Fed Study: Programmer Job Growth Halved Since ChatGPT
Generative AI has altered daily work for programmers more than most occupations. A Federal Reserve Board study now confirms this change appears in US employment figures. Researchers examined monthly data from a major household survey and matched it against a US Department of Labor database. That database sorts jobs by skills and requirements. This method identifies roles heavy on programming, which make up about 3.7 percent of the US workforce.
Sharp Drop in Growth Rates
Prior to ChatGPT's November 2022 release, jobs rich in programming expanded at nearly 5 percent annually. That rate beat the broader job market. Post-launch, growth slowed dramatically. In areas like IT services and software development, where programmers concentrate, expansion has stalled completely.
Critics might argue tech sector woes explain this. Rate increases, the fade of Covid online service surges, and crypto failures all struck in 2022. To account for such factors, researchers created a counterfactual path. It assumes programmers' share within industries held steady, with only industry sizes shifting. Even adjusted, programmer roles decline by roughly three percentage points yearly. Firms appear to reduce programmer proportions intentionally. Tests on AI-immune occupations show no similar trend.
Not Direct Job Losses
Over three years, this difference equals about 500,000 missing positions without large language models. Authors warn against viewing this as pure job cuts. Aspiring programmers often shift to nearby fields. AI redistributes tasks, embedding programming into other jobs. The analysis misses wider economic ripples. Strong AI productivity gains could boost overall labor demand long-term.
Wages show no evident fall. Impact hits job counts, not pay. Indeed reports software developer postings steady since 2024, with recent upticks. Earlier, postings dropped over half in 2022 and 2023.
Programmer Locations and Timing
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Roughly 40 percent of US programmers labor at IT service firms building software for clients, not big tech or startups. This contract area employs most programmers and faces the steepest slowdown. One-third of programmers handle contract development overall.
The employment gap emerged mid-2024, about 1.5 years post-ChatGPT. If AI drives this, firms took time to observe model advances before easing hires. Data unclear if actual gains or anticipated ones prompted cuts.
Other Influences and Checks
AI alone may not explain everything. A 2017 Tax Cuts and Jobs Act rule, effective 2022, spreads research expense deductions over years. Software development qualifies as research, possibly curbing hires. Studies on its effects vary, yet results persist in less-affected sectors.
Causation proves hard amid multiple forces. Researchers tested with history. For bank tellers, it spots ATMs as targeted shock despite bank growth. For seamstresses, it flags offshoring as industry-wide.
Measurement Challenges
No uniform method exists to rank AI-impacted jobs. Half of approaches agree on hardest-hit groups. Programmers stand out: over 98 percent rank highest across all. This aligns with Anthropic Economic Index data, where programming queries exceed one-third of Claude chatbot use.
Future unclear if trends reverse via cheaper coding opening markets, offshore hits, or AI shifting elsewhere. Authors call their work a first step. A Carnegie Mellon and Stanford study notes AI agent efforts center on programming, ignoring bigger economic zones.
The Federal Reserve Board, the US central bank overseeing monetary policy, conducted this analysis using public labor data sources.
Related on Neura Market