Study: Media “bias” based on pre-existing expectations

From The Economist:

Bias can be thought of as a supply-side phenomenon that arises from ideology. Owners’ or employees’ political views will determine how a newspaper or channel slants its coverage of a piece of news. But this does not square with the assumption that readers and viewers value accuracy. If so, then competition should hurt media outlets that systematically distort the news (in any direction). The brouhaha about bias in America, as free a media market as any, suggests something else is going on.

The key to understanding why bias flourishes in a competitive market may lie in thinking more clearly about what readers actually want. Sendhil Mullainathan and Andrei Shleifer, two Harvard economists, argued in an influential paper* that it may be naive to think that people care about accuracy alone. Instead, they modelled the consequences of assuming that newspaper readers also like to have their beliefs confirmed by what they read. As long as readers have different beliefs, the Mullainathan-Shleifer model suggests that competition, far from driving biased reporting out of the market, would encourage newspapers to cater to the biases of different segments of the reading public. A more recent paper** by Matthew Gentzkow and Jesse Shapiro, two economists at the University of Chicago’s business school, set out to test this proposition.

To do so, they first needed a way to measure the political slant of American news coverage. Their solution was rather imaginative. The researchers ran computer programs that analysed debates in Congress and identified phrases that were disproportionately used by Republicans or Democrats. The list of frequent Democratic phrases, for example, included “estate tax�. While talking about the same issue, Republicans tended to use the phrase “death tax�. (This is not just coincidence. Mr Gentzkow and Mr Shapiro quote an anonymous Republican staffer as saying that the party machine trained members to say “death tax�, because “‘estate tax’ sounds like it hits only the wealthy but ‘death tax’ sounds like it hits everyone�.) Having identified partisan phrases, the academics then analysed the news coverage of more than 400 American newspapers to see how often they cropped up in reporting. This gave them a precise measure of “slant�, showing the extent to which the news coverage in these papers tended to use politically charged phrases.

Mr Gentzkow and Mr Shapiro then needed to assess the political beliefs of different newspapers’ readerships, which they did using data on the share of votes in each newspaper’s market that went to President Bush in the 2004 presidential elections, and information on how likely people in different parts of that market were to contribute to entities allied to either Democrats or Republicans. The researchers were now able to look at the relationships between circulation, slant, and people’s political views.

First, they measured whether a newspaper’s circulation responded to the match between its slant and its readers’ views. Not surprisingly, they found that more “Republican� newspapers had relatively higher circulations in more “Republican� zip codes. But their calculations of the degree to which circulation responded to political beliefs also allowed them to do something more interesting: to calculate what degree of slant would be most profitable for each newspaper in their sample to adopt, given the political make-up of the market it covered. They compared this profit-maximising slant to their measure of the actual slant of each newspaper’s coverage.

They found a striking congruence between the two. Newspapers tended, on average, to locate themselves neither to the right nor to the left of the level of slant that Mr Gentzkow and Mr Shapiro reckon would maximise their profits. And for good commercial reasons: their model showed that even a minor deviation from this “ideal� level of slant would hurt profits through a sizeable loss of circulation.

This matches my experience, which, I guess, means that it plays to my pre-existing expectations.

How I vote

Tomorrow’s Election Day. You may not have heard.

Living in the city I cover is normally a good thing: It spurs me on to work harder, exposes me to stories that might not otherwise appear on my radar and makes me more accountable to my readers, whom I can run into at the store, in restaurants or even when picking up a prescription at the pharmacy.

But it makes Election Day a little strange. As a result of my job I am, I think, one of the more educated voters around, but I also have had personal dealings with nearly all the candidates, and while I keep my personal feelings off the page — in 2004, representatives of two rival city council candidates both tried to have me fired, since I was clearly in the tank for the other guy — it confuses things when going into the polling booth.

I should vote for the best candidate, certainly, but what if he’s a complete jerk? Should that make a difference? I will be dealing with him (or her) for four or more years, after all. How about how good of a quote someone is? Or how time-consuming it is to interview them?

In the end, though, it almost doesn’t matter: Every year, Jenn and I go to lunch — this year, we’re going to go to breakfast, to try and beat at least some of the crowds at the polling station — and go through the sample ballot, and she quizzes me about each of the candidates, going beyond what I’ve written about them previously, to the sort of stuff I mention above.

And then she goes and cancels out most of my votes anyway.