Skip to main content

The dynamics of ICT adaptation and the productivity gaps across advanced nations


This paper concerns the relationship between ICT and the emerged productivity gaps across the advanced nations. It is shown theoretically that the standard output elasticity estimation methods may lead to underestimate the role of ICT in generating productivity gaps, when there is a trade-off between speed of introducing technologies and the need to assimilate those as reflected by the inefficiency with which the new technology is used. The trade-off generates leader-follower patterns and mutes the relationship between productivity and ICT. The paper tests several economic hypotheses derived from the theory of technology adaptation using data envelope analysis, a novel growth factorization, and a range of panel econometrics techniques. It finds new evidence for the importance of ICT capital.

This is a preview of subscription content, access via your institution.

Fig. 1


  1. 1.

    The literature finds that ICT elasticity of output estimates are typically low (Stiroh 2002b; Draca et al. 2006; Inklaar et al. 2008), indicating that the technology may have only a limited role as a source of productivity gaps (Stiroh 2002a).

  2. 2.

    Countries make innovations that expand the production possibilities frontier for their ICT-capital intensity. The innovations are commonly available and their sum defines the maturity of technology for a particular ICT-capital intensity.

  3. 3.

    In Basu and Weil (1998) model the production function is of the AK type, which indicates supranormal returns (the ICT capital’s elasticity of output is above the nominal output share of ICT capital).

  4. 4.

    The most productive countries are referred as frontier countries (or ICT-capital intensive frontier countries when they have the highest ICT-capital intensity among all frontier countries).

  5. 5.

    An instrumental variable estimation in the spirit of Guan and Lansink (2006) is used. Countries are AUT, BEL, DNK, ESP, FIN, FRA, GER, IRL, ITA, LUX, PRT, SWE, UK, AUS, JPN, and USA.

  6. 6.

    In the empirical analysis industry-level data is used. Here for the sake of clarity the models are discussed at the country level, but the idea is fully compatible with industry data, when the exogenous productivity term is also thought to include the impact of inter-sectoral dynamics within countries. Furthermore, the analysis focuses on one type of capital good. The effects of the use of other input factors and the way the output transforms into capital goods as well as country-level general equilibrium dynamics are controlled in the empirical analysis.

  7. 7.

    They choose AK model for the sake of simplicity, because it has much simpler dynamics compared to the empirically more relevant case of technological transitions under decreasing returns to scale in the knowledge creation function. The role of this assumption is discussed at the end of this subsection.

  8. 8.

    It also captures how exogenous conditions such as investments in other goods, human capital, institutions, and policies affect productivity.

  9. 9.

    The conclusions regarding group dynamics reflects the results of the numerical analysis in the original article’s section III.

  10. 10.

    As an analogy, consider a group of racing cyclists. In principle, the group can benefit from each other’s slipstream. However, if there is too much heterogeneity in the group, a group of better cyclists may find it useful to break off while leaving the others to try to catch up as a group.

  11. 11.

    In practice, the equilibrium spillover potential is measured as the average spillover of the technology after controlling for short-term macroeconomic shocks, individual characteristics of the countries and industries, as well as potential endogeneity and measurement problems.

  12. 12.

    In particular, the country’s observations have very high B and s, and the knowledge spillovers from the followers are small.

  13. 13.

    Following Los and Timmer (2005), the frontiers are estimated for each period with a dataset that also includes historical observations. Technologies that previously have been used by the leaders should still be available in the future for the followers, and thus lagging behind earlier levels of productivity should be considered as falling further from the frontier, rather than technological regression. Thus, the estimations are based on a set of observations that are current and from the previous 10 years. In the actual estimations, free disposability of inputs was assumed and the frontier was allowed to exhibit variable returns to scale, while being convex [for further information, see Coelli et al. (2005)]. The estimations were conducted using FEAR 1.11 by Wilson (2008).

  14. 14.

    In contrast, Los and Timmer (2005) consider the elements separately (as “assimilation” and “localized innovation”, respectively). Here such distinction is not made for theoretical and practical reasons. First, in the original Basu and Weil (1998) it is possible for countries to operate below the observed technological frontier [when the idiosyncratic productivity term (B) is small], and still contribute to the common technology (innovate). In practice, the distinction would make both elements truncated variables. That is, because below the frontier information gathering contributes to \(g^{eff},\) while on the frontier it contributes to \(g^{tech}\). In any case, it should be emphasized that the negative covariance between the growth component and ICT-capital growth rate remains even if \(g^{tech}\) is omitted.

  15. 15.

    In actual estimations, logs are used as proxies for the growth rates.

  16. 16.

    More generally, they control for interactions between sectors of the economy as separate shocks for manufacturing and services are also considered.

  17. 17.

    The pure catch-up element is missing in the Basu and Weil (1998) model, and it refers to more general model discussed by Los and Timmer (2005).

  18. 18.

    An alternative would be to use stochastic frontier methods. However, since both movement relative to the frontier and along the frontier are (potentially non-linearly) dependent on the ICT capital intensity and its change, the components would be difficult to identify with a parametric model. Furthermore, a two-stage approach would be nevertheless needed to address the endogeneity problem.

  19. 19.

    For instance, manufacturing is much more procyclical than services.

  20. 20.

    When instrumentation is used, the reverse causality affects the elasticity of g\(^{K}\) only when the growth rates of capital are consistently correlated with the errors in the shape of the frontier, not only in the industry i, but also between industries. This effect should be captured by the (year \(\times\) country) controls.

  21. 21.

    March 2008 release.

  22. 22.

    The industries are chemicals and ch. products; electricity, gas, and water supply; transport equipment; rubber and plastics; other non-metallic mineral; machinery, nec; wood and of wood and cork; pulp, paper, printing, and publ.; textiles, leather, and footwear; basic metals and fabr. metal; wholesale and retail trade; transport and storage; food, beverages, and tobacco; manufacturing nec; recycling; construction; real estate activities; business services; hotels and restaurants.

  23. 23.

    See Inklaar and Timmer (2008, appendix B).

  24. 24.

    The reported estimates are averages for the 18 industries listed in Table 1, while the whole set of EU KLEMS countries was used in the estimations of the growth components. The data and algorithms are available from the author upon request.

  25. 25.

    The frontiers are updated relatively frequently at an average annual rate of 1.38 new frontier country observations for each industry. Generally, the frontier countries reflect well CIA world factbook (2000) lists of the most important manufacturing industries, rankings of MFP in manufacturing industries by Fadinger and Fleiss (2011), and listing of MFP leaders in service industries (Inklaar et al. 2008, Table A3).

  26. 26.

    At the same time the average position relative to the observed frontier is 7 % points higher in the US. This may imply that the US industries on average are compared with less mature technology or have on average higher \(B\). Furthermore, it tells that the aggregate comparison hides considerable amount of industry- and country-level heterogeneity across the regions.

  27. 27.

    A further decompositon of \(g^{R}\) shows that \(g^{eff}\) is on average −0.33 % in the EU-15 while this term is negligible in the US.

  28. 28.

    The estimations are made with STATA xtreg program. The standard errors are adjusted for potential clusters in the entity-level as well as heteroscedasticity. The dataset consists of the ICT-using industries in the EU-15 countries and the corresponding US industries.

  29. 29.

    Included countries are the EU-15 excluding Greece: AUT, BEL, DNK, ESP, FIN, FRA, GER, IRL, ITA, LUX, PRT, SWE, and the UK, plus AUS, JPN, and USA.

  30. 30.

    It is noticeable that there are some gaps in the data due to negative capital service measurements. This problem mainly exists in construction and hotels and restaurants. Also, in the earlier period some countries are not included in the data. In particular, SWE, LUX, and PRT enter the panel only in the mid 1990s. However, the panel is found to be fairly balanced before and after 1995.

  31. 31.

    The STATA ivreg2 algorithm is used in the estimation. Standard errors are heteroscedasticity and autocorrelation (bandwidth 3) consistent. The Kleibergen–Paap under-identification test suggests that the instruments are valid. The Sargan-Hansen over-identifying restrictions test is conducted, and the null hypothesis that the instruments are valid cannot be rejected. The weak identification test (Cragg–Donald Wald F statistic and Kleibergen–Paap rk Wald F statistic) rejects the hypothesis of weak identification. Finally, in an alternative specification other inputs were introduced as additional exogenous controls, but the results remained qualitatively the same.

  32. 32.

    As before statistics robust to heteroskedasticity and clustering on country and year were also considered, and the over, under and weak identification tests were conducted.

  33. 33.

    Further details of the estimations are available from the author upon request.


  1. Basu S, Weil DN (1998) Appropriate technology and growth. Q J Econ 113:1025–1054

    Article  Google Scholar 

  2. Basu S, Fernald FG, Oulton N, Srinivasan S (2003) The case of the missing productivity growth: or, does information technology explain why productivity accelerated in the United States but not the United Kingdom? WP 2003-08, The Federal Reserve Bank of Chicago

  3. Basu S, Fernald J (2007) Information and communications technology as a general-purpose technology: evidence from U.S. industry data. Ger Econ Rev 8(2):123–307

    Article  Google Scholar 

  4. Bresnahan TF, Brynjolfsson E, Hitt LM (2002) Information technology, workplace organization, and the demand for skilled labor: firm-level evidence. Q J Econ 117(1):339–376

    Article  Google Scholar 

  5. Coelli TJ, Prasada Rao DS, O’Donnell JC, Battese GE (2005) An introduction to efficiency and productivity analysis. Springer, New York

    Google Scholar 

  6. Draca M, Sadun R, Van Reenen J (2006) Productivity and ICT: a review of the evidence. CEP discussion paper no. 749

  7. Fadinger H, Fleiss P (2011) Trade and sectoral productivity. Econ J 121:958–989

    Article  Google Scholar 

  8. Guan Z, Oude Lansink A (2006) The source of productivity growth In Dutch agriculture: a perspective from finance. Am J Agric Econ 88(3):644–656

    Article  Google Scholar 

  9. Greenwood J, Yorukoglu, M  (1997) 1974. In: Carnegie-Rochester conference series on public policy, vol 46, pp 49–95

  10. Growiec J (2012) The world technological frontier: what can we learn from the US states? Oxf Bull Econ Stat 74(6):777–780

  11. Henderson DJ, Russell RR (2005) Human capital and convergence: a production-frontier Approach. Int Econ Rev 46(4):1167–1205

    Article  Google Scholar 

  12. Inklaar R, Timmer MP, van Ark B (2008) Market services productivity across Europe and the US. Econ Policy 23(1):139–194

    Article  Google Scholar 

  13. Inklaar R, Timmer MP (2008) GGDC productivity level database: international comparisons of output, inputs and productivity at the industry level. GGDC Research Memorandum GD-104, University of Groningen

  14. Keller W (2004) International technology diffusion. J Econ Lit XLII:752–782

    Article  Google Scholar 

  15. Kumar S, Russell RR (2002) Technological change, technological catch-up, and capital deepening: relative contributions to growth and convergence. Am Econ Rev 92(3):527–548

    Article  Google Scholar 

  16. Los B, Timmer MP (2005) The ‘appropriate technology’ explanation of productivity growth differentials: an empirical approach. J Dev Econ 77(2):517–531

    Article  Google Scholar 

  17. O’Mahony M, Vecchi M (2005) Quantifying the impact of ICT capital on output growth: a heterogeneous dynamic panel approach. Economica 72(288):615–634

    Article  Google Scholar 

  18. Oulton N, Srinivasan S (2005) Productivity growth and the role of ICT in the United Kingdom: an industry view, 1970–2000. CEP discussion paper 681

  19. Parente S (1994) Technology adoption, learning-by-doing, and economic growth. J Econ Theory 63:346–369

    Article  Google Scholar 

  20. Simar L, Wilson PW (2007) Estimation and inference in two-stage, semi-parametric models of production processes. J Econom 136(1):31–64

    Article  Google Scholar 

  21. Stiroh KJ (2002a) Are ICT spillovers driving the new economy? Rev Income Wealth 48(1):33–57

    Article  Google Scholar 

  22. Stiroh KJ (2002b) Reassessing the impact of IT in the production function: a meta-analysis. Federal Reserve Bank of New York, Mimeo

  23. Timmer MP, O’Mahony M, van Ark B (2007) EU KLEMS growth and productivity accounts: overview.

  24. van Ark B (2011) Up the hill and down again: a history of Europe’s productivity gap relative to the United States, 1950–2009. Nordic Economic Policy Review, 2/2011

  25. van Reenen J, Bloom N, Draca M, Kretchmer T, Sadun R, Overman H, Schankerman M (2010). The economic impact of ICT. CEP, final report. SMART N. 2007/0020

  26. Venturini F (2009) The long-run impact of ICT. Empir Econ 37:497–515

    Article  Google Scholar 

  27. Wilson PW (2008) FEAR: a software package for frontier efficiency analysis with R. Socio Econ Plan Sci 42(4):247–254

    Article  Google Scholar 

Download references


I thank Matti Pohjola as well as Yuriy Gorodnichenko, Pertti Haaparanta, Pekka Ilmakunnas, Timo Kuosmanen, Antti Ripatti, R. Robert Russell, Marcel P. Timmer, two anonymous referees, and seminar participants in HECER seminars and EWEPA 2010 conference for generous comments.

Author information



Corresponding author

Correspondence to Tero Kuusi.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Kuusi, T. The dynamics of ICT adaptation and the productivity gaps across advanced nations. J Prod Anal 44, 175–188 (2015).

Download citation


  • MFP
  • ICT capital
  • Growth differentials

JEL Classification

  • O3
  • O4
  • C2
  • C4