Time and Newsweek, I believe, skew their polls for desired political effect. Zogby has just flat gone Rat since 2000. Gallup seems pretty solid, and Mason-Dixon is good in state polls. Rassmussen seems to have just decided that they'll keep it a few points either way so they don't get embarassed again.
If Bush is close in Jersey and actually leading Minnesota, Pennsylvania and Iowa, I don't see any way it's a tight race nationwide. Of course, pollsters seldom do representative polling of dead people, dogs and invisible six foot rabbits. This constituency turns out in droves on election day, and is overwhelmingly Democratic.
We have to shoot for a six point win. There was little fraud in Texas in 2000, except in local races, because the Rats can't reasonably fraud out more than 1%-2% of the total vote. They don't bother committing fraud if the race is over 5% to the Republican because they just can't manufacture that many fraudulent votes. We have to have Bush up by 8 by election day if we want to avoid the lawsuits and the dead vote. I think Bush is leading by seven, which means in fraud terms that the race is dead even.
You are right to be concerned. Do you remember in late October, 2000, when several pollsters floated the scenario in which Bush would win the popular vote, but lose the Electoral College?
Something mighty fishy is going on with the polls. I prefer to ignore them, and try to get Republicans out to vote come hell or high water.