Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat
Sign Up
Cited by 1,011Economic Growth and ProductivityEconomic Growth and DevelopmentEconomic Development and Digital Transformation

The DOI matches its Crossref record; 63 of the 63 references it cites were found in a registry or an index; none of the 25 sources its claims cite is checked for support yet.

The Productivity J-Curve: How Intangibles Complement General Purpose Technologies

Open access
Brynjolfsson, Erik, Rock, Daniel, Syverson, Chad·2021·American Economic Journal: Macroeconomics
DOI 10.1257/mac.20180386
Open this paper in CoChat — free accountOpen PDF

Abstract

General purpose technologies (GPTs) like AI enable and require significant complementary investments. These investments are often intangible and poorly measured in national accounts. We develop a model that shows how this can lead to underestimation of productivity growth in a new GPTs early years and, later, when the benefits of intangible investments are harvested, productivity growth overestimation. We call this phenomenon the Productivity J-curve. We apply our method to US data and find that adjusting for intangibles related to computer hardware and software yields a TFP level that is 15.9 percent higher than official measures by the end of 2017. (JEL E22, E23, G31, L63, L86)

Overview

CoChat Summary

TL;DR

The paper develops a Productivity J-Curve model in which unmeasured intangible investments complementary to general purpose technologies cause productivity growth to be underestimated early and overestimated later. Empirically, it finds substantial ongoing J-curve effects for software and, to a lesser extent, computer hardware, with adjusted TFP 11.3% higher than official measures at end-2004 and 15.9% higher at end-2017.

Key findings

  • The model predicts measured TFP growth initially understates true productivity growth when resources are used to build unmeasured intangible capital, then overstates it as intangible capital services generate measured output; in steady state, mismeasurement falls to zero.
  • R&D-related intangibles have negligible net measurement effects: the net adjusted productivity level is 1.6% lower than the measured series as of end-2017.
  • Software and computer hardware show substantial J-curve effects; assuming an intangible multiplier of 10, the software-adjusted TFP level is over 12.4% higher than measured TFP at the beginning of 2017, and overall adjusted TFP is 11.3% higher than official measures at end-2004 and 15.9% higher at end-2017.
  • Intangibles do not explain the post-2004 productivity slowdown under a constant multiplier: measured annual TFP growth slowed by about 1.23% per year from 1995–2004 to 2005–2017, while the adjusted slowdown is larger at 1.49% per year; AI-related effects are small but growing, and explaining the entire 2010–2017 GDP slowdown would require about $107 billion per year of unmeasured intangible investment, implying an AI intangible multiplier of roughly 2.7 to 4.1 times observable investment.

Methods

  • The paper augments standard growth accounting with unmeasured intangible capital, deriving expressions for the difference between the standard Solow Residual and true productivity growth; adjusted TFP growth is expressed as a convex combination of standard productivity growth and the difference between unmeasured investment growth and capital stock growth.
  • It uses q-theory market value regressions to infer intangible correlates: firm market value is regressed on total assets, R&D stock, and capitalized SG&A stock, using a Compustat panel from 1961–2017 merged with Fernald TFP data extended through 2017 and Peters and Taylor R&D/SG&A stock measures.
  • For software and computer hardware, lacking comparable firm-level IT capital data, it applies a range of intangible multipliers λ/z of 10, 5, 3, and 2, guided by Brynjolfsson, Hitt, and Yang (2002), who estimate roughly $12 of market value per unit of observable software and hardware; the AI assessment is described as more speculative.

Limitations

  • The paper states that its method requires intangibles to have a nonzero correlation with measured capital, so completely uncorrelated intangible investments cannot be measured; using firm-level data to infer aggregate intangible-to-tangible ratios requires an aggregation assumption that may be problematic, and it therefore reports multiple multiplier values.
  • The slowdown analysis assumes the intangible multiplier is constant throughout the period; if it were higher in recent periods, mismeasurement would be greater recently, and the AI analysis is early and speculative. The paper also leaves investment-timing effects, such as lower interest rates prolonging the J-curve, to future research.
  • Evident from the text: this version is an NBER working paper and states it has not been peer-reviewed; the empirical estimates also rely on efficient market pricing and external/assumed intangible multipliers.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1257/mac.20180386

Cites

63 of 63 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

0 Valid · 2 Not valid · 19 Not checked · 137 Own results and interpretations

82 more claims

Create a free account to see all 158 claims and to verify the 16 references still open.

See all claims — free account
  • It may state two separate claims as one. 48
  • Probably not a substantial claim. 36

Create a free account to see these claims.

Create a free account to see these assumptions.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Citations

Cites

Show all 57 references

Research works printed but not linked: 6

  1. Brynjolfsson, Erik, Daniel Rock, and Chad Syverson. 2021. "Replication data for: The Production J-Curve: How Intangibles Complement General Purpose Technologies." American Economic Association [publisher], Inter-university Consortium for Political and Social Research [distributor]. https://doi.org/10.3886/E117947V1. · DOI
  2. Byrne David M, 2016, US Federal Reserve Finance and Economics Discussion Series Paper, 2016
  3. David Paul A, 1990, American Economic Review, 80, 355
  4. Fernald John, 2014, Federal Reserve Bank of San Francisco Working Paper, 2012
  5. Solow, Robert M. 1987. "We'd Better Watch Out." New York Times Book Review. July 12, 36.
  6. Waldasin David E, 1984, American Economic Review, 74, 203

Cited by

Items 701–800 of 1,011

Related

See all

Metadata

Authors
Brynjolfsson, E., Rock, D., & Syverson, C.
Affiliations
  • National Bureau of Economic Research (US)
  • University of Pennsylvania (US)
  • University of Chicago (US)
ORCID
Venue
American Economic Journal: Macroeconomics
Publisher
American Economic Association
Published
2021-01-01
Volume
13
Issue
1
Pages
333-372
ISSN
  • 1945-7715 (ISSN-L)
  • 1945-7707
DOI
10.1257/mac.20180386
OpenAlex
W2897697024
Cited by
1,003
References
57
Citations by year
2026: 482 · 2025: 200 · 2024: 138 · 2023: 79 · 2022: 42 · 2021: 31
FWCI
33.97
Type
journal article
Language
en
Open access
Locations
  • American Economic Journal Macroeconomics (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
Topics
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Economic Growth and Productivity
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Economic Growth and Development
  • Social Sciences › Economics, Econometrics and Finance › General Economics, Econometrics and Finance › Economic Development and Digital Transformation
Keywords
general purpose technologies, artificial intelligence, intangible investment, productivity growth, complementary investments, computer hardware and software, national accounts

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-14, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-14 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.