Two of the most discussed issues of this election cycle are whether or not pollsters should control for party identification and whether or not the fortune being spent to register new voters and for GOTV will have an impact on the election.
The Gallup poll has taken the most flack for not controlling for party identification and that failure is seen by some as being responsible for Gallup's results fluctuating wildly and for being at odds with other polling. Ruy Teixeira of Donkey Rising, Steve Soto of the Leftcoaster and Chris Bowers of Mydd have led the charge suggesting that Gallup's results are unreliable because of large swings in the party identification of its samples.
Are they right? Frankly, I am not qualified to say.
A second issue has been whether or not the money spent registering new voters and GOTV efforts in general will have an impact on the race. With various groups spending in excess of $300 million, one would expect that some impact but questions remain. Will the newly registered voters actually turn out? Are the numbers of newly registered voters reported accurate? Simply increasing the number of Democratic leaning voters will not necessarily change the election. If my car accelerates to 90 mph, I still lose ground if yours is going 95. Have Democtratic efforts significantly exceeded GOP efforts both nationally and in the states that matter? Once again, I can't say.
Those questions, however, prompted me to take a much harder look at some recent polling in a few swing states. What follows is what I found.
Wisconsin
Wisconsin is a very important swing state. Its 10 electoral votes might well tip the election either way. Gore won Wisconsin by a razor thin margin in 2000 and both parties are competing vigorously there this year.
The American Research Group (ARG) released a Wisconsin poll conducted fron October 16-19. The top line results showed a dead tie with both Bush and Kerry receiving 47% of the vote, Ralph Nader drawing 2% and 5% undecided.
My initial reaction to those numbers was that the Incumbent Rule would cause the undecided to break for Kerry and that Wisconsin, therefore, looked promising. If the five percent undecided broke 3 for Kerry and 2 for Bush, the final result would be a Kerry win, 50%-49%-1%.
Looking closer at the poll, I noticed that ARG reported the party identification of its sample. They report 35% Democrats, 33% Republicans, and 32% independents.
ARG also reports its cross tabs. Kerry drew the support of 7% of Republicans, 86% of Democrats, and 45% of independents.
George Bush drew 88% of Republicans, 8% of Democrats, and 46% of independents.
Given those levels of support, what would the result be if turnout in 2004 mirrored that of 2000?
ABC News provides exit poll data from the 2000 election. In 2000, the Wisconsin electorate was 37% Democrats, 32% Republicans, and 31% Independents.
I then applied the 2000 exit poll party identification information to the cross tabs of the ARG Wisconsin poll. The results are
Bush 45.5%
Kerry 48.0%
Now that is interesting. Kerry moves from an exact tie to a two and a half point lead if the Wisconsin electorate's party identification in 2004 is identical to 2000. Unless the GOP has gained some ground in party identification in Wisconsin since 2000, or does a much better job of turning out its voters, George Bush is in big trouble in Wisconsin.
New Hampshire
The Wisconsin analysis peaked my interest, so I did the same thing with New Hampshire. The ARG New Hampshire poll has Bush at 47%, Kerry at 46%, Nader at 1% with 6% undecided.
The cross tabs showed Bush with 82%, 11%, and 42% among Republicans, Democrats and Independents, respectively. Kerry's cross tabs show 12%, 84%, and 49% among those same groups.
ABC News 2000 exit poll data, shows a New Hampshire electorate of 32% Republicans, 24% Democrats, and 44% Independents.
Applying the ARG cross tabs to the 2000 exist poll data, I get:
Bush 47.4%
Kerry 45.6%.
Applying the 2000 exit poll party ID to the cross tabs increases Bush's lead from one point to almost 2. Democrats need to hope that demographic change, voter registration efforts, GOTV efforts, or the incumbent rule helps them in New Hampshire.
New Mexico
ARG's New Mexico poll shows Kerry with a two point lead, 48%-46%, with Nader drawing 1% and 5% undecided.
In the cross tabs, Bush draws 88% from the GOP, 16% from Dems, and 43% from Independents. Kerry draws 8% from the GOP, 77% from Dems, and 49% from independents.
ABC 2000 exit poll data shows a New Mexico electorate with 43% Dems, 32% Republicans, and 25% Independents.
Applying the ARG cross tabs to the exist poll data results in the following:
Bush 45.8%
Kerry 47.9%
Like Wisconsin, Democrats do not need to win the registration and GOTV battles to win New Mexico. They do need to hold their own in those efforts.
Ohio
All of that is preliminary, of course. What you really want to know is what is happening in Ohio and Florida. Let's take Ohio first. Bush won Ohio in 2000 by three and a half percentage points even with Nader drawing 2.5% and Gore abandoning the state due to a lack of resources. It is ground zero in this year's election.
There have been lots of recent polls in Ohio. Mark Blumentahl, the Mystery Pollster, has a great post on interpreting those polls in light of the Incumbent Rule. The various horse race numbers range from Bush +5 in a Fox poll to Kerry + 3 in the ABC poll (links available at Mystery Pollster). I wanted to use Fox for my analysis but it did not report cross tabs. I then chose the Ohio Poll.
The top line of that poll has Kerry + 2, 48% to 46%.
The cross tabs have Bush pulling 90% of Republican support, 7% of Democratic support, and only 26% of independents.
John Kerry has 7% of Republicans, 88% of Democrats, and 55% of independents.
ABC's exit poll data shows an evenly divided 2000 Ohio electorate of 37% Repubicans, 38% Democrats, and 26% Independents.
Applying the Ohio cross tabs to the 2000 exit poll data we get:
Bush 42.7%
Kerry 50.3%
Wow. Adjusting the Ohio Poll to the party turnout from 2000 ups Kerry's lead from 2% to more than 7%. President Bush needs to greatly increase GOP turnout, find some way to increase his support among independents or find some way to 270 electoral votes without Ohio.
Florida
Florida, with 27 electoral votes, is the big enchillada of swing states. The least favorable to John Kerry of recent Florida polls is the Mason-Dixon poll. It has a top line of Bush + 3 with Bush receiving 48% support, Kerry garnering 45%, others (Nader) drawing 1% and 6% undecided.
Other polls, such as Survey USA (Kerry +1), Insider Advantage (Kerry + 4), and the University of North Florida poll (Kerry + 1) show a Kerry lead. For all Florida polls, see 2.004k.com and scroll down to Florida.
The Mason-Dixon cross tabs show Bush getting 88% of Republicans, 14% of Democrats, and 39% of Independents. Kerry gets 7% of Republicans, 79% of Democrats, and 49% on independents.
ABC's 2000 Florida exit poll shows a Florida electorate of 40% Democrats, 38% Republicans, and 22% Independents.
When we apply the Mason Dixon cross tabs to the party ID of the 2000 exit poll data, we get:
Bush 47.6%
Kerry 45.0%
If Democrats want to win Florida, Kerry needs to increase his vote share among Democrats and Independents, or demographic change since 2000 needs to favor the Democrats, or Democratic voter registration and GOTV efforts need to change the party ID mix, or the indecideds need to break heavily for Kerry.
Of course, by choosing the Mason-Dixon poll, I favored the Republicans. If we choose the Survey USA poll that has a top line of Kerry 50%, Bush 49%, and apply its cross tabs to the 2000 election poll data, we get:
Bush 46.0%
Kerry 52.7%.
The major difference in those two polls is that Survey USA has Kerry solidifying his base, drawing 91% of Democratic support while Mason Dixon has him languishing at 79% of Democrats. That difference translates into a 4.8% bump when applied to the 2000 exit poll data. Who is right? Who knows.
My guess is that the Democrats have some work to do in Florida both in terms of getting out the vote and in terms of persuading soft Democrats.
Conclusion
Conventional wisdom is that Democrats must win new voter registration efforts and must also win the GOTV battle to prevail in this election. My analysis suggests that if Democtrats simply turn out an electorate in 2004 that is identical in party identification to the 2000 electorate, they will win such swing states as Wisconsin, Ohio, New Mexico, and, perhaps, Florida.
That conclusion is premised on the cross tabs of the selected polls being accurate (a huge assumption), the numbers remaining static over the next 13 days (an even larger assumption), and the incorrect assumption that there have been no demographic changes in the swing states since 2000.
Nonetheless, I entered the analysis with the idea that Democrats needed big wins in both registration and GOTV to win swing states. That may not be the case. Simply not losing ground in those areas may be enough. That said, everybody get back to work in those phone banks.