The importance of absorptive capacities: productivity effects of international R&D spillovers through intermediate inputs^†

† This paper is based on a Background Study for the European Competitiveness Report 2013. The work was done for the European Commission, DG Enterprise and Industry, under the Framework Contract ENTR/2009/033. The opinions expressed are those of the authors only and do not represent the Commission’s official position.

ABSTRACT

Trade in goods and services is likely to be an important channel for international knowledge diffusion. This paper considers the extent of R&D spillovers through intermediate inputs for a sample of up to 40 developed and developing countries at the industry level. Results suggest that such spillovers are present and are economically important. We find that countries and industries initially further behind the technological frontier enjoy stronger foreign R&D spillovers. Furthermore, foreign R&D spillovers are stronger in countries with greater absorptive capacity as measured by average years of secondary schooling and R&D spending.

KEYWORDS:

• R&D spillovers
• intermediates trade
• productivity

JEL CLASSIFICATION:

• F15
• O14
• O19

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

† This paper is based on a Background Study for the European Competitiveness Report 2013. The work was done for the European Commission, DG Enterprise and Industry, under the Framework Contract ENTR/2009/033. The opinions expressed are those of the authors only and do not represent the Commission’s official position.

1. The share of R&D financed by enterprises in advanced countries was 98% in the 1980s and 94% in the 1990s (UNIDO 2002). Even within developed countries, however, R&D is concentrated, with Eaton and Kortum (1999) noting that in the late 1980s, 80% of OECD research scientists and engineers were employed in five countries (US, UK, Germany, Japan, and France).

2. In their preferred specification the stock of foreign knowledge is interacted with the overall import share to take account of the level as well as the distribution of imports.

3. This outcome is not replicated when patent count data are employed, however. Eaton and Kortum (1996) find only limited evidence of a role for imports in facilitating technology diffusion among OECD countries.

4. Engelbrecht (1997a) tests the robustness of the results on the R&D spillover variable to the inclusion of a general human capital variable and a catch-up term. He finds that their inclusion reduces the coefficient on the R&D spillover variable by around 30%. Lichtenberg and van Pottelsberghe de la Potterie (1998) argue that there is an aggregation bias in the construction of the R&D spillover variable and propose an alternative that removes this bias. Results using this alternative still find trade to be an important channel of R&D spillovers.

5. A related strand of literature also points towards the importance of institutional factors as e.g. emphasised in Parente and Prescott (1994, 1999, 2003); Rosenberg and Burdzell (1986), Mokyr (1990) and Coe, Helpman, and Hoffmaister (2009).

6. By doing so, R&D present in final imports is not considered.

7. A major reason for focussing on this first-difference specification is that the levels of the variables included in the regression model tend to be non-stationary, while the first differences tend to be stationary.

8. Specifically, OECD reports R&D expenditure in 2000 prices and 2000 PPP dollars.

9. A list of industries is provided in Appendix Table A.1.

10. Around 10% of observations had to be imputed due to missing values for particular years. Where data was not available we used linear interpolation to fill in the missing numbers.

11. For a detailed descriptive analysis, see Foster-McGregor, Pöschl, and Stehrer (2014).

12. For more detailed descriptive results see the working paper version Foster-McGregor, Pöschl, and Stehrer (2014).

13. See http://www.barrolee.com/. This data has been used as a measure of absorptive capacity in similar studies (see for example Falvey, Foster, and Greenaway 2007).

14. In this paper we do not separate the ‘own industry effect’ of R&D or the impact of the R&D stock itself as the focus is on international spillovers.

15. In the subsequent regressions only the ‘direct and indirect’ measure is included; other results are reported in the working paper version Foster-McGregor, Pöschl, and Stehrer (2014).

16. The time, country and industry fixed effects are jointly significant.

17. In the case of R&D expenditure, this may again be due to the fact that the level of R&D expenditure is missing for a large number of observations, with the missing observations being replaced by $1000. The lack of variation in this variable for such observations may lead to an insignificant coefficient on the variable. Many previous studies have found either an insignificant or even negative coefficient on indicators of schooling in growth regressions (e.g. Pritchett 2001).

18. Falvey, Foster, and Greenaway (2007) present some evidence indicating that the impact of foreign R&D on labour productivity differs according to the degree of relative backwardness.