Adopting GitHub Copilot is linked to modestly higher software-engineer hiring—especially junior hires—and to new hires with slightly more non-coding skills, without evidence of reduced coding ability.
Firms' GitHub Copilot adoption and labor market outcomes for software engineers
Firms that adopt GitHub Copilot are associated with a 3–5% higher monthly probability of hiring software engineers—driven by entry-level hires—whose LinkedIn profiles show about 5% more non-programming skills and no decline in coding skills.
Using LinkedIn and GitHub data, this paper examines how firms' adoption of GitHub Copilot (GHC), a generative AI coding assistant, relates to software engineer (SWE) skills and labor outcomes. GHC adoption is associated with around a 3%–5% higher monthly probability of hiring SWEs, driven by entry‐level hires. New hires exhibit around 5% more non‐programming skills, with no decrease in coding skills. These findings are consistent with, for SWEs, GAI's productivity impacts and creation of new tasks outweighing potential displacement effects from automation of some SWE tasks.
Summary
Main Finding
Firms that adopt GitHub Copilot (GHC) have a 3%–5% higher monthly probability of hiring software engineers (SWEs), an effect concentrated among entry‑level hires. New hires after adoption show about 5% more non‑programming skills and no decline in coding skills. The evidence is consistent with generative AI raising SWE productivity and creating new tasks, which increases labor demand and outweighs any short‑run displacement of coding tasks.
Key Points
- Adoption effect: GHC adoption → ~3%–5% higher monthly hiring probability for SWEs.
- Heterogeneity: The hiring increase is driven mainly by entry‑level hires rather than senior staff.
- Skill composition: New hires post‑adoption display ~5% more non‑programming skills (e.g., collaboration, product, design/analytics), with coding skill levels unchanged.
- Interpretation: Results align with AI acting as a complement that boosts productivity and creates/expands tasks, leading firms to hire more (especially junior) engineers rather than simply replacing them.
- Net outcome: For SWEs in the sample, demand effects from productivity and task creation dominate potential displacement.
Data & Methods
- Data sources: Firm‑level GitHub data to identify Copilot (GHC) adoption and LinkedIn profile/hiring data to measure SWE hiring events and skill composition.
- Outcomes: Monthly probability of hiring SWEs; skill tags/indicators from LinkedIn profiles used to quantify coding vs non‑programming skills among new hires.
- Empirical strategy (high level): Panel comparison of adopting firms before and after GHC adoption versus non‑adopting firms, with controls for time and firm factors and robustness checks (event‑study / placebo checks and alternative samples) to rule out simple confounders. (The paper reports effect sizes above after accounting for such controls.)
- Robustness: Findings reported to be robust to specification checks and are specific to SWEs in the observed platforms/data.
Implications for AI Economics
- Complementarity > substitution (short-to-medium run): For software engineering, generative AI can increase labor demand by raising productivity and enabling new or expanded non-coding tasks.
- Entry‑level opportunities: GAI may lower hiring thresholds (or supervision costs) for junior hires, affecting career pipelines and onboarding/training dynamics.
- Skill demand shifts: Employers value more non‑programming skills alongside stable coding abilities — implying training and education should emphasize communication, product, and analytical skills in addition to coding.
- Wage and inequality implications: Increased demand for junior SWEs could alter wage dynamics, but wage effects are not reported here and may depend on broader labor supply responses.
- Policy and firm strategy: Firms should consider investing in complementary human capital (soft/product skills) and onboarding systems; policymakers should support reskilling and monitor longer‑run displacement risks as capabilities evolve.
- External validity and caution: Results pertain to Copilot users and observed firms on GitHub/LinkedIn; long‑run effects, other occupations, and broader labor market adjustments require further study.
Assessment
Paper Typecorrelational
Evidence Strengthlow — Findings are based on observational associations between firm-level Copilot adoption and hiring/skill outcomes; without a clearly stated exogenous source of variation or quasi-experimental design, results are vulnerable to confounding (e.g., more innovative or fast-growing firms both adopt Copilot and hire more).
Methods Rigormedium — The study leverages large-scale linked administrative/professional-data sources (LinkedIn and GitHub) and measures concrete outcomes (hires, skills), which supports precision and scope; however, the absence of a clearly identified causal strategy, limited information about controls or robustness checks, and potential measurement error in adoption and skill proxies constrain methodological rigor.
SampleFirm-level sample constructed by linking public GitHub organization activity (to identify GitHub Copilot adoption) with LinkedIn employment and profile-skill data; monthly hiring events for software engineers observed over the study window, with skill composition inferred from LinkedIn profile entries and GitHub activity for firms that maintain a public GitHub presence.
Themeslabor_markets skills_training
GeneralizabilityRestricted to firms with a public GitHub presence and LinkedIn-visible hiring/activity (selection toward more digital, software-oriented firms)., Likely concentrated in tech-sector or software-development-heavy firms; not representative of non-tech industries., Findings pertain to GitHub Copilot specifically and may not generalize to other generative-AI tools or broader automation technologies., Potential geographic/language bias if the data are dominated by US/English-speaking users., Limited to observable hires and profile-declared skills; informal or unreported skill changes and internal redeployments may be missed.
Claims (6)
| Claim | Direction | Confidence | Outcome | Details |
|---|---|---|---|---|
| GHC adoption is associated with around a 3%–5% higher monthly probability of hiring SWEs. Hiring | positive | high | monthly probability of hiring software engineers (SWEs) |
around a 3%–5% higher monthly probability of hiring SWEs
0.3
|
| The increase in hiring probability is driven by entry-level hires. Hiring | positive | high | hiring of entry-level software engineers |
driven by entry‐level hires
0.3
|
| New hires at GHC-adopting firms exhibit around 5% more non-programming skills. Skill Acquisition | positive | high | quantity/prevalence of non-programming skills among new hires |
around 5% more non‐programming skills
0.3
|
| There is no decrease in coding skills among new hires associated with GHC adoption. Skill Acquisition | null_result | high | coding skills among new hires |
no decrease in coding skills
0.3
|
| For software engineers, GAI's (GHC's) productivity impacts and creation of new tasks appear to outweigh potential displacement effects from automation of some SWE tasks. Employment | positive | high | net labor effects for software engineers (balance of productivity/task-creation vs. displacement) |
0.05
|
| The study uses LinkedIn and GitHub data to examine firms' adoption of GitHub Copilot and related SWE skills and labor outcomes. Other | neutral | high | data sources / methodological description |
0.5
|