Software and software R&D—linked to AI—explain half of U.S. nonfarm business labor productivity growth between 2017 and 2024, and half of the 1.2 percentage-point acceleration versus the prior five years, implying AI is already materially affecting official productivity measures.
AI as an Innovation in the Method of Innovation: Implications for Productivity Growth
Using an upstream/downstream growth-accounting framework, the paper finds that software products and software R&D—attributed to AI—account for roughly half of U.S. nonfarm business labor productivity growth from 2017–2024 and half of its 1.2 percentage-point acceleration versus 2012–2017.
This paper estimates AI’s impact on labor productivity growth, treating AI as both a general-purpose technology and an innovation in the method of innovation. Using a framework that separates upstream innovation from downstream (other) production suggests that AI boosts both upstream total factor productivity and intangible capital use downstream. We find that AI is already materially affecting official productivity measures in the United States. Software products and software R&D contributed 50 percent of the 2 percent average growth rate in nonfarm business labor productivity from 2017 to 2024 and 50 percent of its 1.2 percentage point acceleration compared to 2012–2017.
Summary
Main Finding
AI is already materially raising measured U.S. labor productivity. Treating AI both as a general-purpose technology (GPT) and as an innovation in the method of innovation, the paper shows AI has increased upstream total factor productivity (TFP) and raised downstream use of intangible capital (notably software and software R&D). Software products and software R&D together account for about half of (1) the 2.0% average annual growth in nonfarm business labor productivity from 2017–2024 and (2) the 1.2 percentage-point acceleration of that growth relative to 2012–2017.
Key Points
- Conceptual framing: AI operates on two margins
- As a GPT that directly raises productivity across many sectors.
- As an innovation in how innovation is produced, improving upstream R&D and idea-generation processes.
- Decomposition result: AI’s impact manifests through
- Upstream improvements in TFP (faster/more effective innovation).
- Downstream increases in the use of intangible capital, especially software products and software R&D.
- Quantitative headline: Software-related items (software products + software R&D) explain ~50% of the recent rise and acceleration in measured nonfarm business labor productivity (2017–2024).
- Periods compared: main window 2017–2024 versus pre-acceleration window 2012–2017.
Data & Methods
- Framework: A two-stage growth-accounting/decomposition approach that separates upstream innovation production from downstream final production, allowing identification of (a) upstream TFP effects and (b) changes in downstream capital composition (intangible use).
- Identification strategy: Attribute changes in official productivity statistics to upstream TFP shifts and to changes in intangible investment/use (here proxied by software products and software R&D).
- Data sources and scope: Uses U.S. official productivity measures for the nonfarm business sector and national-account/R&D/software investment aggregates over 2012–2024 (paper’s reported comparisons are 2017–2024 vs 2012–2017).
- Empirical outputs: Growth contributions and decomposition of the acceleration in labor productivity, isolating the share accounted for by software and software R&D.
- Inference: Interprets the sizable contribution of software-related items as consistent with an active role for AI as both a GPT and an innovation-in-innovation.
Implications for AI Economics
- Measurement: Standard productivity statistics already capture substantial effects of AI through intangible investment channels; analysts should incorporate intangible capital and R&D dynamics when assessing AI’s macro impact.
- Mechanisms matter: Distinguishing upstream innovation efficiency gains from downstream capital-deepening is important for forecasting growth and designing policy.
- Policy levers: Supporting software R&D, digital adoption, and complementary intangible investments can amplify productivity gains from AI; policies that facilitate diffusion could raise the realized macro benefit.
- Forecasting growth: If AI continues to operate as both a GPT and an innovation accelerator, productivity gains may persist or accelerate further, but outcomes depend on diffusion speed, complementary investments (skills, data, institutions), and measurement of intangibles.
- Research priorities: Better sectoral decomposition, firm-level heterogeneity, and direct measurement of AI-specific R&D and diffusion are needed to refine estimates of long-run impacts and distributional effects.
Assessment
Paper Typedescriptive
Evidence Strengthlow — Findings rely on accounting decomposition and attribution of software output and R&D to ‘AI’ rather than on exogenous variation or causal identification; results are informative and consistent with official data but vulnerable to measurement error and attribution assumptions that could bias inferred AI effects.
Methods Rigormedium — Uses a transparent, economically grounded upstream/downstream framework and official productivity/national-account data to decompose growth, which is methodologically sound for descriptive attribution; however, it lacks causal identification (no instruments or plausibly exogenous shocks) and results depend on assumptions about what portion of software and R&D growth is driven by AI.
SampleAggregate U.S. nonfarm business sector productivity and national-account data covering 2012–2017 and 2017–2024 periods, with disaggregation for software products, software R&D (intangible capital), and upstream TFP measures; uses official productivity measures and accounting inputs to estimate contributions.
Themesproductivity innovation
IdentificationGrowth-accounting / decomposition framework that separates upstream innovation (TFP) from downstream production and intangible capital; attributes observed increases in software products and software R&D to AI-driven innovation and computes their contribution to labor productivity growth (comparison of 2017–2024 vs 2012–2017). No exogenous instrument or natural experiment is used.
GeneralizabilityResults are aggregate and U.S.-specific — may not generalize to other countries with different AI adoption patterns or measurement systems., Aggregate averages mask sectoral and firm-level heterogeneity in AI adoption and productivity impacts., Attribution of software and software R&D growth to 'AI' may not hold where software growth reflects non-AI innovations or other forces., Time window (2017–2024) captures early-to-mid adoption; longer-term effects could differ.
Claims (5)
| Claim | Direction | Confidence | Outcome | Details |
|---|---|---|---|---|
| AI functions both as a general-purpose technology and as an innovation in the method of innovation. Other | mixed | high | classification of AI as a type of technological innovation (GPT and method-of-innovation) |
0.03
|
| Using a framework that separates upstream innovation from downstream production suggests that AI boosts both upstream total factor productivity and intangible capital use downstream. Firm Productivity | positive | high | upstream total factor productivity and downstream intangible capital use |
0.18
|
| AI is already materially affecting official productivity measures in the United States. Firm Productivity | positive | high | official productivity measures (U.S. nonfarm business labor productivity) |
0.18
|
| Software products and software R&D contributed 50 percent of the 2 percent average growth rate in nonfarm business labor productivity from 2017 to 2024. Firm Productivity | positive | high | average growth rate in nonfarm business labor productivity (2017–2024) |
50 percent of the 2 percent average growth rate in nonfarm business labor productivity from 2017 to 2024
0.18
|
| Software products and software R&D contributed 50 percent of the 1.2 percentage point acceleration in nonfarm business labor productivity (2017–2024 relative to 2012–2017). Firm Productivity | positive | high | acceleration (difference) in nonfarm business labor productivity growth between 2017–2024 and 2012–2017 (1.2 percentage points) |
50 percent of its 1.2 percentage point acceleration compared to 2012–2017
0.18
|