August 31, 2004
As The Nation Goes, So Do Swing States
Posted by | 10:24 AM
It really is pretty amazing how fast the conventional wisdom can change.
Three weeks ago, most political insiders in both parties gave Sen. John Kerry a slight edge over President Bush. Granted, Kerry's lead appeared to be only 2 or 3 points, but it showed up consistently in the national polls and was corroborated by public and private polling on the state level that showed Kerry ahead in seven or eight of the 10 most competitive battleground states. Experienced Republican operatives, particularly pollsters, were worried. Their Democratic counterparts were
pinching themselves.
Since then, Kerry appears to have lost a point or two, maybe three, and Bush has picked up a point or two. State polls are showing Bush ahead in five or six of those same 10 battleground states.
On one level, this change tells us that even the most subtle shifts can move a state from red to blue, or blue to red. A new CNN/USA Today/Gallup poll shows Bush and Kerry tied in Pennsylvania with 47 percent each; just a couple of weeks ago, Kerry was 5 or 6 points ahead, according to a variety of polls. A pair of Florida newspapers released a poll on Sunday showing Bush and Kerry tied in the Sunshine State; a couple of weeks ago, polls showed Kerry ahead by as much as 4 points.
Senior Bush campaign strategist Matthew Dowd pointed out yesterday that the Bush campaign hadn't taken a national poll in almost two years; instead, it has been polling 18 battleground states. These polls, Dowd said, reflected almost digit for digit what the national polls showed. Rarely do you see one candidate gaining in one state when the tide seems to be flowing the other way.
Bush campaign operatives argue that one cause of this small shift from Kerry to Bush was Kerry's statement that he would have attacked Iraq. My own view is that Kerry has been dinged by the questions raised about his record in Vietnam. Plus, the swift-boat controversy dominated the political news coverage, suppressing other issues. A week when the focus is on the economy and jobs, or on Iraq and casualties, the management of the war, and weapons of mass destruction is a good week for Kerry and a bad week for Bush. When the focus is on almost anything else, it's very likely to be a good week for Bush and a bad week for Kerry.
The point is that in the absence of some major external event or a monumental screw up by Bush or Kerry in this fall's presidential debates, neither candidate is likely to build a significant, sustainable lead. One can look at all the relevant factors in the race and shade it in one direction or the other.
For example, I put great weight in the enormous levels of pessimism among undecided voters and their apparently low opinion of Bush. I think the president's climb is still a bit uphill. My experience tells me that undecided voters invariably break against well-known, well-defined incumbents.
Bush strategists acknowledge that the undecided voters are a tough nut to crack. But they argue that the campaign can offset the undecided voters who will break for Kerry by turning out a pool of conservative and Republican-leaning infrequent voters. Given the experience of 2002, when Republicans were able to elevate voter turnout far above normal in their strong areas, this is a plausible tactic, although it's obviously harder to do in a presidential election when turnout is going to be higher anyway.
The bottom line is that this election wasn't over three weeks ago when
Kerry was ahead, albeit narrowly, and it isn't over now that President Bush is ahead by a comparable margin.
***
Correction: In my Sunday Convention Daily column on presidential election computer modeling, I erroneously reported that the forecast by Yale University's Ray Fair predicted a Gore landslide victory in 2000. In truth, Fair's model predicted that Gore would win 50.8 percent of the major-party vote and said that the race was too close to call. The actual Gore (major two-party) vote percentage was 50.3 percent -- pretty darn close.
This column appeared in the Tuesday edition of National Journal's Convention Daily. Click here for the complete edition.
 |
As a professional poller (I gave up political polling 20 years ago, and now mostly poll for litigation), I find Charlie Cook's comments about the change in the race surprising. I think the key is to understand that there are effectively three sets of pollers operating today, and they are getting three sets of different results. The reason for the differences is almost certainly a combination of how they define likely voters, and how hard they push the undecided to get answers. Thus category 1 which gets the most Republican vote (I'm not saying they are biased in that direction just that their results come out that way) includes, for instance, all those who use Gallup methods. Category 2 is those who have never seen any edge for either candidate, the best exemplars being ABC and Rasmussen. Group 3 is those who have consistently shown Kerry much further ahead, with Zogby being the best example.
Consider polls around the end of July, beginning of August - Gallup had Bush up 4, An average of two ABC polls had a dead heat (Rasmussen's average result was also roughly even), and Zogby had Kerry up 5.
Now consider results at the end of August - Gallup had Bush up 3, ABC had a tie (Rasmussen had Bush up 1) and Zogby/Williams had Kerry up somewhat over 4.
In other words there has basically been no change in the results from most of the specific pollers (the theory holds for among others NPR, Pew, and the Economist). What changes over time is which pollers have produced results - anytime a new poll comes out from the L.A. Times instead of CBS/New York Times Bush looks better and vice versa. In other words, the changes aren't in the race, they are in which pollers poll at specific points in time, and which ones get publicity.
While seemingly all the pollsters and pundits emphasize the small number of undecided voters (as low as 3%), very little attention is paid to what can be called the "soft decideds," those who seem to have declared their intentions one week only to change their minds the following week. These are the voters most easily swayed by campaign commercials and media focus (as with the Swift Boat ads), as well as daily events. I am not a professional pollster, but it seems to me that there must be some way to measure the size and demographics of this fluxional group beyond the rudimentary questions posed in most polls about how strongly an individual is committed to a particular candidate. One can argue that the election may well be decided not by the sliver of truly undecideds but by the substantially larger pool of "re-deciders," those who claim to be decided and are therefore assigned to one candidate's column or the other only to change allegiances at any given time.
All of the analysis and re-analysis and just plain over analysis of these polls (and, increasingly, the polls themselves) are becoming pretty ridiculous and tiresome. Let's just wait until the election is over (note: I did not say November 2).
Posted by: gerald f at August 31, 2004 10:59 PMCharlie, as always you make some good points. Given how steady the results of the various pollsters have been, once you identify and factor in their polling techniques, isn't the best course of action here to simply wait until well after both conventions are over and any bounce has been absorbed to draw any conclusions?
For example, can't it be argued that we should all just wait until about Septembeer 15-20 and then look at the polls to see what we have heading into the debates, before making assessments of this race?
Posted by: Steve Soto at September 1, 2004 11:04 AM
