The persistence and cross-persistence of R&D outsourcing: onshore and offshore strategies

ABSTRACT

Research and Development (R&D) outsourcing has rapidly grown in recent decades but little attention has been paid to the persistence with which these contracts are carried out. In this way, this study contributes to the literature by providing new evidence on the extent of true state dependence in explaining the persistence of outsourcing R&D to a type of contractor. Further, we explicitly distinguish between offshore and onshore outsourcing strategies and study their own- and cross-state dependence. In doing so, we estimate a set of dynamic models of strategy adoption using a panel of Spanish manufacturing firms from 2005 to 2014. The data allow for the distinction between five types of contractors and their domestic or foreign origin in each of them. Our results support the idea that R&D outsourcing with a type of contractor exhibits true-state dependence although moderate, with onshore R&D outsourcing being more state-dependent than its offshore counterpart. Results also indicate the existence of cross-state dependence in both directions.

KEYWORDS:

• R&D outsourcing
• onshore
• offshore
• true state dependence
• cross-state dependence

JEL CLASSIFICATION:

• O32
• F61
• D23
• L60

Acknowledgements

I am grateful for the comments and suggestions received from the editor and three anonymous reviewers. I acknowledge financial support from Grant ECO2017-86793-R funded by MCIN/AEI/ 10.13039/501100011033 and by ‘ERDF A way of making Europe’, and from Grant PROMETEO/2019/095 funded by Generalitat Valenciana. Usual caveats apply.

Disclosure statement

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

Notes

1 We use the term R&D outsourcing – or external R&D – as an indication of contractually paid R&D performed by an external independent knowledge provider (Grimpe and Kaiser 2010).

2 There is evidence that pharmaceutical firms increasingly outsource specific R&D activities that range from basic research to late-stage development, including genetic engineering, target validation, and clinical trials (Morton and Kyle 2012). Likewise, automakers outsource most of the R&D activities formerly done internally to external contractors (Takeishi 2001).

3 While, to our knowledge, there is no prior evidence of the extent of persistence in R&D outsourcing, there are few studies that from a theoretical perspective refer to the dynamic aspect of R&D outsourcing. For instance, Hsuan and Mahnke (2011) and Kamuriwo and Baden-Fuller (2016) stress the role of learning in contractual relationships.

4 Dibbern, Winkler, and Heinzl (2008) argue that firms incur extra costs in managing an offshore project, including post-contractual extra costs in requirement specification, control, and coordination, that increase with the geographic and psychic distance between the residence of the vendor and the client. Larsen, Manning, and Pedersen (2013) discuss that firms additionally incur hidden costs in the implementation of offshore outsourcing strategies due to the complexity resulting from reconfiguring the firm’s value chain.

5 See https://www.oecd.org/sti/rd-tax-stats-spain.pdf for a detailed description of the Spanish R&D tax scheme. According to OECD estimates, the implied tax subsidy rates for Spain are among the highest within OECD countries, despite the decline in generosity of the R&D tax credit regime since 2008.

6 PITEC is sponsored by Fundación Española para la Ciencia y la Tecnología (FECYT) and the Foundation for Technical Innovation (COTEC). Details on the survey can be found at http://icono.fecyt.es/PITEC.

7 Due to the enlargement of the initial sample in 2004, we start the analysis in 2005.

8 We do not consider as outsourcing purchases of R&D from an affiliated foreign subsidiary.

9 This type includes domestic technological centers in the case of onshoring; or other international organizations in the case of offshoring.

10 This approach has been widely used in studies assessing the persistence in innovation activities or in alliances (Badillo and Moreno 2016; Ganter and Hecker 2013; Mañez and Love 2020; Triguero and Córcoles 2013).

11 The alternative would be to use a fixed effects (FE) specification, in which α_ij are treated as parameters to be estimated. However, standard FE versions of nonlinear models, such as the logit and probit models, suffer from the incidental parameter problem, which leads to biased estimates (Albarran, Carrasco, and Carro 2019).

12 Instead of q_i  the original estimator used q_i = (q_i1, …, qx_iT) in equation (3), but time–averages are allowed to reduce the number of explanatory variables (Wooldridge 2005). Moreover, q_it is typically a subset of the time varying variables in x_it.

13 In the probit model framework it is infeasible to estimate a fixed effects model. Thus, the unobserved heterogeneity specified in equation (3(3) ${\alpha }_{ij}={\delta }_0+{\delta }_1{y}_{ij0}+{\delta }_2{\overline{q}}_i+u_{ij}$(3) ) is a compromise between fixed effects and random effects models.

14 Mañez and Love (2020) consider that there are only two sources driving state dependence in R&D: sunk costs and learning effects.

15 We thank a reviewer for suggesting this approach to distinguish the effect of success-breeds-success from that of sunk costs.

16 The TC literature considers collaboration to be a ‘hybrid’ between outsourcing and vertical integration, while other scholars regard collaboration as a strategy in its own right (Añón Higón 2016). In contrast to collaboration, outsourcing implies a more passive role by the focal firm in the generation of knowledge.

17 A detailed description of the variables is provided in Table A1 in the appendix.

18 The approach used by Semykina (2018) differs from the Wooldridge (2005) approach in that, instead of using the within means of all time varying variables in x_it, it takes only the time means of a subset of exogenous variables (q_it) that are theoretically more likely to be correlated with the unobserved individual effects (a_ij).

19 Not distinguishing by type and origin of the R&D contractor, we obtain that a firm outsourcing R&D in t-1 has, on average, a 25 percentage points greater probability to outsource R&D in the current period, relative to one that did not outsource. Results are available upon request.

20 Though all firms (innovators and non-innovators) are asked whether they acquire external R&D or not, there are questions in PITEC that are only answered by innovators; for instance, questions related to cooperation with external agents. Thus, the inclusion of these censored variables may lead to a sample selection bias.

21 We refer to the reader to the online Appendix for a detail discussion of these approaches and the results.

22 Since our variable is lagged outsourcing, we additionally ran the model for the total sample but included an interaction term between lagged outsourcing and a dummy equal to 1 for small firms, and zero otherwise. The results obtained show that this interaction term is negative and significant, which leads us to conclude that small firms are less persistent in R&D outsourcing.

23 Convergence issues may arise given that outsourcing to specific types of contractors (e.g., government agencies) is a very rare event (see Table 1). Therefore, we group outsourcing to government agencies, private NPOs, and other organizations in a single category called ‘outsourcing to other non-profit organizations'.

Additional information

Funding

This work was supported by Generalitat Valenciana: [Grant Number PROMETEO/2019/095]; Ministry of Economy and Competitiveness: [Grant Number ECO2017-86793-R].