The man who rose to prominence by predicting 99 out of 100 states correctly throughout Barack Obama’s presidency is now coming under fire from many of the same people who so vehemently championed his work before. The New York Times Upshot model puts Hillary Clinton’s chances of winning at 84%. Princeton’s handicaps her at 99% odds, one point higher than Huffington Post’s 98%. The betting markets are a little less bullish, as Hillary currently is a 77% favorite, per electionbettingodds.com.
Nate Silver however, gives her just 65% odds on winning the presidency. Part of this is because of the volatility caused by a larger number of undecided voters, but it’s also due to a unique aspect of the 538 model. It’s based off elections since 1972, which clearly has a dramatic effect on their level of confidence in a Trump presidency, as Silver wrote:
What if we changed this assumption? If we calibrated the model based on presidential elections since 2000 only — which have featured largely accurate polling — Clinton’s chances would rise to 95 percent, and Trump’s would fall to 5 percent.
So why not calibrate the model for the new millennium? He continued:
It may be naive to expect the pinpoint precision we saw in polls of presidential elections from 2000 through 2012 — a sample of just four elections — to represent the “new normal.” Going back to 1972 takes advantage of all the data we have, and includes years such as 1980 when there were significant late polling errors.
He has a point. Journalism is dramatically underfunded and polls aren’t cheap. With newspapers dying across the country, we are getting less good polling than we were in past elections.
If garbage goes in to a forecasting model, the same will come out no matter how much mathematical genius one throws into it. The ~10% of undecided/third party voters exacerbates the potential for polling errors, and so Nate Silver's skepticism isn't completely unwarranted.
However, Clinton dropped from 83% on October 26th down to 65% in 538's model. Has the race changed that dramatically since we learned that the FBI is in the middle of a partial mutiny? Doubtful.
There is nothing in James Comey's bumbling announcement to Congress that provides any insight into what Clinton's additional e-mails do or do not say. All this did was reinforce everyone's preconceptions of her: either she's the head of the illuminati imposing the wishes of a globalist cabal on the 99%, or this is yet another attempt to reverse engineer a Clinton “scandal” into a crime.
America is incredibly polarized, to the point where the base assumptions in any poll are affected. Even though pollsters meticulously choose respondents to get a representative sample, if they don't respond, then the sampling will be skewed. As YouGov elucidated in their brilliant write up on this issue:
Although we didn't find much vote switching, we did notice a different type of change: the willingness of Clinton and Trump supporters to participate in our polls varied by a significant amount depending upon what was happening at the time of the poll: when things are going badly for a candidate, their supporters tend to stop participating in polls. For example, after the release of the Access Hollywood video, Trump supporters were four percent less likely than Clinton supporters to participate in our poll. The same phenomenon occurred this weekend for Clinton supporters after the announcement of the FBI investigation: Clinton supporters responded at a three percent lower rate than Trump supporters (who could finally take a survey about a subject they liked).
Dan Hopkins at 538 added to this with his own research of contacting past respondents to see if their votes had changed from earlier this year, and 80% had not (Hopkins did note that his poll was more Trump-leaning due to his respondents skewing older). That figure was 95% of Clinton and 91% of Trump voters in YouGov's poll. Hopkins wrote:
Notice another thing: There is very little movement between the two major-party candidates. 0.5 percent of respondents were with Clinton in January and are now with Trump, while only 1.5 percent of our respondents moved from Trump to Clinton. There is far more movement from not supporting either candidate to backing one than from supporting one major-party candidate to supporting the other. That, in turn, stabilizes the horse race to some degree: Voters who can be persuaded to switch sides are twice as influential, since they affect both candidates' tallies simultaneously.
Despite the insanity of this election season, the electorate has not deviated much from the ~45% for each side and ~10% undecided model that has been largely stable throughout this century. Here are the last four popular votes:
2012: 51.1% D to 47.2% R (D +3.9%)
2008: 52.9% D to 45.7% R (D +7.2%)
2004: 50.7% R to 48.3% D (R + 2.4%)
2000: 48.4% D to 47.9% R (D + 0.5%)
Any generic Republican probably would have topped out around 48% this year. Most national polls in the cycle max out at 44% for Trump, which seems about right when accounting for the size of the #NeverTrump coalition within the GOP. Given the inherent electoral damages enjoyed by the Democratic Party, it's difficult to find where Trump's path to victory resides given that Hillary sits anywhere between 44% and 48% in recent polls. If his Michigan, Wisconsin, Minnesota, Pennsylvania Hail Mary fails, he has no options. Even his most realistically optimistic projection outside those states falls short.
Only eleven states changed their presidential vote from the previous cycle, for a total of 16 times in 200 opportunities since 2000. We will look back on Obama's presidency as transformative not from just a cultural standpoint, but an electoral one as well. 2008 brought ten of those reversals and solidified Nevada, Colorado, Virginia, and New Mexico as likely Democratic states—while also establishing Ohio, Florida, North Carolina, and New Hampshire as the only pure tossups left on the board.
The Democrats have won 19 states in each election since 2000. They account for 243 electoral votes this year. The Republicans have won 22 states that comprise 176 electoral votes over that same period. If all those states hold (a few may not, like Arizona, Michigan, and Georgia), Hillary would need just 22% of the remaining electoral votes to become president, yet 538's model says she isn't even a two to one favorite. The issue isn't that Nate Silver shouldn't be cautious, it's that his model gives credence to far too many scenarios.
It's difficult to imagine that 538's general election model wasn't tweaked after the lumps it took in the primary, predicting that Trump was a sideshow at best. 538 is already assuming this year is unique, so using the lessons gleaned from Trump's Godzilla act over the GOP field could theoretically be helpful, but the general electorate is vastly different from the GOP's. Statistical models became all the rage after Nate Silver nearly threw a perfect game across two elections, but they all have a fatal flaw: They become permanently outdated the moment early voting begins.
Polls are the best tool we use to gauge the intentions of the voter. But you know what tool is better? Their actual vote. It's not that early voting is indicative of where the electorate is going on election day, but that it's simply a huge portion of the overall total. Anywhere between 30 to 40% of the vote is expected to come from early voting, and based on party affiliation, we can get a sense for where those early votes are going. This election may have already been determined before we even learned that Rudy Guliani still has his own unit inside the FBI.
Last week, we saw people stand in long lines to vote. Even though the instinct is to swell with pride when seeing citizens sacrifice that much time to exercise a constitutional right, the result of our failing institutions is right in front of us and we should be angry. Polling stations are being closed across the country in predominantly Democratic (read: minority) districts in states controlled by Republican governors and legislatures.
However, the Republican plot to disenfranchise millions, combined with Trump's fascist rhetoric, may have awoken a sleeping giant. We hear the GOP speak to the “silent majority” all the time, which is mainly a wink to white supremacists. It's also a mathematical theory stating that traditional polling misses a huge chunk of (white) voters who mostly vote Republican. Given what we saw over the weekend, the former portion of that may be right, but the latter is focusing on the wrong group.
If Jon Ralston says Hillary already won Nevada, then she won Nevada. If Ralston reports that she and an army of grey aliens successfully pulled off an Ocean's Eleven-style heist on the Bellagio, well then that's what happened. If you want to know what's going on in Nevada politics, Jon Ralston is the oracle. These statistical models fail to keep up with early voting, which is where we tend to see holes emerge in pollster's assumptions, and this year's theme seems to be a dramatically larger than expected Hispanic turnout.
In general, it is difficult to poll the Hispanic population. Language (nearly 75% report only speaking Spanish or a combination of Spanish and English at home), technology (according to the U.S. Department of Health and Human Services, nearly 60% of Latino adults live in a cellphone-only house, 15% higher than all Americans combined), and the fact that they have been squeezed out of many local political organizations contribute to the difficulty in ascertaining the voting intentions of America's largest minority population.
Given how Trump opened his campaign, it seemed strange how little we heard about the potential for a larger than expected Hispanic voter turnout, but again, this is the oligopolistic media we're talking about. They can't pay attention to anything for longer tha-SQUIRRELL!
One reason behind Silver's pessimism of Hillary's chances is due to his belief that this is not a great map for Democrats.
First off, the Democrats have a big head start thanks to California and New York. Overall, this is a GREAT map for them. There are just more battleground states since we're in the middle (or near the end) of a large shift. White working class states are turning redder while America's fastest growing cities are turning entire states blue. North Carolina is the poster child for this new world. If Donald Trump did inspire an army of minority voters to the polls, his prospects look incredibly dim. And if early voting is an indicator of what may follow on election day in America's southernmost states, Trump can sweep the Midwest, pick up his maximum points in the Northeast, and still lose.
Statistical models are not 100% accurate because polling is an inexact science; that's why they post the margin of error after every poll. Nate Silver is either hedging his bets too hard based on flawed assumptions, or at worst, harvesting free media from America's biggest steaming orange turd. The electorate has remained largely unchanged on a Democratic-leaning map, and there is little reason to believe that any Republican, let alone a godless draft dodger, would be a favorite. 2008 was a transformative moment, and 2016 may complete or at least initiate the second half of this shift. Ohio used to decide elections, but its importance has lessened as states like North Carolina, Colorado, and Nevada have moved left. Over the coming decade, states like Georgia, Texas, and Missouri will follow suit.
Even if volatility is at 1972 levels, the polls are still playing on a 2016 electoral map. Relying on polls from 2000 going forward is logical not just because there is more consistency among pollsters, but also because the electorate has remained fairly consistent from 2000, it's just moved around. Silver's argument in favor of a Trump presidency basically seems to be ¯\(?)/¯. Polling volatility combined with a larger amount of undecideds has scared him into making any confident prediction.
Silver was the punditocracy's wunderkind for the past decade. Politics followed a similar route that sports did, where a bunch of old white men suddenly realized that math was far more complex and helpful than they assumed. Now there are a million different Nate Silvers in the world, so the real Nate Silver has done more punditry than usual this election. He certainly seems to be trying to capitalize on the free heat his model is generating, as he has even carried water for tangerine Mussolini from time to time:
A cynic would see what the rest of big media is doing and call Silver out for following suit and trying to generate easy clicks, but that doesn’t seem to be the case here. His model genuinely is far more cautious than others for reasons that may be unfounded. 99% Clinton is absurd, as her firewall is basically one state—Pennsylvania—and there’s a transit strike in Philadelphia right now. Hillary isn’t a shoo-in, but after seeing the early vote totals in Nevada, Arizona, North Carolina, Florida, and even Georgia, it’s not very much in doubt either. Silver’s skepticism reflects where journalism is at right now. There’s a lot of noise, but instead of drilling down to the facts of the matter, his model refuses to commit to any firm trajectory. Uncertainty means never having to say you were wrong.