Abstract
Using data for twelve manufacturing industries of five developed countries over the period 1980–2002, we perform a dynamic panel estimation of the long-run elasticity of TFP with respect to R&D capital. The highest elasticity is found for the US (0.39), followed by Germany (0.29–32); intermediate values are achieved by France (0.19–0.21) and Spain (0.19), while Italy records the lowest R&D impact (0.08–0.12). The latter finding, supported by an in depth analysis based on a longer time-span and more accurate data, can be ascribed to the declining R&D efforts undertaken, during the Nineties, by Italian manufacturing industries.
Similar content being viewed by others
Notes
According to the European Commission (2010a, p. 3) “the EU has failed to close the productivity gap with leading industrialised countries: total R&D expenditures in the EU expressed as a percentage of GDP only improved marginally (from 1.82% in 2000 to 1.9% in 2008)”. The Europe 2020 Strategy (European Commission 2010b), agreed by the EU Member States at the European Council of June 2010, established five headline targets for the whole EU in 2020. Along with that concerned with R&D, the other quantitative goals refer to employment, climate change/energy, education, and poverty/social exclusion. Within the above strategy, the initiative ‘Innovation Union’ is specifically devoted to R&D and innovation (European Commission 2010c).
For an in-depth analysis of the Italian and Spanish cases see Sterlacchini and Venturini (2011).
According to Schankerman (1981) the double-counting effect on the input side could be counterbalanced by the fact that, on the output side, the value added is not net of R&D expenditures (as it should be as far as R&D is expensed as an intermediate input). However, even when both the R&D double-counting and expensing are taken into account, a substantial downward bias in the R&D elasticity (θ) remains (see also Hall et al. 2009).
The consumption of fixed capital is computed as the gross fixed capital formation at time t minus the absolute variation of net capital stock between time t and t-1. Net and gross fixed capital stocks are built by OECD through a permanent inventory method accounting for the age and efficiency profile of different capital assets. For Italy, France and Spain the period considered to estimate the depreciation rate spans from 1980 to 2002, while for Germany from 1991 onwards. For the US, due to the lack of data on net and gross fixed capital stock, the mean of the industry depreciation rates of the other countries is applied.
Labour compensation is augmented by the remuneration of self-employed workers by assuming that their compensation rate is equal to that of employees (OECD, 2001).
The other reason for using the national source is that OECD data for Italian industries are re-adjusted. The observed discrepancies are not big in absolute terms but, for some Italian industries that spend very little on R&D (see Table 1), they became non negligible.
Both in the aggregate descriptive statistics and in the subsequent regression analysis we exclude the industry of Coke, Refined Petroleum Products & Nuclear Fuel for its erratic and anomalous behaviour in terms of productivity and R&D performances (especially during the first part of the period considered). Depending on the country taken into consideration, econometric results are sensitive to the inclusion of this industry.
We thank Joakim Westerlund for providing us with the GAUSS codes used to implement his panel tests.
It should be reminded that the method used to build the German series for the pre-1991 period excludes any noise due to the reunification process. Industry data before 1991 have been indeed constructed from the levels of the Unified Germany for that year by applying backwardly the annual variations of the corresponding series of the West Germany (linking method; cf. van Ark et al. 2002).
Following Andrews and Monahan (1992), the long-run variance matrix is computed with the pre-whitening kernel-based heteroskedasticity-and-autocorrelation consistent (HAC) estimator.
For UK manufacturing industries Añón Higón (2007) estimates a 0.38 long-run R&D elasticity. Once again, being between the values estimated for Germany and the US, the long-run parameter concerned with the UK is in line with our findings.
Similar results arises when the industry of electrical equipment is excluded, as in Table 6.
See “Annual Growth Survey: Summary of the economic analysis and messages”, released on January 12, 2011 (http://ec.europa.eu/europe2020/tools/monitoring/annual_growth_survey_2011/index_en.htm).
The most appropriate indicator of labour usage is the number of hours worked which, however, is not provided by official statistical sources at industry level and for the time span considered in the present study.
This dataset collects data on a stratified sample representative of the Italian manufacturing firms having between 10 and 499 employees, while it covers the entire population of larger firms. From 1970 to 1989, the white/blue collars’ ratio has been interpolated among four benchmark years (1968, 1973, 1978, and 1984) while for the subsequent years annual data are available. In order to smooth the series, a three-year moving average has been used.
Unreported findings confirm this expectation. With respect to the measure computed by Brandolini and Cipollone (see the previous footnote) we were not able to test its adequacy because, for Italy, data on the educational attainments of workers are not available at the level of manufacturing industries.
ISTAT disentangles gross fixed capital formation into the following nine categories: machinery, office machinery, communication equipment, furniture, vehicles, other transport equipment, buildings, software, and other equipment.
References
Alessandrini P, Presbitero A, Zazzaro A (2010) Bank size or distance: what hampers innovation adoption by SMEs. J Econ Geogr 10:845–881
Añón Higón D (2007) The impact of R&D spillovers on UK manufacturing TFP: a dynamic panel approach. Res Pol 36:949–963
Andrews DWK, Monahan JC (1992) An improved heteroskedasticity and autocorrelation consistent covariance matrix estimator. Econom 60:953–966
Bassanetti A, Iommi M, Jona-Lasinio C, Zollino F (2004) La crescita dell’economia italiana negli anni Novanta tra ritardo tecnologico e rallentamento della produttività. Temi di Discussione Banca d’Italia n 539
Bassanini A, Scarpetta S (2002) Growth, technological change, and ICT diffusion: recent evidence from OECD countries. Oxf Rev Econ Pol 8:324–344
Benfratello L, Schiantarelli F, Sembenelli A (2008) Banks and innovation: microeconometric evidence on Italian firms. J Financ Econ 90:197–217
Brandolini A, Cipolline P (2001) Multifactor productivity and labour quality in Italy. Temi di Discussione Banca d’Italia n. 422
Boeri T, Faini R, Ichino A, Pisauro G, Scarpa C (2005) Oltre il declino. Il Mulino, Bologna
Brandt N (2007) Mark-ups, economies of scale and the role of knowledge spillovers in OECD industries. Eur Econ Rev 51:1708–1732
Breitung J, Pesaran H M (2008) Unit roots and cointegration in panels. In: Matyas L, Sevestre P (ed) The econometrics of panel data, 3rd edn. Kluwer Academic Publishers, pp 279–322
Bussolati C, Malerba F, Torrisi S (1996) L’evoluzione delle industrie ad alta tecnologia. Il Mulino, Bologna
Cameron G (2005) The sun also rises: productivity convergence between Japan and the USA. J Econ Growth 10:387–408
Coe DT, Helpman E (1995) International R&D spillovers. Eur Econ Rev 39:859–887
Cohen WM, Levinthal D (1989) Innovation and learning: the two faces of R&D. Econ J 99:569–596
Del Gatto M, Di Liberto A, Petraglia C (2010) Measuring Productivity, J Econ Surv, doi: 10.1111/j.1467-6419.2009.00620.x
European Commission (2005) The Economic Costs of Non-Lisbon: A Survey of the Literature on the Economic Impact of Lisbon-type Reforms. European Economy Occasional Paper16, Brussels
European Commission (2010a) Lisbon Strategy evaluation document. Commission Staff Working Document, SEC(2010) 114 final, Brussels
European Commission (2010b) Europe 2020: A strategy for smart, sustainable and inclusive growth. COM(2010) 2020 final, Brussels
European Commission (2010c) Europe 2020 Flagship Initiative Innovation Union. COM(2010) 546 final, Brussels
Fachin S, Gavosto A (2010) Trends of labour productivity in Italy: a study with panel cointegration methods. Int J Manpow 31:755–769
Ferrari S, Guerrieri P, Malerba F, Mariotti S, Palma D (1999) L’Italia nella competizione tecnologica internazionale: secondo rapporto. Franco Angeli, Milano
Frantzen D (2002) Intersectoral and International R&D Knowledge Spillovers and Total Factor Productivity. Scott J Polit Econ 49:280–303
Griffith R, Redding S, Van Reenen J (2004) Mapping the Two Faces of R&D: Productivity Growth in a Panel of OECD Industries. Rev Econ Stat 86:883–895
Griliches Z (1995) R&D and productivity: econometric results and measurement issues. In: Stoneman P (ed) Handbook of the economics of innovation and technological change. Blackwell, Oxford Basil
Guellec D, Van Pottelsberghe B (2004) From R&D to productivity growth: do the institutional settings and the source of funds of R&D matter? Ox Bull Econ Stat 66:353–378
Hall B (2010) Measuring the returns to R&D: the depreciation problem. In: Mairesse J, Trajtenberg M (ed) Contributions in memory of Zvi Griliches NBER, pp 341–381
Hall B, Mairesse J (1995) Exploring the relationship between R&D and productivity in French manufacturing firms. J Econom 65:263–293
Hall B, Mairesse J, Mohnen P (2009) Measuring the return to R&D. NBER Wor Pap 15622
Huynh KP, Rotondi Z (2007) R&D spending, local banks, and knowledge spillovers. Working paper available at SSRN: http://ssrn.com/abstract=968666
Inklaar R, O’Mahony M, Timmer M (2005) ICT and Europe’s productivity performance: industry-level growth account comparisons with the United States. Rev Inc Wea 5:505–536
Jorgenson DW, Gollop FM, Fraumeni B (1987) Productivity and U.S. economic growth. Cambr Univ Press
Jorgenson DW (2005) Accounting for growth in the information age. In: Aghion P, Durlauf S (eds) Handbook of economic growth. North-Holland, Amst, pp 743–815
Kao C, Chiang MH (2000) On the estimation and inference of a cointegrated regression in panel data. In: Baltagi BH (ed) Nonstationary panels, panel cointegration, and dynamic panels, advances in econometrics. JAI, Amsterdam, pp 179–222
Maffezzoli M (2006) Convergence across Italian regions and the role of technological catch-up. BE J Macroecon (Topics) 6:15
Mark N, Sul D (2003) Cointegration vector estimation by panel DOLS and long-run money demand. Ox Bull Econ Stat 65:655–680
Pesaran MH (2007) A simple panel unit test in the presence of cross section dependence. J Appl Econom 22:265–312
Romer P (1990) Endogenous technological change. J Polit Econ 98:S71–S102
Sapir A, Aghion P, Bertola G, Hellwig Pisani-Ferry J, Rosati D, Vinals J, Wallace H (2004) An agenda for a growing Europe—the sapir report. Oxf Univ Press, Oxford
Schankerman M (1981) The effects of double-counting and expensing on the measured returns to R&D. Rev Econ Stat 63(3):454–458
Sterlacchini A (2004) Ricerca e alta tecnologia in Italia: le basi per un rilancio. L’Industria 4:675–706
Sterlacchini A, Venturini F (2011) R&D and productivity in high-tech manufacturing: a comparison between Italy and Spain. MPRA paper no. 30048
Van Ark B, Melka J, Mulder N, Timmer M, Ypma G (2002) ICT Investments and Growth Accounts for the European Union 1980–2000. Groningen Growth and Development Centre, Research Memorandum GD-56, revised March 2003
Van Beveren I (2010) Total factor productivity estimation: a practical review. J Econ Surv, doi:10.1111/j.1467-6419.2010.00631.x
Venturini F (2004) The determinants of Italian slowdown: what do the data say? EPKE Working Paper n. 29, NIESR, London
Westerlund J, Edgerton D (2008) A simple test for cointegration in dependent panels with structural breaks. Oxf Bull Econ Stat 70:665–704
Wieser R (2005) Research and development productivity and spillovers: empirical evidence at firm level. J Econ Surv 19:587–621
Zachariadis M (2004) R&D-induced Growth in the OECD? Rev Dev Econ 8:423–439
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1 Data sources
ISTAT, Statistiche sulla ricerca scientifica: Business total Research and Development expenditure by industry (billions of liras in current prices). Reference period: 1967–1987.
OECD ANBERD, R&D expenditure in industry (ISIC. Rev.2) Vol. 2002 Release 1 (millions of national currency at current prices). Reference period: 1973–1980.
OECD, Structural Analysis (STAN) database, 2005: Industry Value Added (at current and constant prices), Gross Fixed Capital Formation (at current and constant prices), Total Employment (employees and self-employed, in thousands), Full-Time Equivalent Employment (available for some countries only), Labour Compensation (at current prices), Net and Gross Fixed Capital (at constant prices). Reference period: 1980–2002.
OECD, Science and Technology Database, 2006: Business total Research and Development expenditure by industry (millions of national currency in current prices). Reference period: 1981–86 and 1987–2002.
Appendix 2 R&D capital stock for selected industries
Appendix 3 Supplementary analysis of the Italian case
In this section the raw measure of TFP employed in previous estimates is adjusted for the degree of utilisation and the quality changes of labour and capital. This is extension is performed exclusively for the Italian case due to the possibility of collecting the required industry data for this country only.
With respect to labour input, instead of employed people, we use the number of full time equivalent (FTE) employed persons, which is a better (though not the bestFootnote 15) proxy for the effective utilisation of labour. Moreover, we adjust such a measure for quality changes by using the white/blue collars’ ratio at industry level, taken from the surveys on Italian manufacturing firms carried out by Mediocredito Centrale-Capitalia.Footnote 16 This correction is aimed at avoiding the risk that the long-run coefficient of R&D might pick up the increase in the workforce skills occurred during the period considered. Our quality-adjusted measure of labour is computed by augmenting the number of FTE employed persons (\( \tilde{L} \)) with the share of white collars on total employment (w):
This correction implies the assumption that white collars have a marginal productivity two times higher than that of blue collars. Although, this hypothesis seems quite strong at a first sight, it must be stressed that the estimate of the long-run R&D elasticity does not significantly change when a different productivity ratio is applied (1.5 or 3 instead of 2). It should be added that this kind of correction is more suitable for the Italian case than that implemented, for instance, by Griffith et al. (2004). Under the hypothesis of perfectly competitive labour markets, they obtain a quality-adjusted measure of labour input by weighting each group of workers with the corresponding wage bill:
where W and B are respectively the number of white- and blue-collar FTE workers, S and 1-S their wage bills. A more elaborated procedure is that used by Brandolini and Cipolline (2001) who computed a quality-adjusted measure of labour input by distinguishing Italian workers in five educational levels, weighted by their respective average wages. For the whole Italian economy, the inclusion of such a measure reduced substantially the contribution of TFP to the growth of value added over 1977–2000.
However, at least until the early Nineties, the high bargaining power of the Italian trade unions did not allow that the wages of white collars were too far from those of blue collars; as a consequence, since the changes in relative wages were modest and not particularly affected by those in relative productivity, the correction proposed by the above authors may produce misleading results.Footnote 17
Moving to capital input, as in Griffith et al. (2004), our measure of capital adjusted for its degree of utilisation is
where Y i,t is the industry valued added at constant prices and \( {\hat{Y}_{i,t}} \) is the fitted value which arises from regressing, for each separate industry, the real value added on a constant and a time trend.
Following Jorgenson et al (1987), an indicator of capital quality (q) has been obtained from the newly available industry series on investment types released by ISTAT (see Appendix 1), computed as the ratio between the capital stock evaluated respectively at rental (R) and market prices (p):
where, in this case, the suffixes i and j denote respectively industries and types of asset,Footnote 18 r is the nominal rate of return, δ denotes the asset specific depreciation rate (obtained as described in Section 3), and π is the rate of change of the investment prices.
The nominal rate of return is estimated internally from ISTAT industry accounts by recurring to the following (ex-post) formula:
which is based on the Gross Operating Surplus (GOS) reduced by the remuneration of self-employed workers.
To be stressed is that the difference between R and p reflects the substitution towards assets with relatively high marginal products as they are characterized by a relatively high physical deterioration and economic obsolescence. This adjustment should avoid that θ captures the qualitative growth of tangible capital, fuelled by the rising adoption by manufacturing firms of technologically advanced capital as computers, communication equipment and software. Accordingly, our measure of quality-adjusted capital is:
Summing up, the adjusted measure of TFP (in logs) is given by:
By using the adjusted measure of TFP, Table 8 shows that the estimated elasticities of R&D capital do not substantially change from those reported on Tables 5, 6 and 7 (Section 5). For the period 1970–2002, the results are almost identical and point to a low impact of R&D. By dropping out the last ten years, the log-run effects of R&D on the adjusted TFP are slightly lower than those arising with a measure of unadjusted TFP. In any case, they are significantly higher than that found for the entire period under examination, suggesting that the low R&D impact is a reflection of the declining R&D efforts recorded during the last decade.
Rights and permissions
About this article
Cite this article
Sterlacchini, A., Venturini, F. Boosting Manufacturing Productivity Through R&D: International Comparisons with Special Focus on Italy. J Ind Compet Trade 13, 187–208 (2013). https://doi.org/10.1007/s10842-011-0114-5
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10842-011-0114-5