R&D spillovers and productivity growth: evidence from Italian manufacturing microdata

Abstract

Recently the focus of R&D impact on productivity moves from internal to external innovation. This paper assesses the link between R&D spillovers and productivity of 1017 Italian manufacturing firms over 1995–2000. Different proxies of R&D spillovers and several specification tests are used to select a well-behaved fixed effect model. As predicted by various new growth theory models, external R&D exerts a positive and highly significant effect on firms’ productivity growth.

Acknowledgement

Earlier versions of this paper have been presented at the International Workshop ‘Empirical Studies on Innovation in Europe’, Urbino (Italy), 1–2 December 2003 and at the 2004 Annual Conference of the Scottish Economic Society, Perth (Scotland), 30 March–1 April 2004. We are grateful to Vincenzo Scoppa and to seminar and conference participants for comments and suggestions. Responsibility for any remaining omissions and errors is solely ours.

Notes

 Mairesse and Sassenou (1991) reviewed much of the relevant empirical literature addressed to estimate at firm level the rate of return of R&D investments and the R&D capital elasticity.

 At macro or industry level, technological spillovers result in being an influent factor for productivity increases (see, i.e., the seminal papers by Griliches and Lichtenberg, 1984; Bernsteinn and Nadiri, 1988; Romer, 1990).

 Weights ω _rs have been calculated using the matrix of interindustry flows of innovations R, whose component R _rs indicates the amount of R&D expenditures of sector r that the sector s uses as input in order to satisfy a unit increase of its final demand. Following Momigliano and Siniscalco (1982) and Leoncini and Montresor (2001), matrix R can be expressed as R = rB, where r is the diagonal matrix of sectoral investment R&D and B = (x)⁻¹ Ld is an operator indicating flows in direct and indirect final demand. x and d are, respectively, diagonal matrices of production and final demand, while L = (I−A)⁻¹ is Leontief 's inverse matrix. Every element α _ij in L measures the amount of output that the ith sector ought to produce in order to allow a one unit increase in sector j. Since matrix R is biased by scale effects, the weighting system has been constructed dividing every element R _rs to the sum by column. In such a way, the matrix Ω has been derived, whose generic element ω _rs indicates the share of R&D expenditures realized in sector r but used in sector s.

 The first cleaning rule concerns the availability of positive value added for at least three years over 1995–2000. Furthermore, the presence of influential outliers in levels or growth rates for every economic relevant variable for the analysis (output, physical capital, R&D capital, labour) has been tested by considering the interval ME ± 6SSM, where ME is the median of the distribution of a variable (expressed in levels or in growth rates) and SSM is the average deviation from the median. Data in level or in growth rates outside this interval have been defined as outliers and excluded from the sample.

 In the procedure for determining R&D capital firm-by-firm, the first year chosen is 1981. Since the data set covers the period 1995–2000, the R&D capital at time t ₀ = 0 (t ₀ = 1981), has been derived by imposing that the growth of R&D investments of ith firm is that observed at sector level. Finally, the rate of obsolescence of R&D capital is supposed to be 15% per year.

 Weights are given by

, where VA _it is the value added of the ith firm at time t (t = 1995, … , 2000) belonging to a group sized N (i = 1, … , N).

 By considering the group of innovators, the value added to employees is 31.2 in the North West of Italy, 27.8 in the Centre, 26.8 in North East and, finally, 25.5 in Southern Italy. This ranking changes when non-innovators are taken into account, since, in this case, the ratio value added to employees is higher for the Centre of Italy (41.2) and lower for the North (about 27.5).

 Spillovers are intra or interindustry. For the ith firm operating in the sth industry, R&D intraindustry spillovers are expressed as

, where ω _ss is the share of technology produced and used inside the sth sector and R_s is the stock of R&D capital in sector s net of the stock of the ith firm. Similarly, interindustry spillovers of firm i are given by the weighted sum of R&D stock existing in sector others than s, that is

.