Last year there was a lot of criticism of pollsters and polling companies because so often two different companies polling on the Presidential race nationally or in different states would come to such different conclusions. This was often treated like it was some sort of new development, perhaps caused by the proliferation of companies active in gauging the state of the 2008 race, and that back in the 'good old days' polling was much more consistent.
That's why I was amused to read this week in the memoir of JFK adviser Ted Sorensen that during a one month span in late 1963 Gallup came out with a national poll showing Kennedy leading Barry Goldwater 53-40 while Harris came out with one showing Goldwater leading Kennedy 54-38: a 29 point discrepancy!
Clearly this is not a new development, and indeed there are few times from last year where I remember polls being this divergent. Given how Goldwater ended up faring against LBJ I'm going to have to guess Gallup was right in this instance.