On May 2, SurveyUSA released some interesting single-state head-to-head primary polls featuring Clinton, Edwards, and Obama against Rudy Giuliani. Unfortunately, SUSA has recently become resistant to putting all their data on a single easy-to-use chart (this data is on 33 different pie graphs!) so I had to do the work. It's listed below, with the Democrat's number first and Giuliani's number second.
These were relatively small surveys -- about 550 people each, with a margin of error a little above 4%.
The set of states sampled is most favorable to Clinton among the Democrats, and probably inflates Giuliani's strength -- New York is in the mix, but neither Obama's IL nor Edwards' NC are. Only two Southern states are surveyed, which probably strengthens Rudy overall.
Obama's poor performance is surprising. He usually does better than this in head-to-head polls. How should we explain the disparity between his showing in national head-to-heads and his showing here? One factor -- though it probably can't be the whole explanation -- is that he has a tremendous level of support in heavily populated Illinois, which props him up in national polls despite his less spectacular numbers elsewhere.
Regarding an Edwards-Giuliani race, there is a spectrum of ways to go with this data. On one hand, you can take the data at face value and go with something the 1976 flipped-regions scenario. In 1976, the Democrats ran Carter and nearly swept the South after losing it massively four years earlier. (The numbers will blow your mind. Carter won by 13% in Alabama and 30% in Arkansas -- states that we'd lost by 47% and 36% respectively in the previous presidential election.) Giuliani's home state advantage will hold and he'll win New York. Edwards' regional advantage will hold in a New-Yorker-fearing South, and he'll make it up down there.
On the other hand, you can assume that as the race progresses, home-state advantages mostly dissipate. Edwards wins New York while Giuliani wins the majority of the South. I'm closer to the second of these poles, but perhaps not as close to it as most people are. I could see us trading NH for AR, and maybe there would be more flippage.
Whether you flip the regions or not, Edwards gets the Midwest and wins the election. He absolutely dominates in the traditional swing states. Edwards crushes Giuliani in IA, has solid leads in OH, MN, and WI, and is the only one to win WI and MO. If you regard Giuliani's new solidly pro-choice positioning (which emerged after this polling) as the end of his quest for the GOP nomination, Edwards' strength in the Midwest compared to the other Democratic contenders is still the major take-home message of this poll.
(Just to express a pet peeve, again: Do you people at SurveyUSA not want anybody to look at your data? Just a couple months ago, you'd put up your monthly 50-state tracking polls of approval ratings all on one sheet like this, sortable across 6 different criteria. When I heard that a new sheet of polls was out, I'd get all giddy. Dissecting those numbers was political strategy nerd heaven. Now every candidate is on a separate page and nobody can make comparisons without actually writing stuff down to remember it all. And putting the data above on 33 different pie charts, none of which are labeled on the page with the state they're from, borders on the sadistic.)