AI as a job creator
The advances in modern AI can enhance productivity by a large amount yet may also endanger some white-collar jobs as I have discussed here. But is AI going to be a net job creator or a net job destroyer? The answer is very uncertain, but based on data from 2011 to 2019, Stefania Albanesi and her colleagues have taken a first stab at it.
They used data collected by Michael Webb from Stanford and Edward Felten et al. to check how advances in software technology and AI have impacted different jobs and what changes in employment and wages have resulted from these technologies. In their new analysis, Albanesi et al. go further than the original analyses insofar as they look at the impact on European countries on a country-by-country level as well as on workers by age and skill level. And there are some clear differences between software and AI.
In the case of software, the impact of software advances on employment and wages was on average zero. Highly skilled workers improved their employment chances a little bit while medium- and low-skilled workers lost on average. But the net effect was practically zero.
However, when looking at individual countries in Europe, we note that the net effect of software on the labour market was negative for countries like Portugal or Greece, but positive for countries like Germany or the UK.
Net effect of software advances on employment and wages in European countries
Source: Albanesi et al. (2023)
Now compare this to the impact of AI technology in robotics and other areas on employment and wages (large language models are such a new invention that there is no data to analyse their impact, yet). The vast majority of countries saw employment increase and substantially so. Wages again didn’t really change, but AI proved to be a job creator, not a job killer. And unlike with software where mostly younger, more tech-savvy workers benefitted from the advances in software technology, the increased rollout of AI benefitted younger and middle-aged workers while older workers saw no change in their employment opportunities.
What drives these differences between countries is up for debate, but it seems that a more educated workforce and a better digital infrastructure are key drivers of the beneficial effects. Countries where the digital infrastructure is more advanced and where there are more skilled workers are early adopters and the target countries for investments in AI technology. Thus, they gain most of the investments and the jobs created from these new technologies while countries that lag in these dimensions tend to be left behind. Furthermore, a more rigid and less flexible labour market can be a hindrance to job creation from AI. Countries with higher workplace protection and lower labour market flexibility saw on average lower employment gains from software technology and AI alike.
Obviously, this analysis is based on the rollout of AI technologies in manufacturing activities, so we have to be cautious about extrapolating these findings to service jobs which are more affected by large language models like ChatGPT. But it seems that the influence of AI on the labour market is much more benign than the influence of software. AI seems to create much more winners than software and other IT developments did in the past.
Net effect of AI advances on employment and wages in European countries
Source: Albanesi et al. (2023)