New Jersey usually polls close early with a high number of undecideds. These people do not even glance at the race until after Labor Day. They then give each candidate some thought. They usually break heavily for the Dems, but not always.
When the polls in 2004 showed the race tight they were not wrong. Polls are not predictors, they are a snapshot of how things look at that moment in time. People laugh them off because Jersey currently tends to break strongly Dem, but it is not a universal truth. NJ is one of the more thoughtful states when it comes to looking at candidates.
There were people on this site declaring that Kerry would take NJ by at least the same margin as Gore did and it would certainly not be closer than 12%. In the end, Bush lost by 6.7%.
You bring up a good point abut undecides breaking strongly for the Dems most of the time, and because of that the poll looks better for the GOp than the final result does. Would like to add to that. As you said it doesn't happen all the time (the strong break towards the Dems), but it does happen most of the time. The exceptions to that tend to be when the GOp candidate is not well known early on (Whitman's 1st run, Franks in 2000), when the GOp candidate is well known those undecideds in those polls tend to break heavily towards the Dems, and kean is a pure example of the wll known kind.
