geo_facet with scatterplots: the relationship between local revenue and student reading proficiency

Scatterplots for each state illustrating the relationship between district local revenue (total and from property taxes) and student reading/language arts profiency.

Amy

Research Question:

What is the relationship between district funding (both total local revenue and local revenue from property taxes) and student reading/language arts outcomes by state?

Design choices:

• Scatterplots enable the audience to observe the strength and direction of the association between two variables, as well as the spread, number of observations, and get a sense for the type of relationship (e.g., linear, quadratic).
• Used geo_facet() to organize the scatterplots in the approximate shape of the United States (instead of alphabetically).
• Improved cognitive load by minimizing graph elements, such as shading, axis breaks, and grid lines
• Selected a muted color for geom_points that was cohesive with the colors in the other plots I created
• Used transparency with geom_point() to allow the audience to gauge the density of observations
• Clear, succinct, and polished title and labels
• Log transformation of the x-axis so that the points weren’t clustered at the lower end of the range

Intended audience:

Researchers and policy makers.

Final Plots

Local Revenue x Reading Proficiency

Property Tax Revenue x Reading Proficiency

Findings and Observations:

• In many states, total local revenue and revenue from property taxes are positively associated with the reading/language arts outcomes of students.
• The strength of the association varies by state.
• There are states where there doesn’t appear to be a relationship between the variables (e.g., Colorado, Texas, Vermont).
• The relationship between local revenue and student reading proficiency appears to be strong in several states, including New Jersey. This state is notable as it has the highest average district local revenue from property taxes. Given more time and bandwidth, I would have liked to create a district-level choropleth of revenue in New Jersey and perhaps a bivariate map showing the correlation between property taxes and outcomes.

Prior Versions

Prior Version 1

Prior Version 2

Prior Version 3

Prior Version 4

Evolution of plots over time:

As both plots are similar in regard to format and design choices and went through the same evolution, I’ve included the previous versions of just one plot above.

• I began with exploring the overall relationship across all states and districts. Prior Version 1 is my initial rough and basic attempt at this visualization. Default settings and elements are used. The points are clustered near zero.
• In Prior Version 2, I made some refinements (alpha, x-axis scale, minimal graph elements, etc.)
• In Prior Version 3, I decided to try color and added stroke = 0 based on the helpful feedback from peer reviews.
• Given that policies of district funding and property taxes vary at the state level, I wanted to be sure to explore scatterplots faceted by state (Prior Version 4). The upside of this version is that only the states with data are included. However, given regional patterns observed with the bar plots and maps, I opted to use facet_geo() so that the plots were arranged roughly in the shape of the US. I also added some color back in and adjusted breaks and labels.