Spatial Filtering, Model Uncertainty and the Speed of Income Convergence in Europe

Abstract

In this paper we put forward a Bayesian Model Averaging method dealing with model uncertainty in the presence of potential spatial autocorrelation. The method uses spatial ltering in order to account for dierent types of spatial links. We contribute to existing methods that handle spatial dependence among observations by explicitly taking care of uncertainty stemming from the choice of a particular spatial structure. Our method is applied to estimate the conditional speed of income convergence across 255 NUTS-2 European regions for the period 1995-2005. We show that the choice of a spatial weight matrix -and in particular the choice of a class thereof -can have an important eect on the estimates of the parameters attached to the model covariates. We also show that estimates of the speed of income convergence across European regions depend strongly on the form of the spatial patterns which are assumed to underlie the dataset. When we take into account this dimension of model uncertainty, the posterior distribution of the speed of convergence parameter appears bimodal, with a large probability mass around no convergence (0% speed of convergence) and a rate of convergence of 1%, approximately half of the value which is usually reported in the literature.

Heterogeneity and Spatial Dependence of Regional Growth in the EU: A Recursive Partitioning Approach

“…This variable has been found important, for example, in Sanso‐Navarro and Vera‐Cabello (). A dummy variable for capital regions (capital) is used. This variable has been found significant in Schneider and Wagner () and has posterior inclusion probability equal to one in Crespo Cuaresma and Feldkircher (), in line with the large literature on core‐periphery effects in new economic geography models (compare Fujita et al ., ). For additional empirical evidence highlighting the importance of agglomeration effects in the EU see, for example, Geppert and Stephan (). We also consider dummy variables for border regions (regborder) and coastal regions (regcoast).…”

“…It is important to note that we allow with this specification for a ‘global’ spatial lag parameter ρ , that multiplies the individual‐specific spatially‐lagged observation , and group‐specific coefficients β g ( i ) for the explanatory variables. In our empirical analysis, we employ the same set of spatial weight matrices as Crespo Cuaresma and Feldkircher (). As already alluded to above, the results are surprisingly robust with respect to the choice of the weighting matrix.…”

“…The differential growth and convergence performance of the PIIGS countries already prior to the crisis is, for example, documented in Ertur et al . (). In our calculations we use 24 of the spatial weight matrices of Crespo Cuaresma and Feldkircher (), including matrices based on nearest neighbors, distance bands, and exponential decay (inverse distance) . Given the remarkable robustness of our result with respect to the weight matrix chosen we focus in the following section on the results obtained with the standard inverse distance weight matrix and only comment upon the differences obtained with other weight matrices …”

Volatility and Regional Growth in Europe: Does Space Matter?

“…We also include in vector X the share of the active population with tertiary education and/or an employment in science and technology as a human capital control. 5 This is particularly important, given the relevant role played by investment in human capital when explaining regional growth in Europe (Crespo Cuaresma & Feldkircher, 2013). Additionally, regional growth patterns may be affected by the possible existence of agglomeration economies (Ciccone, 2002;Fujita & Thisse, 2002).…”

Regional club convergence in the EU: evidence from a panel data analysis

“…However, it does not allow for spatial dependencies. Similarly, Cuaresma and Feldkircher (2013) use a Bayesian model averaging method to detect convergence clubs.…”