Monday eve, 15mi north of the GWB and wondering how Indiana looks for Obama. Man, if he wins both tomorrow, its O.V.A.! Hillary gets to withdraw with dignity which means President Obama becomes more of a reality. I always check RealClearPolitics first, and have done so since last September. Its my favorite poll clearinghouse. Two of the polls included in the average covered 5/3-5/4 and had similar sample sizes (600 & 636). One showed Hillary with a <st>6</st> point lead and the other had Obama up by <st>two</st>. But both cannot be true at the same time as the only way for both Hillary and Obama to win is for there to be a tie.
Now just because you want one poll to be true and the other not doesn't mean that the one you want to be true is the one that is true. There are ways to determine how reliable a poll is, how valid it is. And no, you don't even have to be able to do algebra to do it. You just have to be able to the elementary stuff - read, write, and arithmetic. Oh yea, and google.
Lets jump to it ...
I really don't like the RCP Indiana poll that is showing Hillary with a 6 point lead. I would love to be able to show that this poll is not a reliable indicator for tomorrows Indiana Democratic Primary. If Hillary wins Indiana by 6, the race continues. My goal here is to find at least one reason to consider this poll false - invalid. By comparing the demographics of this poll, the Suffolk University poll, with the expected demographics of the primary we can determine if the sample space of the poll is a good match to reality. That is, do the percentages of sex and race and age in the sample match those of the voters? By now everyone is aware of how the vote in the Democratic Primary breaks along lines of sex, age, and race. If you over/under sample a certain demographic you can produce a sample space that is advantageous to one candidate at the expense of the other. For example, Hillary always beats Obama with whites over 65 years of age. If you employ a sample space that has a disproportionate share of the 65+ voters, your poll, no matter how objective from that point on, has a fatal flaw.
If the percentage of the 65+ primary voters is less than your poll assumed, then this should give Obama a larger margin of the actual vote. If the percentage of seniors voting is more than your poll assumed then this should give Hillary a larger margin than the actual vote. Recently, I've notice that many polls favorable to Hillary have been oversampling seniors. This is just what Suffolk did wrong. Oversampled 65+ year olds.
I had already netted a site with the demographics of Indiana in order to invalidate a different poll than Suffolk so I already knew that 65+ year olds constituted only 12.4% of Indianaians. Checking the age breakdowns in the Suffolk poll reveled that 12% of their sample was 65-75 year olds, and 15% was 75+! Wow! 27% of the Suffolk poll's sample was 65+ while this age group makes up only 12.4% of all Indianains. As 27 is more than twice 12.4 (multiplication is arithmetic), and seniors (and, I assume, the elderly) prefer Hillary, this sample is biased in favor of Hillary. It is also biased at Obama's expense - for a fixed sample size, more seniors (and the elderly) means less young people. You know, the ones who prefer Obama.
To my satisfaction, I have shown that this poll is biased in favor of Hillary, so I expect it to show that Hillary is in the lead. I would be concerned however IF the sample space included too large a percentage of young people and yet still indicated that Hillary was winning. But it doesn't,so f* this IF. I've already biased rated the poll, so its time to kickit2daCurb; however, before doing so I must point out there are other reasons for dissing dis poll. Worse of the worst of Suffolk is that it did not provide any real cross tabulation. 43% of the sample choose Obama and 49% choose Clinton. No breakdown as to what percentage of which age, sex, race group choose Obama and ditto for Clinton. For oversampling seniors and the elderly, in addition to not producing cross tab,s this poll should be troll rated. I'm in too good a mood right now to waste one of my precious five, so I'll let it slide - this time. But Suffolk is now on my s**t list. And anytime that I see a Suffolk poll going against my world, s**t on it I will.
The poll that motivate me to write this diary was the SurveyUSA poll. Currently showing a 12 point lead for Hillary in Indiana, it seems to be an outlier in that most other polls have Indiana competitive, leaning towards Hillary. RCP is not currently including this SUSA poll as part of its average for Indiana, opting for Suffolk instead. But the SUSA poll has a well designed sample space that is a good match of the voter demographics of Indiana. This poll also displays a cross-tabulation of the results, allowing for direct analysis of how each relevant demographic group voted.
Look back at the Suffolk poll and you'll find a flat file of information presented in an isolated linear stylee. Look again at the SUSA poll and you find a two-dimensional table with information presented in a coherent interactive stylee. Though SUSA also has oversampled those 65+ years of age and older, 19% is a lot closer to the state distributon of 12.4% then Suffolk's 27% is. Suffolk's sample space was 15% 75+ year olds, while SUSA's is only 10% older than McCain. As age goes, SUSA's sample space is a better match of reality than Suffolk's. If SUSA is the more accurate poll then maybe the reason its an outlier is that its the only poll that's right. But.That.Just.Cannot.Be.TRUE! Hillary winning Indiana by whatever! means more miles to go before I sleep. And I want this to end tomorrow so Hillary can give Barack the call and concede with dignity.
Now Indiana may be white, be it ain't all white, aaaaight. And as we all know, Obama pulls a larger share of the black, and Hillary needs a significant share of the white, to win. As Indiana is 12.4% black, and blacks always make up a higher percentage of the Democratic Party then they do of the general population (that's snark for you overly sensitive black repubs)the sample space of a poll of likely democratic primary voters should be more than the 10% black that the SUSA poll is. The Suffolk poll was 15% black, and 15% of the democratic party of very, but not totally, white Indiana being black feels more real. More real than 10% at least. Honestly, I think that even 15% black is too small a percentage of the expected Democratic primary vote in a state that is 12.4% black. 18% black should be the minimum sample size in this case.
A quick google netted me this gem , the "online news source for East Central Indiana". That's roots-ee enough for me. The poll concurs my point of contention with the SUSA poll in that ...
Poll: Obama edges Clinton in Indiana
THE INDIANAPOLIS STAR
holds a narrow lead over Sen. Hillary Rodham Clinton in Indiana, with the outcome of the May 6 primary likely in the hands of a large number of undecided voters, according to a new Indianapolis Star-WTHR poll.
The poll showed that Obama -- helped in particular by strong backing from black voters -- is leading Clinton 41 percent to 38 percent among likely Democratic primary voters.
it indicates that the racial demographics of tomorrows democratic primary might have a more determinant effect then age on the outcome. Its also great that this is a local poll, performed by the Indianapolis Star-WTHR . That's Indianapolis, Indiana. As local as local can get.
My breath deepens, my pulse slows. SUSA is an outLIAR, for real. By oversampling older white voters SUSA produced a sample space favorable to Hillary. By undersampling black voters, the sample space was biased against Obama. No wonder RCP choose Suffolk over SUSA. As biased as Suffolk was in its age demographics, the SUSA race-based bias produces a less realistic sample space. I might have to put SUSA on my s**t list with Suffolk.
I don't care about the non-existent IStar-WTHR poll cross tabs. Its validation is derived from the fact that it both compensates for the undersample of black primary voters and clarifies the effect of this undersample in both the Suffolk and SUSA polls. A valid poll without cross tabs beats an invalid poll with. Speaking of validation, nothing could validate my belief that the Obama Phenomena is much bigger than he is more than him winning Indiana by 3. Indiana. Red State. A red state where
Among Hoosiers who said they would vote in the general election -- a statewide sample of voters with a margin of error of plus or minus 4.4 percentage points -- Obama beat McCain 49 percent to 41 percent. Clinton broke even with McCain, with both backed by 46 percent of those polled.
A red state where Obama beats McCain, and Hillary does not.
Then Hillary can bow out gracefully and with dignity. Then we can start focusing on what's really important. But if you think that Obama is what's really important, then you really don't get what the Phenomena is all about.