Essays on Automation and Future of Skills

The adoption of automation in the labour market introduces a complex array of effects
on employment, skill demand, and human capital investment. This thesis encompasses
three distinct but interrelated studies exploring the multifaceted impacts of technological
advancement and automation on the labour market, focusing on different types of au-
tomation and how it affects human capital investment and the demand for skills.
Chapter 1 delves into the differentiated impact of various forms of automation on the
human capital investment. Utilising data from the international survey of adult skills
(PIAAC) across 22 European countries, the study categorises automation based on tasks
and technology (e.g., software, robots, AI) and explores how these factors influence work-
ers’ decisions to invest in training. The findings reveal a nuanced landscape where the
effect of automation on training varies significantly depending on the type of technology
and the level of a country’s technological readiness. For example, while AI is associated
with an increase in the investment in human capital, the reverse is true for other automa-
tion technologies.

Chapters 2 and 3 exploit online job advertisements data in the UK to measure the changes
in the employers’ demand for skills. One of the key challenges when using this type data
is to correctly identify occupation from the advertisement text. In Chapter 2, I address
this issue by proposing a novel methodology for classifying occupations using advanced
language processing models. To implement it, I used web-scraping to collect data on
job advertisements in the UK. Incorporating job titles and descriptions into the classi-
fication process, in particular skill requirements, significantly improves the accuracy of
occupational classification from job advertisements, opening new areas of labour market
research based on job ads.

Building on the previous chapter, chapter 3 uses online vacancy data to examine the
changes in the demand for specific skills and the associated wage premiums within and
across occupations, especially during and after the COVID-19 pandemic. The analysis
is based on the universe of job advertisements in UK, aggregated by online job plat-
form Adzuna. This study uses a novel text classification method I developed by using a
Large Language Model (LLM), specifically the GPT-4 API, to categorise job postings into
skill groups. It finds that workers with ICT and AI skills command significantly higher
wages compared to those with interpersonal skills. Interestingly, the study suggests that
COVID-19 has had no long-term impact on either the demand for specific skills within
an occupation or the wages offered for those skills.

Collectively, these studies offer critical insights into the evolving dynamics of the labour
market in the face of technological change, emphasising the importance of adaptability
in skill development and the potential that the advanced data analysis techniques offer
for informing policy and practice.