I've re-merged and re-analysed 4 times now, and it's undeniable. Very, very unstable voters who indicated a different vote intent every survey, saying after the election they decided a long time ago or were always going to vote that way. Those who posted emotional messages about Bob Hawke dying just two weeks later saying it didn't affect them at all.
I trust the pre-election data. It's the post-election data I think is wrong.
But so much of election study is done post-election. Is it all bad? Is it all not reflective of genuine voter behaviour when they walked into the booth? That would render the entire catalogue of the Australian Election Study since the 1980's bin-worthy.