Manipulated News Model: Electoral Competition and Mass Media

Shintaro Miura
2015 Social Science Research Network  
This paper concerns the distortions in electoral outcomes when mass media strategically distorts the interactions between candidates and voters. We develop an election model where a voter cannot directly observe the policies proposed by two office-motivated candidates. The voter learns this information through media reports before voting takes place, while the media outlet strategically conceals some part of this information. Because incorrect decision-making by the voter is unavoidable (direct
more » ... distortion), the candidates have an incentive to influence the media outlet's behavior through policy settings that are indirectly appealing to the voter (indirect distortion). As a result, policy convergence never occurs if and only if the outlet is sufficiently biased. We then measure the degree of distortion in the equilibrium outcomes by the voter's ex ante expected utility, and characterize the least and most distorted scenarios. This characterization shows that the distortion becomes severer as the outlet becomes more biased. By decomposing total distortion into its components, we also illustrate the tension arising between the direct and the indirect distortion. Journal of Economic Literature Classification Numbers: C72, D72, D82. There are four players in our model: candidates 1 and 2, a single media outlet and a single voter. 6 The players play the following two-stage game. In the first stage, called the policy-setting stage, each candidate simultaneously proposes a policy, and only the outlet observes the proposed policies. In the second stage, called the news-reporting stage, the outlet sends a message about the proposed policies to the voter. After observing the message, the voter casts the ballot for one of the candidates. The winning candidate then implements his proposed policy. Let X ≡ [x − , x + ] ⊂ R be the set of available policies for the candidates with x − < 0 < x + . Let x i ∈ X be the policy proposed by candidate i ∈ {1, 2}, and z ≡ (x 1 , x 2 ) ∈ Z ≡ X 2 ⊂ R 2 describe a policy pair proposed by the candidates. We assume that the information regarding policy pair z is hard information. In addition, we assume that the media outlet, but not the voter, correctly observes policy pair z. Hence, the information about policy pair z is the media outlet's private information in the news-reporting stage. The message space given policy pair z is defined by M (z) ≡ {m ∈ 2 Z |z ∈ m}. That is, the available messages under policy pair z are subsets of policy pair space Z containing the truth z. 7 Let m ∈ ∪ z∈Z M (z) be a message from the outlet. Let y ∈ Y ≡ {y 1 , y 2 } be the action of the voter, where y i represents that the voter certainly casts a ballot for candidate i. We assume that there are two types of candidates: an opportunistic-type candidate and an ideological-type candidate. The opportunistic type is the standard office-motivated strategic type of candidate. Alternatively, the ideological type is a nonstrategic type of candidate that always proposes his preferred policy. We assume that if candidate 1 (resp. 2) is the ideological type, then he always proposes policy r ∈ (0, x + ) (resp. l ∈ (x − , 0)), and |r| < |l|. That is, we assume an asymmetry between the candidates. Let Θ ≡ {O, I} be the candidates' type space, and O (resp. I) represents the opportunistic (resp. ideological) type. We assume that candidate i's type θ i ∈ Θ is candidate i's private information, and θ 1 and θ 2 are independently determined. Let p ∈ (0, 1) be the probability that each candidate is the opportunistic type, and assume this is common knowledge. We define the players' preferences as follows. Define the opportunistic-type candidate i's von 6 Throughout the paper, we treat the candidates and the voter as male and the outlet as female. 7 It is worthwhile to remark that for any subset P ⊆ Z, message m = P has the property that M −1 (P ) = P where M −1 (P ) represents the set of policy pairs under which message m = P is available. That is, information about policy pair is fully certifiable in the sense of persuasion games.
doi:10.2139/ssrn.2571952 fatcat:lzqqkchtx5f6xdstkk5ijnzxri