[Marxism] Class and race in the 2008 U.S. Presidential Vote (was: RE Completed Election Demographics) Joaquin Bustelo jbustelo at gmail.com Sat Nov 15 17:51:40 MST 2008
Previous message: [Marxism] Completed Election Demographics E-Mail Next message: [Marxism] Class and race in the 2008 U.S. Presidential Vote (was: RE Completed Election Demographics) Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] -------------------------------------------------------------------------------- Jeffrey Thomas Piercy writes: "After the election, I started looking at CNN.com's huge list of statistics, copying ones I thought were interesting. I was probably there for 45 minutes before I had to leave and I'd only gone through about half of them. "The more income a person had, the more likely they were to vote for McCain," except, as Jeffrey noted later in his post, for the very top income demographic, $200K or more, where Obama's support increased somewhat. * * * Before I get to the meat of the matter, some background, because the origins, validity, usefulness or seriousness of these numbers has been called into question. The statistics that have been presented are from the general election national exit poll. They're not just CNN's but are done by the Edison/Mitofsky polling consortium for the "National Election Pool," composed of the major TV networks that do news (CNN, FOX, ABC, CBS, NBC [including MSNBC]) and the Associated Press wire service. Other news organizations also subscribe to the poll, I think through AP affiliation. Apparently, all members of the National Election Pool are allowed to call the one and only common exit poll "their own" exit poll. CNN says it's the CNN exit poll and ABC says its the ABC exit poll. But there is one and only one poll. The National exit poll was done at 300 precincts, with an interviewer stopping every nth person as they came out (fifth, seventh, or whatever), asking them to fill out the questionnaire. The number surveyed at each precinct was approximately 50, but the exact figure varied depending on turnout and cooperation rate. I do not know what the cooperation rate was this year; in recent election cycles (2000 and 2004) I believe it averaged just over 50%. In addition, about 2400 phone interviews with absentee/early voters were conducted and integrated into the poll in the right weighted proportion. Those 2400 phone interviews were obtained the same way normal public opinion phone surveys are done, by "random digit dialing." The total sample size of the national poll was almost 18,000. There were also 51 separate and distinct exit polls done in all states and Washington, D.C., with the same basic methodology, except that the number of sample precincts per state varied anywhere from 10 (Utah) to 50 (large "battleground" states like FL, OH, PA). How the number of precincts are assigned to each state is unclear to me, although "battleground" status seems to be a criteria. For example, Virginia had 50 precincts; North Carolina, 45; Georgia had 30, as did California and New York. As in the national poll, where needed state exit polling was supplemented by phone interviews with early/absentee voters using the same questions. These data were then integrated into the overall exit poll in what the polling experts considered to be the right proportion, depending on the proportion of early voters. Oregon was a phone-only poll because all voting is by mail. At a precinct, when someone declines to fill in the questionnaire for the poll, the poll taker takes note of the person's apparent race/ethnicity, age and sex. Exit poll results are "weighted" to take that into account, the overall "weight" of that precinct (obviously, it would be goofy to give the same weight to 50 interviews from a precinct of 300 as 50 interviews from one of 3,000) the actual vote at the polled precinct, and the statewide, regional and/or national projected or actual vote tabulation (depending on the state of the race). One of the uses of the exit poll is to help so-called "decision desks" at the various networks and AP to project who wins each state's electoral votes and a few other important races on election night, but predicting exact percentages of votes in a given state or nationally is not what exit polls are designed for, and in fact, publication or broadcast of the OVERALL exit poll percentages in a race (the so-called "top line") is not allowed, because the design aims not so much at seeing which candidate voters backed, but which voters backed which candidates, and why. Some people say exit polls shouldn't be weighed, or at least that the "raw numbers" should be reported as a check for election fraud. This is not sound. Exit polls aren't truly random samples of the population. They are a sample of precincts, i.e., neighborhoods, anywhere from a 10 to 50 per state. If you want a polling check on how people voted, probably the best instrument would be a phone poll. But even there, you have big problems, especially the low response rate which requires adopting the unproven hypothesis that in THIS election, those who could not be reached and those who refused to cooperate are just like those that you can reach and do cooperate, so that they do not bias the sample. There is no way to prove that, so polls are no substitute for clean, honest, and transparent elections. (It should be noted that, by international standards for free and fair elections, U.S. elections don't even pass the giggle test in every possible way: from the impartiality of top election officials [there is not even a pretense of this], to reasonable access to mass media for all recognized parties [again, not even a pretense] to transparent accounting for each and every vote cast [again, not even a pretense.] Leaving aside, of course, the undemocratic, anti-working-class, racist, winner-take-all by state, electoral college flim-flam.) The weighing of exit poll raw numbers is a GOOD thing from the point of view of the purpose of the poll because the small sample for each precinct and practical difficulties in conducting a poll imply a very large margin of error at that level, and different sizes of precincts and other factors further distort things. The impact of this can be reduced by weighing the precinct, state and regional polls by actual vote counts. One more preliminary matter. Though not usually publicized, exit polls have a claimed margin of error at a confidence level of 95%, depending on the size of the sample and how that sample was put together. For this year's national poll, those percentages are, up to 100, 15%; 100-200, 10%, up to 500, 7%; up to 950, 5%; up to 2350, 4%; up to 5250, 3% and up to 8,000, 2 %. 8,000 or more gives a 1% margin of error. That's worst-case for a 50-50 split. a 95-5 split only has a 6% margin of error even with a sample of 100 or a little less, and only 2% with a sample of only 500+. * * * The overall vote-by-income data that was presented here seems to show a strong correlation between income and vote. Obama, who presented himself as pro-working people, pro-unions, and endorsed as such by the labor movement, moreover, an image re-enforced by the fact he is Black and from a working-class family, as well as by the (admittedly now quite frayed) traditional identification of his party, received the lion's share of the vote of the poor and working people of modest means and 60% of the vote of those earning less than $50K, who represented 40% of the electorate, but only half the vote above that income level. The suggestion is that this represents some degree of class identification or consciousness or instinct --however mistaken on might consider it to be in being focused on Obama-- on the part of a large segment of the voters, using income as a proxy for class. I believe the idea that Obama would be perceived as the more working class or pro-working people candidate is entirely reasonable; other candidates with a more legitimate claim to that distinction were completely ignored by the media, and McCain by contrast re-enforced Obama's "class" image, with everything from losing count of the number of houses he owns to his "economic fundamentals are strong" reaction at the beginning of the financial panic to Palin's $150,000 wardrobe makeover. So the question is, did this matter to people, as might be inferred from vote by class, using income as a proxy for class? I think closer examination of the data --including more detailed breakdowns by race than have been generally available-- show this "class effect" is much more limited than it first appears, and insofar as voting for Obama (however mistaken one might consider such a vote to have been if motivated on this basis) might be taken as an indicator of class consciousness or instinct, what it shows is the ABSENCE of this as a factor among most Anglo (white) working people. Much or most of the differentiation in vote by income was in fact attributable to race/nationality. Obama's most concentrated support came from the Black community, 95% overall. There is no real evidence that the Black vote varied by income AT ALL except perhaps at the very top of the scale, > $200K, where is was 90% for Obama (compared to 95% overall). But the sample of Blacks with such incomes in the national poll is so small (3% of 2300 respondents, perhaps around 70 individuals) that the reason I might believe the poll reflects something real is more because it makes sense to me that a very thin but nevertheless somewhat larger layer of very rich Blacks might be pro-McCain. Statistically the result isn't significant. Add to this that a disproportionate number of those in lower income groups who were Blacks and Latinos. 14% of Blacks were in the less than $15K group, 10% of Latinos, but only 5% of whites. Overall this demographic was 6% of the exit poll. More than half of the Blacks in the national poll (54%) had family incomes of < $50K, and nearly half the Latinos (48%) but only a third of whites (34%) did. Conversely, as you go up the income ladder, you get fewer and fewer Black voters. The result is you'd see a strong correlation between income and vote, simply because of this factor: Blacks, who voted homogeneously regardless of income, are more concentrated in lower income groups. Fortunately, through a friend I was able to see print outs of the white, Black and Latino vote in the national poll, as well as Latino votes in some states. The real question here is white folks, and among them the Obama vote by income nationally was as follows: < $15K 58% $15K-30K 49% $30K-$50K 42% $50K-$75K 41% $75-$100K` 44% $100K-$150K 41% $150K-$200K 44% > $200K 48% Obama gets more support from the poorest whites, the first two categories, a total of 16% of the whites polled -- but even then, far from overwhelming. Beyond that, there is simply no evidence of a "class effect." For what Americans might call "lower middle class" and up, there really isn't any difference. The number bounces around a bit between 41% and 44%, which I believe is "noise" because the sample sizes for each income category are such that EVEN IF exit polls were perfect random surveys (which they are not), the margin of error would be at best +/-3%. So for the bulk of the white respondents, those making between $30K-$200K, 77% of those surveyed, Obama's support averaged 42, and didn't vary significantly. And finally we get to incomes of $200k or more, and Obama's support jumps back up to 48% among white folks. While the 48% who voted for Obama in the top income group isn't strictly speaking statistically significant at the 95% confidence level, it is a big enough sample of such people (around 800), and a big enough difference, that (if we did the math) we could say something like that while not 95% certain this is a good result, it is true at a 90% or some other fairly high level of confidence that more such people voted for Obama than those making between $30K and $200K. A somewhat different pattern of support versus income is presented by Latinos. the more significant findings: The tenth (10%) of poorest Latinos surveyed backed Obama 83-15. That's a small sample, but the figures are so lopsided that the margin of error is small, +/-5%. Looking at a bigger segment, we see the nearly half (48%) of Latinos surveyed making less than $50K were a little less strongly pro-Obama 75-23, but still a crushing 3-1 majority. The third (33%) making $50K-$100K gave Obama a still smaller but nevertheless impressive lead of 19 points, (58-39), as did the fifth (19%) making $100K or more (59-40). There is a weirdness in the more detailed breakdown that I think is a statistical fluke: for Latinos with family income between $150K and $200K, McCain WINS 51%-48%, whereas among those with $200K or more, Obama wipes the floor with McCain 70%-29%. Here is where a having at least a common-sense understanding of statistics is very important. BOTH these samples are 4% of a total national Latino sample of 1350. That works out to VERY small sub-samples of perhaps 54 in each case. With those sample sizes, even the roughly 20-point swing registered is not statistically significant, it is within the margin of error at even a level of confidence much reduced from 95%. For that reason, I don't think I'd even be comfortable saying something like that the exit poll hints that Obama might have had more support among Latinos earning more than $200K than among those earning $150K-$200K. Because if you add those two together you get a 59-40 Obama/McCain split, pretty much the same number as for ALL Latinos making more than $50K. However, looking at the state exit polls for places like Florida, Texas, California, and New Mexico, where there were enough Latinos in the poll for at least partial breakdowns of vote by income among Latinos, my impression is the national figures are a little artificial, in the sense of being a purely mathematical sum of somewhat different patterns, and the evidence from the 4 state polls, limited as it is because they're much smaller, is suggestive of a clear "class effect" in the Latino community even beyond the $50K level. This should hardly be surprising, as a large number of Latinos (albeit not mostly voters) come from countries with strong class consciousness and traditions, and these sorts of ideas, and of self-identification as workers, is therefore much stronger in the Latino community generally than among Anglos, where the dominant ideology has it that everyone, or almost everyone, is "middle class." Complicating things among "Latinos" as a group is the Cuban community, which sociologically and politically is quite distinct from other Latinos. Although only 4% of all Latinos, they're probably at least double that in terms of voters. But when you look at a state with very little Cuban weight in the community, and very large immigrant cultural and social influence, the vote-by-class (income) patterns seems to largely hold true. In California, support among Latinos making less than $50K is 83-12, and then goes down to 69% among 50-100K and, above $100K to some figure in the low 60's (the exact number isn't given, but can be surmised from the fact that among all Latinos making >$50K, Obama's number is 66%, but it is 69% when the cutoff is $100K, meaning those over $100K are significantly lower). In New Mexico (with the largest Latino component of a state exit poll, more than 1000) support among those with income under $15K is 88-11, and $15K-$30K is over 79-21. For the $30k-$50K bracket it drops to 58%, only to bounce back up in the $50K-100K bracket to 64%, and 67% above $100K. Almost certainly, I think, the 58% figure is off, an expected statistical anomaly given that the sample size for that group is less than 200, and the rise to 67% above $100K in income might also be a fluke, but on the high side, since the sample is about 120 people at that income level. And it is important to understand this was a poll in New Mexico of a total of 40 precincts, and close to half were probably close to all-white, so when we're talking about these better-off Latinos, you're talking about three, four or five precincts, neighborhoods, and thus must regard the results as only mildly indicative. * * * Throughout I have counterposed "race/ethnicity/nationality" to "class" to try to separate out what is clearly and solely attributable to "class," using income as a proxy. It should be noted, however, that in a more general political sense, I believe this to be utterly false when it comes to Blacks and (most) Latinos, because their national identities as peoples includes strong undertones or overlays or elements of class. In the case of the Latino community, this could not have been more blatant in the massive immigrant rights demonstrations of a couple of years ago. But I believe it is also true of Blacks generally. Not, of course, of every single individual. But the counterposition is most decidedly true for white/Anglos, where a true manifestation of class instinct begins with rejection of identifying with people like the Clintons and McCains as having the same interests, concerns or needs that you or I have. Nationwide, among whites whose material circumstances make it very hard for those individuals to entertain illusions that they, too, are "middle class," there was a stronger Obama vote. And it should be noted also, that in some states where the Obama campaign especially pressed very hard on issues to appeal to working class voters, like Ohio, there was a differentiation: OH white Obama vote by income < $15K 62% $15K-30K 51% $30K-$50K 47% $50K-$75K 47% $75-$100K` 48% $100K-$150K 41% $150K-$200K 36% > $200K 41% But in other states, like PA, it simply didn't happen: PA white Obama vote by income < $15K -% $15K-30K 50% $30K-$50K 47% $50K-$75K 44% $75-$100K` 53% $100K-$150K 44% $150K-$200K 54% > $200K 55% < $50 K 49% $50K-$100K 47% > $100K 49% To the degree that an Obama vote could be taken as a "class differentiator," the exit poll data, suggests this was truer for Latinos than for any other group. It also appears to have been true only for the lowest socio-economic strata of whites and more broadly in Ohio and a couple of other "battlegrounds" but not generally: on the contrary, the general "lower" to "upper" middle class white vote was surprisingly homogeneous. Blacks, of course, voted almost unanimously for Obama, taking the well-established community tradition of tending strongly towards block voting to an even higher level. Joaquin This message has been scanned for malware by SurfControl plc. www.surfcontrol.com _______________________________________________ Marxism-Thaxis mailing list [email protected] To change your options or unsubscribe go to: http://lists.econ.utah.edu/mailman/listinfo/marxism-thaxis
