Is there a junior penalty?
A working paper testing whether generative AI has cut demand for early-career software developers, using only US public data.
- Role
- Author
- When
- 2026
- Stack
software job postings, late 2022 to mid 2026
of the fall from peak came before ChatGPT
public datasets, fully reproducible
The question
Since ChatGPT’s release in November 2022, has generative AI reduced demand for junior software developers relative to seniors, in postings, employment or pay? I graduated into this market in 2026.
What the data says
- Postings collapsed. Software postings on Indeed fell 54% from late 2022 to mid 2026, the most of 41 occupational categories. But 58% of the fall from the February 2022 peak happened before ChatGPT, alongside interest-rate hikes.
- Employment did not. In BLS data, software developer and QA employment grew 8.3% from May 2022 to May 2025. Pay at the 10th percentile rose faster than at the 90th, so the wage distribution compressed.
- Juniors kept their share. In CPS microdata, the share of developers aged 22 to 25 was essentially flat, ranking 102 of 222 occupations for change.
- One source disagrees. Stack Overflow survey respondents report a junior-senior pay gap that widened 5 to 13% after 2022. It is the only source showing a pay penalty, and it is prone to selection.
Bottom line
Hiring demand for software work collapsed and tilted toward experience, but much of that predates ChatGPT. Representative data through mid-2026 shows no junior employment or wage penalty among employed developers yet.
A note on method
An early draft of this project used a synthetic postings panel. I removed it, rebuilt every result on public data, and published the replication package so each number in the paper can be regenerated.