Australian vs Global Salary Analysis in the AI/ML/Data Science Field¶
Author: Rory O'Connor | Date: 2025-06-27
Introduction¶
This is a fun project I undertook to carry out some exploratory data analysis over a real world dataset form Kaggle
Data Source¶
The dataset includes salaries from various countries and job roles in data science, sourced from Kaggle's open salary dataset. https://www.kaggle.com/datasets/adilshamim8/salaries-for-data-science-jobs
Project Objectives¶
To compare and contrast key characteristics of the Australian data job market with those of the global data job market, identifying trends, similarities, and differences that could inform job seekers, employers, and policymakers.
- Summarise salary distributions
- Investigate role-based, experience-based, and industry-based differences.
- Explore geographic and work arrangement patterns.
- Provide insights into how the Australian data job market aligns with or diverges from global trends.
In [ ]:
# Getting started with the required libraries
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import plotly.graph_objects as go
import plotly.io as pio
from scipy.stats import gaussian_kde
pio.renderers.default = 'notebook_connected'
In [2]:
# load the dataset
df = pd.read_csv("salaries.csv")
# Display the first few rows of the dataset
print(df.head())
work_year experience_level employment_type job_title salary \ 0 2025 SE FT Solutions Engineer 214000 1 2025 SE FT Solutions Engineer 136000 2 2025 MI FT Data Engineer 158800 3 2025 MI FT Data Engineer 139200 4 2025 EN FT Data Engineer 90000 salary_currency salary_in_usd employee_residence remote_ratio \ 0 USD 214000 US 100 1 USD 136000 US 100 2 USD 158800 AU 0 3 USD 139200 AU 0 4 USD 90000 US 0 company_location company_size 0 US M 1 US M 2 AU M 3 AU M 4 US M
Data Quality¶
Since I haven't seen this data before, firstly I'll check to see if it is complete
In [3]:
# check for missing values
missing_values = df.isnull().sum()
# Display the count of missing values in each column
print("Missing values in each column:\n", missing_values)
Missing values in each column: work_year 0 experience_level 0 employment_type 0 job_title 0 salary 0 salary_currency 0 salary_in_usd 0 employee_residence 0 remote_ratio 0 company_location 0 company_size 0 dtype: int64
Great, the data is complete, so we don't need to worry about any missing values skewing our statistical Analysis.
Next lets see what data we are actually playing with here
In [4]:
# Collect the column names
column_names = df.columns.tolist()
# Display the column names
print("Column names:", column_names)
Column names: ['work_year', 'experience_level', 'employment_type', 'job_title', 'salary', 'salary_currency', 'salary_in_usd', 'employee_residence', 'remote_ratio', 'company_location', 'company_size']
Focus: Job Title¶
To start with, I'll do a little dive into the job titles that have the most entries, as these will paint a more statistically robust picture around patterns
In [5]:
# Lets aggreate the job_title and count the number of jobs for each title
job_counts = df['job_title'].value_counts()
# How many jobs are in this dataset?
print("Number of unique job titles:", len(job_counts))
Number of unique job titles: 390
In [6]:
# lets see the top 10 job titles
top_10_jobs = job_counts.head(10)
# Display the top 10 job titles
print("Top 10 job titles:\n", top_10_jobs)
Top 10 job titles: job_title Data Scientist 17314 Software Engineer 15007 Data Engineer 14868 Data Analyst 12381 Engineer 9456 Machine Learning Engineer 8205 Manager 6679 Analyst 4364 Research Scientist 3202 Product Manager 2230 Name: count, dtype: int64
In [7]:
# Lets plot the top 10 job titles salary histograms
for job_title in top_10_jobs.index:
plt.figure(figsize=(10, 6))
job_data = df[df['job_title'] == job_title]
plt.hist(job_data['salary_in_usd'], bins=100, alpha=0.3, color='blue')
plt.axvline(df['salary_in_usd'].mean(), color='red', linestyle='--', label=f'Mean: ${job_data['salary_in_usd'].mean():,.0f}')
plt.axvline(df['salary_in_usd'].median(), color='black', linestyle='--', label=f'Median: ${job_data['salary_in_usd'].median():,.0f}')
plt.legend()
plt.title(f'Salary Distribution for {job_title}')
plt.xlabel('Salary (USD)')
plt.ylabel('Frequency')
plt.grid(True)
plt.show()
Plotly¶
Ok there's clearly some trends. It would be useful to see the density curves associated with these histograms, one on top of the other, for a more interesting comparison. Lets use the Plotly libratry moving forward for nice interactive plots
In [19]:
fig = go.Figure()
for job_title in top_10_jobs.index:
job_data = df[df['job_title'] == job_title]['salary_in_usd']
density = gaussian_kde(job_data)
x = np.linspace(job_data.min(), job_data.max(), 1000)
fig.add_trace(go.Scatter(x=x, y=density(x), mode='lines', name=job_title))
fig.update_layout(
title='Salary Density Curves for Top 10 Job Titles',
xaxis_title='Salary in USD',
yaxis_title='Density',
template='plotly_white'
)
fig.show()
In [8]:
# Now lets look at the country codes used in the dataset
country_counts = df['employee_residence'].value_counts()
# List the unique country codes
print("Unique country codes:", country_counts.index.tolist())
Unique country codes: ['US', 'CA', 'GB', 'AU', 'NL', 'DE', 'FR', 'LT', 'AT', 'ES', 'IN', 'SK', 'PL', 'IE', 'LV', 'MX', 'BR', 'AR', 'IT', 'PH', 'ZA', 'NZ', 'PT', 'FI', 'CH', 'EG', 'CO', 'SG', 'JP', 'BE', 'GR', 'TR', 'CL', 'UA', 'EE', 'TW', 'MT', 'CY', 'IL', 'NG', 'KR', 'CZ', 'LB', 'SE', 'RO', 'PR', 'AM', 'JM', 'PE', 'HU', 'SI', 'NO', 'DK', 'MK', 'HR', 'KE', 'VN', 'CR', 'PK', 'EC', 'HK', 'ID', 'BG', 'RU', 'AE', 'LU', 'RS', 'SV', 'GH', 'DO', 'TH', 'MY', 'SA', 'UZ', 'HN', 'DZ', 'LS', 'PA', 'BO', 'XK', 'JO', 'CD', 'GE', 'VE', 'BA', 'MD', 'CF', 'TN', 'ZM', 'MU', 'RW', 'UG', 'OM', 'BM', 'AD', 'QA', 'IR', 'KW', 'AS', 'CN', 'IQ', 'JE']
Australian Comparisons¶
Shifting the focus, lets see how the data generally trends between Australia and the rest of the world
In [ ]:
# lets do another histogram like before, but looking at Australian (AU) salaries against the rest of the world, normalized by the maximum bin frequency
# load in the data
au_data = df[df['employee_residence'] == 'AU']['salary_in_usd']
other_data = df[df['employee_residence'] != 'AU']['salary_in_usd']
# Calculate means and medians
au_mean = au_data.mean()
au_median = au_data.median()
other_mean = other_data.mean()
other_median = other_data.median()
fig = go.Figure()
# AU Histogram
fig.add_trace(go.Histogram(
x=au_data,
name='AU',
opacity=0.5,
nbinsx=100,
histnorm='probability density'
))
# Global Histogram
fig.add_trace(go.Histogram(
x=other_data,
name='Other Countries',
opacity=0.5,
nbinsx=100,
histnorm='probability density'
))
# Add mean + median lines
for value, color, dashtype, label in [
(au_mean, 'green', 'dot', 'AU Mean'),
(au_median, 'blue', 'dash','AU Median'),
(other_mean, 'red', 'dot', 'Global Mean'),
(other_median, 'black','dash', 'Global Median')
]:
fig.add_shape(
type='line',
x0=value, x1=value,
y0=0, y1=1, # y1 gets auto-scaled later
xref='x', yref='paper',
line=dict(color=color, width=2, dash=dashtype),
name=label
)
# Add annotations if you want labels on the lines
fig.add_annotation(
x=value,
y=1.02, # Just above the plot
yref='paper',
text=label,
showarrow=False,
font=dict(color=color)
)
# Overlay histograms
fig.update_layout(
barmode='overlay',
title='Salary Distribution: AU vs Other Countries',
xaxis_title='Salary in USD',
yaxis_title='Frequencty Density',
template='plotly_white'
)
fig.show()