This Article Is About The News

Understanding media bias

Imagine a straight beach. People come to the beach to swim, and since they find all parts of the beach equally enjoyable, they are evenly distributed over its face. There are two sellers of ice cream, both of whom possess identical costs and productivity. Since their products are indistinguishable, people are fine choosing the one which is closest to them. Where, then, should the ice cream sellers make their shop?

The answer is together, in the middle. Each seller increases their revenue by moving toward the middle, and the only equilibrium is that both will eventually be in the middle. With a limited flight of imagination, the beach can be thought of as being metaphorical. Two companies are producing a product which varies along one dimension, people buy whichever product is closer to their most preferred product, and the only equilibrium is for the two companies to produce the same product.

How might things change if we introduce transportation costs? Everyone receives the same amount of utility from buying ice cream, but if someone faces a sufficiently high cost of walking to the ice cream stand, that person won’t buy anything at all. In that case, the ice cream sellers do not necessarily maximize their profits by moving toward the center, since they will eventually leave some people on the edges who can’t buy. Companies choose different segments in equilibrium.

Again, we can think of this as a set of products, like a newspaper. The dimension which it varies along is “slant” – which political parties its choice of facts favors. Consumers like some facts more than others, and choose the newspapers whose facts are closest to the ones they like. If there is nothing close enough, consumers choose to buy nothing. We assume that the “transportation costs” are quadratic.

In equilibrium, newspapers produce biased news, and divide up the space of possible opinions between them. What is really striking about this is that it does not matter how many newspapers there are. Increases in competition will lead to an increase in diversity, but it will not lead to less biased news. However, an observer who simply wants the truth will be able to get it from aggregating all of the reported news, assuming that the function describing transportation costs is the same throughout the possible space for slanting to occur, and people are evenly distributed in their opinions.

The set up of a linear city, and exploring where companies will be placed is due to Hotelling; the application to the news is due to Sendhil Mullainathan and Andrei Shleifer, in “The Market for News”. I thought it was incredibly elegant. It’s one of the more elegant papers I’ve ever read.

The jury is still out, though, on whether competition increases or decreases bias. Matthew Gentzkow and Jesse Shapiro would show that slanting is possible even if everyone wants unbiased news, provided that people start with different priors, and that this makes competition reduce slanting. People will infer that newspapers which report unsurprising information relative to their own priors are better, and if these priors are rarely challenged, they’ll just accept them. More competition, then, leads to priors being challenged by the world demonstrably not lining up with them more often, and so competition ends up reducing media bias.

Gentzkow and Shapiro would later empirically test that slant is due to reader demand, and found it to be true. Their method was to first construct an index of slant from words that were used by one party or another in the 2005 Congressional Register, and give each newspaper in America a rating. (It is important that their rating have meaningful magnitudes to it – simply giving each newspaper some guesswork ranking would bias the slope of regressions down due to regression dilution). They then compare the differences in circulation between very Republican, and very Democratic, precincts within a city, and estimate how much readers demand slant. With those demand coefficient in hand, they then look at each city and ask, “How much slant would they choose if they were profit-maximizing?”. Since this largely matches up fairly closely with the observed level of slant, we can infer that slant is driven by demand. Owners do not seem to be very ideological, as two newspaper with the same owner, in two different cities, are no more ideologically related to each other than two random newspapers. That is not to say that they might never, on some issues, slant toward their self-interest. (Gilens and Hertzman (2000), which Gentzkow and Shapiro accidentally cite as being from 2009, showed that newspapers were more likely to report favorably on the 1996 Telecommunications Act if they stood to gain from it). Nevertheless, supply is not the primary source of bias.

To me, this means that there are no easy answers to misinformation. We cannot expect either increasing competition, or even trying to produce more truthful content, to change people’s opinions closer to the truth. What we can hope for, however, is for very big shocks to break through to people. Most media reports on things which are not actually happening to you, and won’t ever actually happening to you. People doing deviant things you dislike can always be found, and if it is used to oppress immigrants and poor people in Africa, you personally do not get hit by it. If the madness which the administration indulges in is massive tariff increases, however, you will see this, no matter what the media you consume focuses on. It is your priors being challenged.

I hope that the present administration’s actions are like pyrotherapy. Before the invention of antibiotics, syphilis was an untreatable, terminal disease. All that we had was trying to cook the bacteria, without cooking you. What doctors would do is give terminal syphilis patients malaria, so that the 105 degree fevers would literally cook the disease. Believe it or not, this worked! Obviously you were still likely to die, but it was far better than the inevitable madness, paralysis, and death. Perhaps the cure for illiberal autocracy is its consequences.