points.
To ensure impact, our findings were shared with stakeholders through an interactive webpage and a poster. The code remains open for future researchers to build upon and further efforts to close BUSD’s achievement gap
Final Deliverables
The final findings of this project were shared with peers and district stakeholders through an interactive webpage and a detailed poster presentation.
Includes interactive visualizations
Poster: https://docs.google.com/drawings/d/1oGiShqsR1Ew778qpfLad8v1CW4WqbWv0noClDTACi_I/edit
Wishes
Getting district specific data to enhance our understanding
Unfortunately, challenges in obtaining district-wide data limited our ability to gain a more comprehensive understanding of student performance. Our initial goal was to compare state assessment results with district-wide exams to better evaluate the effectiveness of educational programs. Additionally, this limitation prevented us from accounting for differences in student representation across schools, making it difficult to fully assess variations in achievement gaps.
Learnings
Real-world data is often messy and requires careful validation
Cleaning and validating data was crucial. One challenge was deciphering a code change in CAASPP data, requiring verification with a representative through email.
Incremental progress is still valuable
Given the missing data from 2020 and 2021 and the absence of key district- and school-level details, verifying many findings was difficult. While this may seem like a limitation, it does not mean the project was unsuccessful. I hope that future student researchers can build upon this work, using the existing code, visualizations, and automation scripts—assuming CAASPP maintains consistency in its data reporting—to further advance the study of achievement gaps in Berkeley’s public schools.
Final Mentions
Special thanks to my team and the UC Berkeley Discovery Program!
Each week, I had the opportunity to discuss the data and new findings with a PAC member leading the project, a UC Berkeley professor, and a fellow Data Science major who assisted with final statistics and interactive visualizations. This project was also made possible through UC Berkeley’s Discovery Program, which connects students with data science research opportunities.
Next:
How can we support incarcerated individuals on their journey towards parole?
See case study