Psycho-demographics: Drilling Down on Trump's Campaign Analytics

(Image: rvlsoft/Shutterstock)

(Image: rvlsoft/Shutterstock)

Did psycho-demographics help Donald Trump win the US presidency? Several recent articles, including this one, have chronicled the use of this type of consumer profiling and related marketing tactics as the secret sauce behind Trump's winning campaign against rival Hillary Clinton.

But just what is a psycho-demographic profile? This kind of profile looks at what you like and don't like to determine what you might want to buy, who you might want to vote for, and what marketing messages would most motivate you to take a particular action. For instance, instead of segmenting this middle-aged female suburban resident as a soccer mom, it might profile her according to whether she liked the TV show Glee on Facebook and draw conclusions about her politics and consumer preferences based on that.

Michal Kosinski is an assistant professor at Stanford, and a doctor of psychology and holds a Masters degree in psychometrics. He created a method to profile consumers according to their Facebook likes. Recent reports say that a consulting firm unaffiliated with Kosinski, Cambridge Analytica, leveraged that research to deliver the White House victory for Trump. (Kosinski will be our guest on AllAnalytics radio on May 23, and you can register here now to join us for that show.)

If you want to take Kosinski's algorithm for a test run yourself you can go to this web page and link your Facebook or Twitter profile, or input some text. When I linked my Facebook profile and asked the page to evaluate me, it did correctly guess my Myers Briggs type as INTP, which is a rarer type, especially for women. I thought that was pretty impressive that it got that right. But it also predicted that I am male, which is not the case. When I input the text one of one of my blog posts, it predicted that I was an INTJ, also pretty close. And it had even more confidence that I was a man.

(In fact, two other professional women I know who tried it and reported on their results were determined to be men.)

But the power here is really is more about the fact that this algorithm looks at how you think and what you like in addition to what can be observed about you from the outside -- your age, gender, marital status, and home address. And while Kosinski's algorithm predicts your Myers Briggs type, gender, and other traits about you, it also classifies each person according to a "Big 5 Personality" traits lines.

The first is "conservative and traditional" vs "liberal and artistic," and your Facebook likes will determine where you fit along the continuum. The other trait lines are "impulsive and spontaneous" vs. "organized and hardworking," "contemplative" vs. "engaged with the outside world," "competitive" vs. "team working and trusting," and "laid back and relaxed" vs. "easily stressed and emotional."

By marketing to you based on your psychology and personal preferences rather than what you are supposed to like because you are a 50-year-old married suburban woman, or a 25-year-old single urban man, this type of profiling may offer marketers a more accurate tool to use to craft messaging.

This type of profiling has been in use for years, Cathy O'Neil pointed out in a post for Bloomberg earlier this year. And it's been used not just by Trump's campaign, but also by Obama's and by Hillary Clinton's campaign, too.

You can bet it will be used for years to come as well. I hope you can join us as we welcome Kosinski to our show on May 23. Have questions for him? Post your questions in the comment section of the registration page, and we'll ask them during the show.

Jessica Davis, Senior Editor, Enterprise Apps, Informationweek

Jessica Davis has spent a career covering the intersection of business and technology at titles including IDG's Infoworld, Ziff Davis Enterprise's eWeek and Channel Insider, and Penton Technology's MSPmentor. She's passionate about the practical use of business intelligence, predictive analytics, and big data for smarter business and a better world. In her spare time she enjoys playing Minecraft and other video games with her sons. She's also a student and performer of improvisational comedy. Follow her on Twitter: @jessicadavis.

Success Secrets of Top Omnichannel Retailers

It's a tough and changing environment for retailers. Yet some are enjoying continued success during turbulent times. We take a closer look at how they do it.

A2 Radio: Lean Analytics for 2018

Lean Analytics author Alistair Croll joins AllAnalytics radio to talk about how to apply the process for 2018.

Re: Trust but Verify
  • 5/10/2017 9:58:09 PM

Maryam writes

... analytics do change and your assumptions should be verified making the blanket case that the past will predict the future is just not the case and the Clinton campaign learned it in a very public way. Trust but verify do not take anything for granted analytics are only as good as the data they evaluate.

As I recall, post-election analysis of the polling analytical methodology more or less concluded that the data analyzed was valid, only the analytics were designed to skew the data that were selected as inputs for the predictive models. Which in a way highlights another hazard to watch out for in predictive analytics.


Re: I can not imagine
  • 4/28/2017 12:37:46 PM

Very sophisticated and it apparently worked.

Re: Trust but Verify
  • 4/28/2017 12:36:59 PM

I agree. It just shows you can't predict the unpredictable.

Trust but Verify
  • 4/28/2017 10:35:39 AM

While I am sure that these types of analytics will be used in the future we should also remember that the analytics of the campaign also let Hillary Clinton and the majority of America to believe she would win the election. What we learned is that analytics do change and your assumptions should be verified making the blanket case that the past will predict the future is just not the case and the Clinton campaign learned it in a very public way. Trust but verify do not take anything for granted analytics are only as good as the data they evaluate.

Re: Stereotypes as analytics
  • 4/19/2017 5:31:36 PM

I personally suspect that many of these fake news sites were created to target those psycho-demographics that were most likely to spread them around.  I've seen some Facebook pages where the users could have UFO conspiracies nut jobs rather than political nut jobs by of posts they had. 


It is amazing what little details can reveal about us.  I remember a few analtyical stories that guessed people's political affiliations by their grocery purchases and another one by what type of car they owned. 

Stereotypes as analytics
  • 4/19/2017 1:13:22 PM

In the old days we had stereotypes of racial, economic, religious and other groups based on the analytics of that time.  In some cases, the stereotypes were true of 80% of a demographic group.  In other cases only 50%, or 10%.

Analytics aim to consistently do better than the 50% and 10% levels of accuracy of the past.  But at what level of accuracy, of probability, do they become useful?  At what level do they become ethically acceptable.

Is politics different from mortgage marketing or some other activity?

If mortgage company analytics say 80% of racial group A fit its underwriting guidelines and only 20% of racial group B fit, is it ethically justifiable to ignore the 20% of group B that do fit its underwriting guidelines?

How different is politics?  If politician R or D is better than the opponent, is it ethically justifiable to write off the 45% in the "can't win" districts?

Need a benchmark?  The old Chicago Precinct Captain.  The patronage worker had the anlytical duty to know every person in the precinct.  Many cut/pasted sections of the precinct poll list sorted by address into a binder.  They  made notations beside voters about political preferences and personal attributes.  They wrote in the names of those not registered and whether they should be registered because they would vote the right way ... or whether they should not be registered because their vote could not be trusted.

The lazy precinct captain would see the ethnicity or religion or other stereotype of a voter and calculate:  To keep my job I need to do equal or better than last election.  He played the percentages of the stereotype.

The ambitious precinct captain dug below the stereotypes to find the 20% who did not fit the stereotype and pushed his vote total high enough to get a better patronage job.

Jane Byrne was the Trump style black swan event of Chicago. She caught the overconfident precinct captains napping.  Her unpaid volunteer staffers had better analytics on voters post-DaMare and effectively reached those voters to create the Trump style upset.  As with Trump, she was opposed by the establishment, not due to issues or personality but due to the fact that she beat the establishment at their own game.

Opponents of the black swan respond with inaccurate stereotypes that further destroy their credibility all the while pretending to wrap themselves in analytics.

Re: I can not imagine
  • 4/17/2017 7:15:48 PM

This is just more reason to hate social media: First we create our own echo chambers based on likes, preferences and who we're friends with, then the cash-rich influencers start running their games to, as you note, Jim, help a candidate appear both moderate and conservative.

And we are passive, unthinking consumers of this data and so-called news from these platforms.


Cambridge Analytica connections
  • 4/17/2017 5:55:09 PM


Jessica writes that "... Recent reports say that a consulting firm unaffiliated with Kosinski, Cambridge Analytica, leveraged that research to deliver the White House victory for Trump."

There are some interesting (and perhaps sinister) connections here – although only very indirectly linked to Michal Kosinski's research. This has been discussed in a Feb. 26th article by Carole Cadwalladr in The Guardian titled «Robert Mercer: the big data billionaire waging war on mainstream media». Various reports finger Mercer as the primary wealthy influence that has pushed Alt-Right operatives such as Steve Bannon and Kellyanne Conway into powerful roles in Trump's organization.

Cadwalladr reports that, while doing research into the Alt-Right media activities of the Breitbart outfit, she "recognised Robert Mercer's name: because of his connection to Cambridge Analytica, a small data analytics company." She elaborates:

He is reported to have a $10m stake in the company, which was spun out of a bigger British company called SCL Group. It specialises in "election management strategies" and "messaging and information operations", refined over 25 years in places like Afghanistan and Pakistan. In military circles this is known as "psyops" – psychological operations. (Mass propaganda that works by acting on people's emotions.)

Cambridge Analytica worked for the Trump campaign and, so I'd read, the Leave [Brexit] campaign. When Mercer supported Cruz, Cambridge Analytica worked with Cruz. When Robert Mercer started supporting Trump, Cambridge Analytica came too. And where Mercer's money is, Steve Bannon is usually close by: it was reported that until recently he had a seat on the board.


When I noted the reference to Kosinski's research, I immediately felt this information on Cambridge Analytica might have further relevance to a discussion of the Trump campaign's use of analytics.


Re: I can not imagine
  • 4/17/2017 1:19:12 PM

Right, but the ability to select the right social media audience for a given message was only part of it. It also allowed him (and reportedly others before him) to deliver what could be contradictory messages, making the candidate appear as both a conservative and moderate via social posts that only selected audience groups could see.

Re: I can not imagine
  • 4/17/2017 9:34:10 AM

Jim is right. It was about using your psychodemographic profile (created from your Facebook likes) to determine which message would be the most effective on you.

Page 1 / 2   >   >>