In this paper, the framework of the aggregated Beveridge curve is used to investigate the effectiveness of the job matching process using German regional labour market data. For a fixed matching technology, the Beveridge curve postulates a negative relationship between the unemployment rate and the rate of vacancies, which is efficiently estimated using spatial econometric techniques. The eigenfunction decomposition approach suggested by Griffith (2000, 2003) is the workhorse to identify spatial and non-spatial components. As the significance of the spatial pattern might vary over time, inference is conducted on the base of a spatial SUR model. Shifts of the Beveridge curve will affect its position, and time series estimates on this parameter are obtained. In contrast to findings for the US and the UK, the results provide serious indication that the degree of job mismatch has increased over the last decade. Although the outward shift of the Beveridge curve can be explained by structural factors such as the evolution of long term unemployment, it is also affected by business cycle fluctuations. The role of cyclical factors threatens the stability property of the curve. The relationship might be inappropriate to investigate policy measures directed to improve the mismatch, such as labour market reforms.