[Marxism] Hillary Spin [was: Democrats class war]

[Joaquin Bustelo]

[Warning: this is a fairly detailed discussion of exit poll math, mostly to
challenge the claim (originating with Hillary Clinton's campaign) that she
is drawing working-class support whereas Obama's supporters are more
upscale. I'm sure I could do a better job of this explanation with a lot
more time, but that I haven't got right now. --Joaquin]

Phil writes:

There's an interesting piece by right-wing political columnist Gerard baker
on this.  I read in the Christchurch 'Press' here in NZ, but it was
reprinted from the London 'Times'.

He argues that dunkin' donut Democrats are supporting Clinton and latte
liberals incline more to Clinton.

Here's an extract:
Mr Obama wins disproportionately among people who may be considered the
winners in the global economy: the well educated, the mobile and the
financially secure. Mrs. Clinton's voters are the strugglers, the class that
feels itself left behind by an increasingly unfair global economic system.

Consider the exit poll from California, the largest state to vote on Super
Tuesday. Mrs. Clinton's largest single demographic voting bloc was those who
did not complete a high school education, where she won 82 per cent, against
just 15 per cent for Mr Obama. The more educated you became - from high
school drop-out, through high school graduate then some college, college
graduate and finally postgraduate - the more likely you were to vote for Mr
Obama. The only category he won, in fact, was the propeller heads with
postgraduate degrees.

*  *  *

The Clinton campaign is putting out this sort of "spin" of the exit poll
data at a furious pace, and has been since Iowa, but intensifying greatly in
the past few days. But there are big problem with it. To put it in the best
light for Hillary, it is greatly exaggerated, if not outright false.

The clearest and sharpest demographic group in Hillary's camp are retirees,
those 65 and over. This distorts both the income data and education level as
proxies for class.

Retirees as a whole tend to have lower incomes than people still working.
And there is a huge difference in the percents graduating from high school,
college and getting post graduate degrees by AGE, with a  skyrocketing of
the college population after Sputnik, i.e., in the 1960's, i.e., precisely
when Hillary's support among whites begins to plummet.

You would have to control for age/retirement status using age 65 as a proxy
in order to make the assertions made in this article stick. 

With full access to the exit poll data, one could run cross-tabs (i.e., look
at subgroups of, for example, 18-29 year olds by income level to see how
they voted). However, I've seen some cross-tabs from the exit polls from
last Tuesday (not the particular one I mention) and the size of the
sub-groups is problematic, because the margins of error skyrocket. For JUST
the 18-29 demographic in the California poll, the largest poll of this
election season, you wind up with a sample of just 300, which means a margin
of error of +/- 6%. But you'd have to use very wide income bands for the
further calculations, because you'd need to "catch" at least one-third of
that sample to get at least 100 people in the younger voters by low or high
income subgroup and then see how they voted. Even at a reduced confidence
level of 90%, a 100-person sample has a margin of error of more than 8%; in
addition, that's just the error from statistical theory assuming a perfectly
random sample, which for a variety of reasons exit poll samples are not. 

And there are additional complication with the California Poll.

One is that about 100,000 votes in Los Angeles County weren't counted. In
California, people registered as independent ("declined to state" party
affiliation on the registration form) are allowed to vote in the Democratic
(but not Republican) primary. Los Angeles County instituted a requirement,
apparently tied to their vote-counting technology, that independents fill in
an easy to miss party affiliation box at the top of the ballot for their
presidential vote in the primary to be counted. In addition, poll workers in
at least some cases told voters that was to change their party registration
(because of the way it was worded), not to say they were voting in the
Democratic Party primary.

Exit polls are done at selected precincts, and the results of the poll at
each precinct are re-weighted when the votes come in. So if you have an exit
poll that 100 people filled out at a given location with 1000 voters, and 20
were for Clinton and 80 for Obama, but when the vote comes back it says
Clinton got 200 (40%), and Obama 300 (60%) the poll is adjusted according to
the proportions of those 500 votes, and the 500 annulled votes aren't taken
into account, so the weight of each Clinton voter becomes 2 and that of
Obama voters 0.75. [The actual formulas used are more complicated but this
illustrates this aspect of them. Also, half the votes annulled I'm sure
would catch the attention of the exit pollsters and perhaps that might be
taken into account, but I know lower rates are not. I give a very
exaggerated example to make clear what the effect is.] 

NOTE that typically the total number of valid votes at the precinct will not
influence the weighting. How many voters at that precinct is represented by
each person offered the exit poll questionnaire is controlled in a different
way. The exit poll taker simply asks every fifth, seventh, tenth or
twentieth person coming out (depending on the poll design) and keeps track
of the race and gender and sometimes apparent age of those that decline, so
any under- or over- representation at least by race and sex among those
answering can be adjusted for.

And although it may seem that 100,000 votes out of 4 million is small
change, depending on the luck of the draw --in other words, which precincts
were in the exit poll-- the problem with counting the independent vote in LA
could have drastically affected the final demographic results of the poll.
And it is entirely likely that the uncounted votes are even higher in the
percentage of Obama votes because he has attracted huge numbers of people
that normally don't vote in primaries, and thus more likely to make a
mistake on an obscure requirement like that.

The California poll in particular changed quite a bit over the course of the
night. 

The undercount of independent voters in the California primary is a BIG
problem because in California, as elsewhere, the independents went largely
to Obama. And with Clinton and Obama so sharply differentiated
demographically, an inaccurate counting of the vote at precincts sampled is
going to shift all sorts of numbers around in the poll. Suppose, for
example, that 40 precincts were sampled in California, including three or
four mostly Black precincts, but two of those turned out to be in Los
Angeles where this problem occurred. Obama's vote in his top demographic is
going to take a hit, as well the percentage of the voters the poll says were
Black, because the precinct vote count is going to give him a lower
percentage than the exit poll sample, so Blacks will be weighed down.

I suspect this actually happened, because Obama's percentage of the Black
vote in California was the lowest of anywhere in the country and it is the
only state so far where the percentage of Black voters declined. 

Another problem for saying that class is the issue is the large Latino vote,
which increased from 16% (2004) to 29%. Latinos have the lowest high school
completion numbers in the U.S. due to the large number of immigrants. Saying
that people without a high school diploma in California overwhelmingly
favored Clinton is really the same statement as saying that a lot of Latinos
and retirees voted for Clinton. In this case you're talking about 5% of the
poll who hadn't finished high school, which means 95 people and a margin of
error of over ten percent in each direction, but the figures 80% Clinton,
17% Obama are nevertheless statistically significant. What I'm pointing out
though is that the social significance of the figures might be quite
different. These might be Spanish-dominant older immigrants who have barely
heard of Obama but know Mrs. Clinton. 

And you've really got to be careful at these sort of extreme edges of the
demographics. AFAIK, and I think I would have heard if it were otherwise,
the California poll was done only in English. While it is true that you're
supposed to be able to speak and write English to become a citizen, the
standard is not very high (when I was examined to become citizen nearly a
quarter century ago, my test was to write "I love America"). There is every
reason to think the instrument used to take the poll biases the sample
towards the lowest income/lowest education end of the voters.

Finally, the Black turnout in California was surprisingly low -- only 7% of
the electorate compared to 8% in the previous primary and 11% in 2000, which
was a high-turnout year. Overall turnout was higher this year again, but in
the rest of the country it has been Blacks who have been leading the charge
to the polls, and the California numbers don't seem to fit at all, neither
this year's pattern nor the pattern for high-turnout primaries in the state.


So I think it is possible that the LA County counting difficulties impacted
this estimate of Black participation from this year's exit poll. Reducing
the number of Obama voters is reducing the percentage  of Blacks voting in
the exit poll which leads to reducing the number of poorer people and people
with less formal education counted as voting for Obama in the poll. And when
you're dealing with these sorts of small subgroups in exit poll data, a
little change goes a long way. 

For example, those making under $15,000 a year were 7% of the sample Tuesday
in California, meaning 134 people, give or take 9. Of these, Mrs. Clinton
had around 75, Obama around 55. So the actual numbers involved are small,
which means already a large possible sampling error, and if anything happens
that skews the results, forget it. In  fact, Clinton's 15-point lead over
Obama in this demographic is not statistically significant at a 95%
confidence level, and borderline at a reduced confidence level of 90%,
meaning what we can say, assuming the poll is a perfect sample, with 90%
certainty is that she got at least 1% more of the vote in this group than
Obama did.

The main reason Hilarities are putting out all the "class" spin is to
deflect attention from how she is being humiliatingly *crushed* by Obama in
many overwhelmingly white states, like the caucuses last night in Washington
State and Nebraska, where she was trounced better than two to on. 

The main reason for this is the tremendous enthusiasm that Obama is
generating, i.e., not just the breadth of his support but the intensity of
it. Young people --whites as well as Blacks, but mostly whites in these
states-- are organizing and mobilizing people for those events, and when
they get there they work on others in attendance to convince them. To do the
same thing, the Hillary machine would have to find staff, recruit them, put
them on payroll, and train them. She does have some young women volunteers,
but not nearly as many. 

The OTHER sides of the Hillary spin is that Latinos are too racist to vote
for Blacks, meaning "we don't dare nominate him and lose in November the
Latino vote that the Republicans just gave us with their jihad against
immigrants." And that Obama's wins in heavily Black states must be
discounted because of course they're going to vote for the Black guy, but in
November they'll vote Democrat no matter who the candidate is so there's no
need to give much emphasis to the desires of Black folks.

Joaquín