AI exposure risk equal to opportunity
A report by the European Central Bank (ECB) has stated that projections of AI ending human labour “may be greatly exaggerated”, and that occupations “potentially more exposed” to AI-enabled technologies increased their employment share in Europe during the 2010s.
“Waves of innovation have usually been accompanied by anxiety about the future of jobs,” the report reads. “This apprehension persists, even though history suggests that previous fears about labour becoming redundant were exaggerated.” However, in its surveying of various reports and a sample of 16 European countries, the ECB finds “a positive association between AI-enabled automation and changes in employment shares” regardless of which proxy the report authors used.
Using the work of Stanford University economist Michael Webb, the ECB states that moving 25 centiles up along the distribution of exposure to AI is associated with a 2.6 per cent increase in sector-occupation employment share; the work of Edward Felten, Manav Raj, and Robert Seamans estimates the increase in sector-occupation employment share to be 4.3 per cent.
On the possibility of technology-enable automation affecting relative shares of employment across skill levels and thus impacting earnings inequality, the ECB states that “literature on job polarisation shows that medium-skilled workers in routine-intensive jobs tended to be replaced by computerisation… in contrast, it is often argued that AI-enabled automation is more likely to complement or replace jobs in occupations that employ high-skilled labour”.
“The degree of exposure is as much an opportunity as it is a risk.”
The literature also suggests that AI exposure “does not seem to shake things up significantly” in occupations where average educational attainment is in the low and medium-skill groups. In contrast, a “positive and significant association” is found for the high-skill group: with the distribution of exposure to AI increased by 25 centiles, the boost to sector-occupation employment is estimated at 3.1 per cent by Webb and 6.7 per cent by Felten, Raj, and Seamans.
By age group, AI-enabled automation favours those occupations that employ younger workers, who were found to benefit at double the rate of their older colleagues (when age groups are split into younger, core, and older) regardless of the indicators used.
In their own research, the authors of the ECB report – Stefania Albanesi, António Dias da Silva, Juan Francisco Jimeno, Ana Lamo, and Alena Wabitsch – study the link between AI-enabled technologies and employment shares in 16 European countries between 2011 and 2019, years that “saw the rise of deep learning applications such as language processing, image recognition, algorithm-based recommendations, or fraud detection”. The authors find that around 25 per cent of all jobs in these countries were in occupations highly exposed to AI-enabled automation, “specifically in the upper third of the exposure measure”.
“The degree of exposure is as much an opportunity as it is a risk,” the report states, with the outcome dependent on whether such technologies will substitute or complement labour. Occupations more exposed to AI were found to employ a larger proportion of high-skilled workers when compared to occupations more exposed to advances in software, a fact that “supports the case that AI-enable technologies could be in competition with high-skilled jobs”. With the exposure to technology varying across levels of skills, it is found to be “relatively uniform” across age groups.
These findings mean that AI-enabled automation is “thus associated with employment increases in Europe – mostly for high-skill occupations and younger workers”, a fact which is “at odds with the evidence from previous technology waves, when computerisation decreased the relative share of employment of medium-skilled workers, resulting in polarisation”. The authors state that they found no evidence for this polarisation pattern in their own work, “even when examining the impact of software-enabled automation”, and that the relationship between software exposure and employment changes is null for their pooled sample, with no evidence of software replacing “routine medium-skilled jobs”.
Despite these results for employment shares, the authors also found that neither AI nor software exposure had statistically significant effects on wages, except in the work of Felten, Raj, and Seamans, which “indicates that occupations more exposed to AI have slightly worse wage growth”. The results also show “a mixed picture” across the 16 European countries studied, with the overall positive impact found in the overall research said to “hold true for most countries with only a few exceptions”.
The scale of the impact is, however, said to vary substantially across the country, which “might reflect differences in underlying economic factors” such as technology diffusion, education, and levels of regulation and competition in product and labour markets.
Despite the common fear that technological development brings with it pressure on the labour market and the widespread possibility of jobs being automated into obsolescence, the authors conclude that, during the deep learning boom of the 2010s at least, “occupations potentially more exposed to AI-enabled technologies actually increased their employment share in Europe”, with occupations with a higher proportion of younger and skilled workers gaining the most and neutral to slightly negative impacts on wages.
“These results do not amount to an acquittal: AI-enabled technologies continue to be developed and adopted,” the authors state. “Most of their impact on employment and wages – and therefore on growth and equality – has yet to be seen.”
