I spent a few hours this week digging through the statewide polls and trend estimates on Pollster.com -- and the electoral maps created by others -- to try to resolve the polling argument of the moment: Which of the Democratic candidates for president makes the strongest candidate in the fall?
I come away with a warning: Although it is easy to make states light up as red or blue on a map depending on whatever poll, polling average or estimate you trust, the underlying data indicate a close race regardless of the Democratic nominee. Hillary Rodham Clinton may have a slight advantage for the moment, but the inherent shakiness of horse-race results at this stage of the campaign should give us all pause about reading too much precision into the current poll numbers.
Consider a few important cautions:
1. Lots of variation in national polls
Two recent high profile surveys -- from Newsweek and Gallup -- show Clinton doing slightly better than Barack Obama against John McCain in the national popular vote (as does the Rasmussen Reports automated survey). But other surveys conducted earlier in the month by ABC News/Washington Post, Quinnipiac University, Reuters/Zogby, IBD/TIPP, GWU Battleground and National Public Radio show Obama doing slightly better than Clinton against McCain.
Our trend-line estimates on Pollster.com, based on all national polls, show Clinton leading McCain by three points (47 percent to 44 percent) but Obama in a virtual tie (45 percent each). The RealClearPolitics averages, based on the seven most recent national polls, show Obama leading by 3 points (47 percent to 44 percent) and Clinton ahead by just one point (46 percent to 45 percent).
With individual national polls producing conflicting results within sampling error of each other, and with different methods of averaging producing slightly different results, we ought to be especially careful about seeing any clear advantage for either Democrat.
2. Incomplete state level data
Of course, we choose presidents with the Electoral College, not the national popular vote, and a close contest nationally makes the state-by-state math even more important. Do the state-level polls show one candidate with a pronounced advantage?
One big challenge in this regard is the sparse number of polls in individual states. Even in traditionally competitive “battleground” states like Florida and Ohio, where polling is most plentiful, we still have just two new polls each in May. Moreover, in many states, the only available surveys come from SurveyUSA and Rasmussen Reports, two companies that use an automated methodology rather than a live interviewer.
Whatever your judgment of automated polls, the bigger issue may be placing too much faith in just one survey from just one pollster. In Colorado, Nevada and Arkansas, for example, the only public polls since March all come from Rasmussen Reports.
3. Offsetting advantages?
The Clinton campaign is trumpeting its candidate’s stronger polling performance against McCain than Obama in Ohio and Florida -- two states that George W. Bush carried in 2004. However, the available polls point to similar advantages for Obama in Iowa, Colorado and Wisconsin.
In a report released this week, the Gallup organization found similar evidence. Data from the last two weeks of the Gallup Daily tracking survey showed Clinton running six points ahead of McCain (49 percent to 43 percent) and Obama trailing by three (43 percent to 46 percent) in the swing states where Clinton received more popular votes in the primaries (Nevada, Pennsylvania, Ohio, New Hampshire, New Mexico, Arkansas, Florida and Michigan).
However, Gallup also found Obama with an 8-point advantage over McCain (49 percent to 41 percent) and Clinton at a 1-point disadvantage (45 percent to 46 percent) in the swing states Obama carried in the primaries (Colorado, Oregon, Minnesota, Iowa, Wisconsin and Missouri).
Give Clinton the edge here, if only because her potential advantages extend to more electoral votes than Obama. However, polling shows both candidates as competitive as they need to be, for the moment, in states that can provide the 270 electoral votes needed to win.
4. Snapshots change
The biggest limitation of this exercise is that we are poring over horse-race numbers in May for an election still more than five months away, but the current numbers are far from meaningless. As political scientist Tom Holbrook shows us, polls from May of 2004 were reasonably accurate as predictors of the winner in November. However, not all years show as much stability in vote preferences as 2004.
Consider 1992. At this point in 1992, six different national polls (found in the archives of The Hotline) showed Bill Clinton running third behind both George H.W. Bush and Ross Perot. Clinton received an average of 26 percent of the vote in late May and early June. He was elected president five months later with 43 percent of the popular vote.
And the big caveat to keep in mind is that the Democrats are still in the midst of a highly contentious nomination battle. I cannot put it any better than Hotline editor Amy Walter in her April 2 column:
"Asking Clinton supporters whether they'd pick Obama in the fall is like asking a couple in the middle of a divorce whether they think they'll be friends next week. This question will be much more relevant after the summer is over and everyone's had a chance to, as Bill Clinton would say, 'chill out.'"
(ABC's Gary Langer made essentially the same quote in his column on Tuesday).
So can polls tell us definitively who will be stronger in the general election? As a concise reader on Pollster.com puts it, “Nobody knows, it's freakin' May.”