Data and Economics

The Loop

Matthew Gentzkow and Jesse Shapiro of the Bureau of Economic Research published a working paper a few years back on political media bias in American print newspapers. It was the kind of study that told you something you already knew but couldn’t prove before the dawn of the digital age. Gentzkow and Shapiro took all the digitized text of 400 national newspapers, 70% of all national circulation, and conducted a data analysis comparing the text to language commonly used by Representatives on the floor of Congress. The intent was to measure the degree of print media slant.

The paper uses the term “estate tax” as an example. A Democratic congressman usually uses that term to describe the government’s take on inheritance transferred to heirs. A Republican calls it a “death tax.” A quick comparison of that single term tells you a bunch. The Washington Times used Death Tax about 12 times more frequently than the more liberal leaning Washington Post did.

Gentzkow and Shapiro found that “average news content resembles a left-of-center congressperson.”  Which wasn’t that surprising. The motivation of that slant wasn’t particularly surprising either. What was surprising though, was just how unsurprising it was.

The motivation for the slant was what any student of business or the media would conclude. Money. Not deep state dirty conspiracy money, the way some of the scary folks taking to the air waves these days would have you believe. Just good old fashion profit. Business outcomes…capitalism.

Ironically, it turns out that privatizing media makes for liberal news.

On the whole, data shows that on a national level, the average print media consumer is leftward leaning. Newspapers are published and read more commonly in cities. People who live cities, on average, tend to be more progressive. People who read their news tend to be higher educated. Higher educated people, on average, tend to be more progressive. So, on the whole, the average consumer of print media is more progressive. And so the average newspaper sounds like a “left of center” congressman.

Locally the trend holds up. In conservative areas, the consumers were more conservative and the print media was therefore more conservative.

The truly revealing finding though, was what the data said about deviating from the core belief systems of readers and the business impact related to it. It turns out that the economically optimized slant, the spin that gets you the most readers, just happened to match the exact slant of any given paper. The data also showed that if the slant changed one standard deviation (statistical speak for “just a little bit”), the paper would reduce circulation 3%. Which doesn’t sound like much, but in the business world, losing 3% of business outright, gets people fired.

The model is tried and true. When it comes to news, give the people what they want. Anything else loses business.

No other characteristic of the newspaper determined its level of bias significantly. Not ownership, not changes in elected officials, not competition or pressure from incumbent officials. None of it. As for the deep state conspiracy? Nope. Not that either. Unless of course Gentzkow and Shapiro are a part of it. (Don’t laugh. Someone will comment with that accusation.)

The majority of Americans don’t get their news through print media though. According to a recent Pew Research poll, most watch it on television or their mobile devices. Only about one-in-five Americans regularly read the newspaper today. Increasingly though, that line is being blurred as print newspapers transition their content online.  But the message is still the same. Television news needs ratings to sells ads and cable subscriptions. Online media outlets need clicks and shares to sell their ads. The eyeballs go to the content that agrees with them. And since we live in a world where the news isn’t delivered to our doorstep any more and we can simply change the channel or click a link to find what we want, the cycle of bias following demand perpetuates itself.

The media isn’t driving us to conflict. We drive the media. The media just charges a fee for their service.

So why are we so much more polarized?

If not the media, then surely it’s social media and the increase in availability of echo chambers of information. The constant availability of confirmation bias everywhere we go, in our pockets, has to have an impact.

Right?

Well, it turns out that Gentzkow and Shapiro wrote another working paper a few years after their initial study on print sentiment. It was focused specifically on the scope and characteristics of idealogical echo chambers. Using similar classification of data, they broke down the content found in different mediums and developed an index. They call it the isolation index. The higher it is for a given medium, the lower the likelihood that one will find anyone whose ideology differs from theirs.

Here’s what the data said:

Broadcast television news was the least ideologically isolated medium. Cable television news was next least. Followed by magazines then local newspapers then online news feeds and then national newspapers. The internet and social media, which is often the focus of blame for polarization, does make us more ideologically isolated. About 20% more than if it were completely eliminated.

When the president won’t attend the Kennedy Center awards because he can’t be in the same room with the arts and entertainment industry, it feels like we’re worse than 20% more polarized than we were a generation ago.  There’s no fast way to get to that conclusion by data. But it feels worse. Much.

There’s another interesting impact of social media too. When you look at the isolation index for social media relative to other conventional networks, the internet boogie man starts to get a little less scary. The data tells us that the ideological segregation in any given Facebook feed is slightly higher than the ideological isolation of a given county. Meaning that a random person you run into on your Facebook feed is slightly more likely to share your ideology than a stranger in your county. Which makes sense. Because you choose your Facebook feed and counties are big places in which most of the population was born.

The next part is a little more surprising.

The data tells us, with statistical significance (statistics speak for you can bet your ass on it), that your social media sphere is actually less of an echo chamber than your zip code. It’s way less of an echo chamber than the voluntary groups you join. It’s even less of an echo chamber than where you work. It’s a fraction of the echo chamber your family is. And its a veritable cornucopia of diversity compared to the echo chamber that is your associations with trusted friends and political discussants.

The conclusion you’re forced to draw from the data is that social media isn’t creating echo chambers. It’s brining people together that twenty years ago would never be together.

And the news isn’t conspiring to slant us. We’re slanting it.

So what the hell is happening?

How is Donald Trump president? Why are we excited this weekend because there was minimal violence between white supremacists and Boston Strong (go Sox)? Why does it feel like the country is about to rip apart? If it’s not the media, social and otherwise what’s making it feel that way?

The answer’s pretty simple. It’s us.

This is all on us. And I don’t mean that in the motivational, touchy feely us. I mean literally us.

There’s another paper recently published by Robert Warren Anderson, Noel Johnson, and Mark Koyama of George Mason university that collected data on crop yields and violence against Jews in Europe for the seven hundred year period ending with the turn of the 19th century. What they found was dishearteningly familiar. Things get hard, we take it out on the groups classified as “not us”.  The data showed that if temperatures in a year dropped a third of a degree, an amount that would measurably impact crop yields, violence against Jews increased 50%. The data correlation is dreadfully significant.

This isn’t new. This is us.

History doesn’t repeat itself. We’re neither doomed nor saved based on what we remember or forget. History is read in books and learned in school. Mankind is physics. We’re perpetual. We are motion and direction and force. The more I look at the problem, the more I see a loop; a loop that expands with the relative abundance we gather along the way. But a loop nonetheless.

This is what economic downturn looks like in a world connected by technology. We’ve had wage and productivity stagnation for four decades now. Our version of modern crop yields are bad. And there’s different people everywhere; my TV, my newspaper, my phone…and I can reach right out and hate them with a click.

With that in mind, here’s my ask.

Some of us are really prone to xenophobic behavior. Others need convincing. Others, make it a defining point never to be. And others, don’t care. We’re never going to change the people who live their normal state hating people who are different. And we don’t need to urge on others who naturally come to the defense of those at risk on their own. But there’s a whole bunch of us who have the luxury of not giving a rip because when the crops fail, no one is coming for us.

Here’s my ask.

Start giving a rip. Even if you don’t have to. Because if we don’t, this loop looks like the ones in the past. And those were ugly.

We’re starting to wander into territory where silence is complicit. There’s a million ways to speak out today and let the angry xenophobic minority know that’s just what they are. So let’s get to work.

5 replies »

  1. Fascinating. I found especially intriguing the correlation with [crop yields and variations in anti-Semitism] and [present day economics and “us”]. Hari Selden is somewhere, nodding in agreement.

    Like

  2. Yes. Not caring is a luxury. I wish I could not care sometimes. I really do. Caring takes a lot of energy.

    Like

  3. Very interesting findings & viewpoint. Emotional virtue-signalling at the end was discordant from the main body, tho’. I’m not a fan of “rah-rah”ism.

    Like