between two alternatives. Prior to making a decision, they can receive noisy signals from two news
stations. The news stations decide how to slant their coverage in order to maximize pro t. The
paper rst examines a heterogenous biased population, then examines a homogenous unbiased pop-
ulation. In both settings it is shown that consumers seek biased news, particularly news that is
biased towards their prior beliefs.
There are several recent studies on media bias related to this paper. Mullainathan and Shleifer
(2005) examine the market for news under the assumption that consumers receive utility from
reading news that con rms their prior beliefs. They nd that when readers are homogenous, a
competitive media slants towards their beliefs. When readers are heterogenous, news rms segment
the market and slant towards extreme positions. In Baron (2006), bias arises as journalists slant
their reporting towards their preferred state. Despite competition between pro t maximizing news
sources, bias persists. Additionally, bias can be greater in a competitive news market than in a
monopoly. Both these works di er from this paper by introducing agents with an inherent prefer-
ence for bias. In this paper, agents simply want to choose the correct alternative and rms only
want to maximize pro ts, yet bias arises.
Topically, this work is most closely related to Gentzkow and Shapiro (2006). In their model,
media bias arises as rms distort their reports towards the beliefs of a biased populace in order to
form a reputation for quality. Consumers only wish to determine the truth, yet bias remains and
potentially decreases the welfare of all market participants. When consumers have heterogenous
priors, a segmented equilibrium exists in which consumers only read news biased towards their be-
liefs. This paper di ers in several important respects. First, in this model, rms seek to maximize
viewership, not reputation. If rms were concerned with the number of customers in the Gentzkow
and Shapiro model, the segmented equilibria would disappear as a biased rm would receive a zero
market share when competing with an unbiased news source. Secondly, Gentzkow and Shapiro show
how competition may mitigate bias, yet in this model bias remains no matter how competitive the
market becomes. Finally, and most importantly, by introducing dynamics this work shows that
media bias will tend to arise even when the population is homogenous and initially unbiased.
3
Convention