With the rise of ChatGPT, a fresh wave of concerns about the automation of a huge number of professions has begun again. The previous wave, which was triggered by Oxford research suggesting around 40% of jobs could be automated within a decade, had largely died down after actual evidence failed to support the claim.

While it remains to be seen just how likely jobs are going to be affected by technologies like ChatGPT, the disruptive force of having to change our occupation remains significant. Research from the London School of Economics explores how our careers are impacted when technology encroaches on the demand for our skills.

The researchers examined the way technology affected the employment and career earnings of Swedish workers. They find that the potential losses are relatively modest, with a 2-5 percent decline in earnings and a 1-2 percent decline in employment, but that these losses are much greater for low earners.

Assessing the impact

The researchers analyzed the impact of declining demand for occupations by looking at the way in which workers who don’t experience such a decline fare. They found that these workers tend to gain both directly, through the increased demand for their skills, and indirectly, via the rising demand caused by the general rise in incomes.

They hypothesized that technological change tends to benefit the average worker, which appeared to be the case in Sweden, where the average income rose considerably. Indeed, even those whose incomes suffered as a result of technology encroaching on their livelihood only saw a decline of between 2-5% between 1985 and 2013.

They judged an occupation to have gone into decline if total employment shrank by at least 25% since the mid-1980s. Workers in such professions lost around 5% of their income on average, with a 2% lower chance of employment. Understandably, there was also a significantly lower chance of remaining in that line of work in 2013.

Mobility to the rescue

This suggests that there was a reasonable degree of mobility between livelihoods, with people able to offset the decline in their original income by moving into new fields that were less affected by technology.

This perhaps goes some way toward explaining why those in the lowest earning percentiles were most heavily affected by the introduction of technology. Indeed, they typically saw a decline in their career incomes of between 8-11%. While they were even less likely to remain in their original occupation than higher earners, they were also less effective at finding new forms of work, or at least new forms of work that paid a comparable rate.

This was reflected in the time spent in unemployment during the study period, with unemployment typically contributing to around a third of all lost employment (versus under 10% for retraining). For those toward the end of their careers, there was also a greater likelihood of taking early retirement.

Looking to the future

While the study looked at a generation (or two) of technology before the current wave of generative AI-based tools, the researchers nonetheless believe that their findings are illustrative as we look to help people prepare for the next wave of technological impact on the labor market.

They draw comfort from the relatively low decline in both earnings and employment after the automation of occupations, although this could be because the technologies were impacting a more narrow range of occupations, thus allowing people to transfer into adjacent fields with a degree of ease.

If the boosters of generative AI are to be believed, the impact they will have on the labor market will be much broader and more wide-reaching, with many more occupations thus affected. This may provide fewer safe havens for people to migrate into, and therefore the 8-11% decline in earnings seen by the lower earners in the study might become more reflective across the board.

Gradual decline

Another positive aspect of the research is that occupational decline is a relatively gradual affair, which contributes to its more benign impact when compared to mass layoffs from economic decline. This allowed people to move into other disciplines, retrain, or enter early retirement, and thus mitigate any financial shocks in a way that’s harder to do during mass layoffs.

While it’s possible that the adoption of AI-based tools will be much faster, this would be a break from the adoption curve of pretty much every major technology in human history, so it seems unlikely. To counter this, however, if the more pessimistic projections are true and a much wider range of professions are impacted, it will be harder to transfer from occupation to occupation as human capital investments will be more significant.

Whatever the pace, or indeed nature, of disruption, the cure seems to revolve around helping people to better adapt to that change and pivot into new occupations as smoothly as possible. The data suggests that we’ve been reasonable at doing that in the past, but it’s certainly no time to rest on our laurels.