Enrollment by Grade Level

Figure 1 provides a simple line chart showing enrollment by school year for all grades from Pre-Kindergarten to 12th grade. The red, dashed line marks the last enrollment period prior to the COVID pandemic - the 2019-20 school year.

Most grades demonstrate a relatively steady enrollment rate after the COVID line, with slight drops for a few grade levels. Only the youngest grades demonstrate any type of significant drop at this point in time and are highlighted in the chart as well as called out in the legend. Additionally, Pre-Kindergarten, Kindergarten, and 1st Grade all show signs of recovering enrollment in the last 1 to 2 years.

Half-day kindergarten is an exception here, where the drop in enrollment starts well in advance of COVID. Kindergarten is not mandatory for Washington State, and prior to 2017, the State only mandated that half-day kindergarten be available to all students. Starting with the 2017-18 school year, all school districts were required to offer voluntary full-day kindergarten. Thus, we are seeing a shift away from half-day enrollment as full-day became more available11 See Kindergarten Requirements for the State of Washington https://www.superpages.com/em/kindergarten-requirements-state-washington/.

Enrollment by Race

A breakdown of enrollment by race requires moving to a logarithmic scale due to the extreme differences in enrollment numbers across races. The log scale allows us to focus on the relative change in enrollment and the limited number of races captured allow the introduction of small-multiples in the visualization.

Figure 2 provides definite and somewhat surprising trends. Other than small changes immediately after the COVID outbreak, again marked by the red, dashed line, several race groups show increases in enrollment. American Indian/Alaskan Native shows a drop, but this appears to be a trend continuing from before the pandemic. The White race shows a major drop, though a slight decrease can also be seen prior to the pandemic.

Students of Two or More Races seem to have stabilized with the onset of COVID after a period of marked increase. Black/African American, however, shows a sharp increase in the wake of COVID. Because the White population is a strong majority of the students represented in this data, the decrease in enrollment for White is also reflected in the student total (All Students) and may be responsible for the anecdotal stories of decreases I have heard within my own School District.

Enrollment by Demographic

The demographics provided by the Washington State data, unlike Race, cover several different types of categorization and include overlap of students. For this analysis, several chart types were tested, including density plots and boxplots. Ultimately, the use of line charts in small-multiples and a logarithmic scale proved most useful for the data. These charts allow a simple comparison across student groups and a comparison of each group to the total. In each case, we can see clear patterns Pre and Post COVID-19.

Scanning across the 11 charts in Figure 3, the categories where we see the most evidence of a post-COVID drop in enrollment are Foster Care, Homeless, Low-Income, and Students with Disabilities. Section 504 students seem to have a delayed reduction in enrollment, not occurring until the 2021-22 school year, and may have quickly steadied off. Support for Section 504 students in Washington State is extensive, and with limited options, parents of these students likely opted to continue with the programs.22 A brief comment on the category of Section 504: Section 504 is a Federal law designed to ensure School Districts “provide to students with disabilities appropriate educational services designed to meet the individual needs of such students to the same extent as the needs of students without disabilities are met.” (See U.S. Department of Education FAQ About Section 504 https://www2.ed.gov/about/offices/list/ocr/504faq.html. In the Washington State data, Section 504 refers to students that are specifically in programs designed to meet special educational needs. Although the State Data Catalog is not explicit, the Students with Disabilities category refers to students with disabilities enrolled in regular curriculum.

Both the Female and Male categories demonstrate similar patterns with sudden drops after the red line. However, as these categories overlap with all others, and do not demonstrate a significant difference, we can assume that they are simply reflecting the overall pattern we see with All Students.

The Homeless and Low-Income categories appear to show some recovery since the initial drop in enrollment, and both demonstrated a slight pre-COVID decrease as well. The stand out categories here are Foster Care and Students with Disabilities, clearly staying with low enrollment, and possibly continuing to decrease in the case of Foster Care.

Rate of Meeting Assessment Expectations

The Density Plots in Figure 4 show the percent of students that have met the minimum standards at each school. The line represents the mean percent of students meeting expectations across the state for the given test in the specified school year. By examining the changes over time, we can understand how well students are performing on the standardized tests.

If you are viewing this document as an HTML file, two separate animations are presented, one for each test. The animations cycle through the school years to allow an examination of how the density and mean test results shift from year to year. The static version of the chart, for PDF or Word documents, includes several small multiples illustrating the same information.

As we scan results across years, the density of the percentage of students meeting expectations remains relatively stable until we reach the 2020-21 and 2021-22 school years, highlighted with the dark red fill color. At this point, the data becomes skewed, and the mean clearly shifts to the left, indicating that the percentage of students meeting standards overall decreased in the years after COVID.

We can see similar outcomes for the AIM test. The results appear to support a potential hypothesis that testing results decreased in the wake of the pandemic, and students’ education likely suffered from schooling that shifted online during that time.

Graduation vs Cohort by Year

In this first visualization for graduation rates, we compare the number of students in a cohort to the number of graduates for that cohort. The total number of students can vary significantly from school to school. The boxplots in Figure 5 provide a good perspective of the range of students in each cohort and graduating class from year to year.

As a reminder, the dark red dashed line indicates the school year in which COVID started.

These distributions remain relatively even across the eight school years for which we have consistent data. In each case, the number of graduates is slightly lower than the number of students in the cohort. This is expected based on a certain rate of dropouts and transfers. However, the overall distribution of both the cohort size and number of students graduating remains similar.

We can see evidence of the median number of students and interquartile ranges increasing each year, including the last two years captured: the graduating classes of 2021 and 2022. This seems to indicate that the pandemic did not adversely impact graduation. Though decreased graduation could be offset by population increases, which may appear through the rate of transfers into schools during this time.

Graduation Rate Density Curves

The graduation rate is calculated as Graduated / FinalCohort. Whereas in the previous boxplot, we looked at the distribution of the number of graduates by school, here we will look at the distribution of the graduation rates for each school. As we saw earlier with the rate of students meeting Testing Expectations, a Density plot does a good job of showing us how graduation rates are distributed.

In Figure 6 the graduation rates have steadily increased since the 2014-15 school year, which is good news. It also appears to have plateaued within the last few years in the mid-80 percent rate. It should be noted that a State mandate was issued upon the onset of COVID-19 in the Spring of 2020 that no student would be failed in the 2019-20 school year, which likely impacted these numbers.

The Graduation Rate distribution has remained skewed to the left throughout the school years, strongly indicating a persistence among some schools to have lower rates of graduation. Alternatively, this may also indicate an issue with the data collection.

Graduation Rates by School

In a perfect situation, all students within a Graduating Cohort would graduate. From the Density plots in Figure 6, this is clearly not the case judging by the long tail to the left of each plot. The earliest school year examined had a mean graduation rate of 78%. If we set a threshold close to this mean, we can examine individual schools in a scatter plot and start to understand where students may be falling behind in graduation.

The dark gray straight line in Figure 7 from lower left to upper right represents a perfect graduation rate. Most High Schools within Washington State fall somewhere close to this line. Each point represents a single school for a single school year. The schools highlighted in dark red are schools that had a graduation rate less than 75% for a particular school year. Although alternative thresholds were considered, including standard deviations, I found these alternatives tended to miss the group of schools in the upper right - larger cohorts with lower graduation rates. Schools with smaller cohort sizes may have an inherent bias based on their size, or may have issues with data collection. For example, often these are alternative high schools or schools located in rural or indigenous populations. Both are important populations to consider in this analysis. At the same time, examining larger schools that also exhibit lower graduation rates may provide additional material of interest.

When this document is viewed in HTML, it includes a play button that supports cycling through the School Years. What stands out in this animation is the disappearance of schools with larger cohorts and lower graduation rates during the school years of 2015-16 to 2019-20, and the subsequent reappearance staring in the 2020-21 school year. This encourages further analysis at the school level.

In Figure 8, we are continuing the school level examination, focusing now on those schools that had graduation rates less than 75% at least once over the course of the eight school years included. In the Heatmap below, colors range from the lowest graduation rates, emphasized in dark red, to higher graduation rates above 75% in blue. Graduation rates of 75% are in white, establishing a visual threshold for schools that ranged above and below the 75% mark.

Schools that are consistently low in graduation rates, based on the chart below, include:

• Computer Academy
• New Horizons
• Pace Alternative
• State Street
• Summit View
• Talley
• West Auburn

At the same time, several schools showed an improvement in graduation:

• ACES
• AIM
• Discovery
• Dishman
• Lincoln Hill
• New Market
• Weston

Both sets of schools present opportunities for further exploration at the demographic level, including race and income status. However, another aspect of this data not explored to this point is geographic location.

Mapping Graduation Rates

The first choropleth map shows the graduation rates we just looked at in the previous section. Here we are averaging results across all high schools within a county. Again, we set our threshold to a 75% graduation rate and span colors across that midpoint. Darker blue results are higher graduation rates, while darker red are lower graduation rates.

The results appear in Figure 9. Scanning across the school years, two counties in the Northeast section of the state pop with especially low graduation rates. This is most prevalent during the pandemic years. Figure 10 zooms in on one School Year to label and identify the counties in question.

Note that for the sake of space, not all counties are labeled here, as the primary point is identifying the two counties with the lowest average graduation rates. Figure 10 clearly shows that Douglas and Ferry counties have the lowest average graduation rates. Though it should be noted from Figure 9 that these counties did not report for every school year, indicating there may be missing data that impact this information.

Mapping Assessment Results

Knowing where graduation rates are of most concern, we can now look at Standardized Tests and the percentage of students meeting minimum test expectations. When we examined the density plots for testing earlier, the mean value prior to COVID was close to 50%. This was the case for both the SBAC and AIM tests, indicating that on average, 50% of the students met the minimum test expectations.

After COVID started, though, that mean value dropped to slightly above or slightly below 40%. With this sudden change in testign outcomes in mind, we can use a midpoint of 40% for the next choropleth.

What stands out most in Figure 11 is the sudden change starting with the 2020-21 school year. Prior to this point, most counties showed results above the 40% threshold we set. However, after COVID, the situation is reversed, and most counties are below the threshold. If we are to use standardized tests as an indicator, clearly, students have been impacted by the COVID pandemic.

Ferry county, appearing in the upper right of the state, and one of the two counties that stood out with low graduation rates, consistently shows lower testing results as well. This county shows values in the red spectrum throughout all seven years of test results, with even darker red colors (indicating lower testing outcomes) in the last two years. Ferry county certainly suggests further analysis.

Mapping Enrollment Rates

Finally, we take a look at enrollment rates, the subject matter we started this analysis with. As the actual enrollment does not support the visual design of a choropleth map, the following visualizations examine the change in enrollment from year to year. This allows us to assess whether enrollment is increasing or decreasing within a particular county. Although our data supports several different examinations of enrollment, I have selected three specific visualizations here:

• Figure 12: The change in enrollment from year to year for all students within each county.
• Figure 13: The change in enrollment for the first full COVID school year (2020-21) by race.
• Figure 14: The change in enrollment for the first full COVID school year (2020-21) by demographic.

Note that since we are mapping a change in enrollment, we do not show the first year in our data set, the 2014-15 school year. For these choropleths, I have set the color midpoint at 0. A decrease in enrollment is indicated by the red colors, and an increase in enrollment is indicated by the blue colors.

In Figure 12 we see the appearance of light red across most counties in 2020-21 - a state-wide decrease in enrollment in public schools. The exception is three counties in the Northeastern part of the state, and one county at the bottom center of the state. This is Klikitat county, which appears to increase in enrollment. We will see more of Klikitat in the next visualizations.

In the subsequent two years, that slight enrollment decrease appears to fade again, and we see relative little change in the last two years of enrollment. The one exception here is Columbia county in the lower right, which shows a major increase in enrollment of about 100%.

With 2020-21 providing such an interesting change in enrollment across the state, the next two choropleths zoom in on specific groups of students during that year, starting with a breakdown of race.

The maps in Figure 13 are surprising actually in how little they show us. However, if we remember the trend lines in the first section of this document, most races showed consistent changes from year to year. The exception was the White group. However, that map shows very little color in the choropleth, indicating the average enrollment change by county is relatively low.

The standout here is Klikitat county, the short and wide county at the bottom center of the map. For Klikitat, we see a strong increase in enrollment for the Black/African American, Asian, and Native Hawaiian/Pacific Islander groups. My personal knowledge of the county provides little insight, as there are no significant cities or industries in the area to explain the differences.

Finally we can examine student demographics by county. Figure 14 shows a range of results for the first school year after the start of COVID. The patterns for Students with Disabilities and Low-Income students seem to match the pattern for All Students, providing little information of value. Foster Care stands out with a strong decrease in enrollment across three counties. These counties include Yakima, a highly agricultural county, and Pierce, which includes a large port on Puget Sound. Military Parents show a strong increase in enrollment in Klikitat county, at bottom center. This struck me as unusual as there are no military bases in that county.

Homeless, Migrant, and Military groups show an increase in enrollment for Ferry county in the upper right, one of two counties where we saw decreases in graduation rates earlier. At the other end of the spectrum, Migrant students show an significant decrease in enrollment for Whitman county in the lower right, home of Washington State University.

Appendix A: On the Document Format

I have formatted this document using the Tufte library package (https://bookdown.org/yihui/rmarkdown/tufte-handouts.html). As the course started with a reading from Edward Tufte’s The Visual Display of Quantitative Information, continuing with that theme seemed appropriate. In this case, the Tufte theme applies default formatting to the R-Markdown document that matches the format of Tufte’s books. Although I do not agree with Tufte on all aspects of Data Visualization and design, I find the layout of his books exceptional in supporting readability and the clear communication of information without clutter.

As this is an R-Markdown document, I have included animations using the gganimate and plotly packages. When rendered in HTML, these animations will be available for illustrating some of the key concepts and evidence presented. However, the primary design is intended to be rendered in a more static format, such as PDF or Microsoft Word. My desire is to focus the reader’s attention on the visualizations themselves, and not on the flash of technology. Though dashboards and widgets can support user interaction, they can not replace the value of good chart design and thoughtful analysis.

“Most techniques for displaying evidence are inherently multimodal, bringing verbal, visual, and quantitative elements together.” “…those who reason about evidence often seek to place different forms - verbal or nonverbal - as colleagues in explanation.”

44 Tufte, E. (2006). Beautiful Evidence. Cheshire, Connecticut: Graphics Press LLC. Page 83.

Appendix B: Data Sources and Key Variables

All data has been sourced from the State of Washington’s Data Catalog, available at https://data.wa.gov. The Washington State government provides a range of reports on topics associated to Transportation, Employment, Healthcare, Economics, Demographics, and, of course, Education. Additionally, the State provides a Data Catalog where data can be downloaded in CSV format.

For this project, I have made use of several Report Card data files included in the Education Category of the Data Catalog. For referencing the original files, I have included the search term used on the Washington State Data Catalog in the table below.

Data Sources (https://data.wa.gov)

Subject	Search.Phrase
Enrollment	Report Card Enrollment from 2014-15 to Current Year
Enrollment	Report Card Enrollment 2022-23 School Year
Assessment	Report Card Spring Assessment Data from 2014-15 to 2021-22
Graduation	Report Card Graduation 2014-15 to Most Recent Year
Graduation	Report Card Graduation 2021-22

All data is available from the school year starting 2014 through the school year ending 2022. Enrollment data is also available for the school year ending 2023. All data sets are captured at the level of individual schools, and some data includes more granular results at, for example, grade level. Additionally, aggregates are often provided at the levels of School district and State (i.e. all-up for Washington State). Throughout this analysis, I have often leveraged these inherent levels of information to produce summary results, or to analyze a specific subgroup of students.

The State of Washington Report Card data files include a separate file with Data Notes, providing metadata for all columns included in the downloadable files. The Data Notes is provided with the source files used in this project.