Mapping innovative activity using microdata

Abstract

I discuss the determinants of three alternative measures of innovative output among firms of different size. Moreover, I provide an econometric procedure for testing a Tobit model (Tobin, 1958) against a two-part model. This procedure is alternative to Cragg's (1971) test and feasible for weighted survey data.

Notes

¹ This survey, which was conducted by ISTAT (Italian Institute of Statistics) in 2002, is representative of the entire Italian population of firms with more than 10 employees and active at 31st December 2000.

² Sampling weights ensure that each group of firms is properly represented. Moreover, they correct for sample selection (Deaton, 1997, pp. 44–45).

³ R&D expenditure includes both internally-performed (intra muros) R&D and R&D activities outsourced to other firms or research institutes (extra muros R&D). TA expenditure is the sum of embodied technological change (equipment and machinery) and the acquisition of external (disembodied) technology (know-how, projects and consultancies, licenses, software). R&D and TA are both normalized by a firm's total turnover.

⁴ Firms are considered small if they have less than 50 employees; large otherwise.

⁵ An alternative specification test has been proposed by Lin and Schmidt (1984). They derive a Lagrange Multiplier statistics based only on the Tobit model for testing the restriction β₁ = β₂/σ However, this test is also based on the likelihood function of the Tobit model and for this reason it suffers from the same limitations which affect Cragg's test (1971).

⁶ The empirical analysis includes also cooperation agreements with research institutions (CORES), market cooperation agreements (COMKT), belonging to an industrial group (GP), log of a firm's employees (LOG-SIZE) in year 2000 (Acs et al., 1994; Cassiman and Veugelers, 2002 and Piga and Vivarelli, 2004 discuss the influence of these variables on a firm's propensity to innovate). Sectoral and macro-regional dummies are included in the specification only when they are jointly significant.