How Much Media Bias Is To Make You (The Reader) Happy?

Every academic article I have ever come across on media bias has concluded two points - 1) Media bias is present and 2) It is consumer driven. What #2 implies is that the dominant segments of readers prefer to read something that confirms their pre-existing world view. What is not always clear if this means that there is actual bias by the newspapers. Markets would flush out newspapers that weren't produced in a way that was pleasing to its readers, and the paper that emerges may appear to have have a political bias when it may simply be coincidence. However, new from the NBER:

We analyze the coverage of U.S. political scandals by U.S. newspapers during the past decade. Using automatic keyword-based searches we collected data on 35 scandals and approximately 200 newspapers. We find that Democratic-leaning newspapers -- i.e., those with a higher propensity to endorse Democratic candidates in elections -- give relatively more coverage to scandals involving Republican politicians than scandals involving Democratic politicians, while Republican-leaning newspapers tend to do the opposite. This is true even when controlling for the average partisan leanings of readers. In contrast, newspapers appear to cater to the partisan tastes of readers only for local scandals.

So, this paper shows that even after controlling for reader partisanship, that you have an newspaper bias delivering biased coverage. In fact, the paper presents some evidence that runs against claim #1. For those interested in media bias, the paper looks to be well worth the read.

However, I wonder if this is a result of a prisoner's dilemma for the newspapers with an ideological bent. If you have >1 local paper covering the scandal that shares your point of view, perhaps coverage of the scandal is a competitive point and newspapers go overboard to ensure they have "the most complete coverage." My expectation would be that this would swamp out the regression's ability to pick up the consumer demand result.

Secondly, I am not familiar with the methodology used by the authors (it appears to be some variant of an IV/2SLS approach), and I'm not sure how it separates the reverse causality of the scandals causing the newspapers to be endorsing the opposing party. Since the editorial endorsements are used as a proxy for the the partisanship propensity of the newspaper, the number of scandals would seem to be a causal factor rather than an exogenous one.

P.S. The literature review is well worth opening the pdf on it's own.