Research question

This study sets out to answer the question: In 2024/25, was obesity among Year 6 children more common in English council areas with higher rates of child poverty?

Our hypothesis was that council areas with higher child poverty rates would also have higher Year 6 obesity rates.

The null hypothesis was that there was no link between child poverty rates and Year 6 obesity rates across council areas.

We conducted a literature review, identified suitable datasets, explored these datasets including with visualsations, and then performed statistical analysis to try to answer the research question.

We then evaluate our findings, including limitations, possible impact, and areas for further research.

Literature review

Late 1990s and early 2000s first identified the childhood obesity disparity. (Stamatakis et al., 2005)​

2010s severe obesity in deprived areas (Ells et al., 2015)​

COVID-19 effect on obesity (NCMP, 2021)​

Potential reasons why: ​

Food access costs (Eskandari et al., 2022)​

Physical inactivity (Cheung et al., 2022)​

Stress and parental perceptions of weight (Broadbent et al., 2024)​

Datasets

We used two sets of official statistics to compare child poverty and Year 6 obesity across English council areas in 2024/25.

Year 6 obesity

This dataset shows the percentage of children aged 10–11 who were classified as living with obesity, including severe obesity. The figures come from height and weight measurements taken in mainstream state-funded schools as part of the National Child Measurement Programme.

The programme uses standard measurement procedures, which helps make the figures comparable across councils. Participation can vary, and where fewer eligible children are measured, the figure may be less representative of the area.

We downloaded the 2024/25 CSV for Year 6 prevalence of obesity (including severe obesity) from the Fingertips Compare areas view on 9th November 2026. The original CSV, Year_6_prevalence_of_obesity_compare_areas_2024_2025.csv, is saved in the same folder as index.Rmd. View the source data from OHID Fingertips (indicator 90323) View the programme’s methods and quality notes

Child poverty

This dataset estimates the percentage of children under 16 living in relative low-income families in the financial year ending March 2025. Relative low income means income below 60% of the UK median, after allowing for family size and before deducting housing costs.

The Department for Work and Pensions uses benefits and earnings records to produce local estimates using a common method. This makes it possible to compare council areas. The estimates rely in part on benefit claims, so some low-income families may be missed. Housing costs also vary between areas, and are not considered in this measure.

We downloaded the 2024/25 CSV for children in relative poverty before housing costs from the ONS Explore Local Statistics Compare areas view on 9 November 2026. The data come from the Department for Work and Pensions and cover the financial year ending March 2025. The original CSV, children-in-relative-poverty-before-housing_compare_areas_2024_2025.csv, is saved in the same folder as index.Rmd.

View the source data from ONS Explore Local Statistics View the DWP methods and quality notes

How we used the two datasets

We matched the two datasets using official codes for English council areas. We checked for duplicate codes and missing values.

The obesity figures cover the 2024/25 academic year. The poverty figures cover the financial year ending March 2025. These periods overlap, but they start and end at different times.

The poverty figures cover all children under 16, while the obesity figures cover Year 6 children aged 10–11.

So the comparison is of overall child poverty in each council area in the financial year, with obesity in its Year 6 population in the academic year.

Read and inspect the CSV files

obesity_file <- "Year_6_prevalence_of_obesity_compare_areas_2024_2025.csv"
poverty_file <- "children-in-relative-poverty-before-housing_compare_areas_2024_2025.csv"

missing_files <- c(obesity_file, poverty_file)
missing_files <- missing_files[!file.exists(missing_files)]
if (length(missing_files) > 0) {
  stop("Put these CSV files in the same folder as index.Rmd: ",
       paste(missing_files, collapse = ", "))
}

obesity_raw <- read.csv(obesity_file, fileEncoding = "UTF-8-BOM",
                       na.strings = c("", "NA"), stringsAsFactors = FALSE)
poverty_raw <- read.csv(poverty_file, fileEncoding = "UTF-8-BOM",
                       na.strings = c("", "NA"), stringsAsFactors = FALSE)

data.frame(file = c(obesity_file, poverty_file),
           rows = c(nrow(obesity_raw), nrow(poverty_raw)),
           columns = c(ncol(obesity_raw), ncol(poverty_raw)))
##                                                                      file rows
## 1                Year_6_prevalence_of_obesity_compare_areas_2024_2025.csv  297
## 2 children-in-relative-poverty-before-housing_compare_areas_2024_2025.csv  364
##   columns
## 1      27
## 2       4
names(obesity_raw)
##  [1] "Indicator.ID"                            
##  [2] "Indicator.Name"                          
##  [3] "Parent.Code"                             
##  [4] "Parent.Name"                             
##  [5] "Area.Code"                               
##  [6] "AreaName"                                
##  [7] "Area.Type"                               
##  [8] "Sex"                                     
##  [9] "Age"                                     
## [10] "Category.Type"                           
## [11] "Category"                                
## [12] "Time.period"                             
## [13] "Value"                                   
## [14] "Lower.CI.95.0.limit"                     
## [15] "Upper.CI.95.0.limit"                     
## [16] "Lower.CI.99.8.limit"                     
## [17] "Upper.CI.99.8.limit"                     
## [18] "Count"                                   
## [19] "Denominator"                             
## [20] "Value.note"                              
## [21] "Recent.Trend"                            
## [22] "Compared.to.England.value.or.percentiles"
## [23] "Compared.to.parent.value.or.percentiles" 
## [24] "Time.period.Sortable"                    
## [25] "New.data"                                
## [26] "Compared.to.goal"                        
## [27] "Time.period.range"
names(poverty_raw)
## [1] "areacd" "areanm" "period" "value"
head(obesity_raw[c("Area.Code", "AreaName", "Area.Type", "Time.period", "Value")])
##   Area.Code      AreaName                                Area.Type Time.period
## 1 E92000001       England                                  England     2024/25
## 2 E08000028      Sandwell Lower tier local authorities (post 4/23)     2024/25
## 3 E08000011      Knowsley Lower tier local authorities (post 4/23)     2024/25
## 4 E08000031 Wolverhampton Lower tier local authorities (post 4/23)     2024/25
## 5 E09000028     Southwark Lower tier local authorities (post 4/23)     2024/25
## 6 E07000136        Boston Lower tier local authorities (post 4/23)     2024/25
##      Value
## 1 22.24756
## 2 30.67690
## 3 30.05620
## 4 29.98620
## 5 29.67250
## 6 29.09090
head(poverty_raw)
##      areacd               areanm         period    value
## 1 E06000001           Hartlepool 2024-04-01/P1Y 27.18399
## 2 E06000002        Middlesbrough 2024-04-01/P1Y 34.35237
## 3 E06000003 Redcar and Cleveland 2024-04-01/P1Y 26.23290
## 4 E06000004     Stockton-on-Tees 2024-04-01/P1Y 22.25529
## 5 E06000005           Darlington 2024-04-01/P1Y 24.79042
## 6 E06000006               Halton 2024-04-01/P1Y 23.03207
table(obesity_raw$Area.Type, useNA = "ifany")
## 
##                                  England 
##                                        1 
## Lower tier local authorities (post 4/23) 
##                                      296
table(obesity_raw$Time.period, useNA = "ifany")
## 
## 2024/25 
##     297
table(poverty_raw$period, useNA = "ifany")
## 
## 2024-04-01/P1Y 
##            364

Filter and inspect obesity prevalence before joining

The raw obesity dataset is filtered for year 2024/2025, age 10-11, and area type (lower tier local authorities).

The filtered obesity dataset contains 296 observations of 3 variables.

We can see that 2 local authorities do not have obesity data: City of London and Isles of Scilly. These are excluded from the analysis.

obesity <- subset(
  obesity_raw,
  Indicator.ID == 90323 & Time.period == "2024/25" &
    Sex == "Persons" & Age == "10-11 yrs" &
    Area.Type == "Lower tier local authorities (post 4/23)",
  select = c(Area.Code, AreaName, Value)
)
names(obesity) <- c("authority_code", "authority", "obesity_pct")
obesity <- obesity[order(obesity$authority_code), ]
obesity$obesity_pct <- as.numeric(obesity$obesity_pct)

summary(obesity$obesity_pct)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.     NAs 
##   9.701  18.275  21.491  21.249  24.453  30.677       2
obesity[is.na(obesity$obesity_pct), ]
##     authority_code       authority obesity_pct
## 297      E06000053 Isles of Scilly          NA
## 296      E09000001  City of London          NA

An initial plot lets us see the spread of obesity rates. The horizontal position is simply the order of the local authority codes. We can see that there is considerable variation in obesity rates between areas, from around 10% to around 30%.

point_colour <- adjustcolor("#42586A", alpha.f = 0.55)
plot(seq_len(nrow(obesity)), obesity$obesity_pct,
     xlab = "Council order by area code", ylab = "Year 6 obesity prevalence (%)",
     pch = 16, col = point_colour)

We can also inspect the data using a choropleth map. This draws each council area using GeoJSON data from the ONS May 2023 local authority district boundaries page. This csv file has been downloaded and is saved in the working directory as lad_may_2023.geojson.

library(sf)
## Linking to GEOS 3.13.0, GDAL 3.8.5, PROJ 9.5.1; sf_use_s2() is TRUE
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
boundaries <- st_read("lad_may_2023.geojson", quiet = TRUE)
england <- filter(boundaries, startsWith(LAD23CD, "E"))

obesity_map <- left_join(
  england, obesity,
  by = c("LAD23CD" = "authority_code")
)

ggplot(obesity_map) +
  geom_sf(aes(fill = obesity_pct), colour = "white", linewidth = 0.1) +
  scale_fill_gradient(
    low = "#DEEBF7",
    high = "#08306B",
    limits = c(5, 35),
    breaks = seq(5, 35, by = 5),
    labels = function(x) paste0(x, "%"),
    oob = scales::squish,
    na.value = "grey85",
    name = "Obesity prevalence"
  ) +
  labs(
    title = "Obesity prevalence in England, school year 6",
    subtitle = "Academic year 2024/25",
    caption = "Sources: OHID Fingertips and ONS boundaries."
  ) +
  theme_void()

Filter the poverty data to England

The raw poverty dataset covers the whole of the UK. The first letter E identifies English area codes. We select the financial year starting 1 April 2024. The join below then selects the codes that appear in the lower-tier obesity table.

table(substr(poverty_raw$areacd, 1, 1), useNA = "ifany")
## 
##   E   K   N   S   W 
## 298   1  11  32  22
poverty_england <- subset(
  poverty_raw, startsWith(areacd, "E") & period == "2024-04-01/P1Y",
  select = c(areacd, areanm, value)
)
names(poverty_england) <- c("authority_code", "poverty_area", "poverty_pct")
poverty_england$poverty_pct <- as.numeric(poverty_england$poverty_pct)

nrow(poverty_england)
## [1] 298
head(poverty_england)
##   authority_code         poverty_area poverty_pct
## 1      E06000001           Hartlepool    27.18399
## 2      E06000002        Middlesbrough    34.35237
## 3      E06000003 Redcar and Cleveland    26.23290
## 4      E06000004     Stockton-on-Tees    22.25529
## 5      E06000005           Darlington    24.79042
## 6      E06000006               Halton    23.03207
summary(poverty_england$poverty_pct)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   5.175  12.609  16.696  18.046  22.286  40.553

Again, an initial plot lets us see the spread of poverty rates. The horizontal position is again the order of the local authority codes. We can see that there is again considerable variation in child poverty rates between areas, from around 5% to around 40%.

point_colour <- adjustcolor("#42586A", alpha.f = 0.55)
plot(seq_len(nrow(poverty_england)), poverty_england$poverty_pct,
     xlab = "Council order by area code", ylab = "Child poverty prevalence (%)",
     pch = 16, col = point_colour)

Here is a similar choropleth map allowing visualisation of the spread of child poverty across local authority araes in England:

poverty_map <- left_join(
  england, poverty_england,
  by = c("LAD23CD" = "authority_code")
)

ggplot(poverty_map) +
  geom_sf(aes(fill = poverty_pct), colour = "white", linewidth = 0.1) +
  scale_fill_gradient(
    low = "#FEE5D9",
    high = "#A50F15",
    limits = c(0, 45),
    breaks = seq(0, 45, by = 5),
    labels = function(x) paste0(x, "%"),
    oob = scales::squish,
    na.value = "grey85",
    name = "Child poverty"
  ) +
  labs(
    title = "Child poverty across English councils",
    subtitle = "Children under 16 in relative poverty before housing costs, 2024/25",
    caption = "Sources are ONS Explore Local Statistics and ONS boundaries."
  ) +
  theme_void()

Join the tables and check the match

We keep every council from the obesity table. Joining on the official area code lets us see any missing matches, before excluding the rows with missing values.

stopifnot(anyDuplicated(obesity$authority_code) == 0L,
          anyDuplicated(poverty_england$authority_code) == 0L)

join_table <- merge(obesity, poverty_england,
                    by = "authority_code", all.x = TRUE)
complete <- complete.cases(join_table[c("obesity_pct", "poverty_pct")])
analysis <- join_table[complete, ]

head(join_table)
##   authority_code            authority obesity_pct         poverty_area
## 1      E06000001           Hartlepool     25.6757           Hartlepool
## 2      E06000002        Middlesbrough     25.9162        Middlesbrough
## 3      E06000003 Redcar and Cleveland     24.9123 Redcar and Cleveland
## 4      E06000004     Stockton-on-Tees     25.0000     Stockton-on-Tees
## 5      E06000005           Darlington     22.1277           Darlington
## 6      E06000006               Halton     26.6187               Halton
##   poverty_pct
## 1    27.18399
## 2    34.35237
## 3    26.23290
## 4    22.25529
## 5    24.79042
## 6    23.03207
data.frame(councils_in_join = nrow(join_table),
           matched_poverty_codes = sum(!is.na(join_table$poverty_area)),
           missing_obesity = sum(is.na(join_table$obesity_pct)),
           missing_poverty = sum(is.na(join_table$poverty_pct)),
           councils_in_analysis = nrow(analysis))
##   councils_in_join matched_poverty_codes missing_obesity missing_poverty
## 1              296                   296               2               0
##   councils_in_analysis
## 1                  294
join_table[!complete, c("authority_code", "authority",
                        "obesity_pct", "poverty_pct")]
##     authority_code       authority obesity_pct poverty_pct
## 50       E06000053 Isles of Scilly          NA    8.852459
## 264      E09000001  City of London          NA    6.882022

Exploratory analysis

The plot below shows how obesity prevalence varies across councils with different child poverty rates.

Each dot is an observation of poverty rate and obesity prevalance for a unique local authority area.

There is an initial impression of a positive relationship between obesity and poverty, as expected from the literature.

x_label <- "Children in relative poverty before housing costs (%)"
y_label <- "Year 6 obesity prevalence (%)"
plot(obesity_pct ~ poverty_pct, data = analysis,
     xlab = x_label, ylab = y_label, pch = 16, col = point_colour)

Statistical methods

Fit a straight line

Ordinary least squares fits the line that minimises the sum of the squared gaps between observed and fitted obesity rates.

The slope gives the fitted difference in obesity prevalence per one percentage point difference in child poverty.

model <- lm(obesity_pct ~ poverty_pct, data = analysis)
line_colour <- "#17686B"

plot(obesity_pct ~ poverty_pct, data = analysis,
     xlab = x_label, ylab = y_label, pch = 16, col = point_colour)
abline(model, col = line_colour, lwd = 2)

Add 95% confidence intervals

The shaded band represents a 95% confidence interval for the mean fitted obesity rate at each poverty rate. It shows uncertainty around the line.

poverty_grid <- data.frame(
  poverty_pct = seq(min(analysis$poverty_pct), max(analysis$poverty_pct),
                    length.out = 200)
)
band <- predict(model, newdata = poverty_grid, interval = "confidence",
                level = 0.95)

plot(obesity_pct ~ poverty_pct, data = analysis, type = "n",
     xlab = x_label, ylab = y_label,
     ylim = range(analysis$obesity_pct, band))
polygon(c(poverty_grid$poverty_pct, rev(poverty_grid$poverty_pct)),
        c(band[, "lwr"], rev(band[, "upr"])),
        col = adjustcolor(line_colour, alpha.f = 0.2), border = NA)
points(obesity_pct ~ poverty_pct, data = analysis, pch = 16,
       col = point_colour)
lines(poverty_grid$poverty_pct, band[, "fit"], col = line_colour, lwd = 2)

Results

pearson_r <- cor(analysis$poverty_pct, analysis$obesity_pct)
model_summary <- summary(model)
slope_10 <- 10 * coef(model)[["poverty_pct"]]
slope_ci_10 <- 10 * confint(model, "poverty_pct", level = 0.95)
format_2 <- function(x) sprintf("%.2f", x)

results <- data.frame(
  Measure = c("Councils", "Pearson correlation",
              "Slope per 10 percentage points higher child poverty",
              "95% confidence interval for that slope",
              "R-squared", "Residual standard deviation",
              "Residual variance"),
  Estimate = c(nobs(model), format_2(pearson_r),
               paste(format_2(slope_10), "percentage points"),
               paste(format_2(slope_ci_10[1]), "to",
                     format_2(slope_ci_10[2]), "percentage points"),
               format_2(model_summary$r.squared),
               paste(format_2(model_summary$sigma), "percentage points"),
               paste(format_2(model_summary$sigma^2),
                     "squared percentage points"))
)
knitr::kable(results, align = c("l", "l"))
Measure Estimate
Councils 294
Pearson correlation 0.79
Slope per 10 percentage points higher child poverty 4.29 percentage points
95% confidence interval for that slope 3.91 to 4.67 percentage points
R-squared 0.63
Residual standard deviation 2.43 percentage points
Residual variance 5.93 squared percentage points

In the straight-line model, a 10 percentage point higher child poverty rate across councils was associated with a 4.29 percentage point difference in fitted Year 6 obesity prevalence (95% CI 3.91 to 4.67). The estimate is unadjusted.

Pearson’s correlation describes the direction and strength of the straight-line relationship.

R-squared is the share of variation in obesity rates across councils accounted for by the fitted line. The residual standard deviation describes how far observed rates typically sit from the line. Residual variance is that standard deviation squared.

Evaluation

Strengths:

Large dataset covering all English local authorities.​

Recent 2024/25 data for both variables.​

Pearson correlation and linear regression provide complementary statistical evidence.​

Limitations:

Local-authority-level data cannot be applied to individual children.​

Correlation does not establish causation.​

Other factors, including ethnicity, deprivation, physical activity and food access, may influence obesity rates.​

Local-authority averages may hide inequalities within areas.​

NCMP data excludes some children, including those who are homeschooled or attend independent schools.​

Key findings:

Strong positive association between child poverty and Year 6 obesity (r = 0.79).​

Higher child poverty is associated with higher Year 6 obesity rates.​

The regression model explains approximately 63% of the variation in obesity rates (R² = 0.63).​

Further research is needed to explore other contributing factors.​ ## Further research {#further-research}

References