Basu and Fernald (2007) showed that US industry total factor productivity (TFP) accelerations in the 2000s were positively correlated with 1990s ICT capital growth and negatively correlated with contemporaneous ICT growth, consistent with information and communications technology (ICT) acting as a general-purpose technology whose payoff arrives only after long, unmeasured complementary investment in intangible capital. The authors were explicit that this evidence was exploratory: the lag length is a free parameter, the framework cannot distinguish privately accumulated complementary capital from spillovers, ICT investment is endogenous, and relative-price terms and asymmetric production functions are assumed away. The identification core remains open: later work — the productivity J-curve model, frontier-efficiency and spillover reassessments, and EU panel tests with new intangible-investment data — confirms the mechanism's plausibility and improves intangible-capital measurement, but no study has estimated the lag distribution between ICT investment and TFP growth in US industry data, or credibly decomposed the lagged correlation into private accumulation versus externalities.

Background

Total factor productivity (TFP) is the portion of an industry's output growth left over after accounting for the growth of measured inputs of capital and labor; it is the standard empirical proxy for technical change. A general-purpose technology (GPT) is a technology, like the steam engine or electricity, that is used across many sectors and whose productivity payoff arrives only after users make large, costly complementary investments. For information and communications technology (ICT) — computers, software, and communications equipment — these complementary investments are largely intangible capital: assets such as organizational know-how, business-process redesign, employee training, and co-invention of new products that are not counted as investment in the national accounts. A spillover (externality) exists when these or related investments raise productivity in firms or industries other than the one that paid for them.

The GPT view generates a distinctive empirical signature. While a firm invests in ICT, resources are diverted to building unmeasured complementary capital, so measured output understates true output and measured TFP falls; later, when the complementary capital is harvested, measured TFP rises. Industry data should therefore show a negative correlation between ICT investment and current TFP growth but a positive correlation with the TFP growth that follows it (that is, TFP growth is positively correlated with lagged ICT capital growth). Basu and Fernald (2007) formalized this in a growth-accounting framework in which observed ICT capital intensity and growth serve as proxies for unobserved complementary investment, and found exactly this pattern in data for about 40 US industries (Bureau of Economic Analysis and Bureau of Labor Statistics data, 1987–2004): TFP accelerations in the 2000s were positively associated with share-weighted ICT capital growth over 1987–2000 and negatively with ICT growth in 2000–2004. The companion paper Basu, Fernald, Oulton and Srinivasan (2003) developed the same accumulation framework for the US–UK productivity comparison. The mechanism was later reformalized as the 'productivity J-curve' by Brynjolfsson, Rock and Syverson (2021).

The authors were explicit that this evidence was data exploration, not identification. Their framework leaves the length and distribution of the lag between ICT investment and the productive use of complementary capital undetermined — the 'length of a period' is a free parameter that could be years or a decade or more — and the regressions test sign patterns over period averages rather than estimating a lag profile. Observed industry correlations cannot distinguish whether the lagged payoff reflects privately accumulated complementary capital, intra-industry externalities internalized within the industry, or inter-industry spillovers. The estimation omits relative-price and user-cost terms (the effective rental price of capital, combining purchase prices, interest, and depreciation), assumes for convenience that omitted relative-price effects are small or common across industries, imposes the same production function for complementary capital as for final output (although complementary investment may be intensive in skilled labor), and treats current ICT investment as exogenous: the authors found no credible instruments and reported ordinary least squares, so the estimates carry unknown simultaneity bias from technology shocks and reorganization decisions. Since 2007, industry-level investment in intangibles has become measurable in Europe (the INTAN-Invest/INTANProd databases built by Corrado, Hulten, Sichel and successors), which makes parts of the mechanism testable but has not delivered the US industry-level identification the source paper called for.

Problem Statement

For US industry-level data (the roughly 40-industry Bureau of Economic Analysis/Bureau of Labor Statistics panel used by Basu and Fernald (2007), or a consistent updated extension of it), determine what the observed pattern — industry TFP growth negatively correlated with contemporaneous share-weighted ICT capital growth and positively correlated with lagged ICT capital growth — is made of, and how long the payoff takes. Concretely: (1) estimate the length and distribution of the lags between ICT investment and measured TFP growth, going beyond tests of sign patterns over period averages such as 1987–2000 versus 2000–2004 blocks; (2) decompose the estimated lagged relationship into unobserved private complementary-capital accumulation and externalities/spillovers, distinguishing at least intra-industry from inter-industry spillovers — that is, establish which part of the correlation reflects unmeasured intangible investment being harvested within the investing industry rather than productivity benefits spilling across firms or sectors; and (3) deliver this under identification that addresses the endogeneity of current ICT investment (reverse causation from unobserved technological shocks and simultaneity with reorganization), treats or tests the omitted relative-price/user-cost terms rather than assuming them small or common across industries, and allows the production of complementary capital to use different input intensities (for example skilled labor) than the production of final output. These three components are one identification-and-quantification objective: the lag estimates and the decomposition are jointly interpretable only within a single coherent model of the GPT mechanism, and the identification threats in (3) are the confounds that make (1) and (2) open questions rather than regressions.

References

  1. Susanto Basu and John G. Fernald (2007). Information and Communications Technology as a General-Purpose Technology: Evidence from US Industry Data. German Economic Review 8(2):146–173. DOI: 10.1111/j.1468-0475.2007.00402.x. https://doi.org/10.1111/j.1468-0475.2007.00402.x
  2. Susanto Basu and John G. Fernald (2008). Information and Communications Technology as a General Purpose Technology: Evidence from U.S. Industry Data. FRBSF Economic Review 2008, pp. 1–15. https://www.frbsf.org/wp-content/uploads/er1-15.pdf
  3. Susanto Basu, John G. Fernald, Nicholas Oulton and Sylaja Srinivasan (2003). The Case of the Missing Productivity Growth, or Does Information Technology Explain Why Productivity Accelerated in the United States but Not in the United Kingdom? NBER Macroeconomics Annual 18:9–63. DOI: 10.1086/ma.18.3585244. https://doi.org/10.1086/ma.18.3585244
  4. Erik Brynjolfsson, Daniel Rock and Chad Syverson (2021). The Productivity J-Curve: How Intangibles Complement General Purpose Technologies. American Economic Journal: Macroeconomics 13(1):333–372. DOI: 10.1257/mac.20180386. https://doi.org/10.1257/mac.20180386
  5. Hailin Liao, Bin Wang, Baibing Li and Tom Weyman-Jones (2016). ICT as a general-purpose technology: The productivity of ICT in the United States revisited. Information Economics and Policy 36:10–25. DOI: 10.1016/j.infoecopol.2016.05.001. https://doi.org/10.1016/j.infoecopol.2016.05.001
  6. Ian William Marsh, Ana Rincon-Aznar, Michela Vecchi and Francesco Venturini (2017). We see ICT spillovers everywhere but in the econometric evidence: a reassessment. Industrial and Corporate Change 26(6):1067–1088. DOI: 10.1093/icc/dtx008. https://doi.org/10.1093/icc/dtx008
  7. Carol Corrado, Charles Hulten and Daniel Sichel (2009). Intangible Capital and U.S. Economic Growth. Review of Income and Wealth 55(3):661–685. DOI: 10.1111/j.1475-4991.2009.00343.x. https://doi.org/10.1111/j.1475-4991.2009.00343.x
  8. Carol Corrado, Jonathan Haskel and Cecilia Jona-Lasinio (2017). Knowledge Spillovers, ICT and Productivity Growth. Oxford Bulletin of Economics and Statistics 79(4):592–618. DOI: 10.1111/obes.12171. https://doi.org/10.1111/obes.12171
  9. Keungoui Kim, Ahmed Bounfour, Alberto Nonnis and Altay Özaygen (2021). Measuring ICT externalities and their contribution to productivity: A bilateral trade based approach. Telecommunications Policy 45(2):102085. DOI: 10.1016/j.telpol.2020.102085. https://doi.org/10.1016/j.telpol.2020.102085
  10. Izabela Młynarzewska-Borowiec (2026). ICT Usage and TFP Growth—GPT Hypothesis Verification for the EU Member States. Journal of the Knowledge Economy 17(1):2614–2633 (print issue February 2026; first published online 16 April 2025). DOI: 10.1007/s13132-025-02748-z. https://doi.org/10.1007/s13132-025-02748-z
  11. M. Cardona, T. Kretschmer and T. Strobel (2013). ICT and productivity: conclusions from the empirical literature. Information Economics and Policy 25(3):109–125. DOI: 10.1016/j.infoecopol.2012.12.002. https://doi.org/10.1016/j.infoecopol.2012.12.002
  12. Thomas Strobel (2016). ICT intermediates and productivity spillovers—Evidence from German and US manufacturing sectors. Structural Change and Economic Dynamics 37:147–163. DOI: 10.1016/j.strueco.2016.04.003. https://doi.org/10.1016/j.strueco.2016.04.003