Mapping School Finance Data

Introduction

Maps are often the most effective way to communicate school finance patterns: funding disparities, labor-cost geography, and property-wealth gradients all have strong spatial structure. edfinr does not ship geometries, but its ncesid matches the GEOID used by Census Bureau school district boundary files – the same identifier the package itself uses to join ACS data – so joining finance data to shapes from the tigris package is a straightforward process.

library(edfinr)
library(dplyr)
library(ggplot2)
library(sf)
library(tigris)

# cache downloaded shapefiles across sessions
options(tigris_use_cache = TRUE)

Before you map: district geography caveats

1. Three overlapping layers. Census publishes unified, elementary, and secondary school district boundaries. In states where elementary and secondary districts coexist (Illinois, California, and others), the layers overlap, and mapping only the unified layer silently drops districts. tigris::school_districts() takes a type argument; check your state’s structure before assuming "unified" covers it.
2. Boundaries change. Use a geometry year close to your data year, since district consolidations and boundary changes accumulate.
3. Cartographic boundary files. cb = TRUE returns generalized boundaries that are smaller and draw faster; use the default TIGER/Line files only when you need legal-boundary precision.
4. Not every district joins. Some edfinr districts (certain charter LEAs, for example) have no mapped boundary, and some mapped areas have no finance record. Count the misses after joining rather than assuming completeness.
5. Map area tracks geography, not students. Geographically large, sparsely populated rural districts dominate a statewide choropleth’s ink, while the districts serving the most students occupy the least. A map can invert the visual impression of where most students actually are; consider labeling or insetting major metros when that distinction matters.

Joining finance data to boundaries

Ohio features unified school districts, which keeps the example simple and avoids overlapping geometries. First, pull the finance data and the boundaries, then join on GEOID = ncesid:

oh_2023 <- get_finance_data(yr = "2023", geo = "OH")

oh_shapes <- school_districts(state = "OH", year = 2023, cb = TRUE)

##   |                                                                              |                                                                      |   0%  |                                                                              |=                                                                     |   1%  |                                                                              |===                                                                   |   4%  |                                                                              |====                                                                  |   6%  |                                                                              |======                                                                |   8%  |                                                                              |=======                                                               |  10%  |                                                                              |==========                                                            |  14%  |                                                                              |===========                                                           |  15%  |                                                                              |=============                                                         |  19%  |                                                                              |===============                                                       |  21%  |                                                                              |================                                                      |  22%  |                                                                              |=================                                                     |  24%  |                                                                              |=================                                                     |  25%  |                                                                              |===================                                                   |  26%  |                                                                              |===================                                                   |  28%  |                                                                              |======================                                                |  32%  |                                                                              |=======================                                               |  33%  |                                                                              |=========================                                             |  35%  |                                                                              |============================                                          |  39%  |                                                                              |=============================                                         |  41%  |                                                                              |=================================                                     |  48%  |                                                                              |===================================                                   |  50%  |                                                                              |=======================================                               |  56%  |                                                                              |=========================================                             |  58%  |                                                                              |==========================================                            |  60%  |                                                                              |=============================================                         |  64%  |                                                                              |==============================================                        |  66%  |                                                                              |===================================================                   |  72%  |                                                                              |====================================================                  |  75%  |                                                                              |======================================================                |  77%  |                                                                              |========================================================              |  81%  |                                                                              |==========================================================            |  83%  |                                                                              |================================================================      |  91%  |                                                                              |===================================================================   |  95%  |                                                                              |======================================================================| 100%

oh_map <- oh_shapes |>
  left_join(oh_2023, by = c("GEOID" = "ncesid"))

# diagnostics: how many shapes have no finance record, and vice versa?
sum(is.na(oh_map$rev_total_pp))

## [1] 4

nrow(anti_join(oh_2023, sf::st_drop_geometry(oh_shapes), by = c("ncesid" = "GEOID")))

## [1] 327

Mapping per-pupil revenue

With an sf object, ggplot2::geom_sf() handles the rest. Binned scale usually communicate funding levels better than continuous gradients (which can be skewed by outliers), because readers can attach a dollar range to each color:

ggplot(oh_map) +
  geom_sf(aes(fill = rev_total_pp), color = "white", linewidth = 0.05) +
  scale_fill_viridis_b(
    n.breaks = 6,
    labels = scales::label_dollar()
  ) +
  labs(
    title = "Total Revenue Per-Pupil in Ohio Districts, SY2022-23",
    fill = "Revenue\nPer-Pupil"
  ) +
  theme_void()

Mapping labor costs with CWIFT

Any edfinr variable maps the same way. The CWIFT teacher-wage index makes the metro/rural labor-cost geography visible (see the “CWIFT” article for what the index does and does not measure):

ggplot(oh_map) +
  geom_sf(aes(fill = cwift_est), color = "white", linewidth = 0.05) +
  scale_fill_viridis_c(option = "magma") +
  labs(
    title = "Comparable Wage Index for Teachers in Ohio, FY2023",
    subtitle = "FY2023 values are carried forward from FY2022",
    fill = "CWIFT"
  ) +
  theme_void()

Going further

• Multi-state maps: school_districts() accepts a vector of states; filter get_finance_data(geo = ...) to match.
• Coastal states: tigris::erase_water() improves shoreline maps.
• ACS overlays: the tidycensus package retrieves ACS variables with geometry included (geometry = TRUE); its GEOIDs join to ncesid the same way, which is convenient when you need ACS measures beyond those bundled in edfinr.
• Interactive maps: sf objects work directly with mapview and leaflet for exploratory work.

See also

• The “CWIFT” article for interpreting the wage index.
• The “Community and Economic Context” article for the ACS and SAIPE measures worth mapping.
• The “Data Quality and Comparability” article for panel-composition caveats that affect join completeness.