Monday, November 05, 2012

An Alternative To Nate Silver: Here Are The Real Odds


Back in 2008, a man who established his reputation with Baseball statistics was lionized as a political guru for correctly “predicting” the winner of 49 of the 50 states.  He did this through aggregation of state polls, and a formula which, speaking bluntly, was based heavily on subjective weighting of polls he liked.  On his side, it’s true Silver was accurate in 2008.  Silver, however, doesn’t like anyone bringing up his results in the 2010 mid-term election.  This year, Silver has announced that President Obama has an almost ninety percent chance of winning re-election.  He has also, rather dishonestly, tried to hedge his predictions by saying that turnout will decide the actual odds (hidden well below his bold headline), and admitted that if the state polls used invalid weighting on party participation, then the election results may be very different from his forecast.  His supporters have even gone so far in recent days as to demand that Silver’s system be judged on the process rather than whether his predictions come true.  Leaving all this aside, the question may reasonably be asked, about what the odds really are for Obama and Romney.  There is a reasonable method available for determining those odds.

Whenever a percentage chance is given for something happening, it claims that of a hundred possible outcomes, this one will happen ‘x’ number of times.  So we can figure up the odds by working out what each candidate needs, and how they could get there.

First, President Obama.  By this date, there’s really no doubt that the District of Columbia and the following states will be won by Obama:  California, Connecticut, Delaware, Hawaii, Illinois, most of Maine, Maryland, Massachusetts, New Jersey, New York, Rhode Island, Vermont, and Washington.  That gives Obama DC plus 13 states, for 178 Electoral Votes, meaning he needs another 92 EV to win re-election.

Next, Governor Romney.  There is no doubt that Romney will win the following states:  Alabama, Alaska, Arizona, Arkansas, Florida, Georgia, Idaho, Indiana, Kansas, Kentucky, Louisiana, Mississippi, Missouri, Montana, Nebraska, North Dakota, Oklahoma, South Carolina, South Dakota, Tennessee, Texas, Utah, West Virginia, and Wyoming.  That gives Romney 24 states for 220 Electoral Votes, meaning he needs another 50 EV to win the White House.

There are 13 battleground states plus 1 EV in Maine that is contested, for 140 Electoral Votes.  From the start we can see that neither candidate needs to sweep all the battleground states to win. 

Before going on, there is a controversy which will sort itself out tomorrow.  Many state polls have given heavy advantage in weighting to Democrats, even when previous elections show no such disparity in voting by party.  As a result, there is suspicion among conservatives that the state polls are inaccurate in stating actual voter support.  So the question comes down to whether turnout will be like 2008, 2004, or somewhere in between.  For each of the states, therefore, we can look at the actual part weights from 2004 and 2008, then use that to tell us what we might see this year.  Here’s how each state plays out, starting with the highest EV states still contested:

Pennsylvania:  20 EV, polls either even or show Obama slightly ahead
08:  Obama                  04:  Romney                Mix: Obama
Ohio:  18 EV, polls even or show Obama slightly ahead
08:  Obama                  04:  Romney                Mix:  Romney
Michigan:  16 EV, polls show Obama ahead but close
08:  Obama                  04:  Romney                Mix: Obama
North Carolina:  15 EV, polls show Romney ahead but close
08:  Obama                  04:  Romney                Mix:  Romney
Virginia:  13 EV, polls show Romney ahead but close
08:  Obama                  04:  Romney                Mix:  Romney
Minnesota:  10 EV, polls show Obama ahead but close
08:  Obama                  04:  Romney                Mix: Obama
Wisconsin:  10 EV, polls show Obama ahead but close
08:  Obama                  04:  Romney                Mix:  Romney
Colorado:  9 EV, polls show Romney ahead but close
08:  Obama                  04:  Romney                Mix:  Romney
Oregon:  7 EV, polls show Obama ahead but close
08:  Obama                  04:  Romney                Mix: Obama
Iowa:  6 EV, polls show Obama ahead but close
08:  Obama                  04:  Romney                Mix: Obama
Nevada:  6 EV, polls show Obama ahead but close
08:  Obama                  04:  Romney                Mix: Obama
New Mexico:  5 EV, polls show Obama ahead but close
08:  Obama                  04:  Romney                Mix: Obama
New Hampshire:  4 EV, polls show tie or close either way
08:  Obama                  04:  Romney                Mix:  Romney
Maine: 1 EV contested, polls show tie
08:  Obama                  04:  Romney                Mix: Obama

On the weights, assuming each of the three turnouts is equally likely, then on average Obama would win 70.3 of the EV and Romney would win 68.5.  But since Obama needs 92 to win, he has a net 38.2% chance of winning, while Romney (who only needs 50 more EV) has a net 61.8% chance of winning.

That’s the math.   The votes themselves will decide the fact.

Be sure and vote.

Sunday, November 04, 2012

One Scenario Which Will Definitely Not Happen


As the election race closes to less than forty hours before Election Day, all kinds of stories have come out.  ‘Inside’ stories about Romney preparing for a loss, Obama planning to move to Hawaii next year because he knows he’s done, so-called experts and gurus who promise perfect knowledge of what will happen … there’s no shortage of noise and media flatulence.  While I have my own ideas about what will happen, I don’t pretend to know it for a fact, and frankly I don’t respect anyone who pretend that they have such knowledge.  However the voters think, however early voting went, the race itself will be decided Tuesday by the actual votes. That’s because opinion polls are not actual votes, and however important early voting is, no one has ever had so commanding a lead in early voting that the Election Day was not necessary.   Nate Silver is lying to you, and so is Karl Rove.  It’s what they are paid to do, actually, but you should keep in mind that spin is pervasive in the media, including a lot of people pretending they are objective reporters.  The race could be razor-close, it could be decisive, and it’s even possible that President Obama could win the Popular Vote but lose the Electoral Vote, as Gore did in 2000.  But I have not been able to work out a way for Mitt Romney to win the Popular Vote without also winning the Electoral Vote.

Here’s how that works out:

Obama has a commanding lead by all accounts in polls in five states (54%+ in New York, Illinois, Massachusetts, Connecticut, and Washington).  Obama also won handily in D.C.,  Hawaii, Vermont, Rhode Island, Maryland, Delaware, and Maine, but there are no polls in the last two weeks for those states.  If you take the poll support where known (use those RCP average that get such press), and apply 2008 numbers for states without polls, when you plug it all in, Barack Obama ends up leading Romney 48.9% to 46.6%. 

First, notice that no national poll , well no serious poll by professionals, give Romney support numbers that low nationally.  And Obama is below those numbers in serious national polls.  I don’t think those state polls are kosher, but for here they create a problem for anyone saying Romney would win the PV but lose the EV, because the only way to create that condition would be to add votes for Romney in battleground states … or create ridiculous scenarios, like saying he takes 90% in Texas and 48% in California.

When you add in the necessary votes to give Romney the PV lead, it always creates a condition where he wins the EV as well.  The closest scenario I could find, allowed the 3rd party percentages for each state to be copied from 2008, Obama was allowed the poll support and Romney the rest, and I end up with Romney 49.5%, Obama 49.0%, and Romney wins the EV 318-220.  I’m not saying that’s how it ends up, but if Romney wins the Popular Vote this year, he wins the Electoral College as well.