Aims of project

Unpaid work is often unmeasured and undervalued. It is not included in gross domestic product (GDP) and generally only measured in Time Use Surveys (TUS) which are not standardised nor regularly recorded in all countries (Charmes, 2021). Consequently, the impact of unpaid work on lifestyle, wellbeing, and economic opportunities often goes unrecognised (ILO, 2009).

With ageing populations, the rise of artificial intelligence, and pervasive economic and gender inequality, valuing and analysing unpaid work is more relevant than ever. This project aims to quantify and visualise how unpaid care and domestic work affect time use, daily life, and economic opportunities, and analyse how these impacts vary for men and women over time.

Data and Visualisations

My project visualisations are reproducible and accessible because my data is on GitHub, my sources are referenced in subtitles, and my Colab data cleaning notebooks with comments are linked in-text. For automation, I used tabula, pandas, functions, and loops to scrape data in Colab. For consistency and compatibility in Vega-Lite, I converted all data into long-form.

This visualisation gives an overview of average time spent on work: globally women do approximately three times more unpaid work than men.

To examine whether this is true for all countries or only as a global aggregate, I cleaned data in Colab, selecting values from the most recent TUS, revealing women do more unpaid work than men in every country.

Unpaid work predictions scraped from PDF data in Colab demonstrate large but decreasing gender gaps in all regions 2015-2050.

To examine the scale of unpaid work in daily life, I converted ONS TUS data into long-format and scraped the ATUS data from a PDF in Colab - I aligned the activities and converted AM/PM into 24-hour time.

Visualisation 6

The first bubble chart uses data scraped from the OECD API and IMF data in Colab. Combining this with OECD labour force data in Colab, I then analysed the relationship between the female-to-male ratio of unpaid work and female labour force participation, and performed linear regression analysis using scikit machine learning.

Scraped from a PDF in Colab, ILO calculations of unpaid work’s value demonstrate further gender imbalances.

I considered the future of unpaid work and its potential automation through AI, with PDF data scraped in Colab.

Challenges in Data Cleaning and Analysis

As much of my data is from reports and articles, I had to scrape multiple PDFs. This limited automation via loops, because tables and rows would often differ considerably, requiring manual cleaning. Similarly, I used manual country mappings as the ISO3166 library did not convert all countries to ISO3 codes.

To emphasise unpaid work in visualisation 4, I carefully chose the colours, order of activities, and interactive legend. I could only find time-of-day data for the US and UK, as most published TUS data is aggregated, and unfortunately there is no time-of-day data disaggregated by gender which limited my gender analysis.

For the Unpaid Work and Living Standards bubble chart, I wanted to add a gender parity line. However this caused issues with the interactive scaling, so I set the initial chart axes to show the same domains to emphasise women do more unpaid work than men in every country, while interactivity still allows detailed exploration of the datapoints. Additionally, I was unsuccessful in retrieving OECD API data in JSON format, so I retrieved it in CSV format using the headers dictionary.

The greatest challenge was the lack of standardised TUS and unpaid work valuation data. Nevertheless, my project’s data and visualisations explore the time spent on, value of, and gender inequality in unpaid work estimates, which prompt important conclusions.

Conclusions

My project quantifies that on average, in every country, women spend more time on unpaid domestic and care work than men. Most unpaid and paid work in the UK and US occur during the same hours (09:00-18:00), suggesting a potential trade-off between the activities. This supports the regression result that the burden of unpaid work on women reduces their labour market participation, causing reduced economic opportunities for women. To combat this, policies should be implemented to accelerate progress towards gender equality in paid and unpaid work.

Despite unpaid work constituting a considerable portion of daily life (4h25m for women and 1h23m for men on average per day, visualisation 1), and despite the large estimated values of unpaid work globally (e.g. 41.3% of GDP in Australia and 22.0% in the UK, visualisation 7), unpaid work is not measured or analysed proportionately to paid work. Researching the motivations and socioeconomic factors behind unpaid work could improve policy approaches to increase gender equality and living standards.

Finally, high automation scores of several unpaid domestic activities indicate AI could create another unpaid work revolution, building on the achievements of electrical domestic appliances. Further automation could reduce women’s time spent on unpaid work - this could be an important research field and a profitable market.

Word count excluding titles, headings and visualisations: 800 words (Aims 111, Data 259, Challenges 220, Conclusions 210).

Note: For the purpose of this project, I use ‘gender’ and ‘sex’ interchangeably as binary factors referring to men and women or males and females. In reality, there is a spectrum of intersex and gender identities, but these are not yet reflected in many data sources (NIH, 2023). Considering this topic through sociologic and economic perspectives supported by data inclusive of gender identities could reveal more about how societal and gender expectations affect unpaid and paid work participation and equality.