Skip to main content


Can we Expect AI to Make up for the Labour Force Shortages Caused by an Ageing Population?

While technological advancement can be a major driving force behind economic growth, it also has the potential to cause labour displacement. This fear of ‘technological unemployment’ is not new; many historical figures such as David Ricardo and John Maynard Keynes voiced concerns about technology replacing human labour. These concerns persist today, as shown by the discussions surrounding artificial intelligence (AI). The World Economic Forum (WEF) estimates that 85 million jobs are at risk due to AI.

At the same time, the world population is ageing. According to the World Health Organisation, the ratio of the older adult population to the working age population is increasing and this trend is expected to continue in the near future leading to a reduction in the labour force of many countries. Given this phenomenon and the potential loss of jobs due to automation, a question naturally arises: can we expect automation to make up for the loss of workers in countries affected by population ageing? This blog presents the findings from my research project undertaken to answer that question.

To do so, the project examines the relationship between automation and age distribution within occupations defined in the 2018 Standard Occupational Classification system in the United States. The probability of computerisation values calculated in a paper by Frey and Osborne (Professor Michael Osborne was also my supervisor for this project) are used as a measure of ‘automation’. On the other hand, the proportion of older adults within occupations, provided by the US Bureau of Labor Statistics, is used to quantify the ‘age’ of each occupation. These two datasets are merged and processed to produce a joint standardised dataset covering the years 2011 to 2021.

We found that the construction and extraction occupations experienced the highest percentage increase of proportion of workers aged at least 65 years old at a huge 96.6% from 2011 to 2021. The personal care and service occupations experienced the lowest percentage increase at 13.6%. For workers aged 55 years and above, the farming, fishing, and forestry occupations experienced the highest increase of 46.9%, while the life, physical, and social science occupations actually had a decrease of 3.7%. In general, there was a trend of ageing across most occupations, which was expected.

In addition, we discovered a possible inverse relationship between the proportion of workers aged 55 years and older within an occupation and the probability of computerisation of that occupation. We say possible because we cannot offer guarantees of this trend’s statistical significance without further tests and research. But assuming this relationship is true, it suggests that ‘older’ occupations tend to be less likely to be automated. This seems logical for occupations such as management; Frey and Osborne noted that management occupations are less likely to be at risk of computerisation due to the high degree of social intelligence required for them, and people in management positions would tend to be older since such occupations would be biased towards people with more work experience. There is also evidence to suggest that high skilled workers, who would generally be less at risk of having their jobs automated, tend to retire later than low skilled workers. As aforementioned, we have not confirmed the validity of this trend. The reader is more than welcome to use the joint dataset to further investigate this trend or to find other statistically significant trends.

Unfortunately, we failed to find any other trends. Hence, we tentatively state that we have found no significant relationships between age distribution within occupations and their automatability for the most part. Therefore, we conclude that governments should not expect the loss of jobs due to automation to automatically balance out the shortages in the labour force due to an ageing population. Hence, we recommend that they take more active steps towards addressing these issues. Indeed, job losses are likely to accelerate in the near future due to advances in the technology that produced AI models such as ChatGPT and MidJourney. Goldman Sachs predicts that 300 million jobs around the world are at risk of automation due to generative AI. As a result, many people would need to be retrained and reskilled. A McKinsey report estimates that 375 million people may need to switch jobs by 2030, and this will just be due to automation and AI without taking into account the effects of population ageing.

Nevertheless, there is cause for optimism. While AI will result in job losses, it is also expected to create new jobs according to the WEF and PwC. According to studies by Goldman Sachs, PwC, and IBM, this rapidly advancing technology is also predicted to contribute significantly to economic growth by increasing productivity. Governments around the world have the unique opportunity to strategically leverage the economic benefits derived from AI to support their countries’ ageing populations while simultaneously investing in the reskilling of older workers. Such a two-pronged approach could not only mitigate the consequences of both an ageing population and large-scale automation of jobs, but could also create inclusive and forward-looking societies that embrace technological advancements.

About the Author

Terence Tan is currently an Engineering Science DPhil student at Wadham College, Oxford. 

Under the supervision of Professor Michael Osborne, his fourth-year undergraduate project (4YP) investigated the relationship between the age distributions within occupations in the US and the automatability of these occupations. 

Opinions of the blogger is their own and not endorsed by the Institute

Comments Welcome: We welcome your comments on this or any of the Institute's blog posts. Please feel free to email comments to be posted on your behalf to or use the Disqus facility linked below.