Re: [Marxism] Poll data signals Clinton gaining ground in Pennsylvania

["Joaquin Bustelo" <jbustelo@xxxxxxxxx>]

[We at the bureau of electoral cretinism had decided to give it a rest,
since this campaign seemed to be basically --mathematically-- over some
weeks ago. But this is an excellent example of why you should never trust
the Democrats: someone forgot to put a stake through the heart and the
contest continues, now with the freakish, Frankensteinian aspect of the
undead.]

[We take our inspiration from Fred's subject line: Poll data signals Clinton
gaining ground in Pennsylvania.}

Actually, the poll data "signals" no such thing. 

Both the CNN poll of polls, which is quite selective, limited to 3-5 polls
with known methodologies that fit certain guidelines, as well as the
somewhat more (but not irresponsibly) inclusive RealClearPolitics poll of
polls (that one is here:
<http://www.realclearpolitics.com/epolls/2008/president/pa/pennsylvania_demo
cratic_primary-240.html>) have been showing a 5%-7% Clinton lead over Obama
for the last week and a half or so with little movement.

The polls cited in the article Fred posted are in some cases about as good
as the 50 cents a call polls that TV stations used to run in the 1990s. For
example, ARG reported about 12 days ago that Obama was tied with Clinton; a
week later, that she was ahead by 20%, at a time when no other poll
registered any significant shift in the race. (see the ARG PA poll results
here in all their nonsensical glory:
<http://www.americanresearchgroup.com/pres08/padem8-705.html>)

I heard someone at one of these famous three-letter tv outfits heavily
involved in their polling effort suggest --only half jokingly-- that they
add a laugh track when reporting poll results because these have been so
wildly off this year. Thus in South Carolina, what was to have been an
impressive Obama win of 12% with nearly 40% of the vote on election day
turned into a crushing 2-1 humiliation of the Clintons with 29 percentage
points separating Barack from Hillary.

And it hasn't been all in Obama's direction. In California, a predicted 1%
Obama squeaker became a serious setback for the Illinois Senator, with
Clinton almost 10% ahead of Obama.

In these and other cases, the exit poll demographics show the reason for the
discrepancy. In South Carolina, the young vote tripled and the Black vote
doubled. In California, the percentage of Latino voters was DOUBLE that in
primaries in previous years, and they voted 2-1 for Clinton. 

Are the polls really that bad? Well, yes, because they're trying to measure
something quite squirmy and slippery using a play-dough ruler covered with
the slickest oil imaginable.

First, consider the voters. In a typical contested primary, turnout might be
10-15%; this year has seen extraordinary turnouts of 20%-25% in the
Democratic contest. There is a world of difference in what happens in U.S.
elections depending on WHO votes. U.S. bourgeois elections are as much or
more about which people vote than who "the people" vote for. 

Now finding out who is really going to vote is extremely difficult. In the
normal monthly population surveys that the census bureau does, after every
general election they ask in December whether you voted. And every time
their survey says there were millions more voters than were really there on
election day. Now if it's hard AFTER the fact, imagine BEFOREHAND.

Now consider those polled. The percentage of respondents to phone polls is a
closely-guarded secret of polling organizations, but it is believed to be no
higher than 1/4th or 1/5th. And there is NO GUARANTEE, none whatsoever, that
the ones you're actually REACHING are the ones that are really VOTING.

To try to approximate the vote on election day, pollsters use "likely voter"
screens as well as weighing of results from different sectors according to
what they thing the electorate will look like. Thus if your phone calling
results are that Blacks are 20% of the sample, but history or your analysis
or Ouija board says they're going to be 10% of the voters, you "weigh" each
Black in the sample as 1/2. 

But things get even more complicated. Suppose your "Likely Voter" model has
weighing by race, gender, income group and education level, factors that
supposedly correlate with how people have voted in the past.

Now suppose a couple of those supposed correlations were false or are simply
not the driving correlation THIS year.

I have argued several times here that the alleged attraction to Mrs. Clinton
of "blue collar" or "working class" white democrats is exaggerated or even
false, that this seeming correlation is to a large degree GENERATIONAL, with
just about every exit poll showing Obama winning among the youngest whites
and losing among the oldest. The significance of this to the two proxies
used for class --income and education-- is that retirees have much lower
incomes, and that the percentage of young people who completed college in
the late 40's and 50's was much, much lower than among later generational
cohorts. 

Controlling for all those demographics is hard with a small survey of a few
hundred or even the 1250 needed to reduce the margin of sampling error to
3%. Young people are typically harder to reach. The most reputable firms
keep calling until they get the demographics roughly right for the
population being sampled. That is more expensive and tends to make it a 2-3
day survey, which, a few days before an election, tends to blur any real
shift in opinion. 

Given all this, how much credit should the argument presented in the article
here: that recent PA surveys which press people harder to choose between
Clinton and Obama, and thus have less undecideds, tend to show higher
Clinton numbers?

Not much, for a couple of reasons. First, the real evidence for this
phenomenon is to be found in earlier primaries, where Clinton did better
among those who said they'd decided at the last minute ("today").  These
have tended to be more than 10% of the voters. And those that say they
decided over "part three days" have been one-fifth to one fourth of the
voters. Among those, Clinton has often done better, but sometimes Obama has.


These figures have no resemblance to the undecided in the pre-election
polls, and reflect that the people actually voting are a much greater and
different cross-section of the population than the "likely voters" of
polling models. 

The other reason is a well-known political phenomenon, which is that "going
negative" --as Clinton has done with some vigor in the past week-- drives
down voter participation but disproportionately among younger and minority
voters. In such climates much fewer "undecideds" actually decide. There is
some question whether this quitessentially Republican general election
strategy will work within a Democratic Party primary, or whether it will
provoke a backlash. That appears to have been the case in South Carolina,
where last-minute attacks on Obama by the Clintons were credited with
provoking a tsunami of Black and young voters that almost sunk the Clintons
then and there.

Perhaps the Clinton camp has decided it is worth the crap-shoot: a Clinton
victory by a few percent will do so little to close her overall vote and
pledged delegate deficits that its effect will be a renewed round of calls
on her to quit, as this was her golden opportunity to cut Obama's lead
significantly and she failed to do it.

The benighted and mathematically challenged members of the bourgeois
punditocracy seem to be settling on "at least double digit" as the
definition of a Clinton victory, but if you do the math, a 2-1 trouncing of
Obama is what she needs to be on a path to closing the delegate and voting
gap, and something like a 15-20% victory margin something that would allow
her to argue she was on her way to recapturing the lead, but just ran out of
primaries.

On the other hand, a 10% margin is what you'd expect on election day given
current polling numbers and that on average she has closed a little more
strongly than Obama. And its what she's done in Ohio and California. Which,
is, of course, why the bar is set there. This drawn-out fight has been good
for the ratings of TV News, no one wants to see it declared over and done
with so even a result that basically leaves her even worse off than before
is going to be proclaimed ok.

Now the reason the result will leave her worse off than before is that Obama
is going to win North Carolina two weeks from now. Right now the polls say
by 15%, but ALL the polls in southern states have been WAY off in
underestimating Obama's support. I suspect the reason for that is that there
tends to be in the South very well organized Black political networks all
over the place -- and they have been all pulling out ALL the stops for
Obama. But even with a 15% margin, Obama would completely undo Clinton's
gains if she wins by 10% in PA, or as close to as makes no difference. 

(The Math: Although not a straight proportional representation of the
statewide vote (delegate allocation is actually done by districts, usually
Congressional Districts, as well as statewide) assuming a straight statewide
proportional representation is usually pretty close. Clinton's 10% lead in
PA would give her a 16 delegate edge (out of 158 delegates at stake); a 15%
lead by Obama in North Carolina would give him a 17-delegate edge (out of
115 delegates at stake). Even throwing Indiana in the mix with a 10% Clinton
lead (right now the RCP average is a 2% Clinton lead), works out to a total
Clinton net delegate advance of 23 between the two states. If Obama takes NC
by 20%, which I suspect is more likely than 15%, that wipes out those 23. 

And that's not even looking at the superdelegates. Every couple of days, the
Obama camps announces one or a couple of new adherents. 

Today there were three -- all heavyweights. Former Oklahoma Sen. David Boren
and Georgia Sen. Sam Nunn, from the more conservative/hawkish/southern wing
of the party, and Robert Reich, Clinton's labor secretary in the 1990's. 

These hurt Clinton's case before the superdelegates tremendously. Her "3 AM
phone call" argument is basically shot. Boren is the former head of the
Intelligence Committee and Nunn of the Armed Services Committee, real top
ruling class cadre in their field, and as striking a confirmation as one
could ask for that the ruling class is entirely comfortable with Obama's
ability to answer the phone.

And since Reich is a white guy I doubt greatly that James Carville is going
to call him Judas like he did to NM Gov. Bill Richardson, but Reich's
heartfelt cry, "my conscience won't let me be silent any longer," must have
been the unkindest cut of all. He's been pals with the Clintons since the
60's, having been classmates (him at Oxford; her at Yale). He is the
Democratic Party's REAL top liberal policy über-wonk, and quite neatly (and
symmetrically) undercuts Hillary Clinton's claims on that score.

Very nicely packaged and announced on a Friday to serve as fodder for the
gasbags on the weekend political talkfests to chew over.

*  *  *

We at the Bureau of Electoral Cretinism wish Obama would outright win the PA
primary for two reasons: a) we're bored, bored, bored with this contest and
b) Mrs. Clinton has run a truly disgusting campaign of late and OUGHT to be
punished. However, our Ouija board (the crystal ball seems to have clouded
over permanently) says it's going to be Hillary in PA by 10%-12%, enough
--given her delusional state-- to keep her going. 

The big, big question is whether there's much of a movement left behind the
Obama campaign in the Black community and among young people. I've not seen
any signs of that lately, but then again it's not the sort of thing the
bourgeois press likes to report on. 

As for the relevant demographics, Blacks are 11% of the population of the
state and were 13% of the voters in the 2004 general election, but
Pennsylvania hasn't had a contested Democratic primary in recent cycles,
which means there is no clear guideline as to what percentage of the primary
vote might come from African Americans. Ohio, which is 12% African-American,
had an 18% Black turnout in the primary. PA does have a very large college
student population --but there is a question about how many are actually
registered in Ohio as opposed to their home state. And PA's retirement age
population, Clinton's bastion, is 15%, a touch more than Ohio's 13%.
Clinton, of course, took OH by 10%, which her side hailed as a tremendous
victory but which wasn't nearly enough to significantly reduce Obama's lead.


Looks like PA is going to be an OH rerun.

Joaquín.



________________________________________________
YOU MUST clip all extraneous text when replying to a message.
Send list submissions to: Marxism@xxxxxxxxxxxxxxxxxxx
Set your options at: 
http://lists.econ.utah.edu/mailman/options/marxism/archive%40archives.econ.utah.edu