David Sirota on the Clinton Campaign

He meditates on this graph:

The Clinton Firewall -- In These Times: The Race Chasm may sound like a conventional discussion of the black-white divide, but it is one of the least-discussed geographic, demographic and political dynamics driving the contest between Clinton and Obama... you are left with 33 elections that best represent how the black-white split has impacted the campaign... when you chart Obama's margin of victory or defeat against the percentage of African-Americans living in that state, a striking U trend emerges. That precipitous dip in Obama's performance in states with a big-but-not-huge African-American population is the Race Chasm--and that chasm is no coincidence.

On the left of the graph, among the states with the smallest black population, Obama has destroyed Clinton. With the candidates differing little on issues, this trend is likely due, in part, to the fact that black-white racial politics are all but non-existent in nearly totally white states.... On the right of the graph among the states with the largest black populations, Obama has also crushed Clinton.... "in the Democratic primary the black vote is so huge [in these states], it can overwhelm the white vote," says Thomas Schaller, a political science professor at the University of Maryland--Baltimore. That black vote has gone primarily to Obama, helping him win these states by big margins.

It is in the chasm where Clinton has consistently defeated Obama. These are geographically diverse states from Ohio to Oklahoma to Massachusetts where racial politics is very much a part of the political culture, but where the black vote is too small to offset a white vote racially motivated by the Clinton campaign's coded messages and tactics. The chasm exists in the cluster of states whose population is above 6 percent and below 17 percent black, and Clinton has won most of them by beating Obama handily among white working-class voters...

Comments

[John Howard Brown]

I have a lot of difficulty with the following sentence.

"....but where the black vote is too small to offset a white vote racially motivated by the Clinton campaign's coded messages and tactics."

I am a white man in Georgia who voted for Obama on the grounds of electability. (Anyone who would not vote for him because he is black has probably not voted Democratic in 30 years or more.) I, however, have trouble reading racial codes, etc into almost any of the Clinton campaign's statements. (E.g. No citing a "crime problem" as an anti-African American strategy.) To suggest otherwise makes the process more poisonous than it hase already become. I JUST WANT IT TO BE OVER!!!

[anne]

"These are geographically diverse states from Ohio to Oklahoma to Massachusetts where racial politics is very much a part of the political culture, but where the black vote is too small to offset a white vote racially motivated by the Clinton campaign's coded messages and tactics."

Absolute rubbish; and I have supported Barack Obama from the beginning. This sort of hate-filled deceiving rubbish really disturbs and shocks me. There is a etermined attempt to destroy Clinton that makes me ever more sympatheitic to her and worried about supporting Obama.

[derek]

There's something really bogus about that graph (apart from the usual mistake of encoding it as a JPG instead of a GIF or PNG). The states march from left to right in an orderly progression, allegedly from less to more black, with even spacing. And I'm supposed to believe that's the percentage? And where is the percentage scale on the bottom?

[anne]

John Howard Brown:

"I have a lot of difficulty with the following sentence.

" '....but where the black vote is too small to offset a white vote racially motivated by the Clinton campaign's coded messages and tactics.' "

Agreed; this is intolerable hate-filled slander; and even supporting Obama, I admire Clinton and will not be insulted and slandered for my admiration. I will vote for who I wish and support who I wish and admire who I wish, and the heck with such slander.

This is horrid. I admire Hillary Clinton, that is my right.

[JRoth]

Good catch, derek - if nothing else, a properly scaled graph would let us see just how clear the "chasm" is (e.g., are DE and AL really just two and three steps away from TN?).

Anyway, Sirota does the same condescending thing that all anti-HRC commenters do - ascribe to straight-up racism what can just as well be ascribed to demographics. HRC is the *type* of candidate that working class white Dems prefer (practical-minded, detail-oriented, familiar name) - this was never in question before January. The only questions before the voting began were whether working class white Dems were anti-HRC* enough to vote for Obama anyway, and whether HRC could get any of the black vote. It turns out that working class whites aren't misogynist - so they must be racist. Nice work, Democratic elites.

*Everyone had a story about their union-member uncle who would never vote for "that woman."

[derek]

I see I've done Sirota an injustice, and the original is indeed a 22KB GIF. Whose bright idea was it to turn it into a JPG, bloating it to 125KB while degrading the quality?

[Karl Bilawski]

Call me provincial, but I noticed immediately that California was not on the chart.

The 2000 census says that the African-American population of California is 7.4% of the population, equivalent to that of Oklahoma.

It appears that the percentage of Obama's win/loss was 42.5% - 51.9%, or -9.4%

This would have put it roughly equal to NV on the chart above. I wonder why it was left out? Did OK just look better, down below -20? Are they just leaving out data in the quest for clean lines?

2000 Census, the Black Population: http://www.census.gov/prod/2001pubs/c2kbr01-5.pdf
2008 California Democratic Presidential Primary Results:
http://politics.nytimes.com/election-guide/2008/results/states/CA.html

[anne]

J Roth:

"It turns out that working class whites aren't misogynist - so they must be racist. Nice work, Democratic elites."

Perfectly parodied.

[Adam]

Let me see if I understand this correctly. The 60% of white voters who support HRC in ethnically diverse states are racists, but the 90% of African Americans that support Obama in ALL states aren't.

Isn't there something terribly condescending about this line of argument, as if Sirota is saying something that makes white people racist pigs (i.e. voting for "one of your own") is somehow normal for African Americans?

[Jeffrey Harris]

The graph is interesting, but there is always the question of what the REAL independent variable is. It would be useful, for example, to plot it again with "percentage of work force that is unionized" as the x-axis variable, and also with some measure of the average age or per cent of population under 30 or over 65. jhh

[anon]

Clinton ran the professional, big-business (big politic) type of campaign - she circled big states and built an orgz there years ago - california, texas, florida, ny. this is why the win in texas was not all that much of a surprise - she has been on the ground longer and with more resources in states such as TX. Hillary clinton may have spent less time, money and attention on smalller states (is there a way to check this?). Obama won the small states and Illinois.

It is possible that Obama would have done better in bigger states if he had more money and more orgz earlier (years ago) vs. only recently having popularity and resources.

[stefan]

How White Attitudes Vary with the Racial Composition of Local Populations: Numbers Count
MC Taylor - American Sociological Review, 1998 - JSTOR

http://links.jstor.org/sici?sici=0003-1224(199808)63%3A4%3C512%3AHWAVWT%3E2.0.CO%3B2-V

[JRoth]

OK, following up a bit on my own question: I'm not sure these data are right. Unless there's been a surprising amount of population movement since the 2000 census, WI has more blacks than RI (6.1 vs. 5.5), and DE has (slightly) fewer than VA (20.1 vs. 20.4). There are, however, gaps between TN, VA/DE, and AL (16.8%, 20.1/20.4%, 26.3%).

So it looks like the shift does exist at the right side of the graph - once you cross the 20% black threshold, Obama wins every time. Given that a 20% black state has probably a Dem primary electorate that's 40+% black, I'm not sure that's saying much (or at least nothing that Bill Clinton didn't say in SC).

But at the left side of the so-called chasm, the percentages don't fall where Sirota says they do - according to the 2000 Census, the order should be MN (4.1%) AK NE CO (4.4%) RI (5.5%) WI (6.1%) KS MA NV (7.5%). Aside from the fact that laying out the numbers in this order doesn't give you the neat "chasm" that Sirota wants to tell us about (it meanders around the line for a a few contests there), you're talking about tiny margins. So a state that's four and a half percent black is a post-racial paradise, but five and a half percent is Fort Apache, The Bronx?

All of which is not to say that there's no there there. Just that the graph doesn't quite tell the story it purports to.

[Juris]

I am embarrassed for David Sirota. He appears to have taken selected data to prove a pre-determined narrative. Most caucus states did swing heavily to the Obama side. Two states that I know of -- Texas and Washington -- had both a caucus and a primary; and whereas both caucuses went for Obama by huge margins the primary votes went for Hillary or were essentially a draw. Washington was roughly 68% for Obama in the caucus vote and 51% in the primary.

Only 4 of the states shown on the left of the "race chasm" are from primary states while ten are caucus states, or only show the caucus vote result. Sirota has also conveniently left off Texas and New Hampshire. His conclusions about "race chasm" are based on nothing more than his wild imaginings.

All in all a very sloppy analysis, in my opinion. I say that with some sadness, because I generally have liked his writings and thinking.

[misplaced trust co.]

So when you throw out all the states (AR, AZ, CA, IA, IL, NH, NM, NY, TX) where this relationship doesn't hold, it appears to hold up quite well. Looks even better when you add FL, where the contest was a sham but the result fits the chart.

Likewise, Miami tied for first place in the NBA, counting only games it was leading at the end of regulation.

[Bernard Yomtov]

Not sure this is fair. I know a bit about TN, at least. Remember that Harold Ford came very close to winning a Senate seat there. I think it is unlkely that Clinton's win can be attributed to racial code words. I suspect rather that this is just a case of Clinton being seen as somewhat less liberal than Obama.

[meno]

Geez guys

You Americans are getting very toxic in your presidential campaign. And that's just against the rival candidates from your own party, who mostly have near-identical policies with the candidate you support.

Brad - do you really want to keep participating in this toxicity? Speaking as a foreigner who's read your blog for many years, I must say that I would have thought your willingness to wade into this cess-pit is surprising.

[Colin Danby]

It would be nice to see the scatter-plot properly scaled (anyone want to do that?) and without that silly red line. NH also seems missing and might disrupt the preferred story.

But there's an analytical point which I expect will still emerge from a more scrupulous presentation of data. Sirota suggests that states have distinct political cultures, and that people with similar individual-level characteristics may vote differently because of the way state-level political culture has accustomed them to process race. (As a WA resident with spells of residence in states across the x axis, this seems at least a plausible hypothesis.)

It is also empirically testable. In fact I can't imagine that someone hasn't already done this, and in a way that would allow us to say whether simple demographic differences in age or something else account for the variation instead. The availability of county-level data would also allow all kinds of interesting further tests. (If you have any geographical feel for parts of the U.S. , poking around in the county results available on the NYT site is fascinating.)

The "coded messages and tactics" part would be harder to assess empirically. But it seems far more productive to dive into the data and think about what it tells us about the world than to throw a fit any time someone raises race.

[Colin Danby]

Thanks Stefan for the ref while I was typing the above. It shows that the local-level effect is real. What we have less data on is how that effect influences national races.

[JRoth]

What toxicity, meno? Here? It's mostly kvetching about graphing data (and graphics formats, of course). We are all very earnest about it, tho.

[Colin Danby]

"ridiculously low number of voters"

You might find it salutary to compare turnout figures for this year's caucuses with turnout figures for those caucuses in previous years.

In any case, all someone needs to do is set up the regression, including a dummy variable for caucuses to capture that effect. No need to be yelling.

[jerry]

"white vote racially motivated by the Clinton campaign's coded messages and tactics."

Going to Paul Graham, Brad, linking to Sirota's nonsense is an unjustified name calling DH1 smear and should have been beneath you.

[Juris]

Colin -- "ridiculously low number of voters"

I can only speak to the Washington State results. We had a caucus and then a little more than a week later we had a primary vote (which was non-binding on the Dem Party, but still illustrative).

Total caucus voters for Dems was 31,984 of which Obama got 68%

Total primary voters for Dems was 691,381 of which Obama got 51%

I was a precinct officer and our caucus numbers were up by a factor of 3 from 2004; the other precincts that I was aware of had turnouts that were an order of magnitude higher than in 2004 -- to put it in some perspective.

Washington was only one of two states that I'm aware of that had both caucuses and primary voting -- and I think that they both tell pretty much the same tale -- that primaries and caucuses are very, very different.

If I remember correctly Texas results were:

Primary voters total approx. 2,800,000 of which Obama got 47%

Caucus voters total approx. 53,000 of which Obama got 56%

[Colin Danby]

I don't think we differ, Juris (I hope it was clear I was responding to Didier). I participated in the WA process too, and as I note above the primary-caucus differences are worth factoring in to an effort to empirically assess Sirota's argument. "Ridiculously low" just seemed weird given how high caucus turnout was here and in many other caucus states.

[p.lukasiak]

"but where the black vote is too small to offset a white vote racially motivated by the Clinton campaign's coded messages and tactics. "

I'm sorry, but this is complete BS.

If you look at the SC pre-election polling, what you find is that well before the whole "coded messages" controversy happened, blacks overwhelmingly supported Obama -- and the accusation of race baiting being made against the Clinton camp had little impact on the black vote...

PERCENTAGES OF AA VOTE FOR CLINTON AND OBAMA (SURVEY USA POLLING DATA LINKED FROM http://www.pollster.com/08-SC-Dem-Pres-Primary.php )

Dec 7-9 Clinton 39% Obama 56% (this is well before there were serious accusations of 'racially coded' messages)
Jan 4-6 Clinton 23% Obama 69% (right after the Iowa caucuses, where Obama proved he was a 'viable' candidate)
Jam 16-17 Clinton 20% Obama 74% Edwards 3% (this was at the height of the "race card" controversy)
Jan 23-24 Clinton 18% Obama 73% Edwards 6% (right before the primary--and after the controversy had died down)
Exit Polls (Jan 26) Clinton 19% Obama 78% Edwards 2%

Obama had the support of well over half of SC's African Americans well before the primary season began, and it appears that it was increasing "viability" as a candidate in Iowa that lead to his real gains among SC's blacks. The "controversy" itself provided Obama few additional black votes at Clinton's expense, and Obama's final total reflects the abandonment of Edwards by 2/3 of his black supporters.

In other words, black people aren't stupid -- Clinton lost less than 1/6 of her black support during the controvery, and black voters voted as an affirmation of their own identities -- Edward had spent 4 years talking about issues of importance to the black community, had gotten a plurality of the black voter in SC in 2004 (37%), but could gain no traction with black voters because of Obama's presence on the ballot as a viable candidate.

But, the fake controversy did have an impact on white voters in SC -- and I think it because they too recognized the false nature of the accusations, and found themselves identifying with Clinton's predicament. For a while, Clinton became the embodiment of the fear of white Democrats of being unjustly accused of racism...but once the controversy died down, she lost that status.

Dec 7-9 Clinton 46% Obama 18% Edwards 32%
Jan 4-6 Clinton 38% Obama 29% Edwards 28%
Jam 16-17 Clinton 50% Obama 22% Edwards 26%
Jan 23-24 Clinton 37% Obama 24% Edwards 34%
Exit Polls (Jan 26) Clinton 36% Obama 24% Edwards 40%

As with black voters, white voters significantly increased their support for Obama when he became perceived as a viable alternative to Clinton. But as a result of the controversy, Obama lost a quarter of his white support, all of which went to Clinton, who had the support of half the white voters thanks to the false accusations. White voters didn't go back to Obama once the controversy died down, however -- they understood what had happened, and how the Obama campaign was responsible for the false accusations. When Clinton's percentages returned to 'pre-controversy' levels, Edward was where the white voters once there was no longer a need to identify with Clinton's status as a victim of false accusations of racism

[p.lukasiak]

"Washington was only one of two states that I'm aware of that had both caucuses and primary voting -- and I think that they both tell pretty much the same tale -- that primaries and caucuses are very, very different."

this is an excellent point -- and while Obama supporters argu that the 'non-binding' nature of the WA primary skewed the results, on the GOP side, where both the caucuses and equally meaningful/meaninngless (half the delegates were awarded through each process, and the caucuses took place a week after Super Tuesday, when McCain had established himself as the presumptive GOP nominee) it becomes glaringly obvious had bad an indicator of popular support caucuses are.

McCain got 25%, and Paul got 22%, of the delegates elected in caucuses on Feb 9.
McCain got 50%, and Paul got 8%, of the popular vote in the primaries held on Feb 17.

But since you're from Washington, can you explain Romney's support. He'd quit the race right after Super Tuesday (and well before the caucuses) but got 16% of the delegates selected on Feb 9, and the same 16% of the votes on the primary a week later.

Are there lots of mormons in Washington, or don't Republican Washingtonians pay attention to the news, or what?

[Juris]

Colin -- I didn't mean to imply that I disagreed with you in general.

You were right on about the caucus turnout in Wa (and elsewhere). I just wanted to emphasize that while the caucus turnout was huge, it was still an order of magnitude or two lower than the primary vote turnout.

I was very impressed with your views on this issue -- so much so that, despite my better judgement, I decided to put the data on a spreadsheet and do a scatter plot with the same ordinate that Sirota used, but an abcissa that was linear with % Black in the state (from the 2000 census). So far, it looks wildly different than what Sirota published. I'm now getting the data for the states that Sirota excluded to see how the whole thing looks.

[p.lukasiak]

Juris --
I'm actually working on data from the 50 state Survey USA poll conducted at the end of February -- my emphasis so far has been on the gender gap, misogyny and sexism as factors in the November election, and my next (and final, I hope -- I want to examine the impact of racism and 'racial identity voting' on the landscape next) part of that series (which I published over at correntewire.com) looks at how race impacts the gender gap -- and a couple trends are already clear. (the series was published over at correntewire.com)

1) As the percentage of blacks among registered voters increases, the level of white support for Obama decreases.
2) Because of this, the gender gap actually decreases in states as the percentage of african american voters increases -- but the decrease is most significant in those states where blacks comprise at least 5% of the electorate, and the rate of decrease begins to fall off as you approach 20% (the greatest rate of decrease in the gender gap is between 5% to 14% of AA voters.)

The phenomenon can best be described is that a certain percentage of men are both sexist and racist -- they won't vote for either a woman or a black person. Women (belaboring the obvious, here) don't seem to have a problem supporting women, but a certain percentage of them are racist -- the gender gap closes as women vote more like men when the choice is between McCain and a black candidate (as opposed to a choice between McCain and a woman candidate.)

fun stuff, huh?

[Colin Danby]

Brad, if yer tuning in, maybe you could post Juris' scatterplot, when ready, as an update to the main post. Juris, if there's a way to use a different symbol for the caucuses in the plot that might help bring that effect out.

[p.lukasiak]

re the juris scatterplot...

the caucuses really do have to be kept separate from primary states for two reasons

1) as we've seen in Texas and Washington, caucus results are not reflective of the actual universe of people would vote if given a chance..

2) There is only one state with over 6% of african american voters in the overall electorate -- that's Texas, which held both a primary and a caucus. In fact, there are only two caucus states where (according to the Survey USA poll) blacks constitute more than 4% of registered voters; Kansas (with 5%) and Nevada (with 6% -- and in Nevada, Hispanics, not African Americans, are the "dominant" minority, comprising 15% of voters.)

[Chris Rose]

This was a great read -- measured and almost no flaming.

In any case, what would be really interesting would be predictions rather than a retrospective analysis. Folks who do prediction (in the financial industry, for instance) are well aware that retrospective models can easily be hot air. So, for any given theory, why not just calculate and post predictions (numerical! -- anything else is hand-waving) and see where they fall out after the contest in question?

Same thing with all the pundits and polls.

Overall, it would be really nice if someone kept a running tally of predictions versus reality.

Or maybe someone's doing this already?

[Juris]

Paul --

Looking at my plot, I sort of agree with you. But the scatter is so wicked below about 20% that I wouldn't try to make any conclusions about anything below that level. Sirota was able to stretch his analysis by relying on the weight of caucuses at the low end. Sadly, there are only two states that did both types of votes: WA where the swing was +32 to Hillary in a Primary vote; and in TX where the swing to hillary was +16 ------ relative to the caucus votes. It's obvious that Caucus votes are way higher than corresponding primary votes -- but by how much is the question. From the two data points (WA & TX) one could estimate that difference to be between 16 and 32.

COLIN --

I'd love to send the graphic If I know what format (I'm assuming gif) and where to send it. Or....... is there somewhere I could post it for general access. I could send/pst the Excel file as well. It's been fun -- but it does seem a bit obsessive doesn't it?

[Chris Rose]

Oh, and one other thing and it's the usual analytic complaint -- one needs to be very careful about drawing causal arrows from statistics. Perhaps many use the terms "racism" and "sexism" here as physicists use the term "God" -- it's used agnostically as a tag rather than a mechanism. That is, ascribing racist/sexist motives based on these data seems more than a little analytically shaky.

[Colin Danby]

Obsessive is good, Juris. I would e-mail it to Brad and hope that he finds time to post it.

--

Just two points on the more contentious bits. One is that "caucus results are not reflective of the actual universe of people would vote if given a chance.." strikes me as a category error. Conventions, caucuses, and primaries are various ways different state parties have devised to choose delegates, period. None of these means is necessarily representative of any other or larger group or supposed to be. Primaries certainly give you a bigger population to study, though, so they have advantages for testing certain questions and there is a plausible argument that the caucus and primary groups differ sufficiently in how they think about politics that one should be careful about mixing them in one regression.

Second, just to extend Chris' point, you can make the empirically-obvious point that race affects how lots of people see the world without imputing evil to them -- that was part of the point of BHO's speech the other day! We'd need methods closer to ethnography or psychology to work out how people think about their political choices.

Check out http://www.youtube.com/watch?v=r9IldaegAB0&e and just *listen* to what folks say: race is something produced in the here and know, not an immutable inheritance. And people will make or unmake race in different ways depending on circumstance. Having the conversation, and maybe changing the terms of conversation, means not flipping out or throwing fits the moment the subject arises, and if those kids can handle it so can we.

--

Oh and Chris, Al Giordano of The Field, probably the most interesting new bloggy voice this cycle, has been predicting pretty decently based on regional demographic breakdowns and and so forth: http://ruralvotes.com/thefield/

[Juris]

Colin:

Thanks for your good words. They gave me the incentive to do up a plot of the data and send a Gif to Brad.

I heartily appreciate your cautions on imputing race as the reason for differences in polling, caucus, or primary results of voting for whatever reason. In all my years I have come to the conclusion that most people very sincerely vote their honest, compassionate beliefs. They may see the world very differently than someone else, but that doesn't mean that it's based on anger. prejudice, or evil intent.

Thanks for your comments. As my late father used to say as his highest compliment -- "not bad."

[anne]

We are passing through an especially difficult time politically, when it is essential to be able to gain a sense of the perspective policies of candidates. David Sirota's boorish essay is a means of intimidating those who would seriously question and challenge policy positions of candidates, of a candidate for President in particular. I have questions I think important for Barack Obama, who I have vote for, but the questions are repeatedly buried by intimating Obama supporters and I am thoroughly angered by the intimidation and wondering if that is why there has been relatively more success in caucus settings.

I am not pleased enough with Obama's approaches on Social Security, health care insurance, tax cutting, increases in troop levels, willingness to leave Iraq, willingness to respect the territorial integrity of Pakistan and Somalia, policy in Afghanistan, and on. So I look for clarification and change, to strengthen my support. I will not be intimidated.

This was a horrid essay.

[anne]

Enough with intimidation; explain why Samantha Power, who surely knew, told us that Obama would only be leaving Iraq as a "best case scenario." Will Obama really be leaving Iraq completely and immediately? That is what I want to know to sustain my support.

Nothing coded, just explain policy to me.

[anne]

http://krugman.blogs.nytimes.com/2008/03/31/why-edwards-hasnt-endorsed-yet/

March 31, 2008

Why Edwards Hasn't Endorsed Yet
By Paul Krugman

John Heilemann * on why John Edwards hasn't endorsed anyone yet:

"So appalled was Edwards at Clinton's gaudy corporatism...that he'd essentially called her corrupt."

But then why didn't he endorse Obama after dropping out?

"According to a Democratic strategist unaligned with any campaign but with knowledge of the situation gleaned from all three camps, the answer is simple: Obama blew it. Speaking to Edwards on the day he exited the race, Obama came across as glib and aloof. His response to Edwards's imprecations that he make poverty a central part of his agenda was shallow, perfunctory, pat. Clinton, by contrast, engaged Edwards in a lengthy policy discussion. Her affect was solicitous and respectful. When Clinton met Edwards face-to-face in North Carolina ten days later, her approach continued to impress; she even made headway with Elizabeth. Whereas in his Edwards sit-down, Obama dug himself in deeper, getting into a fight with Elizabeth about health care, insisting that his plan is universal (a position she considers a crock), high-handedly criticizing Clinton's plan (and by extension Edwards's) for its insurance mandate."

* http://nymag.com/news/politics/powergrid/45604/

[Shannon]

it would be interesting to see your prediction of how the remaining primary states will vote on this basis ???

[marc sobel]

http://thenexthurrah.typepad.com/the_next_hurrah/2008/02/obama-won-caucu.html
Obama won caucuses, Clinton won primaries

Color coding caucuses vs primaries would make this graph a lot more interesting but less (whatever it is when you accuse your opponent of playing the race card)

[Juris]

This should be the final punctuation on this topic.

An excellent, fairly rigorous, and balanced analysis of Sirota's "Race Chasm" plot can be found on Brendan Nyhan's site:

http://www.brendan-nyhan.com/blog/2008/04/need-to-be-care.html

It covers most of the concerns in this discussion thread.

I salute you Brendan Nyhan !

[Chris Rose]

Colin: Thanks much for the comments and great pointers.

Now, here's a challenge to the folks doing the stats. Make a quantitative prediction (right now!) based on your analysis and post it for later deconstruction!

Yes, these are certainly non-stationary processes and most of our machinery presumes stationarity. But without predictive power, the modeling is essentially a discussion of angels on the head of a pin.

So, to be specific, how about making a prediction for N.C., Penn and others based on the Sirota claim (which though it suffers from all the various problems mentioned, does seem to indicate some correlation of Obama support -- or White lack thereof --with percentage of Black population) and then making a perhaps more measured prediction using the corrections to the data (which seem to result in the same trend, though the data look scattered enough that I doubt anyone sane would stake a pinkie finger -- amputation, like in the old short story -- on it :) ).

Any takers? Or maybe I missed the memo -- I'm an infrequent visitor.

[Tonymil]

One major flaw in all this analysis is that it assumes race is the only factor in this primary when in fact gender has been a huge factor. These charts completely fail to take gender into account and are terribly skewed as a result. The assumption is that Obama has been losing the white vote in states where the black population is greater than 6 percent. But, if you separate white female voters from white male voters, the assumption falls apart. Obama has lost the white female vote in almost every state, so that is a constant. But Obama has won the white male vote in many states, including states like California, Connecticut, Georgia, Maryland, Massachusetts and Virginia, states with black populations greater than 6 percent. So these charts don't give a fair picture of what has happened in these democratic primaries. And they are an even worse predictor for a general election between Obama and McCain. In fact, the white vote may flip-flop in the general election. White women, given a choice between Obama and McCain, may overwhelmingly choose Obama, and white men may overwhelmingly choose McCain. So I don't see the utility of these charts in either analyzing the democratic primaries or predicting results in a general election.