Technology Spillovers and International Borders: A Spatial Econometric Analysis

ABSTRACT

The borders of the EU are open for the movement of resources but still there can be some strong negative effects of international borders on productivity and knowledge spillovers compared to the internal regional borders. These negative effects could be due to language barriers, cultural differences, local rules and regulation, legal issues, property rights, etc. These effects of international borders have an economic significance that need to be controlled when analyzing the regional knowledge spillovers. This aspect related to international borders has not been fully taken into account in the existing literature related to knowledge spillovers, therefore, ignoring this effect might under- or overestimate the effect of knowledge and technology spillovers. The results show that technology and knowledge spillovers are mainly coming from internal neighbor regions only, whereas spillovers across the international borders are statistically insignificant. Moreover, the results show that not properly incorporating border effects will lead to inaccurate estimates of the spillovers.

Notes

1. Some studies on border regions have discussed the conflicting effects of border on trade and community of border regions. For detail, see Alm and Burkhart (2013), Krugman, Obstfeld, and Melitz (2011), Millimet and Osang (2007) and Topaloglou et al. (2005).

2. From a methodological point of view, ignoring the neighborhood effect will create two problems, spatial dependence and spatial heterogeneity. Ignoring spatial dependence violates the basic assumptions of least square estimates which causes the results to be biased and inconsistent, while spatial heterogeneity causes the instability or non-stationarity of economic relationships over space (see Arbia, Basile, and Piras 2005; Anselin 1988).

3. They have not incorporated two different types of spatial weight matrices in spatial regression analysis, by doing so one can get more reliable and additional information regarding different type of spillover effects.

4. Earlier studies do not differentiate between the border neighbor region and internal neighbor regions.

5. The econometric specification of the production function implements extended versions of spatial econometric models like Spatial Autoregressive model (SAR) and Spatial Durbin Model (SDM) by using maximum likelihood method of estimation. The extended versions of SAR and SDM models can simultaneously incorporate more than one type of spatial dependences. The results show that more spillover is coming from internal neighbor regions than the border regions. Moreover, the results show that not properly incorporating border effect will lead to inaccurate estimates of the spillovers.

6. The detailed methodology and empirical specification is available upon request to the authors.

7. Where and representing the neighbor regions within the country and across the international borders respectively.

8. The maps of regions are shown in the Appendix figure A1.

9. The construction of the TFP index can be provided on request.

10. We are using Queen Contiguity which means two regions are neighbors in the sense if they share any part of a common border, no matter how short (one region share a common side or vertex with the regions of interest).

11. These tests are called Moran’s I test, LM, LR and Wald (for detail, see Arbia, Basile, and Piras 2005).

12. We also have estimated the same models separately from 1999 to 2010 and the results are available on request.

13. The detail about SAR and SDM is given in the Appendix section A1.

14. The available software (MATLAB) allows the SAR, SEM, SDM, and Extended SAR and ESDM models. Some of them are available on the following website.

http://www.spatial-econometrics.com. These models are estimated via maximum likelihood methods using a pseudo likelihood definition and the approximate non-linear maximization method.

15. They also provide dispersion measures for the direct, indirect and total effects, which allow us to draw inference on their statistical significance, for careful interpretation we follow LeSage and Pace (2009), and Marrocu, Raffaele, and Stefano (2013).

16. As mentioned by LeSage and Pace (2009) most studies compare estimates of the SAR and SDM models without finding the own and cross partial derivatives for impact estimates.

17. See Anderson and Wever (2003), Topaloglou et al. (2005), Millimet and Osang (2007) and Krugman, Obstfeld, and Melitz (2011).

18. We construct both weight matrices manually with the help of ArcMap of each country and regions at NUTS 2 level.

19. We estimated all the models for all the years from 1999 to 2010 and the results can be provided on request.