Tuesday, November 6, 2012

Richard Charnin's Latest Update: Wisconsin & Other Swing States Go Obama

Sensitivity analysis is a fancy word for models and predictions that acknowledge uncertainty -- we don't have perfect information.  This does not mean we have to give up -- it merely means we should construct models and projections to overcome our lack of perfect information.  Richard Charnin, a highly qualified quantitative computer programmer, builds uncertainty into his election projections by considering different scenarios -- sensitivity analysis -- more on what this means for Wisconsin after the jump.

One of the most useful pieces of information used to predict an upcoming election is the actual behavior of voters in previous elections -- returning voters.  This should intuitively make sense -- market researchers quickly learn that people may not be direct, accurate, or even honest when asked to describe their behavior -- especially when social expectations and norms are involved.

Market researchers know that if you can observe a behavior, this will result in more meaningful data than simply asking people to describe their behavior (NOTE:  observations do not explain WHY people behave the way they do).

Charnin understands and his statistical analysis confirms this -- while honest and accurate pre-election and exit polling has its place in his True Vote model, looking at returning voters preferences is a gcritical indication of how those voters will vote again.

Sure, some voters die, new voters come in, others drop out, and others change their mind -- but the law of large numbers assures that these changes are not, by themselves, skewed towards the left nor right.  Even if something does skew "new" voters entering or leaving the election pool, patterns that result in one party getting more support are measurable and observable in other ways -- all part of Charnin's True Vote model.

So today, Richard Charnin expanded on his final forecast:  fraud-free results and expected results (meaning WITH fraud) with a detailed look at what we can expect to see in swing states like Colorado, Florida, Iowa, North Carolina, North Carolina, New Hampshire, Nevada, Ohio, Pennsylvania, Virginia and the Badger State -- Wisconsin.

He begins by pointing out:
  • In 2008, 
    • Obama won the recorded vote [with fraud factor] by 52.9%-45.6%.
    • He won the unadjusted [without fraud factor] state exit poll weighted aggregate by 58-40.5%
    • The True Vote Model [no fraud] indicated that he won by 58.0-40.3% 
  • In the 10 states, Obama’s…
    • Average 2008 recorded share [with fraud factor]: 53.3%
    • Average 2008 unadjusted exit poll [without fraud factor]: 57.8%
  • 2012 True Vote Mode [no fraud factor]: 54.9%
  • 2012 Poll-based projected share [with fraud factor]: 51.6%
All models, predictions, and projections are only as good as the assumptions that were made to make the analysis manageable or even possible.  Unlike the media 's projects and speculation, Charnin states these upfront -- anyone that shares models and analysis without discussing assumptions is not being direct and honest about their analysis and may actually be trying to hide what underlies the things they want you to believe.

Charnin's assumptions for all 10 swing states are clearly stated and they are reasonable -- to some, even generous:
  1. The number of returning voters is based on the 2008 state exit poll.
  2. There is 5% voter mortality (1.25%/year)
  3. Turnout: 95% of Obama 2008 voters; 97% of McCain voters.
  4. Obama wins 92% of returning Obama voters and 5% of McCain.
  5. Romney wins 95% of returning McCain voters and 8% of Obama.
To responsibly acknowledge and take-into-account uncertainty, Charnin always uses sensitivity analysis -- this is what quantitative analysis and scientists are expected to do.  It is the only way to accurately model and project.  It is important to note that media pundits NEVER do this -- actually proving they are not qualified to speculate or may even me misleading.

We can acknowledge uncertainty here with 2 lines of analysis:
  • Obama shares of returning Obama voters (previously voted for Obama) is in the range of 87-97% of those voters and his share of returning McCain voters (previously voted for McCain) is in the range of 0-10%.
  • 2008 turnout of Obama voters is in the range of 92-98% and turnout of 2008 McCain voters is in the range of 94-100%.
For Wisconsinites, it looks like Obama will win -- even with the fraud factor.   With these 2 lines of analysis and 9 senario combinations in each swing state -- there are potentially 18 outcomes in each.  For the full analysis, please see his post HERE (and follow the links it contains).

But the short take of what this all means for Wisconsin is this:  

Regardless of which uncertainty model you use (less generous or more generous), all 18 scenarios in Wisconsin indicate Obama will prevail in the Badger State unless the election is stolen -- actually a coup-d'etat against a sitting administration.

And while Richard Charnin does not get into the down-ballot races -- this bodes well for Tammy Baldwin, Mark Pocan, and every other Wisconsin democrat in today's election.

Paul Ryan and Tommy Thompson probably have sads today.

No comments: