||The objective of the paper is an analysis of the service life of dwellings in Germany. The service life of dwellings is important for different reasons: It is necessary for an estimation of the future development of the building stock and quality as well as for projections on new constructions. In addition, asset durability is one of the most important features of the housing market. An analysis of this feature may provide further insight into the market. Furthermore, the service life of the investment good is an important characteristic of the decision to invest in dwellings. Hence the service life is also important in the context of housing valuation: Since the initial rent in the housing market is usually less than in other real estate markets (e.g. office or retail), the payback period of the investment is quite long. The aim of the paper is the comparison of actual service life times with projections of how the service life of buildings will develop assuming economically rational investors. Since many, formerly publicly owned, housing companies in Germany are subject to increased economic pressure as a result of their sale to private investors, future predictions on the development and composition of the housing stock may be improved under this assumption for owner behaviour. The service life times of buildings have internationally been analysed (among others) by Johnstone (2001), Meikle&Connaughton (1994) Gleeson (1985) and Needleman (1965). In Germany, the service life of the building stock has been analysed by Bradley et al. (2005), Bradley (2004) and estimated by Voss (2001) and Eekhoff (2002). Among these, two different approaches can be identified: The analytic approach is based on macro data and estimates building loss rates over the whole stock and the micro/life table analysis identifies loss rates in certain subsets of the total stock within regional, age or other boundaries. The estimation of service life and loss rates of buildings by using the Kaplan-Meier-Estimator conducted is one standard method among the authors of the second (life table) approach: It was already conducted by Johnstone and Bradley, both using data on a different stock. However, the idea of the calculus of the investor as the most important determinant of service life has not been articulated among these assessments. The work is based on observations of houses of a big dwelling company in the Ruhr region of Germany with approx. 15,000 buildings. The buildings date from all over the 20th century. They are of lower quality. Data on loss and cash-flows is available from 1991-2004, over 14 years. The normal service of a building is defined as the time that elapses before 50% of the objects under observation are knocked down. Using the Kaplan-Meier-estimator for estimating the survival function and extrapolating these estimations with the Weibull-function yields an estimated service life. This is remarkably longer than expected using conventional data on service life. This result for service life will be compared to the service life that is obtained under the assumption of an economically acting investor. To obtain this, more information about the development of rent and running costs of the buildings is necessary. This is done by analysing the patterns of the cash-flows of the dwellings using data from 1991 to 2004 (14 years). The results are obtained by towns as boxplot diagrams (cash-flow over time) and mean values over time. Using the results from this analysis a replacement time may be identified. Down to the present day the analysis has not been finished yet. However, preliminary results are available: Even when extrapolating the extraordinarily high demolition rates (in relation to the one for the whole stock of Germany) from the last decade the estimation for the service life of the dwellings included in the analysis is approx. 140 years. This means that the German society will live in these low-quality buildings for a longer time than perceived at present. Another preliminary result is concerning rent: It is decreasing as the building is ageing. There does not seem to be a strong correlation of rent and age, though. These findings are only partly in line with considerations of Sotelo (2001) for the German dwelling market, the results of Baum (1991) for the London office market and the Filtering Theory. No dwelling included in the analysis is older than 107 years though, hence validity over the total German stock is limited. On the other hand, there seems to be a similar correlation of running, especially maintenance costs, with age. In opposition to other investment goods, the maintenance costs of dwellings seem not to follow the bathtub curve. This curve with initially falling, in the second phase constant and then rising maintenance costs is presumed to be typically for investment goods. However, this seems not to be the case for housing. After a short period of low maintenance costs the costs are rising quite rapidly. After 10 years approximately the annual increase is starting to slow down. The pattern can be described with a initially slowly growing logarithmic function. Further analysis needs to be conducted. These two observations of slowly decreasing rents and increasing running costs need to be further assessed for their stability. If the findings should hold it may be possible to calculate an optimal time for replacement of dwellings. Comparing this calculated economic replacement age with the observed replacement age will yield results that allow for the estimation of future development of the necessary construction rate, demolition rate and of the size, state and age of the building stock.