Do the polls really tell us anything?
Now… Reuters/C-SPAN/Zogby 10/27 - 10/29, 1179 LV, 50-43 Obama (+7)
Then… Reuters/Zogby 10/29 - 10/31, 1208 LV, 48-47 Bush (+1) … and Bush won by 6.5%
Now… GWU/Battleground 10/23 - 10/29, 1000 LV, 49-46 Obama (+3)
Then… GWU/Battleground 10/31 - 11/1, 1000 LV, 50-46 Bush (+4)
Now… ABC News/Wash Post 10/25 - 10/28, 1316 LV, 52-44 Obama (+8)
Then… ABC News/Wash Post 10/28 - 10/31, 2904 LV, 49-48 Bush (+1)
Now… IBD/TIPP 10/24 - 10/28, 894 LV, 47-44 Obama (+3)
Then… TIPP 10/30 - 11/1, 1041 LV, 50.1-48.0 Bush (+2.1)
Now… FOX News 10/28 - 10/29, 924 LV 47-44 Obama (+3)
Then… FOX News 10/30 - 10/31, 1200 LV 46-48 Kerry (+2)
Now… Gallup (Traditional) 10/27 - 10/29, 1825 LV, 50-45 Obama (+5)
Now… Gallup (Expanded) 10/27 - 10/29, 2437 LV, 51-44 Obama (+7)
Then… CNN/USA/Gallup 10/29 - 10/31, 1573 LV, 49-49 (TIE)
Depending on whether you’re an optimist or a pessimist, 2004’s results show that either the pollsters underestimated the strength of the poll leader by 2.5% to 8.5%, or they undercounted the Republican vote by those margins. Which means either Obama’s headed for a landslide win, or he’s going to win by 1.5% or lose by as much as 5.5% if one of those pollster’s margin of error is the same as before.
That’s what you can depend on from most pollsters: confusion because they’re wrong. The most accurate of those pollsters last time was GWU/Battleground and if they err the same this time, Obama will win by 0.5% or by 5.5%. TIPP was second most accurate. Its same error would have Obama winning by 7.4% or losing by 1.4%.
Confusion. And since all have upgraded their models, their error rates will differ. So what’s an analyst to do?
Mark Blumenthal displays the latest trends which shows little or no narrowing of Obama’s lead.
Steve M. explains that Fox News has cooked its numbers to display a closer race than the other pollsters have… except for TIPP and GWU/Battleground, who were closest in 2004.
Nate Silver directly questions TIPP’s current methodology while asserting that it’ll be harder for every pollster to deliver results as accurate as 2004’s. Yet he also indicates McCain’s too far behind in the states that count to make up the margin he needs to win, barring some external event.
Still. Confusing.
So what can we say for certain about what the polls tell us?
Simple: go vote. Your vote might make the difference, especially in one of the most contested states. I mailed in my ballot last night so it should be recorded on Monday.
Have you voted? Let there be no excuse this time to prevent yur vote this time. If all new registrants vote, there’ll be a Democratic president. If they don’t show up, Sarah Palin could be our president. Which only guarantees that in the next four years, you won’t get attacked by Bullwinkle.
October 30th, 2008 at 7:56 pm
Actually, Bush won by 6.5% in the Electoral College — he won by 2.46% in the popular vote.
October 31st, 2008 at 8:48 am
Get Out the Vote. That’s all. If the Obama ground game is better than the repubics, he wins.
So get off your computer and knock on doors, drive voters, and get them to the polls…
November 1st, 2008 at 12:48 am
You’re right Steve. My error.Which means the pollsters TIPP, Zogby, WaPo and Battleground were closest, respectively. And that means, if heir error rates hold, Obama should win by 3.36% or 8.46% or 9.46% or 1.46%. The (mathematically rough) average of those 4 would be +5.69% for Obama.