This paper uses simulation methods to investigate the accuracy of vehicle speed measurements derived from anonymous tracking of cellular phone calls. In particular, these measurements are tested for their sensitivity to positioning algorithm (an angle-angle and an angle-hyperbola system are tested, both engineered to a level of accuracy consistent with FCC requirements for Extended-911 use), angle-of-arrival accuracy, time-difference-of-arrival accuracy, sample size, variance of true vehicle speeds, and sampling distance. The results suggest that such a surveillance system would be capable of stratifying observed vehicle speeds (and congestion levels, insofar as one is a predictor of the other) into at least three categories, such as low, medium, and high. While many other parameters and error sources can and should be investigated, individually and jointly, these results provide a qualitative foundation for a more thorough investigation of the efficacy of such systems.