Policies based on (flawed) exit polling are suspect, imho. Better to remain true to principle.
Michael Barone was right in 2000; he was right in 2002; and I'm willing to bet he's right in 2004.
You may call him flawed because it doesn't jive with your agenda, but you do so at the risk of undercutting your argument.
Michael Barone is one of the most respected political "statisticians" in this day and age.