: it is too subject to interviewer bias. Since there is to be no up-to-date list of households, simple random sampling is out. Besides, if attitudes are to be affected by the distance from the proposed site, the households should be broken up into strata, with households of comparable distance in the same strata. That leaves as choices "stratified, multistage cluster sampling" and "stratified random sampling". But if there is no up-to-date listing of households for the city, there are none for the individual strata, either. So random sampling within each strata is out, too. This leaves stratified multistage cluster sampling as the best of the choices.