Understanding the diversity of cooperation on innovation across countries: multilevel evidence from Europe

Abstract

Much has been written about innovation cooperation, but little research has been done to explain the national differences thereof. Using macro and micro evidence from the fourth Community Innovation Survey, we econometrically investigate the extent to which national framework conditions account for the propensity of firms to cooperate on innovation at home and abroad. The results indicate strong differences across countries in the latter. Firms operating in countries with less developed research infrastructure are shown to be more likely to cooperate with foreign partners, supporting the thesis that in this context the foreign linkages tend to be diasporic. The size and openness of the economy matters too. But the characteristics of firms that explain cooperation are not found to differ much across country. The results furthermore draw attention to the limitations of the existing micro data sets on innovation cooperation.

Keywords:

• innovation
• cooperation
• innovation system
• multilevel model
• Europe

JEL Classification:

• D21
• L16
• F23
• O23

Acknowledgements

Source of the micro data is European Commission, Eurostat, Community Innovation Statistics and Fourth Community Innovation Survey. I am grateful to Eurostat for providing the firm-level data. Eurostat has no responsibility for the results and conclusions in this paper. Financial support from the Czech Science Foundation (GAČR) project on ‘Innovation, productivity and policy: What can we learn from micro data?’ (Grant agreement P402/10/2310) and VINNOVA Core Funding of Centers for Innovation Systems Research on ‘Transformation and Growth in Innovation Systems: Innovation Policy for Global Competitiveness of SMEs and R&I Milieus’ (Grant agreement 2010-01370) is gratefully acknowledged. Earlier versions of this paper have been presented at the Globelics Conference 2012, Hangzhou, China, 9–11 November 2012, and the Conference in Memory of Lennart Hjalmarsson, University of Gothenburg, Gothenburg, Sweden, 7–8 December 2012. The paper has benefitted from comments and suggestions from participants at these and other events as well as from the referees of this journal. All usual caveats apply.

Notes

1. CIS4 data on innovation cooperation by location of the partner is not available for Iceland, Ireland, Switzerland and the UK. Aggregated data by location of the partner is also not reported by Eurostat (2013) for Slovenia, even though the required information is present in the micro data by Eurostat (2009), and hence has been imputed from the latter source.

2. The number of firms cooperating exclusively with foreign partners can be computed by deducting the firms cooperating at home from the total number of cooperative innovators. The number of firms cooperating exclusively with domestic partners, however, cannot be computed, because information on how many firms cooperated with partners abroad, regardless of the foreign country, is not reported in the aggregated data. Since a closer look at the micro data set reveals that most firms cooperating abroad have partners simultaneously in other European countries and elsewhere, and the degree of overlap between these categories highly differs by country, we cannot simply add them to get the general prevalence of foreign cooperation.

3. Closer scrutiny of the data does not reveal any particular reason for this lower loading, such as an outlier; however, the observations for the debit licence payments are more dispersed around the trend line than observations for the other two indicators.

4. Eurostat (2009) data availability dictates the country composition of the micro sample. Romania is not included in the analysis, because of missing data about the location of cooperation partners.

5. In the harmonized CIS4 questionnaire, innovation cooperation was defined as active participation in innovation activities together with other enterprises or non-commercial institutions; pure contracting-out of work without active collaboration was emphasized not to be regarded as cooperation.

6. Since one of the micro predictors accounts for the information whether the firm is affiliated to a group, for more see below, internal cooperation arrangements with other firms affiliated to the same group are excluded from the definition; hence the dependent variables only refer to external cooperative linkages.

7. Unfortunately, the questionnaire does not allow us to identify whether the domestic-based group has operations in other countries, so that this dummy covers not solely domestic groups, but also home-based multinational corporations.

8. Table 4 and Table A2 do not contain exactly the same figures because observations with missing records are excluded from the micro sample and because the aggregated data are adjusted for the sampling fraction, non-response and no longer existing enterprises. Nevertheless, the correlation is 0.93 and 0.87 for total cooperation and exclusively foreign cooperation, respectively; thus, the difference is small.

9. The Brant test and the approximate likelihood-ratio test of proportionality of odds across response categories clearly indicate that a multinomial model is appropriate given the data.

10. Note that there are no terms for the level-1 residuals, because for a multinomial outcome in the multilevel framework the variance is completely determined by the population means, so that these residuals are not separate terms to be estimated.