"Understanding the determinants of real estate investment performance is a necessary precondition for successful portfolio management. In particular the relationships between real estate returns and key economic factors such as GDP, total investment as well as unemployment, bank and inflation rates have been a key concern for researchers and practitioners. For the first time, this study investigates these relationships on a regional level, employing NUTS2 real estate return data for office buildings in the UK for 27 regions over the time period from 1981-2004. Real estate return data has been provided by IPD, regional economic data by Cambridge Econometrics. Conducting regression analyses for each region we identify the respective return determinants and, in a second step, use quadratic programming to construct diversified investment portfolios based on the identified relevant economic factors. Our findings clearly suggest that the relationships between economic factors and real estate returns found on a national level cannot simple be assumed to hold on regional levels. Constructing portfolios based on the analysis of national data can therefore be severely misleading due to extreme interregional differences with regard to the respective relevant factors. While strategies based on single economic factors such as GDP lead to only a minor loss in risk adjusted performance, any combination of economic factors severely reduces it. This holds true regardless of whether total return, income return or capital return data is used. Consequently, depending on the employed investment strategy (e.g. value added, opportunistic), portfolios have to be composed very differently.""