Weak and Strong cross-sectional dependence: a panel data analysis of international technology diffusion

This paper provides an econometric examination of geographic R&D spillovers among countries by focusing on the issue of cross-sectional dependence, and in particular on the different ways – weak and strong – it may affect the model. A preliminary analysis based on the estimation of the exponent of cross-sectional correlation proposed by Bailey et al.(2013), a, provides a very clear-cut result with an estimate of a very close to unity, not only indicating the presence of strong cross-sectional correlation but also being consistent with the factor literature typically assuming that a = 1. Moreover, second generation unit roots tests suggest that while the unobserved idiosyncratic component of the variables under study may be stationary, the unobserved common factors appear to be nonstationary. Consequently, a factor structure appears to be preferable to a spatial error model and in particular the Correlated Common Effects approach is employed since, among other things, it is still valid in the more general case of nonstationary common factors. Finally, comparing the results with those obtained with a spatial model gives some insights on the possible bias occurring when allowing only for weak correlation while strong correlation is present in the data.