Bigger turnout favors Obama (0.00 / 0)

The polling has mostly underestimated turn-out for Obama. I go with Obama +12.

---

by: cmpnwtr @ Tue Feb 19, 2008 at 12:12:32 PM CST

Dunno (0.00 / 0)

what to think about this one.  I have a hard time seeing Obama lose.  It would be a complete reversal of everything that we've currently seen, all the trends.

So I'll go with Obama +8.  I don't think he will come out all that far ahead on pledged delegates, but it would be nice if he could increase his lead by 10 tonight.

---

by: Cycloptichorn @ Tue Feb 19, 2008 at 12:23:40 PM CST

12 delegate net gain for Obama (4.00 / 1)

Obama will get 56% of delegates in WI for 8 or nine delegate increase and 60% of Hawai'i for 4 more, for a net gain of 12 delegates.

However the win will spell a big drop in popular support for HRC. Look to her losing Texas and a disappointing small net gain from Ohio when that primary comes.

This race is over I think, we are just waiting to see how much damage the 'contest' is doing to the party.

Obama needs a great early evening victory speech, calling for unity and common purpose. Setting forth a vision of the shining city on the hill. (Duh)

He cannot run against Hillary any more. It is time fort Obama to start running against Bush's third term in office.

Change
"We must break up the banks and never again let them get so big that they distort our politics and take down the economy.

---

by: HousesofProgress @ Tue Feb 19, 2008 at 12:41:38 PM CST

[ Parent ]

On Ohio, new poll, Obama closing gap. (0.00 / 0)

Two Weeks to Must-Win Ohio, Clinton 9 Points Atop Obama: In a Democratic Primary for President of the United States in Ohio today, 2/16/08, two weeks till the votes are counted, Hillary Clinton defeats Barack Obama, 52% to 43%, according to a SurveyUSA poll conducted exclusively for WHIO-TV Dayton. But her lead has narrowed. Just one week ago, Clinton held a 17 point lead at 56% to 39%.

http://www.whiotv.com/politics...

From a 17 point lead to a 9 point lead in one week. At that rate, (I know it's not scientific) Obama wins by 7 on The day.

Ohio was the last, last, last firewall for HRC. This kind of drop is significant.

Change
"We must break up the banks and never again let them get so big that they distort our politics and take down the economy.

---

by: HousesofProgress @ Tue Feb 19, 2008 at 13:14:29 PM CST

[ Parent ]

You are right on here (0.00 / 0)

and your other comments as well.

If Obama wins WI and HI, and wins at least delegates in TX and comes close in OH, wins VT and RI, it would seem to be all but over.  

John McCain--He's not who you think he is.

---

by: Mimikatz @ Tue Feb 19, 2008 at 15:01:20 PM CST

[ Parent ]

It's an open primary. (0.00 / 0)

Aside from New Hampshire, it seems like Obama tends to get an extra boost in states with open primaries, due to the influx of indies/Republicans/college kids, etc., that tend to break in his favor.

We'll see.

---

by: nautilus1700 @ Tue Feb 19, 2008 at 12:42:46 PM CST

Polls and African-American support (4.00 / 1)

Thought about a diary but this'll do as a comment here, I think.

We expect to see both candidates overperform their percent support in the last polls available because undecideds pick a candidate, and for the most part, they have been splitting their votes between Clinton and Obama (favoring Clinton somewhat).

But, looking through exit polls and the data from SUSA and PPP, Clinton actually underperforms in many states - almost all of those states with an African-American electorate of about 30% or more - DE, AL, GA, MD, SC, VA (but not TN).  This is the same list as Chris has above...

Looking closer at the data, we see two reasons for this: the polls overestimate the support of Clinton among African-Americans, and underestimate the turnout of African-Americans.

In the following graphs, if polls were perfect, we would see all the dots fall on the line.  If polls had random error, we would see an equal number of dots on each side of the line.

Click to enlarge.

The above shows more African-Americans showed up on election day than the pollsters thought, on average, as most of the dots are above the line.

Click to enlarge.

The above shows, on average, African-Americans gave less support to Clinton than pollsters thought, as most of the dots are below the line.

What does this mean for Wisconsin? Obama could do a point or two better than polls based on the history of polling African-Americans this cycle alone.  PPP polled Wisconsin, so we will know tomorrow.  However, a swing towards Clinton in other demographic groups could offset this advantage.

---

by: dreaminonempty @ Tue Feb 19, 2008 at 12:49:06 PM CST

I'm not sure your numbers bear you out (0.00 / 0)

That's an interesting analysis, and thanks for doing it. But it looks the statistics don't bear out your conclusions very well. Just looking at your scatter plots, it looks like the p-values for your conclusions are in the 0.25 range or so. Not very strong.

The truth about Saxby Chambliss

---

by: nstrauss @ Tue Feb 19, 2008 at 13:05:34 PM CST

[ Parent ]

Hmm, I probably stated the case too strongly. (4.00 / 1)

Thank you for your criticism.  Folks, pay attention: I should have put more modifiers in the original post.  The data are suggestive, but not strongly conclusive.

But one thing I didn't show is the regression of Clinton's underperformance (or overperformance) versus percent African-American voters, which has R2 value of 0.56 - pretty decent.

---

by: dreaminonempty @ Tue Feb 19, 2008 at 13:18:15 PM CST

[ Parent ]

Interesting (0.00 / 0)

Just did a back-of-the-envelope calculation, diving the above data into three groups: pre-2/5, 2/5, and post-2/5. My theory is that his rate of outperforming the polls has been increasing through time.

And it's correct. The pre-2/5 average is Obama +3.2; the 2/5 average is Obama +5.6 and the post 2/5 average is Obama +9.7.

It's obviously not statistically significant given the low number of observations but it suggests that he has been doing better and better as the process has continued.

Based on this, and a gazillion other factors, I'll go Obama +14.

---

by: Matt H @ Tue Feb 19, 2008 at 12:50:30 PM CST

Also, (0.00 / 0)

The leader in the final polls has won the actual vote 22 of 28 times, Obama faltering only in NH and NM.  Clinton lost the outcome when leading in the polls 4 times.  But generally, it seems that numbers break in favor of the person already ahead in the state.

---

by: Ron in Chicago @ Tue Feb 19, 2008 at 12:53:07 PM CST

caucus variation is to be expected (0.00 / 0)

It makes sense that caucuses would have a lot of variation because the percentages are actually the percentage of delegates to the state convention won rather than the raw popular vote.

---

by: redwards95 @ Tue Feb 19, 2008 at 14:01:09 PM CST

Wisconsin Win (0.00 / 0)

Well based on this graph, Hillary could still squeak out a win of around 2%(gaining 10% over the weekend) with people suddenly getting cold feet. But Obama has won 8 in a row convincingly - so I don't think people feel they are going out on limb the way they did in N.H. So I don't expect that kind of reversal. On the other hand, there was also a reversal in N.M. But there she only jumped plus 7% which get her a slim loss here.  

---

by: C.B. Todd @ Tue Feb 19, 2008 at 14:43:12 PM CST

My Comments (0.00 / 0)

1. I think it should be noted that many of the states that Obama heavily over performed were states with large black populations. This shows that either the pollsters are doing a bad job predicting the turnout of blacks or the blacks aren't being entirely truthful. Perhaps with a black man on the ticket, they're coming out to the polls in greater numbers than they have in the past. Also, there could be a different kind of reverse Wilder effect going on, whereby the blacks tell the person on the phone that they're going to vote for the white person, but then when they get into the voting booth, they switch.

2. The opposite is happening in largely hispanic states. California, New Mexico and Arizona all broke for Clinton.

3. Standard deviation might add something to the table. In South Carolina the polls varied tremendously, while in Virginia, they all basically showed the same thing.

---

by: Saddlebags12 @ Tue Feb 19, 2008 at 14:55:54 PM CST

awesome work - one suggestion (0.00 / 0)

there is one part of the data that is not clear from this, and that is when the polls not just underestimated the victory, but where predicting a victory for the other candidate. it would be cool to see those cases, like Newhampture colored red or some other color so we could see those distinct from those that unpredicting the victor but still picked the right winner.

Michael Bloomberg, prince of corporate welfare

---

by: Will @ Tue Feb 19, 2008 at 15:20:35 PM CST

GoBama!!! (0.00 / 0)

I hope your projections turn out to be correct!

My sense is that more and more folks are starting to become dis-illusioned with the slame old "slick and tired"paradigms.

The way that Obama is expanding the base of the Democratic party is exciting; I pray he's rewarded with the nomination.

Clinton had this sewed up a year ago; all she had to do was to admit being conned into an erroneous vote, get really tough on Bush and his enablers, and adopt a genuinely populist message; she did neither...and, thereby, left herself open/vulnerable to the appeals of Edwards and Obama.

I hope that the recent negative campaigning doesn't blunt the turnouts or turn off the Obama "new blood".

---

by: Nick Lento @ Tue Feb 19, 2008 at 15:50:10 PM CST

early voting (0.00 / 0)

It would also be interesting to see which of these states had early voting. The early voting would not be subject to the late swings that happened in several states, notably California.

---

by: Saddlebags12 @ Tue Feb 19, 2008 at 16:49:39 PM CST