That was then, this is now

The data came from Facebook:

During Obama’s initial 2008 bid for office, his team had already embraced technology in a greater capacity than any before it, assembling massive email lists and other targeted initiatives that earned Obama historic fundraising tallies. But for 2012, campaign manager Jim Messina wanted to take things even further.

To get there, his staff needed to link what had previously been disjointed databases of voter information (collected by volunteers, pollsters, and other campaign workers) into a single, comprehensive pool unrivaled in detail and scope. Whereas most voter logs used by campaigns often list only names and telephone numbers, Obama’s advanced tool dove into specifics like age, race, district, and voting history: it allowed field workers to rank voters intelligently and not waste time chasing unlikely votes.

Here is the article from The Verge, via the excellent Ben Thompson.  I recall similarly glowing coverage from The New York Times.


This is what peak performance looks like???

I think Tyler's point is that intrusive big tech is still described in glowing terms when used in service of a Democratic campaign. If it were a Republican campaign, it would be something to be feared. Reporting on big infusions of outside cash into a campaign breaks down along the same lines.

Note the apologists below, seeking to limit the discussion only to what is legal and to avoid considering what may or may not be socially desirable.

Heh. Illustrating my point, the W. Post had a glowing article today about Buttigieg staying afloat with lots of celebrity donations.

So what you are saying is Obama colluded

"But the Obama campaign wasn't alone in its obsessive hunger for statistics, percentages, and other figures. In the weeks leading up to Election Day, Nate Silver's fivethirtyeight blog became an essential destination for politico diehards. Silver's supremely-detailed analysis of polling data led him to predict (perhaps flawlessly, pending the final outcome in Florida) a race many dubbed too close to call as voters took to the polls. "Data is vindicated," tweeted the Poynter Institute in reaction to Silver's incredible accuracy. Barack Obama has yet to be sworn in for a second term, but the question has already become whether data can again prevail so decidedly in 2016."

Nope. Trump did a head fake and an end run around the clueless lefty media.

Brilliant and delightful!

Delicious even!

Nate Silver gave Trump a 28.6% chance of winning in his final results. His modelling (and fivethirtyeight's reporting on it) recognized that Trump had a very real chance of winning, when most of the media did not.

The fact that Trump won doesn't prove him "wrong". There is a huge amount of uncertainty in elections. Events with a 28% chance of happening happen quite frequently.

Silver is a liberal, but he is a straight shooter when it comes to poll analysis, and I believe in his methods. He is the best in the business.

Nearly agreed. Silver is nearly wholly trustworthy in this regard. I do however sense he pulls his rigour in a bit when he runs into uncomfortable data (This may not be entirely his fault. 5:38 has a breaking liberal readership that doesn't like to hear uncomfortable data; they've had at least 1 case of discontinued analysis / correspondent because they didn't like where it was going...).

But Mr Silver is indeed very honest, and competent. (See, liberals, it can be done if you get out of your little bubbles!) However, as we say in statistics, there is no such thing as chance. If Silver had been as good as people like to think he is, he would have called Trump 100%, not 28%.

"I do however sense he pulls his rigour in a bit when he runs into uncomfortable data (This may not be entirely his fault"

Silver has admitted that he is prone to that but is honest enough to confront it and push back against his own biases. He deserves praise for his work.

Maybe he pushes back just a little before giving in, like this ...

But what's the point of Nate's 28.6% chance prediction? What does it mean to say he isn't "wrong"? Any 2 horse race will have one of 2 winners, and if I give the smaller chance to the eventual winner, I think I'd say I've got it wrong.

"But what's the point of Nate's 28.6% chance prediction?"

I tell you that the odds of rolling a '6' on a 6 sided die is 1/6 or 16.7%, but I roll a '4', am I wrong?

Please read a basic book probability and statistics.

Is this parody? You think Silver observed the number of possible election outcomes and divided by ... what, exactly? You think Silver ran the 2016 election thousands of times and observed how often each candidate won? There was no probability or statistics at work in Silver's estimate, and it's embarrassing for you to have asserted that there was.

Silver's whole shtick is that he weights the polls based on his personal judgment, which he advertises as being better than other people's. The precision implied by his use of a decimal point is faux. He was less wrong in 2016 than most people, but he was still wrong.

His defense, by the way, was that the polling data was too imprecise to make an accurate prediction. In other words, his defense was that the work product he puts out was always inherently useless.

Correct. Like all humans, he is biased, but he is the darling of the left, their Oracle at Delphi, because he tells them what they want to hear.

I don't believe that there has been a single case where the Democrats have outperformed Silver's prediction. Sometimes he's spot on, sometimes the Republicans do better than he said. If you read him with that hermeneutic in place, you can learn something. I always try to learn from people even if they are not as smart as I am.

Democrats outperformed nearly all of his predictions in the 2018 elections.

"Silver is a liberal, but he is a straight shooter when it comes to poll analysis, and I believe in his methods. He is the best in the business."

You are a true believer, with all that implies.

I think this article is overstating the effects of campaign tech while also missing what Obama’s advantage was. I worked on Republican state legislative campaigns in 06 and 08. They had a massive database called Voter Vault (I think that was the name) open to all Republican campaigns with these data. It would produce walking lists for campaign workers based on whatever criteria you had mind such as vote history on particular races or likelihood to vote based on those data, etc.

Where Obama has a bigger advantage was with employing experiments to test out the effectiveness of all sorts of campaign approaches and media. They had a top experimental political scientist leading these efforts.

But, the evidence suggests that even without these campaign advantages, Obama would have still won. “It’s the economy stupid” and all that. Trump loving, Romney hating Republicans love to believe that Romney lost the election. But it was a by the book outcome that political scientists predicted months in advance. See Sides and Vavreck’s book, The Gamble.

For more on campaigns’ use of voter databases, see Eitan Hersh’s book, Hacking the Electorate. It turns out that it’s hard to do. The most important info is if the state tracks voters party registration. Turnout history helps too. But that’s about it. I have friends who work for big data firms and try to predict vote choice with hundreds of variables and fancy AI and Bayesian models. It turns out there models are just barely better than only knowing someone’s party ID. There’s also a danger with micro targeting political ads in that you might send the wrong message to the wrong people and actually hurt yourself more than helping. Again, we overstate how much these campaigns matter or are persuading/tricking voters. It’s hard to do this right. Maybe Russian trolls and ingenious social media ad campaigns mattered in 2016 because the election came down to 78000 votes, but even then, I’m skeptical. It probably had more to do with lower minority turnout and some former Obama voters who wanted an outsider to shake things up. Those factors didn’t need fancy micro targeting ads.

"Trump loving, Romney hating Republicans love to believe that Romney lost the election. But it was a by the book outcome that political scientists predicted months in advance. "

There was little chance of Romney winning. However that makes the media's demonization of Romney even more despicable. And Romney's attempts to gracefully deal with people who were constantly attacking him by enaging them on their own terms made him look weak.

It's one of the critical reasons Trump won. Trump doesn't engage the media on their terms. The media (and most politicians) are left apocalyptic with rage but impotent.

The Dems taught us, via the Romney debacle, that nice guys finish last, as Pelosi knows very well. What's the title of her book.

Trump is a brawler.

Yes: *very* few of X million voters are going to change their decisions based on the equivalent of play-the-record-backward messages.

It's hard to move the needle long-term between Coke and Pepsi through brilliant marketing, so it's probably hard to move the needle between Republican and Democrat through marketing as well.

A consultant does not needs actual performance to invoice a client. All you need is that the client believes he's getting something for the money. This is dishonest, but it's marketing.

Amen! Of course Cambridge Analytica claimed to have won the 2016 election, and of course the liberal media thought that was scandalous when the right was playing this game

A lot of times what the client is getting is a "outside expert" stamped report that justifies what he wanted to do anyway.

I've seen the same thing done with nominally unbiased internal working groups where the recommendations somehow pretty well match what the CEO wanted to do anyway. It's all in who you invite to give input.

Suddenly in 2016 a bunch of Russian doofuses with tiny budgets outsmarted all the Democratic Party geniuses who had won in 2008 and 2012.

Or maybe not ...

Well, Cambridge Analytica got a pink haired guy and quizzed people the Big 5, so obviously that was a game changer of psychological power and insight.

It is psychologically intolerable that clever liberal opinion should have completely misjudged national mood, especially the concerns of flyover country and those rubes still stuck in jobs that don't use MacBooks.

So, obviously, like, it was a conspiracy.

Yes, national mood. That's definitely what drives presidential races. Nothing to do with weird state-by-state voting that effectively disenfranchises 90% of the country, whose mood thereby becomes irrelevant.

Odd, still no mention of these people - It isn't about a single campaign, it is about the bounty that B-B brings us all.

Like this - 'Your volunteers can now reach voters more effectively with i360 Text.

i360 is always ahead of the campaign technology curve,and i360 Text is no exception. Our closed-loop, peer-to-peer texting feature enables you to cut through the clutter and reach your target audience quickly and effectively.

With i360’s closed-loop system, there’s no need to import and export your data through various systems, wasting time and money.

Instead, get the full picture of your grassroots activity and monitor all your voter outreach seamlessly in the i360 Portal.' And you have to admire that someone wrote 'grassroots' without any apparent irony. Assuming one gets into the outer party space, you can even request a demo -

Truly, we are living in a golden age of privacy, brought to us by B-B. One where facebook is merely a supplier to a company that claims the following - 'i360’s comprehensive data is a unique combination of hard data points and predictive modeling. Our dataset incorporates extensive political identification, coalition and membership information collected by way of in-person, phone and online surveys, as well as through partner relationships, in addition to lifestyle and consumer data collected from multiple top-tier providers. Our data is further enhanced by our suite of predictive models, filling in gaps and helping us build the most complete profile for every individual possible.'

There was a big difference. The extent of Obama's Facebook campaign was basically a get-out-the-vote app that had to be downloaded along with targeted ad buying while Cambridge Analytica was mass harvesting user data pretending to be a personality test. Facebook deserved their $5 billion dollar fine, probably more.

snopes: Although Obama for America did collect data on users’ friends, it was at the time in line with Facebook policy. A Facebook spokesperson told us both candidates Obama and Republican Mitt Romney had access to the same tools. In 2015, Facebook changed the rules so that apps could no longer target the friends of users who downloaded them.

The Big O's campaign did mass harvest friend data as well, not just willing app downloaders. The "moral difference" between CA vs Obama campaign depends on how much weight you put on compliance/non-compliance with facebook's internal policy.

'depends on how much weight you put on compliance/non-compliance with facebook's internal policy'

Give or take a FTC consent decree involving such compliance, as reported on in 2011 - '“It is ordered”, says the FTC’s proposed settlement with Facebook, that the social network “shall not misrepresent in any manner … the extent to which it maintains the privacy or security” of its users.

And with that order, the battle between the FTC and Facebook has reached a settlement, as was foreshadowed earlier this month.

The FTC has announced the proposed settlement, under which it had alleged that Facebook “deceived consumers by telling the they could keep their information on Facebook private, and then repeatedly allowing it to be shared and made public”.'

The legal documents are here -

I agree that violating ToS is an important distinction but your "moral difference" excludes the difference between the consent of the person using an app whose very purpose is to communicate with your friends and therefore requires this data versus the fraud of acting as a fake quiz to grab those same details. Without consent many acts that have the same end result risk crossing a legal line so no I don't think painting it as "both sides did it so it boils down to ToS" shows the whole picture.

i'm sure it made sense in your head

It does to me, and not only that, Tyler isn't too clear about what he is trying to clear by implication.

A big data operation using public or legal information is legal.

You certainly don't need to say "but other people use computers!" to make that case.

the same end result

Are there other kinds of results? If so, why do we not ever see terms such as "the interim result" or "the temporary result"? If not, what does "end" add to the concept?

"the interim result" gets used regularly, see also "short term" vs "long term" as opposed to just "later".

anonymous: between the consent of the person using an app whose very purpose is to communicate with your friends and therefore requires this data versus the fraud of acting as a fake quiz to grab those same details

See, that's a difference between "Bob gives away the data of his friends" and "A quiz takes Bob's friends without asking Bob". Who the fuck is Bob to decide on that question? And let's note Bob's probably does not even give informed consent in the first place.

To boil that down to somewhat less opaque language than you have used to describe what the Obama campaign did (no, they did not just take it to facilitate you communicating with your friends).

So that seems to me a minor distinction of ethics compared to directly asking and seeking active consent from all Bob's friends, with full and exhaustive description of implications, careful curation of purpose, etc.

Both in Paragraph A are fairly close in the moral ballpark, compared to the distance of either with the option in Paragraph B.

If they were morally equivalent, they why did the FTC fine Facebook and not the millions of Bob's? Why did Cambridge Analytica go out of business if they were so innocent asking personality questions?

You can't really argue that "If X and Y were the doing same activity, X and Y would both have been prosecuted", in lieu of actually understanding what they did and describing it.

I don't think that 1) what the Obama campaign did with user data was in line with FB policy, and 1) that this data was also available to the Romney campaign are true.

My understanding (and I unfortunately cannot find the source where I read this) was that the Obama campaign did violate FB policy, but FB just decided to slap them on the hand by saying "don't do that again" and allowed other campaigns to do what the Obama campaign did. They just didn't advertise it.

Vaunted Obama campaign advisors Axelrod and Messina parlayed their success in the US into ultra-lucrative gigs in European elections. It didn't work out quite as well.

Whataboutism at its finest! #VoteTrump #TaxCuts

Thinking hypocrisy is cool if you just call it 'whataboutism' at its finest

Of course Hillary's 2016 run showed what happens when you take this too far.

Hillary lacked basic information about the rust belt that was available out there. Trump was outraising her there, both in number of donors and total money counts. She misspent a large percentage of her ad money.

If anything, 2016 taught us that today it's relatively easy to know a whole lot about individual voters and that a candidate is at a severe disadvantage without those tools. Having yet another layer on top, trying to predict which ads have a chance of swaying each voter, and how to make it less likely that opponents will show up to the vote, might be table stakes pretty soon.

Obama’s data science ops are nothing new. Acxiom Corp. is one of the nation’s largest data brokers and got its start in the late 60s as a Democratic database outfit.

"The data came from Facebook"

[citation needed]

I don't see that stated in the linked article

I am convinced by Tyler's six-word analysis proving that everyone is as bad as everyone else, so it's fine to let Trump/Thiel run things.

I'm convinced by internet rando that everything Obama did is fine because Trump is bad

Comments for this post are closed