OFDM is a promising digital communications technique for high data rate transmissions. In this paper, we have analyzed the performance of mobile OFDM systems in the presence of channel estimation error. A new channel estimation error model is presented. Based on the estimation error model, new and simple expressions for the average bit error probability of M-QAM OFDM systems are derived for zero-forcing, minimum mean square error and maximum likelihood receivers. A complex correlation coefficient is used to quantify the quality of the channel estimation scheme. We have shown that if the complex correlation coefficient is a function of the signal to noise ratio, then the inter-carrier interference caused by channel variations is the dominant source of performance degradation. On the other hand, if the channel estimation scheme produces a constant value for the complex correlation coefficient (estimation quality is not a function of the signal to noise ratio), then the channel estimation error will cause an error floor that is larger than the one caused by the inter-carrier interference, and is considered as one of the dominant causes of degradation. And channel estimation using pilot symbols are discussed in terms of different pilot symbol placement schemes.