Non-contact skin moisture measurement based on near-infrared (NIR) spectroscopy is proposed in the spectral range from 1300 to 2000 nm. A gap is introduced between the optical fiber probe and the skin surface in order to avoid occluding surface vapor. In vitro and in vivo experiments for measuring the water content of skin are implemented. The measured absorbance spectra are processed by multivariate analyses. Processed results are compared with the water content values obtained by a capacitance method. The correlations between the optical method and the capacitance method obtained by partial-least squares regression are higher than those obtained by multiple linear regression. In addition, a Monte Carlo simulation is implemented to evaluate measurement depths of the optical methods. It is presented that the measurement depth of the optical method depends largely on the water absorption. The simulation result also shows that the measurement depth of the optical method is much deeper than the depth of the capacitance method, especially in the spectral range where water absorption is relatively weak.