This is a useful time to look at changes in employment growth in different parts of the country and compare these with the pre-pandemic economy, since overall job growth is currently only a touch lower than it was in 2019. US employment grew 1.2% in April from a year earlier, close to the 1.3% growth seen in 2019.
Let’s start with the reduced migration from cold states to warm states that is showing up in the employment data of the metros most affected. House prices are a useful proxy for interstate migration. The two large metros with the weakest home price growth are Dallas, Texas, and Tampa, Florida, according to the S&P CoreLogic Case-Shiller indices. Employment in Dallas grew 1.4% in March from a year earlier, roughly half its 2019 pace, while in Tampa, it grew just 0.9%, a third of its 2019 pace. Reduced outmigration from the Northeast and Midwest, on the other hand, helps explain why job growth is above the pre-pandemic pace in Buffalo, New York, and Pittsburgh, Pennsylvania.In theory, less interstate migration should also boost the fortunes of the San Francisco Bay Area, a region famous for people leaving to get away from high housing costs. Yet employment in the region has been declining since its post-pandemic peak in November 2022 — coincidentally, the month OpenAI’s ChatGPT launched.
This could be chalked up to tech companies laying off workers in 2023 and early 2024 to preserve profitability after a hiring binge in the years previous. Meta Platforms Inc.’s headcount fell by 22% on a year-over-year basis in its “year of efficiency” push in 2023. But that post-Covid culling is behind us — big tech companies are collectively investing hundreds of billions of dollars into AI, and Meta’s headcount has grown 10% over the past year. So, it’s noteworthy that employment in the country’s tech heartland continues to decline at a time when, in theory, the Bay Area has the tailwinds of massive investment and fewer people leaving the region.
The most likely culprit is the impact that AI is having on labor demand. The types of white-collar jobs shrinking the fastest in the US include categories such as software engineers, data engineers and application engineers. On recruitment site Indeed, job postings for software engineers are currently lower than they were in the spring of 2020. Some of this drop is likely due to efficiency gains from AI, while some may be due to companies needing to control headcount where they can while they pile huge sums of money into advanced chips and data centers.Boston is another knowledge job-centric metro area showing signs of labour market softness, with its unemployment rate rising from 3.5% to 4.5% over the past year, the highest level outside of the pandemic period since 2015. The education hub may be more vulnerable than some other cities to AI adoption since the technology tends to be good at the kind of work that would typically be offered to young college graduates.These are dynamics we should expect to see, according to “From San Francisco to Savannah? The Downstream Effects of Generative AI,” a paper by researchers at Louisiana State University and Massachusetts Institute of Technology. Metro areas with both high educational attainment and high costs of living are the most vulnerable to AI disruption. These metros have many of the kinds of jobs AI can replace and cuts there help reduce headcount expenses more quickly. Authors Scott Abrahams and Frank S. Levy highlight parallels with the 1980s when manufacturing-centric communities with a low share of educational attainment were hit hardest by the negative shock of automation and globalization.
This doesn’t mean that San Francisco will turn into a technological Rust Belt — the region is at the forefront of AI innovation, after all, and even if it ends up with fewer software development jobs, those could be replaced by different kinds of AI-enhanced knowledge work. Abrahams and Levy point out that the manufacturing shock benefited the southern US, not just Asian exporting countries, and it’s possible an AI shock will lead to more balanced economic development across the US. That may take time, though, particularly while interstate migration remains constrained by a dysfunctional housing market. Given our experience with manufacturing job losses and signs that another labor-market reordering is coming, it’s time policymakers at the local and national level start to take the potential for AI disruption seriously.
