We assess the noise characteristics in time series of daily position estimates for 23 globally distributed Global Positioning System (GPS) stations with 3 years of data, using spectral analysis and Maximum Likelihood Estimation. A combination of white noise and flicker noise appears to be the best model for the noise characteristics of all three position components. Both white and flicker noise amplitudes are smallest in the north component and largest in the vertical component. The white noise part of the vertical component is higher for tropical stations (¿23¿ latitude) compared to midlatitude stations. Velocity error in a GPS coordinate time series may be underestimated by factors of 5--11 if a pure white noise model is assumed. ¿ 1999 American Geophysical Union