In the Capital Asset Pricing Model (CAPM), beta coefficient is a very important parameter to be estimated. The most commonly used estimating methods are the Ordinary Least Squares (OLS) and some Robust Regression Techniques (RRT). However, these traditional methods make strong as sumptions which are unrealistic. In addition, The OLS method is very sensitive to extreme observations, while the RRT methods try to decrease the weights of the extreme observations which may contain substantial information. In this paper, a novel fuzzy regression method is proposed, which makes less assumptions and takes good care of the extreme observations. Simulation study and real word applications show that the fuzzy regression is a competitive method.