We present a robust and reliable methodology to calibrate and test the Constant Elasticity of Variance (CEV) model. Precisely, the parameters of the model are estimated by maximum likelihood, and an efficient numerical method to maximize the likelihood function is developed. Furthermore, a consistent and effective goodness-of-fit test of the CEV model is obtained using the Rosenblatt probability transformation and the ? 2 analysis. The novel procedure is employed to investigate the performances of the model on the Italian market. This analysis reveals that the CEV model does not offer a correct description of equity prices.