One might argue that, according to Bureau of Labor Statistics data, information sector jobs and ChatGPT launch are correlated.
Which is to say, information sector jobs, which had been on an upswing since 2020, reversed course about the time ChatGPT was launched.
One also might be quite tempted to suggest coding jobs were most affected. That might not be correct, though.
source: Econ Reporter
The BLS “information” sector includes:
In the graph below, reddish areas are the content industry category job losses. In blue are information sector losses. “Hollywood” seems to be the area where the most jobs were lost, but a clear majority of the attrition came in content-related industries.
source: Seeking Alpha
Correlation is not necessarily causation. To be sure, slowdown in hiring seems to have occurred. Perhaps a number of changes in the economy are correlated. Many observers would agree that “over-hiring” happened after the Covid epidemic. Some amount of correction seems to have occurred. Higher interest rates which discouraged business startups or expansions could have played a role as well.
Digital initiatives might be allowing firms to restructure their business processes in ways that reduce labor demand, as well. And demographic change might be an issue as well.
Possible cause | Timing | Industries most affected | How it could reduce employment |
Post-COVID normalization | 2022 onward | Leisure, retail, transportation, logistics, professional services | Companies that massively expanded hiring in 2020–22 no longer needed the same workforce |
Fed tightening / high interest rates | 2022 onward | Housing, construction, startups, finance, tech, business services | Higher cost of capital reduces investment, startups and expansion; companies become reluctant to hire |
Tech-sector over-hiring in 2020–22 | 2022 onward | Software, internet, media, telecom | Firms unwound extraordinary pandemic hiring regardless of AI |
Labor-market normalization | 2022 onward | Economy-wide | Job openings fell from extraordinary 2021–22 levels; hiring slowed without mass layoffs |
Temporary-help collapse | 2022 onward | Staffing, business services | Temporary workers are often the first variable labor input cut when demand softens |
Manufacturing/industrial cycle | 2022 onward | Manufacturing, durable goods | Higher rates, inventories, weak goods demand and capital-cycle changes reduce labor demand |
Retail restructuring | 2022 onward | Retail | E-commerce, productivity, store rationalization and post-pandemic normalization reduce labor requirements |
Telecom structural decline | Long-running | Telecommunications | Wireless/fiber consolidation, automation, declining legacy services and network efficiencies reduce headcount |
Demographics | Increasingly important | Economy-wide | Baby Boomers retire; fewer workers are available, changing both hiring and measured employment growth |
Immigration changes | Especially 2025–26 | Construction, hospitality, agriculture, services | Changes labor supply as well as employers' ability to fill positions |
AI/GenAI | 2023 onward | Information, software, customer service, professional services | Some work can be automated or performed by fewer employees |
Tariffs/trade/geopolitical shocks | 2025–26 especially | Manufacturing, logistics, retail | Higher input costs and uncertainty discourage hiring |
The point is that there are many potential correlations between job loss overall and trends in technology and the economy. Some of those, or the ensemble of trends, might be “causal.”
But the point is that use of language models as the “cause” of job declines is too simplistic. There are lots of other potential explanations.
No comments:
Post a Comment