The Prediction of Close Elections: Comments on Some 1960 Polls

An examination of the results of 36 measurements of electoral preference or voting intention. The data indicate that 'the partisanly sponsored polls overestimated the actual vote for the candidate of the party for which the polls were done;' perhaps the pollsters desire to please their partisan employers may help account for this. It is seen that some of the less sophisticated sampling techniques may be quite satisfactory for many purposes. 'Most importantly it is suggested that for the really close elections - which tend to be also the really important elections for the health of a 2-party democracy - the pollsters, within the 6% points range they ask to be allowed, may have to rely as much as the politician himself on hunch & luck.' I. Taviss.