Saturday, November 5, 2016

Better than polls, which get volatile at this point

No doubt, polls show a tightening of the race, causing much gnashing of teeth.  Here's some reassuring news.  Polls notoriously get volatile this close to the election, especially this year with many late-deciding voters not wanting to vote for Trump or Clinton but also not committed to one of the alternate candidates.

Some of the polls coming out now are of questionable reliability.  And it's particularly troubling that guru Nate Silver's prediction of Clinton's win has dropped significantly.   But it's still at 63%.   That's pretty good, if we hadn't gotten used to it being in the 80s.  Both the Princeton Election Consortium prediction and the Huffington Post model put her chances above 90%.

So let's calm our fears and look at what David Plouffe calls the fundamentals in the race that are not based on volatile polls.

1.  The first is from Moody's Analytics Election Model, which has correctly predicted the winner in each presidential election since it began in 1980 (see footnote below).  It does not includes polls at all.   They look at six factors, three economic and three political.

Economics, done individually for each state:   (1)  growth of family income;  (2)  increase in  home value;  (3)  price of gasoline.

Political factors:  (1) the incumbent's share of vote in the last election;  (2)  incumbent's approval rating;  (3)  fatigue factor -- people usually want a change of party after two terms.

By these measures assessed for each state and adding up electoral votes, they predict that Clinton will win 332 electoral votes to Trump's 206.

2.  In early voting, each county election board puts out a daily list of the people who have voted that day.   Not how they voted, just their names.  A tracking survey, done by TargetSmart and the College of William and Mary, then contacts a sample of these early voters to do an exit poll on how they voted.   It's far more reliable than the random exit polling done on election day, because they have the whole list and can construct a representative sample, which can then be interviewed and extrapolated.

In Florida, as of Monday evening, almost 4 million people had voted, either by absentee ballots or early poll voting.   Surveys of voters found that 48% had voted for Clinton and 40% for Trump -- in Florida!!

But then here's the startling thing.   They found that 28% of registered Republicans in their sample had voted for Clinton.

These are votes already cast, and they won't change, no matter what new leaks may come from the rogue Trumpsters in the FBI.  If she wins Florida's 29 votes, especially if it's anywhere near this 8%, then it will be a landslide.  Never mind those recent polls showing a tie in New Hampshire with its 4 votes;  Florida has 29.

3.  This one is not scientific but it has the advantage of having picked the wrong winner only twice since 1940, and one of those was the year even the next-morning newspaper headlines were wrong, proclaiming that Dewey had beat Truman, it was that close. 

This is a nation-wide educational exercise with school kids, the Scholastic Student Vote.  It combines information about democracy, our civic process, and the election.   Then kids are asked to pick their choice for president.   If you want to know more, here's the link:

Whatever its worth, Clinton wins overall with 52% to 35%, and she wins every battleground state except Iowa and Missouri.  She even wins Utah, Georgia, South Carolina, and Texas.

One can certainly be skeptical, because it is non-representative and non-scientific.   But look at their record:  two wrong predictions in 75 years.  My guess is that her win won't be that big but that she will win decisively.   And it is in line with another straw vote taken through schools here in Georgia, which includes parents and teachers as well as children.  In that, Clinton won 54% to 29%, compared to the Scholastic Student Vote, where she won Georgia by 48% to 37%.

Another thing to remember, if those recent close polls scare you:   a good percentage of voters have already voted, before Comey dropped his bombshell, before these polls came out.   Besides, the bad news was Thursday;  on Friday, polls from the highly respected PPP put her ahead 5% in NH, 3% in NV, 7% in WI, 4% in PA, 2% in NC, 5% in CO, 5% in MI, and 5% in VA.   Good news.   It seems the FBI leaks caused a slight dip for a day, and then a rebound.


fn 1:  Professor Alan Abramowitz, a political scientist at Emory University, has a similar model based on fundamentals.  His is based on three predictors:  the incumbent president's approval rating at midyear, the growth rate of real GDP in the second quarter, and whether the incumbent president's party has held the White House for one term or more than one term.   This model also has an excellent record of correct predictions since it's beginning in 1988.

In mid August 2016, Abramowitz's model predicted a narrow victory for a generic Republican candidate.   However, Abramowitz gave a nuanced discussion that cast doubt on the efficacy of his model with such a non-typical candidate as Donald Trump.

No comments:

Post a Comment