Research on hedonic office rent prediction is well established and numerous regression models have developed by various researchers. Proposed regression models incorporate linear or log-linear solution procedures or, in some research work, both. Mostly, comparisons between linear and log-linear models indicate a slight improvement in the explanatory powers of the log-linear representation, though the improvements are rather trivial [1, 2]. Some models prefer change in rent, as the dependent variable, while others specify the rent value itself, by making use of asking rent, contract rent or effective rent data. The sample data is, asking rents, as they may be obtained from ads or real estate firm reports, cannot produce reliable results. Instead, the use of contract rent, or even more appropriately, the use of effective contract rent, considering the leasing terms and concessions, is imperative for modelling more accurate rent predictions [3, 4, and 5]. However this necessity brings about the difficulty in obtaining sufficient undisclosed contract data from real estate firms [5]. The aim of this paper is to examine problem with constructing an office rent prediction model for Istanbul. For this purpose, a proposed regression model will be constructed with using contract data and asking data between 2000-2006. The proposed models will also incorporate the linear and log-linear solution and will investigate a whole range of parameters, classified as econometric, contract, building and location. The study was conducted as part of research for developing a viable office rent prediction model for Istanbul.