| Abstract: | ABSTRACT This paper reports on the findings of the Internet component of the 2001 BES and compares them with those of the other BES pre-election surveys. Part 1 outlines the rationale that underpins the introduction of Internet polling as a supplement to more traditional methods of assessing mass public opinion. Part 2 describes the marginal distributions on the key dependent variable-the projected vote shares of the main political parties-of the three pre-election polls that were conducted using BES questionnaires. Intriguingly, the (unweighted) Internet-based poll provided a better guess of the actual vote shares in the subsequent election than either of the two conventional polls. Part 3 provides a more detailed comparison of the profiles of the face-to-face and Internet-based polls. It shows how the Internet poll, compared with the face-to-face poll, was skewed demographically towards the professional classes and politically towards the Conservative Party. Part 4 explores the extent to which the use of the Internet poll might result in spurious causal inferences being drawn about the sources of voting preferences in the 2001 UK election. A simple, direct-effects causal model is estimated using both the face-to-face probability sample data and the Internet survey data. The results suggest that, although the raw probability and Internet samples differ significantly, the relationships among the key variables do not differ significantly across the two samples. We conclude that Internet polling has an important part to play in gauging and analysing public opinion in future UK elections. |
