R&D vs. Other Factor Inputs in a High-Tech Industry

Abstract

According to knowledge-based growth theories, catch-up with the global technology frontier calls for a shift from physical investment to innovation. This prevailing premise in Finland is tested with an unbalanced panel of firms in Finnish information and communications technology (ICT) manufacturing, over a period of rapid growth, 1990–2003. Stochastic frontier model estimation results are overwhelmed by scale elasticity associated productivity growth. Contrary to beliefs, R&D productivity was relatively low, regardless if measured by efficiency impacts of R&D intensity, technical change or R&D elasticity of output. Results are consistent with industry outcomes and reveal internal causes to the subsequent downfall of ICT manufacturing in Finland.

Keywords:

• Scale
• technical change
• efficiency
• R&D
• size

Notes

 ¹ Though per capita income would be slightly above average, labor productivity would be clearly below.

 ² In 2001, for example, the International Institute of Management Development (IMD) rated Finnish technological infrastructure as the third best in the world. The United Nations Development Programme (UNDP) has considered Finland as the most technologically advanced country in the world and in 2001 the World Economic Forum (WEF) rated Finland first in its growth and current competitiveness rankings. In 2002, the IMD placed Finland second in its competitiveness ranking, while the WEF ranked Finland third in its technology index and first for public institutions. Top rankings have persisted, although with a declining trend over the 2000s, for example, to sixth (WEF) place in 2008 or to ninth place (IMD) in 2009. See also Sabel and Saxenian (2008).

 ³ SIC 32 classification.

 ⁴ Defined as firms with less than 200 employees operating in high-tech sectors.

 ⁵ Radical (breakthrough or revolutionary) innovations render existing technological knowledge, competences or products obsolete. In contrast, incremental innovations are small, with gradual improvements that build on and strengthen existing knowledge and competences.

 ⁶ For instance, McGahan and Silverman (2001) found no proof of a decline in patent applications, or of a shift from product to process innovation, or of market leaders playing more important roles as industries matured.

 ⁷ Coelli et al. (1999).

 ⁸ If the official categorization shifted back and forth between services and manufacturing, a firm was not excluded nor was the time series interrupted, provided that the firm was categorized in manufacturing at the beginning and end of the sample period.

 ⁹ LDPM/Teollisuustilasto.

¹⁰ Tilinpäätöspaneeli.

¹¹ After the exclusion of firms with less than 20 employees, the Statfin longitudinal database covers 35–70 per cent of all firms in the period (Maliranta, 2003: 295).

¹² Since external R&D expenditure has a one-time impact on the company knowledge stock and ranged at about 5 per cent when available at all, it was excluded.

¹³ The use of a stock measure raises the R&D input too high only if the R&D depreciation rate is too low. Hence, the (high) 30 per cent depreciation rate applied to the knowledge stock is appropriate, although asymmetric and imperfect information on future returns make it impossible to value it precisely. In contrast to the R&D as an investment view, a flow measure equals 100 per cent depreciation of knowledge acquired over the period.

¹⁴ In 2004, the sector employed more personnel in subsidiaries abroad than in Finland as a consequence of outsourcing, but the sample figures include only activities in Finland.

¹⁵ For 1998–2000, the R&D panel suffers from several missing observations, overcome by imputing two-year averages. This is justifiable by the high serial correlation of R&D relative to other investment, since firms avoid laying off R&D employees during downturns (Hall, 2002).

¹⁶ The depreciation of the R&D stock (δ) is often fixed arbitrarily at 10 or 15 per cent, which is consistent with the estimates obtained by Bosworth (1978) on the basis of patent renewal data. However, Pakes and Schankerman (1984) estimate an average rate of 25 per cent, also from patent renewal data, and recently Bernstein and Mamuneas (2006) estimate industry-specific rates that range from 18 per cent for chemicals to 29 per cent for electrical products.

¹⁷ Size classes 1 and 2, and age classes proved statistically insignificant in OLS regressions.

¹⁸ Statistics Finland, Science, Technology and Information Society Statistics, 2005.

¹⁹ Statistics Finland, Quality Description: R&D, 2005.

²⁰ The search for new Nokias has been a constant feature of the Finnish public debate.

²¹ A recent study has shown that 39 per cent of the value-added of a standard Nokia phone (N95) has on average been produced in Finland (Pajarinen et al., 2010).

²² Relative to success stories such as Hong Kong or Singapore, Finland's scanty record in services, its location and taxes promise no leapfrogging.

²³ In 2008, though well above the total economy average, the value-added per employee in ICT services (sector J) relative to ICT manufacturing (sectors 26–27) was less than 60 per cent.

²⁴ Nokia and Microsoft cooperation plans (published in winter 2011) have made Nokia R&D stocks and technologies obsolete to the extent they have been replaced by Microsoft technology. Such a downgrading of the largest accumulated Finnish R&D stock would obviously slash future returns to R&D. Moreover, the destruction may not be altogether creative, since R&D stocks take the physical form of researchers whose highly specialized expertise is not necessarily useful to other industries.

²⁵ Approaches favored by the national technology agency, Tekes, in its reporting.