Introduction to Learning Analytics
Dr. Joey Huang
December 6, 2025
In 2025, Multilingual learners make up 11% of school age children in North Carolina. According to the Migration Policy Institute, in 2024 21.9% of children in North Carolina had at least one immigrant parent.
But are our Multilingual learners being served equitably in schools throughout the state? North Carolina has a wide variety of counties, from mountainous regions to rural farmlands to metropolitan cities to coastal beaches. Are our English learners making similar growth in each area?
In this study, I will examine the North Carolina’s Multilingual Learners English Language proficiency growth since 2018 in three different types of areas: rural, urban, and suburban.
___
A multilingual learner is an English as a Second Language student.
“Multilingual Learner” is now used instead of ESL as many language learners speak and are learning more than one language.
An annual English Language Proficiency (ELP) assessment for Kindergarten to 12th grade ML students.
Measures the percent of Multilingual Learners who made adequate progress that year. “Adequate progress” is determined by a formula, but can be approximated to 1 English Language Proficiency level higher per year.
Around 42 states including North Carolina use the ACCESS test as a federal accountability measure.
ACCESS measures English Language Proficiency in 4 domains: Listening, Speaking, Reading, and Writing.
I used 2 datasets from the NC School Report Cards from North Carolina Department of Instruction and one dataset from the US Census Bureau Population Division.
Dataset 1, rcd_location: Basic information and category code about each Local Education Agency (LEA).
Dataset 2, Rcd_acc_elp: English Learner growth based on the WIDA ACCESS English Language Proficiency test.
Dataset 3, NC_Urban_Rural: The relative urban/rural qualities of counties in North Carolina.
{r}
library(dplyr)
library(tidyverse)
library(readxl)
{r}
rcd_location <- read_excel("~/rcd_location.xlsx")
rcd_acc_elp <- read_excel("~/rcd_acc_elp.xlsx")
NC_Urban_Rural <- read_excel("C:/Users/katie/OneDrive/Desktop/NC_Urban_Rural.xlsx")
For this study, there are 3 types of necessary information:
Year: attached to the school’s agency code and county name in rcd_location and the school only in rcd_acc_elp
County type (urban, rural, or suburban): only on NC_Urban_Rural from Census.gov, attached to county name
Percent of English Language Proficiency Growth: only on rcd_acc_elp, attached to school’s agency code, status of students as English learners, and year.
I removed vocational, laboratory, federal, and regional schools to focus solely on traditional public schools.
{r}
rcd_location_elp <- rcd_location |>
filter(year >= 2018,year <= 2024, school_type == "Regular School")|>
select("year", "agency_code", "county") |>
arrange(year)
View(rcd_location_elp)
View(NC_Urban_Rural)
The Census Bureau identifies counties with several characteristics, so for the purposes of clarity in this study the columns and values were mutated from 3-column characteristics like “Urban-Metropolitan-Outlying” to a 1 column format like “Suburban”.
{r}
NC_Urban_Rural <- NC_Urban_Rural |>
mutate(CENTRAL_OUTLYING = replace_na(CENTRAL_OUTLYING, "Rural"))|>
select(COUNTYNAME, URBAN_RURAL, CENTRAL_OUTLYING) |>
rename(county_type = CENTRAL_OUTLYING)|>
mutate(county_type = recode(
county_type,
"Outlying" = "Suburban", "Central" = "Urban"
))
{r}
urban_rural_suburban <- NC_Urban_Rural |>
select(COUNTYNAME, county_type)
Columns for the year, agency code of the school, subgroup of students, and the percent of English Learners that made progress on the ACCESS test were selected on the English Language Proficiency score datasheet “rcd_acc_elp”.
Results were limited to only rows that applied to students listed as English Learners.
{r}
elp_MLs <- rcd_acc_elp |>
select("year","agency_code", "subgroup", pct) |>
filter(subgroup == "ELS")
The next step in this study was to join the county data from the census with the information from NC School Report Cards. The two datasets were joined using the county names.
counties_merged <- rcd_location_elp |>
inner_join(urban_rural_suburban, by = c(county = "COUNTYNAME"))
View(counties_merged)
Next, the new dataset was joined with the English Language Proficiency dataset.
{r}
elp_growth <- counties_merged |>
inner_join(elp_MLs, by = c("year", "agency_code"))
View(elp_growth)
This final merged dataset was cleaned to eliminate the column “subgroup”.
{r}
elp_growth_final <- elp_growth |>
select("year", "county", "agency_code", "county_type", pct)
View(elp_growth_final)
In this analysis we are looking at 2 key metrics:
We are using a box plot to examine how Multilingual Learners perform differently or similarly in each of these areas, and comparing these results by year.
{r}
library(ggplot2)
ggplot(elp_growth_final, aes(y = pct, x = county_type, fill = county_type)) +
geom_boxplot(size = .5) +
labs(title = "Multilingual Leaners Language Growth",
x = "Type of County",
y = "Percent ELP Growth") +
facet_wrap(~year)
There is a distinct difference in growth in rural and urban counties from 2022 to 2024, after the interrupted schooling from the COVID Pandemic.
A wide disparity exists now between Multilingual Learners in rural and urban counties that did not exist in the pre-pandemic data.
Median English Language Proficiency scores of all county types plummeted from above 37% to below 25% after the interrupted schooling of the 19-20 and 20-21 COVID pandemic school years.
Urban schools are showing consistent signs of recovery in English Language Proficiency scores post-pandemic, gradually approaching pre-pandemic scores.
Suburban schools are showing some growth, although at a slower rate than urban schools.
English Language Proficiency growth in rural schools has been affected the most by the COVID Pandemic. From 2022 to 2024 there was minimal change in language proficiency scores in rural areas.
Prior to the COVID Pandemic, rural, suburban, and urban schools had close median scores, with less than a difference of 7 percentage points. By 2024, there was a difference of 15 percentage points between rural and urban English Language Proficiency scores.
Teacher Allocations: NCDPI should consider adjusting the ML teacher allocations for rural school districts. The state provides position allotments to districts based on student needs, in addition to funding. As there is no mandated teacher to student ratio for Multilingual Learners, the delivery of ML services throughout the state varies greatly. In some rural counties, teachers sometimes cover 3 to 5 schools, and have a caseloads of up to 100 students. In urban counties, ML teachers typically teach at 1 to 2 schools and have caseloads of less than 60 students. Additionally, in some rural counties ML services only happen virtually.
Training: English Language Proficiency can be improved in more ways than just ML teachers. Providing classroom teachers with high impact, effective training in working with MLs and developing language proficiency will be beneficial.