<p>This paper proposes a single-index varying coefficient spatial autoregressive model which has measurement errors in the index covariates. We combine a local-linear smoother based simulation-extrapolation (SIMEX) algorithm, the estimation equation and the profile maximum likelihood method to estimate our model. Under some regular conditions, the estimator for the nonparametric part is proved to be asymptotically normal at any fixed point, and the estimators for the parametric part are derived to be asymptotically normal as well. Some simulation studies indicate our estimation method performs well. Finally, our method is illustrated with the real dataset of Boston Housing Price.</p>