Direct and indirect knowledge spillovers and industrial productivity

Abstract

This paper analyses the importance of externalities related to the spread of innovation across sectors. Such spillover effects arise from R&D activities and input–output (IO) linkages among sectors in the country. We borrow Spatial Econometrics techniques to make consistent estimates of the impact of these systematic direct and indirect spillovers on sector’s productivity and the possibility of other types of productivity spillovers in the error term. We find that direct spillovers emanating from IO horizontal linkages determine sector’s productivity, while the indirect effects prove to be negligible. Furthermore, the technological intensity of IO linkages and the productive structure of the underlying economy are key factors determining the effectiveness of economic policies focused on increasing total industrial productivity.

Keywords:

• Direct knowledge spillover
• indirect knowledge spillover
• industrial productivity

JEL classification:

• C21
• C67
• O33

Notes

¹ The ‘Annual Growth Survey: Summary of the economic analysis and messages’ 12 January 2011 (http://ec.europa.eu/europe2020/tools/minitoring/annual_growth_survey_2011/index_en.htm).

² These studies argued that, rather than innovation output in neighbouring regions, it is the research effort made in neighbouring regions that should be regarded as the catalyst generating spatial spillovers on innovation activity in the target region (i.e. Moreno, Paci, and Usai 2005).

³ Subindex C indicates that the corresponding variable is a column vector including all its sectors’ values. County c’s matrix W_c can be written in terms of column vectors: . Considering that there are zero values in its principal diagonal, Wc will be where w_ijc, is the generic element for i and j sectors in the country c.

⁴ Terleckyj (1974) provides an estimation of inter-industry technology flows by constructing a technology flow matrix. The relevance of this inter-industry technology flow is analysed in Schmookler (1966).

⁵ Funk (2001) finds that bilateral exports used as weights perform as well as bilateral imports and so we take this approach.

⁶ O’Mahony and Timmer (2009).

⁷ Data on R&D stock for market services are provided for the whole sector in the EU KLEMS database.

⁸ If we assume that on average each person engaged in production works the same number of hours as the rest, both shares of human capital over the total supply of labour factor will be equal.

⁹ The countries in our sample are: Austria, Belgium, Canada, Denmark, Finland, France, Germany, Italy, Japan, the Netherlands, Spain, Sweden, United Kingdom and United States of America. Sectors included in our sample are: Food prod., beverages & tobacco; Textiles, textile prod., leather & footwear; Wood & prod. of wood & cork; Pulp, paper, paper prod., printing & publishing; Coke, refined petroleum & nuclear fuel, chemicals & chemical prod.; Rubber & plastics products, Other non-metallic mineral products, Basic metals & fabricated metal products, Machinery & equipment NEC; Electrical & optical equipment, Transport equipment; Manufacturing NEC, recycling; Electricity, gas and water supply; Construction and Market & non-market services, post & telecommunication services. NEC acronym stands for ‘not elsewhere classified’. For further details see Table A1 in the Appendix 1.

¹⁰ We focus on the years 1990, 1995, 2000 and 2005 without imposing assumptions about the evolution of IO relationships over the rest of the years.

¹¹ See Griliches (1980) as an example.

¹² We also checked the possible endogeneity of the sectoral-lag of the endogenous variable. We separately estimated robust GMM estimations for each 2005 and 2000 equations in the SURE model and instrument the sector-lagged endogenous variable regressor with its proper t-5 lag. The relevant tests for endogeneity problems and validity of instruments (Durbin–Wu–Hausman endogeneity test and Hansen–Singleton test for orthogonality of instruments) are provided at the bottom of Table 2. All specifications satisfactorily pass both tests.

¹³

¹⁴ We used different non-linear specifications for both the direct and indirect spillover variables. In all cases the results were robust for the non-significance of indirect spillover effects. For the sake of brevity, we do not include these results here, but they are available on request.

¹⁵ We assume that firms’ innovation performance and productivity in each sector depend on firm attributes as well as on the positive externalities from knowledge spillovers within sectors.