Abstract
We augment the existing literature on regional convergence by uncovering a number of stylized facts on the heterogeneity of regional convergence processes in the absence of currency devaluation as a key policy instrument, and use them to highlight reform strategies that are most likely to be conducive to a successful catching-up of the periphery countries of EMU. We show that regional convergence processes in Europe were extremely heterogeneous, highly discontinuous and strongly concentrated during the last two decades. These stylized facts question the focus of the traditional literature on average (β-)convergence and suggest substantial nonlinearities in regional convergence processes that have yet to be understood in detail. Our results further suggest that growth strategies based on increasing human capital investments and innovation capacities are the most likely to be successful in triggering convergence of lagging regions in currency unions.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
We refrain from filtering business cycle effects from the data because within countries, business cycle synchronization is usually high for data on an annual frequency (Montoya and de Haan 2008). This implies that the risk of identifying effects from asymmetric regional cycles relative to the national average is relatively low and is likely to have a lower impact on results than the ambiguities necessarily induced by trend-cycle decompositions.
These results are available from the authors upon request.
The fact that there are more regions located to the left of the country average than to the right is interesting in its own right. It illustrates the skewness of the GDP per capita distribution documented in many other empirical descriptions of regional disparities in Europe (e.g. Juessen 2009).
In a conditional convergence framework a poor (rich) region can fall further behind (forge further ahead) despite convergence if it converges to a steady-state that is below (above) average. In case of unconditional convergence, instances of regions falling behind or forging ahead are only possible for a short period of time and in the existence of temporary idiosyncratic “shocks”.
For instance one could argue that short periods of above and below average growth could be expected in a world in which temporary idiosyncratic “shocks” repeatedly shift individual regions above or below their respective steady state growth paths.
For this purpose we recalculated the average annual growth rate over the period if the year with the highest (lowest) growth rate were replaced by the region’s average growth rate: Given that the average growth rate of a region is the geometric mean of the annual growth rates this hypothetical growth rate for region i can be calculated as \( \hat{g}_{i}^{hyp} = [ {\bar{g}_{i} /g_{i}^{max} } ]^{1/(t - 1)} \) with \( \bar{g}_{i} = ( {GDP_{t} /GDP_{0} } ) \) and with \( g_{i}^{\hbox{max} } \) the maximum annual growth rate of the region in t periods. The contribution of the year with the strongest growth in percentage points \( ( {Cont_{i}^{max} }) \) can be calculated by \( Cont_{i}^{max} = 100( {\bar{g}_{i}^{1/t} - \hat{g}_{i}^{hyp} }). \) The same calculations apply to the year with the weakest growth performance.
For education levels data is only available from 1999 on, so that we use the earliest available observation in both the descriptive and the econometric analysis below.
This index is half of the sum of absolute changes in sector employment shares from period t − 1 to t. It ranges between zero and one, with zero (one) indicating no (maximum) structural change in employment.
We give preference to using probit regressions rather than more conventional convergence regressions because they are more apt to capture the substantial nonlinearities in convergence behavior implied by our descriptive results.
The spatial lag is based on the product W × GDP using the contiguity matrix W, with element wij = 1/n if region i borders on region j and is located in the same country, and wij = 0 otherwise, with n being the number of same country neighbors of region i. A dummy variable taking the value of 1 for regions lacking neighbors of the same country (i.e. islands and Northern Ireland) is included in the regression to account for the missing spatially lagged GDP for such regions by definition.
Analyzing poor regions, Bulgaria has to be dropped from the sample, because all three regions below the country mean level of GDP per capita were converging during the period observed. In the regression analysis for rich regions, Sweden and Slovakia only had one region above the mean during the period observed and in Belgium, Germany, Finland and Portugal, all regions above the mean converged, so that these countries had to be excluded from these regressions.
Since the set of poor regions includes only one region hosting a national capital (Berlin), specification (6) that adds the dummy variable for capital city regions to the set of variables of specification (5) is not estimated for poor regions.
Since we were concerned that the high volatility of investments may weaken the results we also used the three-year and five-year average investment rates rather than single initial year investment rates as explanatory variable in additional regressions not reported. This, however, did not have an impact on the significance levels of the coefficient on the investment rate. Furthermore, in additional robustness checks we also experimented with additional definitions of the dependent variable by using different cutoff points for the level of GDP defining the rich and poor regions and using the growth rates of independent variables as additional controls. None of these alternative specifications (which are available from the authors upon request) led to substantial changes in results for the main variables.
There may be concerns that in richer regions (in particular in large cities) service orientation is much higher than in the poor regions and that this could lead to a bias of the coefficients on patents in rich regions on account of the low propensity of service enterprises to patent their innovations. This potential bias, however, should be of minor importance because, as shown in Table 3, the industry shares of rich regions are not smaller on average than those of poor regions.
The high explanatory power of this variable as a predictor for above average growth is not only driven by the developments of the new EU member states. Also ten out of the fourteen capital regions of Western European countries that are found in the group of rich regions were diverging from their national average between 1991 and 2009.
This finding is supported by t-tests of the null-hypothesis of equality of the coefficients of the estimates for poor and rich regions (see Table 7 in the appendix). These tests suggest that among those variables that are significant either for poor or for rich regions significant differences in marginal effects between rich and poor region only exist for initial GDP per capita, the industry share and in some specifications the share of primary educated as well as unit labor costs.
References
Abreu M, de Groot HLF, Florax RJGM (2005) A meta-analysis of β-convergence: the legendary 2%. J Econ Surv 19(3):389–420
Barro RJ, Sala-i-Martin X (1991) Convergence across states and regions. Brookings Pap Econ Act 1:107–182
Barro RJ, Sala-i-Martin X (1992) Convergence. J Polit Econ 100(2):223–251
Bartlett MS (1937) Properties of sufficiency and statistical tests. Proc R Soc Lond A 160(901):268–282
Baumol WJ (1986) Productivity growth, convergence and welfare: what the long-run data show. Am Econ Rev 76(5):1072–1085
Crespo Cuaresma J, Feldkircher M (2013) Spatial filtering, model uncertainty and the speed of income convergence in Europe. J Appl Econom 28(4):720–741
Crespo Cuaresma J, Foster N, Stehrer R (2011) Determinants of regional economic growth by quantile. Reg Stud 45(6):809–826
Crespo Cuaresma J, Doppelhofer G, Feldkircher M (2012) The determinants of economic growth in European region. Reg Stud. doi:10.1080/00343404.2012.678824
Dobson S, Ramlogan C, Strobl E (2006) Why do rates of β-convergence differ? A meta-regression analysis. Scott J Polit Econ 53(2):153–173
Durlauf SN, Johnson PA, Temple JRW (2005) Growth econometrics. In: Durlauf SN, Aghion P (eds) Handbook of economic growth, vol 1. Elsevier, Amsterdam, pp 555–677
Easterly W (2006) Reliving the 1950s: the big push, poverty traps, and takeoffs in economic development. J Econ Growth 11(4):289–318
Easterly W, Kremer M, Pritchett L, Summers LH (1993) Good policy or good luck? Country growth performance and temporary shocks. J Monet Econ 32(3):459–483
Ertur C, Koch W (2007) Growth, technological interdependence and spatial externalities: theory and evidence. J Appl Econom 22(6):1033–1062
Faini R (2003) Migration and Convergence in the regions of Europe: a bit of theory and some evidence. FLOWENLA Discussion Paper 9, HWWA Hamburg
Hausmann R, Pritchett L, Rodrik D (2005) Growth accelerations. J Econ Growth 10(4):303–329
Jones BF, Olken BA (2008) The anatomy of start-stop growth. Rev Econ Stat 90(3):587–592
Juessen F (2009) A distribution dynamics approach to regional GDP convergence in unified Germany. Empir Econ 37(3):627–652
LeSage JP, Fischer MM (2008) Spatial growth regressions: model specification, estimation and interpretation. Sp Econ Anal 3(3):275–304
Magrini S (2004) Regional (di)convergence. In: Henderson JV, Thisse JF (eds) Handbook of regional and urban economics, vol 4. Elsevier, Amsterdam, pp 2741–2796
Mankiw GN, Romer D, Weil DN (1992) A contribution to the empirics of economic growth. Q J Econ 107(2):407–437
Montoya LA, de Haan J (2008) Regional business cycle synchronization in Europe? IEEP 5(1–2):123–137
Rodrik D (2013) Unconditional convergence. Q J Econ 128(1):165–204
Acknowledgments
The authors thank two anonymous reviewers, Jesus Crespo Cuaresma, Peter Mayerhofer, Johanna Vogel, and the participants of the 1st ISW-Workshop on Applied Economics in Salzburg, the WWWforEurope Area 5 Meeting in Vienna, the NOeG Conference 2013 in Innsbruck, the Euroframe June 2013 conference in Warsaw, and the 53rd ERSA congress in Palermo for helpful comments. The usual disclaimer applies. The research leading to these results has received funding from the European Community’s 7th framework Programme FP7/2007–2013 under Grant agreement No. 290647.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Firgo, M., Huber, P. Convergence as a heterogeneous process: what can be learnt about convergence in EMU from regional experiences?. Empirica 41, 129–151 (2014). https://doi.org/10.1007/s10663-013-9242-y
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10663-013-9242-y