刘项原来不读书:大数据的笑话

万得福

To What Extent Are Clinton and Trump Campaign Strategies Driven by Analytics?News

• Author: Carlos Alberto Gómez Grajales
• Date: 04 November 2016
• Copyright: Image appears courtesy of Getty Images
  

This is it. We are only a few days away from one of the most important elections on the planet, the U.S. presidential election. Even considering the impact and attention this election gathers every four years, this particular time the campaign has been… peculiar, to say the least. The rather shocking nomination and campaign of the republican candidate Donald Trump has dominated the news worldwide, flooding media with each one of his statements, insults and scandals. It’s T.V. gossip along political gossip, a jackpot for the media.

But I’m not here to discuss the speeches of the candidates or the many, many differences between them, which are very well documented by now. Instead, I’m going to focus on a major difference between the two campaigns, one that will have a profound effect on the outcome we’ll see in a few days and that is often neglected in media coverage: their data operations.

Let’s start discussing the democratic campaign, since there’s more to talk about. To put in perspective the amount of statistics used in Hillary Clinton’s campaign, it is only necessary to mention that a former Obama staffer declared that he had never seen a campaign that’s more driven by analytics than this one [1]. The democrats use statistical analyses to decide things such as voter contact, where the organizers spend their time, which voters to call, which to reach by social media and which to knock on their doors [1].

Hillary Clinton has a full-time internal department of analytics [1]. She hires about 60 analysts, all under the guidance of Elan Kriegel, a statistician who runs the campaign’s statistical efforts. He is one of the top campaign advisers and one of the highest-paid ones as well. Kriegel is responsible for many of the top campaign decisions: he and his team were responsible for deciding where and when to place each of the 60 million dollars that Clinton invested on T.V. ads during the primaries. To allocate the resources, the analytics team devised a model that optimized the “cost per flippable delegate,” an interesting KPI that ensured that Clinton always had an edge on the race against Bernie Sanders. All this has been possible in part due to Clinton’s access to a huge Democratic National Committee dataset of voters [2] [3]. This database has been gathered since the 2008 election, with information about which magazines voters subscribe to, whether they like to vote early, how likely they are to open certain emails, among others, precious information that Kriegel and his associates have surely been using to push their candidate.

Now it is time to discuss the other side of the coin. The analytical efforts of Trump's campaign have been as unstable as his candidate. As early as May, Trump declared to the Associated Press that data was “overrated” and that he intended to win the campaign solely based on the strength of his personality [4]. However, since his poll numbers started to crumble this summer and due to the insistence of his closest advisors, Trump decided to make a final push on analytics [5].

Trump hasn’t been as open as his rival about his campaign’s data operations, but financial reports show the campaign paid, this July and August, a good amount to Cambridge Analytica for a dataset of voter information paired with consumer data [5]. He has also requested the Republican National Committee database, a 100 million dollar effort the party started four years ago. But even with these massive sets of information at his disposal, the campaign has only had a mere 4 months to put it into use, as Trump did not invest a penny in analytics during the primaries. As such, their digital data-driven operation has made some hilarious hiccups. Some notorious Democratic donors reported to have received invitations to donate to Trump’s campaign [5]. Even Vicente Fox, a former Mexican ex- president, who is not allowed to contribute to American politicians as he is not a U.S. citizen, tweeted earlier in the year that he had received Trump fundraising emails [5].

This difference between the two campaigns may prove to be crucial for the November 8th outcome. But don’t take my word on it, listen to Zac Moffatt, Mitt Romney’s digital director during his 2012 campaign, a republican who already expressed his worries about the gap they have with the Democratic analytics operation [1]. A few months ago, Moffatt feared that whoever emerged as the GOP nominee would be perilously handicapped when it came to data analytics just as Romney had been compared to President Barack Obama who, like Clinton, had honed an analytics operation more than a year in advance, even taking all their analytical operations to the mid-term elections in 2014 [3].

This year's election will certainly be pivotal for the use of intensive statistics-based political operations. The many statisticians working on this campaign will surely develop a whole new set of tools that will redefine election cycles for years to come. Sadly, with all the remarkable differences between Clinton’s and Trump’s campaign strategies, consultants and the media will have so many theories to explain the election that the value of the most important data-driven operation ever seen in politics may be overlooked, once again.

References

[1] Goldmacher, Shane. Hillary Clinton’s ‘Invisible Guiding Hand’. Politico Website (September, 2016)
http://www.politico.com/magazine/story/2016/09/hillary-clinton-data-campaign-elan-kriegel-214215
[2] Clinton to get help from Obama campaign machine. The Associated Press (July, 2016)
http://bigstory.ap.org/article/3b1308e6139d405d8660c20f8f414fcf/clinton-get-help-obama-campaign-machine
[3] Grajales, Carlos A. Gomez. How statisticians have changed elections. StatisticsViews Website (Nov, 2014)
http://www.statisticsviews.com/details/feature/6931581/How-statisticians-have-changed-elections.html
[4] Trump's questioning of the value of data worries Republicans. The Associated Press (May, 2016)
http://bigstory.ap.org/article/6d588a38061c4657a557d1dde86782ec/trumps-questioning-value-data-worries-republicans
[5] Trump, once a data skeptic, spending millions on data. The Associated Press (October, 2016)
http://bigstory.ap.org/article/608cc0a7c83c48a88e65ccacac35931b/trump-once-data-skeptic-spending-millions-data

万得福

look back, how wrong MSM polls were

万得福

美利坚大学历史系教授Allan Lichtman的预测——他此前成功预测了30年未失手，这次又成功预测川普当选。如此高的准确率，方法其实并不复杂——看以下13个条件，如果有6条或以上是不成立的，那么目前正在执政的党派将会失利。




1. 中期选举之后，执政党比上一次中期选举后占有更多的众议院席位。

2. 执政党内部初选阶段没有激烈的竞争。

3. 执政党的候选人就是目前的在任总统。

4. 没有重要的第三党或独立候选人。

5. 竞选期间，经济没有处于衰退。

6. 这一届政府的实际人均经济增长超过或等于前两届。

7. 执政党进行了重要的全国范围内的政策改革。

8. 这一届政府没有持续的社会不满。

9. 这一届政府没有重要丑闻。

10. 这一届政府没有外交或军事上的重要失败。

11. 这一届政府在外交或军事上取得了重要成绩。

12. 执政党候选人很有个人魅力。

13. 挑战者候选人没有个人魅力。




对照这个列表，1、3、7、11、12都是铁定不成立的，4很有可能不成立（因为有Gary Johnson的存在）。所以共和党肯定会赢。




可以看到，这种预测方式基于的是美国政治中两党轮替的基本特征。当一个党派连续执政8年之后，要再连任是相当困难的。

万得福

http://www.zerohedge.com/news/2016-11-04/