Hamsher proves old adage: lies, damned lies and statistics

This morning, I woke up to a Politico story about Democrats sparring with the notorious Grover Norquist ally Jane Hamsher over a series of polls her blog commissioned in some Republican leaning but Democratically held Congressional districts.  The poll is nothing more than an attempt to rile people up about the individual mandate in the current health care bills.  At the request of the Democratic Congressional Campaign Committee, Emory University Professor Alan Abramowitz released a memo with what are in my judgment pretty strong and unavoidable criticisms.  Let me go through the memo (which can be found here), and test if each of its criticisms are valid.  We will take, for convenience's sake, the data for the polling of Indiana's 9th CD.

The first question Prof. Abramowitz raises is that the questions asked about the health care bill is biased.  Well, let's have a look at the questions.
5. There are many pieces to the health care reform bill that is being considered by Congress. One change, called the individual mandate, would require every American to carry health insurance, whether they want insurance or not. Are you familiar with the individual mandate?
This is, of course, a bold faced lie.  Every American will not be required to carry health insurance under any bill.  If one cannot find health insurance for less than 8% of their income in premiums, they are NOT required to buy health insurance.  I have prepared a nifty table and explanation of the caps (both premiums and out-of-pocket expenses).  The House bill's caps are a little different, but in the same range.  I scoured the questions, and yet I cannot find any explanation of the hardship exemption.  The truth of the matter is that the bills require one to buy health insurance only if one can afford it.  Hmm.  I wonder why Jane Hamsher would choose to leave that out.  Perhaps Jane has an interest in driving a certain narrative regardless of the whole truth.

Medicaid is expanded to 133% of FPL in the Senate bill, 150% of FPL in the House bill.  People who will fall under this are obviously not required to buy health insurance.  Again, completely left out of Hamsher's poll.  Even people for whom the bill would deem insurance affordable are not required to carry insurance.  If they do not, they are merely required to pay a small fine, err, tax to the tune of 2% of their income.

Which brings us to their next question:
Under one proposal, if a person does not carry health insurance from a private insurance company, they would be fined up to 2% of their income. Is this fair? Or unfair?
Once again, the 2% "fine" applies only to those for whom insurance is available under 8% of their income in premiums.  The poll questions conspicuously and conveniently hide this fact.

While we're going along with pumping propaganda to voters, here's another classic one:
If the bill required individuals to have health insurance, but gave them the choice of buying health insurance from a private insurance company, OR buying into the government run Medicare program ... would you be strongly in favor, somewhat in favor, somewhat opposed, or strongly opposed?
Uhh.  Anyone bother to tell the voters that the Medicare buy-in that was under consideration would only be available to people aged 55-64 who could not obtain insurance through an employer?

Oh, by the way, where are the questions about whether or not the people who were polled know about the minimum benefit requirements in the Democratic health care legislation, which will force insurance companies to provide comprehensive coverage?  Where are the questions about whether people know that their insurance company will no longer be able to drop them or jack up their rates because they get sick?  Where are the questions about affordability credits?  In other words, how are the very valid criticisms of Professor Abramowitz addressed:
First, with regard to the wording of the health care questions, the questions asked in these polls are clearly biased and loaded. Respondents are asked their opinions about imposing fines on individuals who fail to obtain health insurance without any explanation of the benefits that such individuals would obtain such as the ability to obtain insurance regardless of pre-existing conditions, the availability of lower cost policies in the proposed insurance exchange, or, most importantly, the availability of subsidies to help those with low to moderate incomes purchase insurance.
I will give you a hint.  They are not addressed.  In my judgment, that is because the poll's funding source had an interest in not finding out the answers to those questions.

Next, Jane Hamsher and Survey USA severely under-represent young voters.
Second, the samples in all three of these House districts appear to severely underrepresent younger voters and overrepresent older voters. The percentage of voters under the age of 35 in the samples range from zero to five percent while the percentage age 65 and older range from 35 to 45 percent. These percentages are way out of line with the known age distribution of the electorate in recent midterm elections.

For example, according to the 2006 national exit poll, voters under the age of 35 made up about 25 percent of the electorate while those age 65 and older made up about 20 percent of the electorate.
Interesting to watch how self-proclaimed leaders of the progressive movement don't care about under-representing young voters.  But perhaps it should not be a surprise coming from a funding source whose blog traffic is dominated by (estimated) a decidedly older, white and childless household demographics.  Be that as it may, my job here is to do statistical analysis.  The sample is completely out of whack.  Yes, more older voters tend to vote in midterms, but not that much more.  Yes, less younger voters vote in midterms, but not essentially zero!  In the Indiana 9th polling, for example, Survey USA assumes young voters to be 5% of likely voters and older (65+) voters are estimated to be 32% of likely voters.  But the Professor says that in 2006, voters under 35 made up a quarter of the electorate and and voters over 65 about a fifth.  Is he right?  More importantly, is he right in the case of Indiana?  It turns out that as far as Indiana is concerned, the actual numbers fall right in the middle.
Notably, only 12% of Indianans of voting age are over 65.   Essentially, Survey USA is underestimating the youth vote in Indiana by about 300%.  A glaring error like that can by itself make a poll so biased that it serves no useful purpose.

So far, we have established the following in Jane Hamsher's poll:
  • Facts about the health care bill was suppressed and only biased information was given to voters.
  • Younger voters were severely under-represented in the poll.
So the absolute numbers of the results are essentially (and quite conclusively) invalid.   However, the before and after numbers do add a twist.  And the twist is this: If Hamsher and FDL's goal was to show that these Democrats are more endangered because of the individual mandate to purchase health insurance, she utterly failed.  Let's look at the second and last questions:
Assume Baron Hill votes to pass a version of the health care bill that does NOT include a fine. If there were an election for US House of Representatives and the only two candidates on the ballot were Democrat Baron Hill and Republican Mike Sodrel, who would you vote for?

Hill (D): 40%
Sordel (R): 48%
Which is the exact same margin at the beginning of this poll:
If there were an election for US House of Representatives today, and the only two candidates on the ballot were Democrat Baron Hill and Republican Mike Sodrel, who would you vote for?

Hill (D): 41%
Sordel (R): 49%
So, essentially, Jane Hamsher just spent a bunch of money to mislead voters, under-represent young people, and find out that the presence of the mandate makes no difference whatsoever in terms of a Democrat's position in conservative districts.  Congratulations, I guess.

Like what you read? Chip in, keep us going.

FixitAndPassIt.org cleans up act, sort of

Lawrence O'Donnel has been wrong before