Abstract
This paper looks at unemployment rate convergence among the EU Member States. In particular, it examines whether the unemployment rate differences diminished between 1991 and 2014 among the old EU countries, among the new EU countries, among the euro countries and among the no-euro countries, and whether the new EU countries were converging in this sense to the group of old EU countries. Three alternative concepts of convergence are considered and tested for on each group of countries: σ-convergence, stochastic convergence and time-series β-convergence. The results provide strong support for EU unemployment rate convergence, although within each group there are large differences between the countries and during the recent economic downturn, there were clear signs of temporary divergence as well.
Similar content being viewed by others
Notes
http://ec.europa.eu/eurostat/statistics-explained/index.php/Unemployment_statistics, accessed 06 December 2016.
Euro countries are meant to be the Eurozone countries at the end of 2014.
This issue was pointed out by Perron and Yabu (2009, p. 369) who designed an alternative, but similarly asymptotic, test for shifts in trend based on two sets of critical values that are much closer to each other.
In the literature cross-sectional β-convergence and time-series β-convergence are both known as β-convergence as they are both related to the slope parameter of some regressions. Still, to avoid any confusion, in this paper we are going to refer to them as cross-sectional β-convergence and time-series β-convergence, respectively, unless the context makes this distinction unnecessary.
In the first case xi,t and \( x_{i,t}^{*} \) are both integrated of order zero, I(0), while in the second case they are integrated of order one, I(1), and cointegrated of order one and one, CI(1,1). This definition of stochastic convergence has been generalised by Bernard and Durlauf (1995): xi,t and \( x_{i,t}^{*} \) converge to each other in a stochastic sense if their long-term forecasts are equal at a fixed time (Definition 2.1); and they have a common trend if their long-term forecasts are proportional at a fixed time (Definition 2.2). Consequently, xi,t and \( x_{i,t}^{*} \) are either stationary with zero (or at least equal) means, or are cointegrated with cointegrating vector [1,-1] (Definition 2.1) and [1,-α] (Definition 2.2), respectively.
Both requirements of time-series β-convergence are satisfied when \( 0 < t \le \; - \mu_{i} /\beta_{i} \).
About bootstrap tests see e.g. Godfrey (2009).
A useful review and comparison of these tests are provided by Palm et al. (2008).
In principle the break dates could be considered known and the same for all series, but we prefer considering them unknown and estimating them with a grid-search for every group of countries and for every country within each country group individually prior to testing for unit roots for the following reasons. (i) Common exogenous break dates would impose some kind of uniformity for all countries and if these strong but untested restrictions are false they can potentially falsify the convergence test results. (ii) Even if a particular event, like the introduction of the euro or the latest recession, triggered a structural break in every EU country, it is still questionable whether the actual breaks occurred instantaneously and simultaneously in all data series. (iii) If there were common break dates in the data series, we would expect the grid-search procedures to pick them up. However, as the results in Tables 2, 3, 4, 5, 6 and 7 indicate, the revealed break dates vary by countries and groups of countries.
Following Hayashi (2000, p. 594), we set the maximal lag as \( pmax = \text{int} \left( {\hbox{min} \left( {\frac{T}{3},12} \right) \times \sqrt[4]{{\frac{T}{100}}}} \right) \).
The ILO unemployment rate indicates the proportion of the labour force that does not have a job and is actively looking and available for work. It is harmonised to account for differences in national data collection and tabulation methodologies, but for the sake of brevity, we are going to refer to it as unemployment rate. The unemployment rate (UR) and labour force (LF) data were extracted from the ILO database by the 9th Edition of the KILM (Key Indicators of the Labour Market, 2015) software.
These country groups are shown in Table A in the Appendix.
See e.g. http://ec.europa.eu/economy_finance/euro/adoption/who_can_join/index_en.htm about the euro convergence criteria, also known as the Maastricht criteria..
This way we also manage to take into consideration the huge differences between EU countries in terms of labour force. The most striking difference is between Germany and Malta. During our sample period the labour force of Germany was about 225 to 270 times larger than that of Malta.
To save space, we do not present these trend regressions here, but they are available on request.
These dates coincide with the 1997 Asian financial crisis and the 2007-2008 global financial crisis, but whether the potential structural changes are indeed linked to these events is a different issue.
To save space we focus on these two groups. As mentioned above, the EURO and NONEORO average unemployment rate series appear to be mainly driven by the OEU group and the NEU group, respectively.
For the sake of simplicity, we refer to NEU2 too as a weighted cross-sectional standard deviation.
It does not matter whether these deviations are calculated from the group mean or from the mean of the other group members. Suppose that \( y_{i,t} = x_{i,t} - \bar{x}_{t} \), where \( \bar{x}_{t} = \frac{1}{n}\sum\nolimits_{j = 1}^{n} {x_{j,t} } \) is the mean of all n group members in period t, and \( y_{i,t}^{*} = x_{i,t} - \bar{x}^{*}_{t} \), where \( \bar{x}^{*}_{t} = \frac{1}{n - 1}\sum\nolimits_{j = 1,j \ne i}^{n} {x_{j,t} } \) is the mean of all group members but i in period t In this case \( \bar{x}_{t} = \frac{n - 1}{n}\bar{x}^{*}_{t} + \frac{1}{n}x_{i} \) and \( y_{i,t} = \frac{n - 1}{n}y_{i,t}^{*} \), so up to a constant factor the \( \left\{ {y_{i,t} } \right\} \) and \( \left\{ {y_{i,t}^{*} } \right\} \) time series are equivalent.
To save space, the URD time-series plots are not shown here, but they are available on request.
The relevant t-ratio and p value are 0.4355 and 0.6677, respectively.
In fact, the EURO group has the largest sample mean and the smallest sample standard deviation out of all groups.
References
Bai J, Perron P (1998) Estimating and testing linear models with multiple structural changes. Econometrica 66:47–78
Bayer C, Jüßen F (2007) Convergence in west german regional unemployment rates. German Econ Rev 8:510–535
Bernard AB, Durlauf SN (1995) Convergence in International Output. J Appl Econom 10:97–108
Blanchard OJ (2006) European unemployment: the evolution of facts and ideas. Econ Policy 21(45):5–59
Breuer JB, McNown R, Wallace M (2002) Series-specific unit root tests with panel data. Oxford Bull Econ Stat 64:527–546
Breuer JB, McNown R, Wallace M (2006) Misleading Inferences from panel unit root tests: a reply. Rev Int Econ 14:512–516
Bühlmann P (1997) Sieve bootstrap for time series. Bernoulli 3:123–148
Carlino GA, Mills LO (1993) Are U.S. regional incomes converging? A time series analysis. J Monetary Econ 32:335–346
Carrera, D.R. and Rodriguez, G. (2009) Have European Unemployment Rates Converged? Working Papers, Banco Central de Reserva del Perú, No 2009-007
DeLong JB (1988) Have productivity levels converged?—productivity growth, convergence, and welfare in the very long run. NBER working papers, no. 2419
Estrada Á, Galí J, López-Salido D (2013) Patterns of convergence and divergence in the Euro area. IMF Econ Rev 61:601–630
European Commission (2014) Employment and social developments in Europe 2014, ISSN 2315-2540 (web)
Glynn J, Perera N, Verma R (2007) Unit root tests and structural breaks: a survey with applications. Revista de Métodos Cuantitativos Para la Economía y la Empresa 3:63–79
Godfrey L (2009) Bootstrap tests for regression models. Palgrave Texts in Econometrics. Palgrave Macmillan, Basingstoke
Larsson A (1998) The European employment strategy and the EMU: you must invest to save. Econ Ind Democr 19:391–415
Maddala GS, Kim In-Moo (1988) Unit roots, cointegration, and structural change. Cambridge University Press, Cambridge
Ng S, Perron P (1995) Unit root tests in ARMA models with data-dependent methods for the selection of the truncation lag. J Am Stat Assoc 90(429):268–281
Palm FC, Smeekes S, Urbain J-P (2008) Bootstrap unit-root tests: comparison and extensions. J Time Ser Anal 29:371–401
Perron P (1989) The great crash, the oil price shock, and the unit root hypothesis. Econometrica 57:1361–1401
Perron P, Vogelsang TJ (1992) Nonstationarity and level shifts with an application to purchasing power parity. J Bus Econ Stat 10:301–320
Perron P, Yabu T (2009) Testing for shifts in the trend with an integrated or stationary noise component. J Bus Econ Stat 27:369–396
Rosiek J, Włodarczyk RW (2012) Comparative analysis of the EU-27 countries labour markets’ convergence. Econ Manag 17:216–222
Saint-Paul G (2004) Why are the European countries diverging in their unemployment experience? J Econ Perspect 18:49–68
Solow R (1956) A contribution to the theory of economic growth. Q J Econ 70:65–94
Solow R (1957) Technical change and the aggregate production function. Rev Econ Stat 39:312–320
Swan TW (1956) Economic growth and capital accumulation. Econ Rec 32:334–361
Tomljanovich M, Vogelsang TJ (2002) Are U.S. regions converging? Using new econometric methods to examine old issues. Empir Econ 27:49–62
Van Rie T, Marx I (2012) The European union at work? European employment strategy from crisis to crisis. J Common Market Stud 50:335–356
Vogelsang TJ (1998a) Testing for a shift in mean without having to estimate serial-correlation parameters. J Bus Econ Stat 16(1):73–80
Vogelsang TJ (1998b) Trend function hypothesis testing in the presence of serial correlation. Econometrica 66:123–148
Vogelsang TJ, Perron P (1998) Additional tests for a unit root allowing for a break in the trend function at an unknown time. Int Econ Rev 39:1073–1100
Zellner A (1962) An efficient method of estimating seemingly unrelated regressions and tests for aggregation bias. J Am Stat Assoc 57:348–368
Acknowledgements
We are grateful for the helpful comments of two anonymous referees of this journal.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations'' (in PDF at the end of the article below the references; in XML as a back matter article note).
Appendix A
Appendix A
See Table 13.
Rights and permissions
About this article
Cite this article
Kónya, L. Did the unemployment rates converge in the EU?. Empir Econ 59, 627–657 (2020). https://doi.org/10.1007/s00181-019-01678-5
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00181-019-01678-5