by Kathleen, Thomas, Meg, and Dieter
Our aim in this project was to search for relationships between expenditure, revenue, and standardized test scores in the United States. We planned to find out if different kinds of expenditure have notable effects on test scores, as well as how expenditure per student affects test scores. While standardized test scores are not the most holistic approach to gauging academic success in a school district or state, these scores are the most objective and accessible gauges of measurement to use in a study of this nature. We approached this analysis from several angles, creating visualizations that address each state within them, as well as narrowing in on specific locations. We began our exploration with the creation of a correlation matrix, continuing on to create a visualization of the state average reading scores by year. Washington D.C. stood out in this visualization for consistently having much lower scores than other states, which led to a deeper dive into the city’s scores and total revenue. We proceeded to investigate a possible relationship between expenditure per student and test scores, which is visualized in a plot as well as a map. Visualizations of support services expenditure as well as instruction expenditure did not prove to have any correlation to an increase in test scores.
We found that there is a positive correlation between school expenditure and the test scores of students. The relationship, however, is not linear. At relatively low per-student expenditures, we find that increasing per-student expenditure results in a larger proportional increase in student test scores. There appears to be an inflection point that occurs after the national average. At which point, increasing per student expenditure has a drastically reduced impact on student test scores. It is important to note that this result is true if we do not include the District of Columbia in our calculations. D.C., as a result of its unique circumstances, such as its considerably lower number of schools and the urban nature of its school districts, is a true outlier. By including D.C in our data analysis, we see that the increase in spending per student, beyond the national average, actually has a negative correlation with test scores. The maps presented help to visualize these disparities in national test score distributions as well as national school expenditure distributions. It is important to note the inherent limitations of our data sets and the conclusions we present. D.C is illustrative of the fact that the demographics and economic factors of a region play a considerable role in the performance of students. These factors are intrinsically linked to the data we present, as such, while we do argue that there is a correlation between student expenditure and student performance, the data we have does not allow us to rule out or quantify the impact these other factors have. Hence, we recognize that correlation does not imply causation and we advocate for a more comprehensive analysis of how some of the mitigating factors mentioned above impact student performance.
Our presentation can be found here.
The dataset used in this presentation is found in the website, Kaggle and it is named “U.S. Education Datasets: Unification Project”. This dataset is a combination of three different sources: the National Center for Education Statistics, the United States Census Bureau, and The Nation’s Report Card. This dataset has an extensive amount of information that allows us to explore different aspects of education, more specifically how, and if, the total expenditure of states for schools affected the scores of its students. This was possible due to the fact that the dataset contains information from all 50 states regarding the total expenditure of each state and their average test scores. The average scores were obtained by the National Assessment of Education Progress, also known as NAEP. It is important to note that the average test score is an estimate gathered from a representative sample of students in both public and nonpublic schools rather than all the student population. With this information, the group was able to look at different statistical analyses of the data to help us answer our question.
Project.” Kaggle, 13 Apr. 2020, https://www.kaggle.com/noriuk/us-education-datasets-unification-project.
Data is sourced from the U.S. Census Bureau and the National Center for Education Statistics (NCES). # Enrollment https://nces.ed.gov/ccd/stnfis.asp # Financials https://www.census.gov/programs-surveys/school-finances/data/tables.html
https://www.nationsreportcard.gov/ndecore/xplore/NDE
Garrard, Roy. “U.S. Education Datasets: Unification Project.” Kaggle, 13 Apr. 2020, https://www.kaggle.com/noriuk/us-education-datasets-unification-project.