Is East Germany catching up? A time series perspective

Abstract

This article assesses whether the economy of East Germany is catching up with the West German region in terms of welfare. While the primary measure for convergence and catching up is per capita output, we also look at other macroeconomic indicators such as unemployment rates, wage rates and production levels in the manufacturing sector. In contrast to existing studies of convergence between regions of the reunified Germany, our approach is based purely upon the time series dimension and is thus directly focused on the catching up process in East Germany as a region. Our testing set-up includes standard ADF unit root tests as well as unit root tests that endogenously allow for a break in the deterministic component of the process. We find evidence of catching up for East Germany for most of the indicators. However, the convergence speed is slow, and thus it can be expected that the catching up process will take further decades until the regional gap is closed.

Notes

 1. See Barro and Sala-i-Martin (1992) and Mankiw et al. (1992) for a formal derivation of an empirical neoclassical growth-convergence equation from theoretical models.

 2. For further objections see Evans and Karras (1996). Bernard and Durlauf (1996) show that cross-sectional tests tend to spuriously reject the null of no convergence when economies have different long-run steady states.

 3. This allows for different steady-state levels defined by a fixed constant. This may be due to differences in regional characteristics such as demographics, regional policy and agglomeration effects.

 4. This means that the existence of a time trend alone does not necessarily imply convergence. Additionally, the time trend must lead to a reduction in the differences in the observed variables. Otherwise, when differences tend to increase over time there is evidence of divergence.

 5. As a robustness check we also consider additional unit root tests. These are the Phillips-Perron (PP) test, the Kwiatkowski–Phillips–Schmidt–Shin test (KPSS) and the Elliot–Rothenberg–Stock (ERS) test. While the PP test is similar to the ADF except for the way autocorrelation is taken into account, the KPSS reverses the null hypothesis. The ERS test is considered because its power properties are preferable under certain conditions compared to the ADF test when deterministic components are present.

 6. According to Perron and Ng (1996) this strategy may be superior compared to information criteria in finite samples to determine the optimal lag length K. For quarterly data we choose K _max = 5, for monthly data K _max = 13. We assume a significance level of 5%.

 7. There are also tests available that test for two or more break points. But since our time span is generally rather short (around 15 years) we allow only for one potential break point. Considering more than one break also increases potential problems of overfitting.

 8. See Brautzsch and Ludwig (2002) for methodological background. The data set is regularly updated and available under http://iwhd:3129/c/start/prognose/prognose.asp?lang = e.

 9. Data for the population of East and West Germany are obtained from the national accounts data base from the German Länder. We have converted population data from yearly to quarterly frequency by linear interpolation.

10. We use ARIMA-X12 as provided by EViews 6.0. Alternatively, we also augmented the unit root tests by seasonal dummy variables, which leads to qualitatively the same results as the seasonal adjustment procedure.

11. Monthly seasonally adjusted unemployment rates as well as the index of nominal wage rates (defined as per man-hour worked) are taken from the data base of the Deutsche Bundesbank (keys: US02CC, US0CC2, US08RB, US0RB8). The information about wage levels is taken from the Federal statistical office (Fachserie 16 Reihe 2.3). For manufacturing production we have seasonally unadjusted data that were adjusted using ARIMA-X12. One break due to methodological changes has been removed from this data series. The index series has been obtained from German regional statistical offices. The reference level of the production values is obtained from an input–output table for Germany. As with per capita output, we divided this series by a quarterly interpolated population measure. All data series are available on request.

12. We also applied other unit root tests. The PP test confirms the results based on the ADF test and rejects the unit root hypothesis at a 10% significance level. The KPSS test does not reject for any standard significance level, which suggests that the series is stationary. Also the EPS test reports evidence in favour of stationarity. This test rejects the hypothesis of a unit root even at the 1% level.