Skip to content

Social Media, Advertising, 2016 election – Part 3

Looks like FB and the impact of social media advertising in the 2016 election is making headlines. In this post, I’ll try to lay out a possible next logical stepping stone, along one possible line of reasoning.

I’ll temporarily assume that the self-reinforcing dynamics of FB advertising were a top factor in determining the outcome of the election, rather than a “straw that broke the camel’s back”. In reality, I think the camel was pre-loaded with a ton of other factors. Most prominently economic inequality, but also very significant disillusionment with both the Republican and Democrat parties, as exemplified by the Bush and Clinton names. If you think in terms of advertising, that’s a major disadvantage in branding.

But I do think it is entirely appropriate to focus specifically on the workings of internet-age advertising, so lets do that. Now that FB has to defend itself a little bit, I think we will find the following 2 things:

1. The toolkit of social media advers/manipulation was equally available to both sides in the election.

1a. It wouldn’t surprise me if it was used with equal or greater effort by Clinton, simply due to superior Democratic fundraising in the 2016 election.

2. The interplay between Democrats’ conflicting social media advertising objectives during the primaries, and during the general election.

2a. During the primaries, Clinton managed to convince Democratic voters that Clinton would be best able to appeal to Independents and Swing-State voters. (i.e., “electability”). This was in spite of polls showing that her general-election appeal was at best equal, and most often, lower compared to Sanders – against all potential Republican opponents.

It wouldn’t surprise me that there was a very successful advert / media campaign specifically to buttress the faith of Clinton’s general electability among Democratic primary voters. The necessary side effect of this campaign would be to de-emphasize any attention paid to the desires of disillusioned voters in swing states. This applies especially to superdelegates – experienced in US politics – who should’ve known better.

2b. The “pied piper candidate” angle. Unlike the pundits on TV, actual campaign managers do pay attention to the polls, from early on. And they have the benefit of all this social media analytic data we are now talking about.

Democratic campaign managers were surely disturbed by the polls. According to the numbers during the 2016 primaries, Clinton had only the slimmest little lead over the “better” Republican candidates. In fact there was only one likely opponent she was consistently beating by more than the polling error – Trump.

I believe that there was a deliberate media campaign, by Team Clinton, to encourage the Republican party to nominate the opponent most favorable to Clinton, namely Trump. This was entirely conventional wisdom of course. Nearly everyone, including me, thought that Trump’s nomination was a colossal blunder for the R party. And to be fair, on the flip side, it’s equally plausible that the R. party would support media campaigns within the Democratic primary attacking Clinton.

But now lets mix in what we’ve learned recently about social media advertising and how it amplifies the snowballing effect of group behavior in popularity contests / branding etc.

Consider the possibility that Team Clinton’s efforts to line up a more favorable general election opponent involved social media?

Recall that FB advertisers receive an increasing discount to reward them for adverts that are actually popular. This mechanically amplifies the self-reinforcing group popularity effect that is already there just from human nature.

So during the primaries, social media on the Republican side builds a mass of people clicking on Trump stuff, liking it, bringing down the price of Trump ads. Democrats spending social media ad money prefer Trump to Cruz, so they fuel this.

By the time of the convention, you get a mass of people whose facebook accounts are now trained to show them Trump content.

Now, for the General election, if you’re a Democratic campaign manager, you want this all this to flip, but it’s too late. The advertising pricing system senses that there is a mass of accounts programmed to be in Trump mode, and so Trump ads get a discount. The Democrats at this point stop paying for Trump ads, but the Republicans unify after the nomination and step up their funding, so the commercial side of facebook is still pumping the content.

The advertising machinery shows most of the pro-Clinton material to people who already like Clinton, it shows most of the pro-Trump material to people who already like Trump. Otherwise, these groups of FB users would have an unpleasant experience viewing unfriendly ads and might stop playing with their phones. That would be bad for business as far as FB is concerned. In any case, those two groups have their minds made up.

What about independents? Well guess what – the advertising machinery already sorted them out, during the primaries! Back when both convential and social media was encouraging Trump supporters to gel as a group.

The mechanical amplification of the popularity-snowball-effect we’ve been going over recently, gave the group of accounts associated with Trump content a head start. I suspect Democratic spending during the primaries contributed to this. Millions of accounts pre-programmed, and the whole thing reaching self-reinforcing critical mass by the time of the 2016 election nominations. Big Oops. They are never going to admit this happened. (unless they squeeze FB’s Mark Zuckerberg too hard, but I seriously doubt it).

As for Cambridge Analytica? A little more about their parent company, here. Turns out they’re very much a part of the UK establishment, on the conservative side.


( via  <– via NakedCapitalism )


Facebook, Cambridge Analytica, and Trump 2016 – The Story of Chris Wylie [Guardian]

Consider the article below a follow-up to the previous post, on FB advertising meets politics [this blog Feb 27th].

Main link (read first):

The subtext here is that the outcome of the 2016 election can be attributed to potent new advertising techniques (aka propaganda/PR, aka social media campaigns, aka psy-ops – call it what you like!). Such a belief continues the direction set by the FBI’s extraterritorial indictment of the St. Petersburg internet advertising / propaganda / social media posting firm.

This time the focus is on mainly domestic advertisers and political figures, with additional involvement by researchers from Canada, UK, and it seems both US and Russian national security organizations who were in love with the possibility of using commercial advertising tech to do their hard work for them.

Central to all this is Facebook – who built one of the premier businesses of our time – by creating “profiles” of users and selling bundled groups of “eyeballs” with similar profiles to advertisers. Clients of course including political campaigns of all stripes. Again, see the Wired article linked in the Feb 27th post on FB/advertising/politics.

In the case of Chris Wylie in the current piece, it seems Cambridge Analytica used what was supposed to be an academic dataset, thereby violating disclosure rules. The data correlated self-tested Big-5 personality factors (psych 101 stuff), with other features not described in the article – I’m guesssing routine A/B testing of ad materials, and lining that up against both the academic data set in question, and the the wealth of “standard” profile information FB would sell them. This concept is right out of the 20th century – the only remotely novel thing that I can tell is the application of “big data”, and that too is pretty much a stock-in-trade, on the 5-year time scales typical of the tech world (i.e., both sides in the election had it). Once again, read the previous Wired article.

Bottom line:

1. It seems a bit exaggerated to imply that there was some kind of magic bullet formula here to make an advertising campaign of unprecedented potency. Certainly the scale of this operation is impressive, but for the middle-american audience, the fundamentals favored Trump (relatively speaking). For the simple reason that his last name is neither Bush nor Clinton. This little bit is consistently missing from the story. Still.

2. This is what FB (and Google) has been doing with all the you data you give it. All that stuff they give you “for free” …..

3. It’s cool when we benefit from these tools and tactics, but when they are turned against you? Double edged swords … cut both ways. When will we ever learn?

Italy votes

UPDATE – Looks like Berlusconi’s Forza Italia actually got fewer votes than their northern regional partner, the right-wing Lega. This could take the formation of a coalition government in a different and more troubling direction.

UPDATE2 – added 2 more links for deeper context

Italy had their national election. (On a Sunday, take note. Not Tuesday.) Turnout was just over 70%, which is a postwar low for Italy.

Preliminary results show none of the 3 major blocs high enough to form a government. More telling are the choices themselves:

First, Berlusconi’s center-right bloc, consisting of his personal Forza Itaila party, joined in coalition by Northern and Southern regional right-wing parties. I always said he was the prototype for Trump. In a lot of ways, he’s way ahead of our own fearless leader. His “liberal-conservative” program: A sweeping package of tax cuts. (sound familiar?) Deportation of migrants. Increase minimum pensions. Oh and he’s currently banned from holding office (til 2019) due to prior conviction. His right wing coalition partners add support for the Euroskeptic package, and explicitly nativist flavor.

Next, the populist M5S (Movimento 5 Stelle). Highest single party result, around 30% per early polls. General theme: mass discontent. Program: Universal Basic Income Minimum Income.  Deficit Spending. Repeal liberal labor market reforms. Euroskeptic – demand continental authorities exempt Italy from austerity treaty obligations, or else quit the EU. “Anti-Corruption”. Raise taxes on finance and energy business (most doubtful). In the past, claimed as a matter of principle that they will not form coalitions with other parties.

Last of the major groups, the PD (Partito Democratico). Typical centrist-liberal fare. Major selling point is sanity. Just to show where the country’s politics are, the program also includes raising minimum pension, minimum wage, subsidies for families with young kids. Willing to “negotiate” with continental authorities on austerity imposed due to treaty obligations vs Italy’s high debt.

Just for background — Italy’s cumulative GDP growth since the financial crisis is about 15% behind the EU overall. Realistic next steps: Berlusconi negotiates for maximum possible concessions / personal favors, threatens to form coalition with M5S, but ends up in coalition with PD when the payoff is high enough.

see also: Yanis Varoufakis interview (blog) on the outcome of the Italian Election.

see also: 18 minute audio interview (youtube) with Mark Blyth and David Kertzer

US Politics: TCJA, roads not taken

Since it’s getting close to tax time, I stumbled across something that dropped entirely off my radar – the 2017 Tax Cuts and Jobs Act (TCJA). Aka the Trump tax cut. Classic trickle-down economics. (It doesn’t actually affect me until the NEXT tax filing).

The details are many, but it’s basically front-loaded, and most of its benefits go to upper income individuals and businesses. Over the next decade, it represents a transfer of close to a trillion dollars from the lower-middle and middle class to the upper class.

A typical/median-income level US household will see something like a $1000 reduction in 2018. However, the formulas for future adjustments, and future expirations of parts of the law, result in the whole package tilting further in favor of upper incomes and big businesses, in future years.

The TCJA barely passed (51-48 in the Senate), along party lines – give all Democrats credit for voting against it. You would think this would be a valuable talking point. However, since its passage, silence.

Perhaps as a result, public perceptions are drifting and the TCJA appears to be gaining in popularity, breaking even in approve/disapprove as of mid February.




None of this applies until the 2018 tax year, so what you’re going to file in April is unchanged (with a handful of obscure exceptions).

Rates — Under the TCJA, tax rates themselves for most people’s brackets go down by 3% of adjusted income (or 4% for “upper-middle” incomes).

Personal exemptions eliminated, rolled into Standard Deduction and Child/Dependent Credits – A bunch of shuffling replaces personal and dependent exemptions, with a combination of raised standard deduction and additions to child and dependent tax credits. Resulting net change for most typical individuals and families is something like a 1% tax increase vs top-line income, partially counteracting that 3-4% above.

Brackets – The formula for inflation-adjustment of the brackets is changed to a weaker flavor of the CPI. This will make the brackets rise slower (a tax increase over time).

Small Businesses and “Self-Employed” – This is a major item for me. Individual income from pass thru’s (e.g., LLP’s, self-employment) is now 20% deductible (up to $315k/$157k). For a self-employed middle-income earner, this means an additional tax cut of 5-6% (as a % of top-line income). This represents relief for the ever-expanding “gig economy”. For professionals paid via 1099’s, this is a fantastic bonus – an extra half month’s pay in after-tax $ for typical income levels.

Big Businesses  – Another big item. Corporate tax rate drops from 34-35% to 21%.  Businesses with earnings parked in foreign jurisdictions can repatriate them for 8%/15%. Corporate AMT eliminated. Business Interest deductibility now limited to 30% of EBITDA.

Mortgage Interest Deduction – Upper limit on mortgage interest deduction is reduced from $1MM to $0.75MM (max loan balance on which interest is deductible). Second mortgage interest no longer deductible. May have effects.

State and Local Tax Deduction – Deductibility of state/local taxes (incl. property taxes!) capped at $10k. Hits high-tax states like NY/CA hard enough that Republicans in these states voted against the TCJA.

Obamacare  – “individual mandate” penalties eliminated starting 2019.

Estate Tax – Exemption from Estate/kid-gift tax doubled from $5.6MM to $11.2MM. Even fewer people will now be paying any such tax.

AMT – Exemption from Alternative Minimum Tax raised to $109k/$70k, from $84k/$54k. Fewer people will be paying AMT now.

What does this add up to?

  1. There are at least some goodies for most everyone. However, the bulk of the benefits goes to big businesses, small businesses, higher income individuals / households, and self-employed individuals / households, in roughly that order. (not sure about the middle two).
  2. Impact on US tax revenue? CBO estimate: additional budget deficit of $1.45 tril over 10 years.

follow-up [nonsense / stream of consciousness]

follow-up to the previous post about on how the FB advertising targeting/pricing model amplifies…. 

The neat side effect of this ad-targeting/pricing system is that you can train it quickly, for yourself at least.

I had the luck/misfortune of being in a long-distance relationship recently. When the gf would spend the weekend at my place, I would see her ads mixed in with mine for a few days after.

It was easy to tell they were her ads – models who looked very very approximately like her wearing her clothes, basically. My first reaction was actually- “Wow! the internet really does have me figured out!”.

I think the reason the models looked like her is because she already “trained” the ads to show her models that look like her. As in, she would think “that looks like it might look good on me” — Click! Or maybe her online purchases were sufficient for the algorithm to home in on that information.

Anyway, I have an ad-blocker on my “work” laptop, but not on my phone or beater/sketchy-website/blogging laptop. And becuase of this political blog, I end up in parts of the internet inhabited by a different demographic – older, occasionally right-wing since they appropriated some libertarian and progressive issues like anti-war. Therefore I get quite a few ads I personally don’t want beaten into my subconscious mind – nor maybe shown to guests using internet at my house. One guest in particular.

Solution? Click on her clothing ads. Pretty women, nice clothes. You don’t like clothes? No problem, they sell underwear too. The ads know her exact body type…. [aside: Is it possible to determine the shape of someone’s butt from the undies they buy? Did general-purpose advertising algo’s manage this entirely on their own? Subject of future study, I suppose.]

TMI. Lets get back to ads, and let’s take it to the next level – enter into a very un-creative corner of the dream-world. More of a wandering mind than a dream… there’s plenty to entertain me in the shallow end. Ok:

It would be cool to make some “ad-training” software. Clicks on an ad, opens up a window in the background for a few seconds so you don’t have to see it. Maybe opens a shopping cart and puts something in it, they seem to like that. Save you the trouble of doing all that manually. You can set the software up to train for whatever … maybe you want to get all purple and yellow flower prints in your ad boxes from now on. Yes, I know you can just get an ad-blocker, but that’s just an ugly gray box. This is better.

Next level:

When running a political campaign in the 2020 primaries, instruct your followers to click on the ads, to train the ad-targeting/pricing model. Or make it really easy – distribute the training software to your followers. Cel phone app. [not serious! joke! joke! although you know this is coming…]

Next level:

I want the wife or girlfriend (or husband/bf!) to start wearing purple-and-yellow flower prints… or more[less] revealing clothing… or whatever… Solution? Ad-training software. [JOKE JOKE JOKE JOKE NOT SERIOUS DONT KILL ME!!!!]

Seriously, do online ads and their targeting/pricing model really function as a sortof (degenerate) nervous system for a larger organism? To what degree do they facilitate group thought and behavior? Paranooooiiiiaaaaayayayaya. It’s so nice and sunny outside, time to replace the morning rage read with some warm-ups and who knows…

How FB advertising works, and how it applies to political campaigns [A.G. Martinez / Wired]

Article with the sensational title “How Trump Conquered Facebook – Without Russian Ads”, really about how the advertisement-targeting machinery works in. Easy read, and important. Author is a FB manager or exec.


[via NC]

In the specific case of the 2016 election, my take-home’s are:

  1. FB and Google knows everything you’ve ever done on the internet and with your credit card and have the two neatly cross referenced (but you already knew that)
  2. The FB pricing model rewards advertisers for engaging ads, by giving them a discount. Swing-state social media users found Trump ads more engaging. This in turn gave a cost advantage to further social media ad spending by Trump, for those specific populations. Presumably, there is a psychological effect where once the audience is trained, their engagement with further ads is reflexive. Self-reinforcing loop.
  3. Team Clinton had a similar self-reinforcing loop of ad-spending-effectiveness for coastal-urban audiences, which apparently her advertising team chased. She destroyed Trump in CA, for example. Unfortunately, not a swing state.
  4. Even besides the dynamics of the automagic ad targeting/pricing model, swing state audiences cost less to reach than coastal-urban, simply because they’re less valuable eyeballs in the overall ad market. However this doesn’t favor any candidate in particular.

Now since I’ve weaned myself off FB some years ago, I guess none of this affected me. But just a thought — if they have all this super-powerful data about how engaging particular candidates are, mapped out by population and geography, could the DNC not have made a more intelligent choice in the 2016 primaries? Hmm… I bet they had it all.



Richard Falk on Trumpism, Liberalism, and restoring a brighter, rather than less dismal political vision.

4800 word essay subtitled The Psycho-Politics of Geopolitical Depression.


Richard Falk is a professor emeritus at Princeton and was Special Rapporteur for the UNHRC.