Real convergence in some Central and Eastern European countries

Abstract

This article examines the real convergence hypothesis in some Central and East European countries (both towards the German and the US economies) by means of using time series techniques during the period 1950 to 2003. No evidence is found of real convergence for the whole period. However, when one allows for structural breaks, evidence is found of a catch-up process during the 1990s to 2003 period for Poland, the Czech Republic and Hungary towards Germany and only for Poland towards the US economy.

Acknowledgements

The authors gratefully acknowledge the valuable comments from an anonymous referee.

Notes

¹ Although there are several papers which study the economic growth process in Central and Eastern European countries (see for example Estrin et al. 2001; Kocenda 2001; Dawson and Hubbard, 2004; Dritsakis, 2004; Janicki and Wunnava, 2004; Kominek, 2004), little attention has been paid to the real convergence process in the post-1950 period.

² See Van Ark (2000) and Barbone and Zalduendo (1996).

³ The study uses as benchmark economies both Germany (as an economy representing the European Union standards) and the US economy (in a more global context, one of the leader international economies used as benchmark in most of the analysis on real convergence).

⁴ According to neoclassical models, unconditional or absolute convergence holds when per capita GDP of the different countries convergence to the same steady state. In contrast, conditional convergence applies when per capita GDP of each economy converge to its own steady state. In this last case, the constant α₀ measures the differences in the steady state of each of the economies.

⁵ Carlino and Mills (1993), for example, use this methodology in order to allow initially low income countries to grow faster than higher income countries.

⁶ Recent empirical application of these tests can be found in Leybourne et al. (2003), Atkins et al. (2004), Sen (2004), Hacker et al. (2005) or Narayan (2005a,b), Sen (2004), Atkins et al. (2004) or Narayan (2005a, b) among others.

⁷ This two-break LM unit root test is used and not the one-break LM test because an examination of the empirical results reveals that the two structural breaks included in the model were significant in all the five analysed countries for the period 1950–2003.

⁸ In the empirical analysis, and as in Lee and Strazicich (2003), we consider Model C which allows for two changes in level and trend.

⁹ See Lee and Strazicich (2003) for a more detailed description of the test. The computation of the LM unit root test statistic has been carried out using the Gauss codes provided by Junsoo Lee and available on the web site http://www.cba.ua.edu/~jlee/gauss.

¹⁰ The same unit root tests have been applied to an alternative dataset obtained from the Penn World Table, version 6.1 (Heston et al., 2002). The analyzed variables are real per capita GDP adjusted for PPP and the sample periods range from 1991 to 2000 for Bulgaria, 1990–2000 for Czech Republic, 1970–2000 for Hungary, 1979–2000 for Poland, 1960–2000 for Romania and 1970–2000 for Germany. The results, which are consistent with those obtained with the Maddison dataset, are presented in the Appendix.

¹¹ We have carried out our empirical analysis for these five economies since Maddison (2001) data base has data only for five Central and Eastern European economies begining in 1950.

¹² In all the cases, the dummy variables which account for the structural breaks were significant.

¹³ The one-break LM minimum statistic has not been applied in the cases of Bulgaria and the Czech Republic because the sample period in these countries range from 1990/1 to 2000. For the rest of the countries one has only allowed for one-break since the sample period range from 1970/9–2000.