In　this　paper,　it　is　aim　to　propose　an　instrumental　variable　rank　estimation　method　for　varying　coefficient　spatial　autoregressive　models.　The　newly　proposed　method　provides　a　highly　efficient　and　robust　alternative　to　the　existing　quasi-maximum　likelihood　estimation　or　GMM　estimation,　and　can　be　implemented　using　the　existing　R　software　package　conveniently.　Under　mild　conditions,　the　consistency　and　asymptotic　normality　of　the　resulting　estimators　are　established.　The　finite　sample　properties　of　the　proposed　method　are　investigated　through　Monte　Carlo　simulation　studies.　Finally,　the　Boston　house　price　data　and　crime　data　of　Tokyo　are　analyzed　to　illustrate　the　usefulness　of　the　proposed　estimation　method.