This paper describes the results of some simulations designed to study properties of confidence intervals based on the application of cross-validation in robust regression. Two procedures, namely the jackknife technique and the wave-interval procedure, were used to construct confidence intervals. We find that the robust regression procedure based on cross-validation is sensitive to multicollinearity. Moreover, we find that the wave-interval procedure performs very well when the data is not multicollinear, while the jackknife performs very poorly.