21
As with Verdoorn’s law the spatial effects can be splitted into common components and
map patterns. From Moran's I it becomes obvious that the spatial components capture
the largest part of the cross-section dependencies present in the change of the unem-
ployment rate. On the average, the spatial correlations of the residuals are reduced by a
factor of about 12 by spatial SUR estimation compared with non-spatial GLS estima-
tion. In six out of eight subperiods Moran’s I of the spatial SUR residuals lies close to
zero (< 0.05).
Figure 6.4 exhibits that again both west-east and east-west trends dominate the map
patterns of the spatial components. With Okun’s law the Moran scatterplot trends are
pronounced (MI∕MImax≥0.9) in five years. For 1998 the MI∕MImax value indicates a
strong MC trend and for 1993 and 1993 noticeable MC trends. It is conspicuous that the
east-west trend inclinations run considerably smoother than the declining trend surfaces.
In the latter case always several moulds in the western part of the country can be lo-
cated.
The difference between the unemployment and employment threshold is noticeable. The
gap is likely to be attributed to the growth of the working population, mainly due to a
perceptible rise in the participation rate of women, which increased by 1.1 percentage
points in the sample period (Statistisches Bundesamt, 2003). Also, the unemployment
benefit system has to be considered. In periods of economic upturns new jobs are par-
tially filled with people from outside the labour force. Thus, employment is affected,
while unemployment is eventually stable. In case of downturns, job losses will imply a
simultaneous rise in unemployment to receive financial support.
7. Conclusions
In this paper, we have estimated thresholds of output growth needed for a rise in em-
ployment and a fall in the unemployment rate for the unified Germany on the base of
the laws of Verdoorn and Okun. A spatial analysis of 180 German regional labour mar-
ket regions reveals that the relations between employment, unemployment and produc-
tion may be distorted by strong spatial dependencies. To capture spatial autocorrelations
a feasible spatial SUR techniques is proposed. It turns out that minimum output growth
sufficient for a rise in employment is below the level which is needed for a drop in the