https://medium.com/startup-grind/how-the-trump-campaign-built-an-identity-database-and-used-facebook-ads-to-win-the-election-4ff7d24269ac is purely journalistically descriptive, and while this second link is more analytical it does not contain any solid academic references. You may be able to track down Kosinski and the other names for hard statistical  research, such as links to academic papers or theses/copies of their research - but I imagine Cambridge Analytica will be holding their cards close to their chests... https://motherboard.vice.com/en_us/article/how-our-likes-helped-trump-win

I noticed your question a few days ago and I have been thinking about it quite a bit. I think that the growing consensus is that the Trump campaign's investments in digital marketing and the use of social media platforms provided him an edge against the Clinton campaign. In a recent interview in the Recode's Code Conference in Silicon Valley, Hillary Clinton lamented that the Democratic National Committee had failed to make the necessary investments in its data operations. Her allegations has created quite a storm in Democratic circles as the DNC data people are pushing back that their databases were updated and helped many candidates -- Governors, Senators, Representatives to the House -- win elections. The new book Shattered by Jonathan Allen and Aime Parnes did not mention this issue - which I find surprising. But the authors do note that Clinton's data operations team felt that they had very strong data and that they had been able to identify her supporters. Indeed, Roby Mook, the campaign manager, believed that Clinton would win the election by turning out her supporters. He did not however believe that he could persuade undecided voters. So it seems that the campaign made a choice to focus their advertising to support turnout operations. In many ways, the strategy was not a complete failure. The campaign used 2012 as good standard to model their strategy. Take Pennsylvania as an example. Obama got 2,907,448 votes in 2012 and Hillary Clinton 2,926,441 in 2016. Trump won the state by less than 50,000 votes. It is important to note that 5.7 million Pennsylvanians votes in 2012 to 6.1 million in 2016. This story repeats itself in several other places. In the end, Trump edged Clinton in Midwestern states and he spoke to the needs of poor White Americans in a way that others have not. In fact, Bernie Sanders did the same thing -- though he also appealed to White progressives. In Shattered, the authors document some fights between Bill Clinton and Roby Mook on whether they should persuade White working class voters to support Clinton. Mook decided that it would be a waste of money and his data may have suggested that he was correct. But in the end the turnout operations failed to get the Obama coalition out to vote. In a new piece in the Washington Post's Monkey Cage, Fraga et al (2017) note that African American voter participation declined by 4.7% and this happened especially in Pennsylvania, Michigan and Wisconsin. They use the voter file to arrive at their numbers and it is an interesting study. For my part, I have downloaded all of Trump's and Clinton's Facebook posting to try to understand how each campaign communicated with their "followers". I don't have any conclusions at this time. But what is clear is that Trump was more effective at getting followers to interact with his messages. I have not ran the numbers yet -- hopefully I will do so in the next week or so -- but the Trump follower was more willing to "share" Trump's post than the Clinton follower was willing to share her posts. This is important. Because Trump's post may have been seen by more voters than Clinton's. If we believe the Pew Research Foundation's numbers that 70% get their political news via Facebook, Twitter, etc.... then it stands to reason that Trump was more adept at communicating his ideas. And that he was also better at persuading undecided voters. Indeed, the number of undecided voters by late October was quite high and it seems that they broke to Donald Trump's favor. I wish I could calculate this with existing polling data but I first I would have to find some reliable numbers. I am including a link to an article on Facebook and the election. SORRY FOR THE LONG ANSWER... but your question is one that I have wondered myself and I have use this opportunity to organize my thoughts.... GOOD LUCK with your work....

https://www.washingtonpost.com/news/monkey-cage/wp/2017/05/08/why-did-trump-win-more-whites-and-fewer-blacks-than-normal-actually-voted/?utm_term=.72697fba0bfa

Thank you very much for your answer !

It is very complete and addresses the strategies of digital communication, and particularly the use of social networks in campaigning. But what I notice from your answer is also the observation that I make about this field of research : there are several interpretations, which are opposed or complement each other ... but we can not do anything else to use other variables ( Participation, etc.) to talk about digital campaigning.

As a result, there is a little bit of a myth, and we can not do research with more details because we do not have data. This is especially true for targeting. My goal is to verify the value of the targeting (as a technique), which is often seen as a "new electoral science". Literature never speaks of it precisely, but rather from canvassing, and sometimes tends to assimilate the two ... which is understandable but imprecise.

Thanks specially for the Clinton talk about the DNC database.

Do you have some links for this :  " Her allegations has created quite a storm in Democratic circles as the DNC data people are pushing back that their databases were updated and helped many candidates "  ?

I completely agree on the lack of data. We have a lot of reporting on the use of these techniques to microtarget voters. In the past, we could see some of the mailers used, etc.... The Trump team mentioned after the elections that they were crafting multiple adverts every day and using Facebook to reach particular audiences. Facebook allowed them to test taglines, themes and ideas. But all these observations are merely anecdotal so it is difficult to assesses what is real and not. I suspect that the story could be different if Trump lost -- or at least I am guessing that is the case. Still Facebook data could be useful in the sense that we can measure followers' reactions to certain ideas, themes, taglines.....There are of course limitations analyzing the data. The Facebook API does not allow me to access all the demographic data I would like. Plus they limit the amount of comments I can access. So there are posts were more than 50,000 people have commented but I can only access around 24,000 comments. Facebook is providing the top comments comments but there is no way you can assess whether these downloaded comments are representative sample. But we can get a better picture of people's attitudes and these comments sometimes are more insightful than polls, given that polls limit people's choices to particular answers. 

Indeed, my main interest is followers' reactions to specific issues - trade, terrorism and foreign policy - which are my main lines of research interests. So, I am not only studying Trump's posts but followers' comments to the posts. And what I am noticing is that the followers' comments largely echo Trump's rhetoric, ideas and worldview. My analysis, which is in the early stages, is not a scientific given that there may be problems regarding replication -- I am not sure whether the comments are downloaded are the same another person is downloading.

I am doing this now in terms of his decision to put out of the Paris Agreement. My guess is that he and his advisers believe that it is beneficial to talk directly to his supporters. Indeed, his personal Facebook account (and also his Twitter account) has been telling his supporters that he was thinking about the Paris Agreement for the last week or so. And people have been providing comments. I am not sure whether the White House is reading these comments but I am guessing that his team knows that the decision would resonate with his base. And I am also guessing that part of the reason that the he is doing this is because it will drown out the Russia controversy, get his base excited, and more importantly get the Sunday news, show which tend to set the media's agenda in the week, to focus in this divisive issue.

Also, it is important to note that the White House has barely used its official Facebook page to talk about the Paris Agreement, though it did have a Facebook Live link to yesterday's speech.

One final thought. Trump and his team have done a great job questioning the mainstream media's truthfulness. And his supporters, longing for good news, tend to depend on his posts. At a time when Trump's popularity is dipping among Americans who voted from him, his team knows that they need to keep up his numbers with the core supporters and that is what the Facebook account and the tweets do.

On Clinton's allegations and the brewing debate in the DNC. Here is a piece from CNN published earlier today.

http://www.cnn.com/2017/06/02/politics/hillary-clinton-dnc-data-pushback/index.html

Does anyone have anything new about political technology data?

The lack of research on this object, or the absence of research, is very interesting.

Indeed, it is one of the most covered objects of the media when it comes to dealing with elections. But yet, very little is known about it in science, and in political science in particular.

I don't think I've ever seen, for example, research on the effectiveness of micro-targeting algorithms for electoral micro-targeting? data quality? etc.

And still no international database that would synthesize the different characteristics of digital campaigning... ?

All this at a time when these are the most important changes of the last 15 years and are becoming increasingly important.

Hi,

Some time later, is there any news on this issue of the electoral targeting datasets or database ?

We have written here (https://www.researchgate.net/publication/336777794_Digital_Campaigning_in_France_a_Wide_Wild_Web_Emergence_and_Evolution_of_the_Market_and_Its_Players) an article on the situation in France.

Are you aware of any other recent case studies on electoral micro-targeting?

The Internet Policy Review has released a very interesting special issue on Data-driven elections (2019, DOI: 10.14763/2019.4.1433), but still no comparative database/dataset (of any kind).

Best regards,