Reply #55: the missing link (and GSS tables online) [View All]

[OnTheOtherHand]

(it wasn't from foo_bar at all -- my confusion -- foo_bar has been posting snippets of NES codebooks; this was posted by Internut)

http://www.icmresearch.co.uk/reviews/2002/failure-of-the-polls-1997.htm

Tough sledding. Table 4 provides some figures on faulty memory. (In this article, ICM is actually defending their practice of weighting based on reported past vote, which conceding the existence of faulty memory. The gist of the argument seems to be that the other firms' samples are so faulty that any anchor is useful, and besides, that in Britain the net effect of faulty memory seems to be decreasing.)

You can generate your own evidence for the U.S. at
http://webapp.icpsr.umich.edu/GSS/

Click on "Analyze," then select "Frequencies or crosstabulation (with charts)"

Enter Year as the column variable, pres88 pres92 pres96 pres00 (or your choices) as row variables.

Note that Clinton beats Dole by 19 points in 1998, but by 31 points in 2002. Bush beats Dukakis by 8 points in 1989 (close to the actual margin), but by 41 points (yes, 41) in 1993. Poor Mike Dukakis. All anyone seems to remember about him was that he looked stupid riding around in a tank.