The joint impact of different types of innovation on firm's productivity: evidence from Italy

ABSTRACT

This paper explores the firm-level relationship between product, process, organizational and marketing innovation activities and firm productivity. We propose a structural model that relates R&D decisions, innovation activities, and productivity by using a version of the model developed by [Crépon, B., E. Duguet, and J. Mairesse. (1998). “Research, Innovation and Productivity: An Econometric Analysis at the Firm Level.” Economics of Innovation and New Technology 7: 115–158] and empirically analyze the drivers of firms’ innovation strategies as well as which combination has effects on firm's economic productivity. Results show that R&D expenditures are an important predictor of all types of innovation as well as an important indirect driver of firm productivity through innovation activities. Both process and product innovation have positive effects on firm's economic productivity, especially when they are jointly conducted. Organizational activities are beneficial also for other types of innovation and especially for process innovation. The introduction of a new product on the market may raise productivity if complemented by marketing innovations. Results are driven by firms that have invested the most in new equipment and machinery.

KEYWORDS:

• Innovation strategy
• product innovations
• process innovations
• marketing innovations
• organizational innovations

JEL CODES:

• O31
• O32
• D22
• L20

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Complementarity between product and process innovations has been found correcting for unobserved individual time-invariant heterogeneity (Martinez-Ros and Labeaga 2009).

2 The study of complementarity, previously focused on process and product innovations, has been extended to organizational innovations even though full complementarity is never obtained and the complementarity between pairs of strategies depends on the country examined (Ballot et al. 2015).

3 Using microdata on individual enterprises, several variants of the CDM model have been estimated for different countries. Among others, see Lööf and Heshmati (2002) for Sweden; Janz, Lööf, and Peters (2004) for Germany and Sweden; Griffith et al. (2006) for Germany, Spain, the UK and France; Parisi, Schiantarelli, and Sembenelli (2006), Hall, Lotti, and Mairesse (2009) and Conte and Vivarelli (2014) for Italy; García-Quevedo, Pellegrino, and Vivarelli (2014) for Spain.

4 When endogeneity or selectivity are not corrected for, the significance of the estimated parameters drops, pointing to an error in variables problem, probably related to the subjectivity of the answers to some of the questions that generated the data, rather than a simultaneity problem (Mairesse, Mohnen, and Kremp 2005). When the endogeneity and selection are taken into account for, the results are rather robust to different estimation methods (Mohnen and Hall 2013).

5 For further development on the CDM literature by linking the firm's R&D investment, innovation and productivity in a dynamic framework see Aw, Roberts, and Xu (2011) and Peters, Roberts, and Vuong (2017).

6 Even though Hall, Lotti, and Mairesse (2009) and Musolesi and Huiban (2010) do not report a great difference in the estimation results when comparing a sequential IV estimation with a maximum likelihood estimation approach.

7 Several versions of this model have been used in the literature such as, among others Hall, Lotti, and Mairesse (2009, 2013).

8 As there is no direct question asking whether the firm has decided to invest in R&D or not, we indirectly assume that those firms with positive values of R&D investment have decided to conduct R&D.

9 The inclusion of the predicted R&D intensity in the regressions accounts for the fact that all firms may have innovative efforts but only a few of them may actually report it (Griffith et al. 2006). Moreover, using the predicted value instead of the realized value is a sensible way to instrument the innovative effort in the knowledge production function in order to deal with the simultaneity problem between R&D and the expectation of innovative success (see Hall, Lotti, and Mairesse 2013).

10 Differently from physical capital which is introduced in the regression in log, we use the variables human and ICT capital in levels as otherwise we would have too many missing values (taking the log, firms with zero human and ICT capital would turn into missing observations).

11 We do not have information on the share of sales coming from innovation in the total amount of turnover (see Raymond et al. 2010). Therefore, we use sales per employee as a measure of firm productivity (see among others, Hall, Lotti, and Mairesse 2009, 2013).

12 Pavitt's taxonomy is a classification of the product sectors based on the sources and the nature of technological opportunities and innovations, the intensity of research and development (R&D intensity), and the type of knowledge flows (see Pavitt 1984). On the basis of the above-mentioned criteria, Pavitt identified four large sectoral groupings: (PAVITT 1) Supplier dominated – ‘dominated by suppliers’ – which includes: textiles (textiles); footwear (footwear); food and beverage sectors (food and beverages); paper and printing (paper and printing); timber (wood). (PAVITT 2) Intensive scale – ‘scale-intensive’ – which includes: base metals (basic metals); motor vehicles and related engines (motor-vehicles, trailers, and semitrailers). (PAVITT 3) Specialized suppliers – ‘specialized suppliers’ – which includes: agricultural and industrial machinery (machinery and equipment); office machines (office, accounting, and computing machinery); optical, precision, and medical instruments (medical, precision, and optical instruments). (PAVITT4) Science based – ‘science based’ – which includes: chemistry (chemicals); pharmaceutical (pharmaceuticals); electronics (electronics). Each grouping is considered characterized by internal regularities regarding: the potential sources of innovation; the type of innovations; their degree of appropriability; the height of barriers to entry; the average size of the companies.

13 The robustness using firm's size, their location as well as Pavitt sectors reports the estimates when innovation activities are proxied by the predicted probabilities of introducing jointly process and product innovations, at least one of the activities (either process or product), organizational and marketing innovation. Results when using predicted probability of introducing only process or product innovation as a proxy for innovation activities as well as other combinations when using marketing and organizational innovations are available on request.