The backstory

It’s a touchy subject for many. Yet this question—how much do you get paid?—is at the core of the online questionnaire that was recently administered by Aiden Oakley (@ajoakley on Twitter). And it is a question that was willingly answered by over 100 sport scientists and strength & conditioning coaches who currently work for football (soccer) clubs in the UK.

Why?

This job advert says it all…

Blackpool FC was roundly criticised for advertising such a position for a measly £12,000 per annum. The UK Strength & Conditioning Association then withdrew this job posting from their website. Regrettably, this is not a one-off. Blackpool happened to rightly draw the ire of the Twitter crowd on this occasion, but many of us working in sport have seen too many job ads just like this in our global industry.

So what is the current state of play when it comes to salaries in UK soccer?

Aiden set out to explore this question via a simple survey:

After a fortnight, 100+ responses were received. I put my hand up to help with visualising the data. This post is the first in a series where I will show how I created the plots that Aiden published on Twitter here and here, alongside some commentary about the choices I made throughout the process and my own interpretations of the results.

Getting started

Let’s load the cleaned data in and load the necessary libraries. Note: Some variables were omitted during the data cleaning process to protect the anonymity of respondents.

# Load libraries
library(readr)
library(knitr)
library(kableExtra)
library(ggplot2)
library(dplyr)

# Read the data into R
dataSource <- "https://raw.githubusercontent.com/jacquietran/soccerSalariesDataPublic/master/soccerSalariesDataPublic.csv"
soccerSalariesData <- read_delim(dataSource, delim = ",", col_names = TRUE, col_types = NULL)

Now that we have the cleaned data loaded in, we can begin to examine it. The data set has 16 variables and 102 responses. For pedants like me, I’ve also included a data dictionary below.

# Examine first few observations of the data set
head(soccerSalariesData)

## # A tibble: 6 x 16
##   responseID country leagueOfFirstTeam levelOfWork position dept        partTime
##        <dbl> <chr>   <chr>             <chr>       <chr>    <chr>       <chr>   
## 1          1 England Championship      1st Team    Coach    Sport Scie~ <NA>    
## 2          2 England Championship      1st Team    Head     S&C         <NA>    
## 3          3 England Championship      1st Team    Coach    S&C         <NA>    
## 4          4 England Championship      1st Team    <NA>     <NA>        <NA>    
## 5          5 England Championship      1st Team    Head     Sport Scie~ <NA>    
## 6          6 England Championship      Academy     Head     Sport Scie~ <NA>    
## # ... with 9 more variables: academyCategory <chr>, ageGroupAllPhases <lgl>,
## #   ageGroupPDPhase <lgl>, ageGroupYDPhase <lgl>, ageGroupFPhase <lgl>,
## #   reportedSalaryGBP <dbl>, salaryGBP <dbl>, highestQual <chr>,
## #   highestQualCategory <chr>

Salaries by league

In this post, I will attempt to answer one simple question: What are the reported salaries of performance support staff working in UK soccer clubs when categorised by league?

(It’s worth noting here that this refers to the league in which the club’s first team plays, so it may not be the league that the respondent may predominantly work within, e.g., if they support the Academy programs.)

After a quick exploration when I first started working with this data set, I noticed that salaries varied considerably between respondents. To capture this, I decided to build box plots (to represent ‘typical’ salaries within a particular category), overlaid with a scatter plot where each dot represents one respondent. I also chose to present the the x axis categories such that the leagues progress from higher tier to lower tier as your gaze moves left to right.

# Set factor levels
soccerSalariesData$leagueOfFirstTeam <- factor(
    soccerSalariesData$leagueOfFirstTeam,
    levels = c(
        "Premier League", "Scottish Premiership", "Championship", "League 1", "League 2"))

# Create object to store axis labels with line breaks
labelsLeague <- c(
    "Premier\nLeague", "Scottish\nPremiership", "Championship", "League 1", "League 2")

# Create a list to store common theme options for use across plots
plotFeatures <- theme(
    plot.title = element_text(size = 16),
    plot.subtitle = element_text(size = 14),
    axis.title.x = element_blank(),
  axis.text.x = element_text(size = 11),
    axis.title.y = element_text(size = 16),
    axis.text.y = element_text(size = 11),
    legend.position = "none")

# Build the plot
p <- ggplot(soccerSalariesData, aes(x = leagueOfFirstTeam, y = salaryGBP))
p <- p + geom_boxplot(outlier.shape = NA, coef = 0)
p <- p + geom_point(alpha = 1/4, size = 4, aes(colour = leagueOfFirstTeam))
p <- p + scale_y_continuous(
    limits = c(0, 60000),
    breaks = seq(0, 60000, by = 10000))
p <- p + scale_x_discrete(labels = labelsLeague)
p <- p + labs(
    title = "UK soccer salaries of support staff",
    subtitle = "By league that the first team plays in",
    y = "Salary (GBP)")
p <- p + plotFeatures
salaryByLeague <- p

# Display the plot
salaryByLeague

What about median reported salaries?

# Calculate median salary by league in which first team plays
medianSalariesByLeague <- soccerSalariesData %>%
    group_by(leagueOfFirstTeam) %>%
    summarise(median = median(salaryGBP, na.rm = TRUE))

# Rename variables for aesthetic purposes
colnames(medianSalariesByLeague) <- c(
    "League in which first team plays", "Median salary (GBP, £)")

# Display the table
medianSalariesByLeague %>%
    kable("html") %>%
    kable_styling()

Salaries by league and competitive level

We can drill down further, plotting the data by league and by competitive level; that is, whether respondents work for their club’s first team or academy.

# Build the plot
p <- ggplot(soccerSalariesData, aes(x = leagueOfFirstTeam, y = salaryGBP))
p <- p + facet_wrap(~levelOfWork)
p <- p + geom_boxplot(outlier.shape = NA, coef = 0)
p <- p + geom_point(alpha = 1/4, size = 4, aes(colour = leagueOfFirstTeam))
p <- p + scale_y_continuous(
    limits = c(0, 60000),
    breaks = seq(0, 60000, by = 10000))
p <- p + scale_x_discrete(labels = labelsLeague)
p <- p + labs(
    title = "UK soccer salaries of support staff",
    subtitle = "First Team vs. Academy",
    y = "Salary (GBP)")
p <- p + plotFeatures
salaryByLeagueAndCompLevel <- p

# Display the plot
salaryByLeagueAndCompLevel

Meanwhile, median salaries now look like this:

# Calculate median salaries by league and competitive level
medianSalariesByLeagueAndCompLevel <- soccerSalariesData %>%
    group_by(leagueOfFirstTeam, levelOfWork) %>%
    summarise(median = median(salaryGBP, na.rm = TRUE))

# Order by $levelOfWork, then $leagueOfFirstTeam
medianSalariesByLeagueAndCompLevel <- medianSalariesByLeagueAndCompLevel[order(
    medianSalariesByLeagueAndCompLevel$levelOfWork,
    medianSalariesByLeagueAndCompLevel$leagueOfFirstTeam), ]

# Revalue duplicates in $levelOfWork for aesthetic purposes
medianSalariesByLeagueAndCompLevel$levelOfWork <- ifelse(
    duplicated(medianSalariesByLeagueAndCompLevel$levelOfWork),
    "", medianSalariesByLeagueAndCompLevel$levelOfWork)

# Rename variables for aesthetic purposes
colnames(medianSalariesByLeagueAndCompLevel) <- c(
    "League in which first team plays", "Competitive level", "Median salary (GBP, £)")

# Display the table
medianSalariesByLeagueAndCompLevel %>%
    kable("html") %>%
    kable_styling()

What’s next?

In the second post of this series, I will delve into whether staff salaries differ when the sample is categorised into sub-groups by level of job seniority. Stay tuned!