Вы находитесь на странице: 1из 8

Jacob Horowitz

Gordon Lindsay Brown

Advanced Placement United States History

May 25, 2017

Dazed and Confused

A note on fivethirtyeight.com:

Fivethirtyeight.com is a website that focuses on statistical analysis of

current event and sporting news through the usage of opinion polling and

statistical models.1 Editor Nate Silver had his rise to fame after correctly

predicting the outcome of every state in the 2012 political election.

Fivethirtyeight.com closely covered the 2016 election and used the same

predictive model used in the 2012 election. Nate Silver goes into much detail

about the inner workings of the model here2. To summarize his explanation: the

model works by compiling and then weighting by accuracy many state and

national polls. Each poll is then adjusted for various biases and demonstrated

inaccuracies. Then the polls are compiled into two (or in this case three, one for

Gary Johnson) percentage probability of winning the electoral college. Nate

Silvers model did not predict a Trump victory, but it did give Trump one of , if

1
"FiveThirtyEight." Wikipedia. Wikimedia Foundation, 23 May 2017. Web. 30 May

2017.

2
Silver, Nate. "A User's Guide To FiveThirtyEights 2016 General Election

Forecast."FiveThirtyEight. FiveThirtyEight, 05 Aug. 2016. Web. 30 May 2017.


not the highest, chances of victory compared to other predictive media.

Nonetheless, because of his models failure to predict correctly and the general

shock of the electorate, Silver wrote a series of eleven articles examining what

caused Trumps victory to appear so improbable to most people, informed or not.

On the night of November 8th, the majority of the American electorate found itself dazed

and confused. Donald Trump defeated Hillary Clinton in the electoral college, 306 to 232. The

vast majority of predictive models had Hillary Clinton with above an 85% chance of winning

and most mainstream media outlets discussed the coming election with even more certainty.3

How could this happen? What brought about this gross error in predictive confidence? And

finally, why did she lose?

Lets begin by looking at what was most likely the tipping point of the election: the

October 28 release of then FBI Director James Comeys letter to congress stating that more

investigation into Hillary Clinton and her use of unsecured email servers was warranted.

Previous to this release, Clinton led in most national polls by three to five percentage points and

led in fivethirtyeight.coms modeling by six percentage points.4 Then, if we take another look at

fivethirtyeight.coms forecast seven days later (this takes into account the polls published

post-letter) we have a different story entirely. Trumps numbers were higher than at any other

point during the campaign in both the vast majority of national polls and fivethirtyeight.coms

3
Katz, Josh. "2016 Election Forecast: Who Will Be President?" The New York Times.

The New York Times, 19 July 2016. Web. 30 May 2017.

4
Silver, Nate. "The Comey Letter Probably Cost Clinton The Election." FiveThirtyEight.

FiveThirtyEight, 03 May 2017. Web. 30 May 2017.


own forecast. Even more importantly, Trumps polling numbers went up in swing states. On

election night, Trump had close victories in the following swing states: Florida, North Carolina,

Pennsylvania, Wisconsin, and Michigan. Once again, polls after October 28 gave Trump his

highest chances of winning these states.

On the 29th, the day after the letter was released, the print version of The New York

Times was entirely comprised of stories about Comeys letter, headlined by NEW EMAILS

JOLT CLINTON CAMPAIGN IN RACES LAST DAYS. 5 Immediately, almost all other

media outlets followed suit. Between the 29th and the 4th of November, stories detailing

Comeys letter and its impact were the lead stories in multiple national news outlets, six days out

of seven.6 Suddenly, an undecided voter was hard-pressed to find an article criticizing Donald

Trump and it was less than a week until election day.

On election day the two candidates were separated by about three percentage points.7 But

this really isnt that significant a gap. The average polling error of general elections since 1968

is 2 percent and in 2012 the error was 2.7 percent.8 This means that in no way was Clintons

lead safe. Furthermore, fivethirtyeight.coms model had Clinton at a only 64% chance of

winning the electoral college -- ultimately the only measure that matters. (Its important to

5
Front Page of New York Times on October 29th, 2016. Digital image. New York

Times. N.p., n.d. Web. <http://www.nytimes.com/images/2016/10/29/nytfrontpage/scannat.pdf>.

6
Silver, The Comey Letter

7
Enten, Harry. "Trump Is Just A Normal Polling Error Behind Clinton." FiveThirtyEight.

FiveThirtyEight, 04 Nov. 2016. Web. 30 May 2017.

8
Enten, Normal Polling Error
remember that national polls attempt to reflect the popular vote, not the electoral college.) Then

why did The New York Times, whose website alone received 502 million visits in the month of

October, 2016, say in an article released on election day that Clinton had a clear if not

insurmountable advantage over Donald Trump? 9 Its no surprise that much of the electorate

went into election night on November 9, 2016 expecting news of a Clinton victory.

But lets back up for a second. Lets say that Trumps victory wasnt just a polling error

and that there was something fundamentally wrong with polls on the national and state level and

that Clinton shouldve become the 45th President of the United States? So then what did go

wrong? Why were the polls so misleading? This has a lot to do with undecided voters. The

number of undecided voters on election night (12.5 percent) was more than the percentage of

undecideds from the past three general elections combined and the 18.5 percent of voters

undecided in the last 100 days was the largest percentage since 1996.10 It makes perfect sense

that a bunch of minds that arent made up can compromise any poll. A pollsters main job is to

give a preview into a popular vote held in a given population. The problem is, some polls dont

even report participants who answered undecided and when they do, its often just an asterisk.

This can easily make a lead look much more secure than it actually is. (Its worth noting that one

of the unique aspects of fivethirtyeight.coms model is that Nate Silver did take into account, to

an extent, the unusually high number of undecided voters in his model.) Furthermore, the vast

9
Cohn, Nate. "What Time Will the Presidency Be Decided?" The New York Times. The

New York Times, 08 Nov. 2016. Web. 30 May 2017.

10
Silver, Nate. "The Invisible Undecided Voter." FiveThirtyEight. FiveThirtyEight, 25

Jan. 2017. Web. 30 May 2017.


majority of statistical models other than fivethirtyeight.com dont even take into account

undecided voters, or at least to a much lesser extent. If you made up any of those 502 million

views of The New York Times that I mentioned earlier youd have read on that fateful night that

Clinton had a 85 percent chance of winning the electoral college, its no surprise that the loss

was so shocking.11 Oh, you used a different source? It mustve been The Huffington Post; well

they reported a 98 percent chance of a Clinton victory. Or did you use predictwise.com (89

percent chance) or maybe dailykos.com (92 percent chance)12? I really hope you didnt use The

Princeton Election Consortium, an excellent source really any year other than 2016, which gave

above a 99 percent chance of a Clinton victory. Ouch.

Its now pretty clear where all of these undecided voters went: to the current President.

Earlier, I talked about how the Comey letter effected the polls during election week but even

after that, 12.5 percent of the electorate was still undecided. And what happened with this

portion of the electorate is what decided the election. The vast majority of these undecided

voters decided to Make America Great Again. Why? I suspect that it had a lot to do with the

Comey letter still dominating much of the news cycle that week. But, maybe in those final days

of decision making people simply found something in Donald Trump that resonated with them.

That brings us back to the all-important question: Why did Clinton lose? Well, according

to the map, she lost Michigan, Wisconsin, and Pennsylvania, states that were in her supposedly

impenetrable Blue Wall, as well as swing states North Carolina and Florida. Many

post-election articles have pointed the finger at her lack of attention to the ground game in these

11
Katz, 2016 Election Forecast

12
Daily Kos Elections. Daily Ko, n.d. Web. 30 May 2017.
states. For example, she didnt spend any time in Wisconsin.13 But if you look at her

involvement, or lack of involvement, in other states this argument falls apart. First off,

remember that even allowing for a losses in Wisconsin and Michigan, the states in which her

lack of involvement is most often cited, she still wouldve won. So lets look at some of the

other states. In Pennsylvania, both candidates focused extensively on the ground game, and

Trump won by the exact same margin as he did in Wisconsin. Furthermore, Trump spent

significantly more in Colorado, another swing state with a large number of undecided voters, and

still lost it.14 There is no consistent pattern among these state losses that reveals an answer to why

she went fell short.

So if the ground game didnt seal the deal, what did? Well, this essay has been focused

on the We know the inaccuracy of polling during the election meant no one really saw a Trump

victory coming. So then, lets go back to the polls. The states Clinton lost had a majority

population of white, non-college educated voters.15 This chunk of the electorate were

unwavering Trump supporters throughout the campaign: from the the beginning until right

before they walked into the booth. They voted for Trump on a larger scale than anyone expected.

Pennsylvania is a great state to see this phenomenon t. Going into election day, Clinton had

more decided voters in Pennsylvania but critically there were plenty of undecided voters to

13
Arp, Jessica. "Why Hasn't Clinton Come to Wisconsin? Here Are Some

Theories."WISC. N.p., 15 Dec. 2016. Web. 30 May 2017.

14
Silver, Nate. "Clinton's Ground Game Didn't Cost Her The Election." FiveThirtyEight.

FiveThirtyEight, 23 Feb. 2017. Web. 30 May 2017.

15
Silver, Clintons Ground Game
swing the vote to Trump. Also, demographically, we now see that the majority of these

undecided voters were non-college educated whites and, as we now know, the vast majority of

that demographic voted Trump. While that demographic does often vote red, it hasnt done so

on such a scale is a very long time. In the last election cycle much of the idea of the blue wall

was established because of how many non-college educated votes Obama was able to reel in.

If you remember anything from this paper, I dont want it to be why Clinton lost or just

the examples of the gross predictive errors weve identified. I want you to remember to always

take a closer look when evaluating at any statistical model. Figure out why a certain candidates

chances are so high or so low, but dont just look at one source; make sure to read many different

opinions. Otherwise, youre going to end up like most democratic voters in the early hours of

November 9th, 2016: dazed and confused.


Works Cited

Arp, Jessica. "Why Hasn't Clinton Come to Wisconsin? Here Are Some Theories." WISC. N.p.,

15 Dec. 2016. Web. 30 May 2017.

Cohn, Nate. "What Time Will the Presidency Be Decided?" The New York Times. The New York

Times, 08 Nov. 2016. Web. 30 May 2017.

Daily Kos Elections. Daily Ko, n.d. Web. 30 May 2017.

Enten, Harry. "Trump Is Just A Normal Polling Error Behind Clinton." FiveThirtyEight.

FiveThirtyEight, 04 Nov. 2016. Web. 30 May 2017.

"FiveThirtyEight." Wikipedia. Wikimedia Foundation, 23 May 2017. Web. 30 May 2017.

Front Page of New York Times on October 29th, 2016. Digital image. New York Times. N.p.,

n.d. Web. <http://www.nytimes.com/images/2016/10/29/nytfrontpage/scannat.pdf>.

Katz, Josh. "2016 Election Forecast: Who Will Be President?" The New York Times. The New

York Times, 19 July 2016. Web. 30 May 2017.

Silver, Nate. "Clinton's Ground Game Didn't Cost Her The Election." FiveThirtyEight.

FiveThirtyEight, 23 Feb. 2017. Web. 30 May 2017.

Silver, Nate. "The Comey Letter Probably Cost Clinton The Election." FiveThirtyEight.

FiveThirtyEight, 03 May 2017. Web. 30 May 2017.

Silver, Nate. "The Invisible Undecided Voter." FiveThirtyEight. FiveThirtyEight, 25 Jan. 2017.

Web. 30 May 2017.

Silver, Nate. "A User's Guide To FiveThirtyEights 2016 General Election Forecast."

FiveThirtyEight. FiveThirtyEight, 05 Aug. 2016. Web. 30 May 2017.

Вам также может понравиться