If you're like me, and most Democrats, or just people who agree with my friend's facebook encapsulation: "You cannot be for Trump and for common decency," you may be spending this weekend freaking out each time you refresh Nate Silver's prediction website and see Hillary's chance of winning the election declining by a fraction of a percentage point.
But as a political scientist & specialist on elections & parties--as well as a junkie of this election, and a dab hand at quantitative analysis in politics, I wanted to share a few observations that I hope might help preserve some people's arterial walls in the next three days.
Here are three big reasons why the closer we get to the election, the less sense this seemingly precipitous decline seems to make (and, keep in mind, Mr. Silver & co would still say the chances of Hillary winning are almost double what Trump's are):
- Math of the electoral college: Hillary Clinton could lose Florida, Nevada, Iowa, & Ohio (that Obama won), plus North Carolina, where she's been ahead in polling (and which Obama won in 2008), and still be president, as long as she wins Pennsylvania & New Hampshire (last won by a Republican in 1988 & 2000, respectively--are Keystone & Granite staters really going to let you-know-who lure them over to the Republican side after having turned their noses up at more respectable Republicans like Mitt Romney, John McCain, and the Bush family?)
- Early voting doesn't seem to have a meaningful place in Silver's forecast model, and in two of Trump's must-win states, the story for the Republican looks much grimmer than Silver's outcast predicts:
- Nevada, which Silver's projections give Trump a 50.5-51.1% chance of winning, seems like a lock for Hillary, considering early voting, which ended there yesterday. Probably at least two thirds of all votes that will be cast in the state have already been cast, and 73,000 more Democrats have cast ballots than Republicans. In 2012, Obama won Nevada by about 68,000 votes (a comfortable, if not huge 6.5% margin), after Dem-registered voters cast 71,000 more early votes than Republicans did. So Hillary is farther ahead of Trump in Nevada than Obama was ahead of Mitt Romney (the first Mormon candidate--NV's electorate has one of the largest Mormon blocs after UT). To imagine that Trump actually has a slightly better chance of winning Nevada than Hillary now is to imagine a rather preposterous degree of uncertainty in what will happen on Nov 8.
- In Florida, Democrat-registered voters have a lead in early voting, too, although it doesn't compare nearly as well with 2012. They lead by just about 7,000 votes, as opposed to 104,000 four years ago, when Obama won the state by just 74,000 votes. Of course, if that were the only info we had, that difference of 97,000 votes is more than enough to erase Obama's final margin. But from Florida, we also have information about the ethnicity of early voters, and the big change in Florida this year is the huge growth in the Latino vote: although early voting is still going on today, as of yesterday, Latinos had cast nearly 600,000 early votes, or 14% of all early votes (compared to just 300,000 in 2008--the last year with comparable data, when Obama won Florida by 240,000 votes). In 2012, Latinos accounted for less than 10% of early votes, on their way to account for 17% of all votes in the state that year.
And this year, there are a lot more Latino voters in Florida, and most of the newcomers are US citizens from Puerto Rico (who lean Democrat, as opposed to the stereotypical Cuban-Floridian voter, who leans Republican), while the total number of white-only voters in the state has declined has declined by 1.5% from 2012.
While the early voting picture in Florida is not as contradictory to the idea of Trump winning the state as in Nevada, Silver's projection of Trump with a 52.4-53.3% chance of winning the state should not make us simply cast it on Trump's pile in our mental electoral-college maps - I could go on with similar stories about places like Arizona, Iowa, and North Carolina where already-in early-voting numbers that aren't built into Silver's model tell a much more hopeful story for Hillary than the numbers on Silver's map (Trump with a 74.1-76.1% chance in AZ, 68.7-70.0% in IA, and 51.7-53.7% in NC), but more meaningful is an investigation of how the model's calculations are currently "grade-inflating" Trump's polling numbers in states across the country.
- Dubious aspect of the FiveThirtyEight model: One big part of Silver's calculation of the final vote-estimate is based on how much each state's vote-outcome correlates with with national polls. A big part of this is the trend-line adjustment that Silver adds on top of his polling averages for each state. That is to say: if a candidate seems to have gained in national polls in the past weeks or months, then Silver's model if the polls in a given state haven't moved much in the period, those state polls must be under-counting that candidate's likely final vote outcome.
The upshot is that in calculating the current estimates (estimates, we should hasten to emphasize, as Silver would) of the final-outcome vote shares, which he then uses for his percent-chance calculations, Silver's model adds on some "missing" voters for Trump to whatever the adjust poll averages say: about a 2% "bonus" for Trump across all the states.
Thus, if the polling average (adjusted for the partisan leans of the polls) in, for example: - NH is 43.9-40% (Clinton +3.9%), the "Trend Line Adjusted" average is 43.9-41.9% (Clinton +2%).
- OH is 42.8-43.2% (Trump +0.4%), the TLA avg is 43.0-45.6% (Trump +2.6%)
- NC is 45.5-43.6% (Clinton +1.9%), the TLA avg is 45.3-45.5% (Trump +0.2%)
Thus, Trump's projected chances in all these races are much higher than if Silver didn't believe all the state-poll averages were under-counting Trump. NH looks much tighter for Clinton, OH looks like a decent lead for Trump, rather than a neck-and-neck, and NC looks neck-and-neck for Trump, rather than a tight lead for Clinton. While this built-in calculation of the model might normally make sense, the yawning 2% Trump bonus that the current set of polls produces for it seems way too big to bank on.
Now, maybe the all these state polls really are "missing" the "Trump-bump" that Silver's calculations find in the national polls, but I find big (cardiac-muscle-saving) solace in doubting that these and other well-polled states are all under-counting Trump's final strength, come Tuesday night.
A final big reason to breathe easy is organization: The Clinton campaign has something like four times as many people organizing Get-Out-The-Vote (GOTV) phone banks, door-knocking, and other activities as the Trump campaign does. And they have been at it longer. Clinton's side began staffing up in swing states 7-8 months ago, building lists of volunteers to door-knock the data-chosen voters, networks of civil-society allies, and armies of organizers (including some great vets of Bernie's campaign!) to make it all happen. Trump has nothing like this, and the Republican party has barely been scrambling in just the past month or two to come up with some semblance of ground game, and fallen far short. Having been out there for months for Bernie, and known dozens if not hundreds of others who've done that, and also who are out for Hillary, this matters. Not enough to create a swamping landslide, but by enough (2-4%?) in a close election like this that it should hearten us.
So we should all Keep Calm, and Carry on--and direct any restless energy we might have into some GOTV calling!
No comments:
Post a Comment