In 2000, the networks called states won by Bush about 15 minutes slower on average than states won by Gore even after controlling for other factors that might have affected call lags. There were no states that Gore won by six or more points that received delayed calls (i.e., were not called immediately upon poll closings) but there were nine states that Bush won by six or more points that received delayed calls. (See my article with Wilson Mixon and Amit Sen in the January 2004 issue of Public Choice for details.)

I haven't had time to run the numbers from last night, but ocular least squares (a nifty term I first heard from co-blogger Bob) suggests CNN might have been up to its old tricks. There were a number of states that Bush easily carried such as NC, SC, VA, and MT that received delayed calls. I'll update after I get a chance to fully analyze the call times.

UPDATE: I've now had a chance to analyze the call lags and have found a pattern similar 2000 (so much for the networks fixing the system).

Some quick descriptives:
States won by Bush with margins of 5-10 percentage points: 5
Number of those states receiving lagged calls: 5

States won by Kerry with margins of 5-10 percentage points: 5
Number of those states receiving lagged calls: 2

States won by Bush with margins of 10+ percentage points: 22
Number of those states receiving lagged calls: 6

States won by Kerry with margins of 10+ percentage points: 8
Number of those states receiving lagged calls: 0

Probit and Tobit analyses that control for each state's number of electors, pre-election poll margin (absolute value), and actual election margin (absolute value) find that, ceteris paribus, Kerry states were significantly less likely to receive lagged calls and that, ceteris paribus, Kerry states received significantly shorter call lags. ("Significantly" in the previous sentence refers to the 1% level.)

What to make of the results? There are at least two possible interpretations (other than random happenstance). First, the apparent media bias may really reflect systematic problems (deliberate or otherwise) in the methodologies (e.g., exit polling) that networks use to call states. Second, it could be a deliberate attempt on the part of networks to dampen voter turnout in states which still have open polls. Consistent with this interpretation is the fact that many of the Bush states with healthy margins but delayed calls are early closers NC, SC, VA, FL, MS, LA, and AR.