Innovation, complementarity, and exporting. Evidence from German manufacturing firms

Abstract

This paper assesses whether there might be complementarities between different types of innovation activities (product, process and organizational) and how these effects may be linked to the likelihood that a firm will export. Complementarity is addressed through the properties of supermodular functions, and firm heterogeneity by export destination is explored. A new econometric strategy to test for pairwise complementarity in a function with three independent variables and a binary dependent variable is proposed. Exogenous and endogenous innovation variables are considered by using bootstrapping for hypothesis testing, propensity score matching and treatment effects models. The empirical analysis shows that complementarity relationships between innovation strategies are more likely to exist when firms export to multiple foreign markets.

Keywords:

• Export propensity
• complementarity among innovations
• multiple hypothesis testing
• binary choice model

JEL classification:

• C12
• C25
• F14
• O31

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Aw, Roberts, and Winston (2007), Aw, Roberts, and Xu (2008), Bustos (2011), Lileeva and Trefler (2010), Van Long, Raff, and Stahler (2011).

2. Caldera (2010), Cassiman, Golovko, and Martínez-Ros (2010), Damijan, Kostevc, and Polanec (2010), Van Beveren and Vandenbussche (2010), Cassiman and Golovko (2011), Becker and Egger (2013).

3. As organizational/marketing innovations concern new organizational or marketing methods, they may play a role both in the supply of products on foreign markets and in the productive process.

4. On this point, see Melitz (2003).

5. See Wagner (2007) and Becker and Egger (2013) for detailed references.

6. Still in a theoretical setting, the formal model of Athey and Schmutzler (1995) shows that the firm’s short-run choices about process and product innovations are complements and that this in turn leads to complementarities between the firm’s long-run decisions on process and product flexibility.

7. Among others, see Veugelers and Cassiman (1999), Bocquet, Brossard, and Sabatier (2007), Cassiman and Veugelers (2006), Cozzarin and Percival (2006), Percival and Cozzarin (2008), Gomez and Vargas (2009), and Schmiedeberg (2008).

8. ‘[…] the two types of innovations should be seen as “brothers” rather than “cousins”’ (Reichstein and Salter 2006, 677).

9. Firms self-select into export markets if their productivity is high enough to afford the entry costs and competition of the export market. For an exhaustive literature review on this subject, see Bernard et al. (2007) and Wagner (2007).

10. See, for instance, Yeaple (2005), Bustos (2011), Constantini and Melitz (2007) and Lileeva and Trefler (2010).

11. Through both the production of new and the significant improvement of existing goods or services (Mairesse and Mohnen 2010).

12. ‘A firm must find and inform foreign buyers about its product and learn about the foreign market. It must then research the foreign regulatory environment and adapt its product to ensure that it conforms to the foreign standards (which include testing, packaging, and labeling requirements). Firms can export their products to any country although entry into each of these export markets requires a fixed investment cost’ (Melitz 2003, 1706).

13. They involve variations in workplace organization, in business practices or in firm external relations (Mairesse and Mohnen 2010).

14. This way of dealing with process innovation is in line with Reichstein and Salter (2006).

15. Following Topkis (1995, 1998), Milgrom and Roberts (1990, 1995), and Milgrom and Shannon (1994), two variables x′ and x″ in a lattice X are complements if a real-valued function F(x′, x″) on the lattice X is supermodular in its arguments. That is, if and only if: .

16. where k = 1, 2, … , K denotes the kind of practice.

17. In Cassiman and Veugelers (2006), devoted to the empirical analysis of complementarity between the firm’s internal R&D and external knowledge acquisition, great emphasis is given to the analysis of the contextual variables affecting the supermodularity of the performance function that allows one to understand the conditions under which innovation strategies are complementary.

18. The first condition is met if the third innovation practice is 0; the second, if the third practice is 1.

19. The first condition is met if the second innovation practice is 0; the second, if the second practice is 1.

20. The first condition is met if the first innovation practice is 0; the second, if the first practice is 1.

21. Following Amir (2005, p. 656), a function f: S × A → Rhas the single-crossing property in (a, s) if : and .

22. For a survey, see Nichols (2007).

23. The results are available upon request.

24. Actually, in our sample, besides exporting to only EU markets or to both EU and non-EU markets, firms can choose to export to only non-EU markets.

25. New organizational methods involve changes in workplace organization, external relations and business practices. New marketing methods concern changes in product promotion, pricing, design, packaging and placement.

26. Aw, Roberts, and Xu (2008) show that the firm’s observable revenue share is strictly linked to a theoretical measure of relative firm profitability in a dynamic model of exporting that shares many features with Melitz (2003) and Constantini and Melitz (2007). Such relative profitability depends on firm’s productivity level, capital stock, mark up and return to scale parameters.

27. The combined test indicates complementarity when z* is the critical value is the following:

.

Given a significance level for the combined hypothesis of α = 0.05, the critical value is z* = 2.241 given that we test 4 constraints. As for the substitutability test, we evaluate the cases in which the following conditions are satisfied:

28. Following Savin (1980), let L be the set of linear combinations of b₀₀₀, … , b₁₁₁ and f_i be every linear combination in L, i = 1, … , m. An induced test can be used to test H₀: f₁ = f₂ = … = f_m = 0 against the alternative H₁: f₁ > 0, f₂ > 0, … f_m > 0. Given a significance level for the combined hypothesis of αα and a total of m constraints, the Bonferroni procedure suggests a significance level for separate hypotheses of αα/m, with H₀: f_i = 0 against the alternative H₁: f_i > 0, i = 1, … , m. When α = 0.05, the Normal critical value for each hypothesis is z_α/2 = 1.96 with 2 constraints and z_α/4 = 2.241 with 4 separate tests.

29. For the other two innovation pairs the following specifications are estimated:

for product and organizational/marketing innovation and.

for process and organizational/marketing innovation.

30. The complementarity hypothesis cannot be rejected when , where z* = 1.96. As for the substitutability test, we evaluate when the following conditions are satisfied: .