MEDIA POWER by Andrea Prat Discussion by Tim Bresnahan An Intriguing Point about voters who see multiple media • Reach • Π(K=1) ≡ ≡ Pr(A voter sees this medium) f(Pr(A voter saw this medium most recently)) • Key idea: a consumer who sees multiple media is less influenced by any one of them • In the case of K=1 (only recalling one message (wlog most recent message) this comes down to “attention share” Three Elements of an Index … Audience Statistics Voter Response Election Outcomes Three Elements of an Index … Audience Statistics Voter Response Election Outcomes Reach MP Index (Π) Media Usage Data Robust w/in Class or Calibrated Quality of Candidate Majority of all voters … mean more assumptions. Voter Response Assumptions • Bayesian voters. • Potentially limited bandwidth. • Potentially “sophisticated” about biased media. • Defining the class within which approach is “bounded,” “robust” • Obvious first class to consider. Alternatives (from advertising and PE) Audience Statistics Media Used Frequency of Use Other intensity Segments Demographic Psychographic Ideological Prior Moralizing Geographic Swing States Media Markets Ads by Candidates Voter Response Bayesian limited bandwidth. sophistication about bias Other rational inattention Emotive Moralizing Competitive Interference Vote / Prefer / Know of Election Outcomes Quality of Candidate Issues (heterog. qual. assessment) Coalition (mult. issues) Majority of all voters Swing States Turnout: Mobilize Base Who is Pivotal? Picking from this menu (1) If voters assess issues, not “quality” • Coalition formation and diverse interests • e.g. Budget conservatism and social conservatism • Heterogeneous responses to specific messages • Are there additional constraints on media? • What to tell centrist voters if their responses are different sign from base? • “MP” can’t capture this kind of heterogeneity • All voter reactions are the same direction • Strength potentially muted by ideology (prior), “sophistication” or high bandwidth Picking from this Menu (2) If Moralizing/Ideological appeals are like emotive appeals in product advertising • Let total exposures =r*f= (reach*frequency) • Frequency of exposure, in some contexts, works to create emotional bond so that ∂response(r,f)/∂r < ∂response(r,f)/∂r * (f/r) • If political messages work like this, it weights up intensive use of one medium by voters • Or of media transmitting similar messages (Fox, WSJ is big in paper) • Application to positive political messages? • Application to negative ones? Picking from this menu (3) Advertising by candidates or 2 evil media moguls Picking from this menu (4) Getting out the vote in the base vs. moving swing voters • Opposite signs in power index for stronger media that segment by hitting base vs stronger media that segment by hitting swing voters • Robustness is, ahem, challenging Some Breathless Empirical Finding Reports Highest-Ranked Newspaper is ~10th • NYT is also ~10th in reach The Top Three Media Conglomerates could swing an election • Fox & WSJ + CNN + Comedy Central + MSNBC + NBC • Translation??? “The Top Two Parties could swing an election” Finished … the alternatives Audience Statistics Media Used Frequency of Use Other intensity Segments Demographic Psychographic Ideological Prior Moralizing Geographic Swing States Media Markets Ads by Candidates Voter Response Bayesian limited bandwidth. sophistication about bias Other rational inattention Emotive Moralizing Competitive Interference Vote / Prefer / Know of Election Outcomes Quality of Candidate Issues (heterog. qual. assessment) Coalition (mult. issues) Majority of all voters Swing States Turnout: Mobilize Base Who is Pivotal? Fin Old slides follow If it were advertising -Metrics • Reach • Frequency • Impressions • Segments • Various targeted versions (“Reach into swing voters….” “… into swing states”) Impacts • On sales (voter turnout), on knowledge of product (candidate), on perception of product (candidate) on intentions (to vote, buy) • Impacts of reach, frequency the same or different? • Competitive Interference • Etc. An Index of media power • Move from media markets to individual voter information sources (in the simplest case, Attention Shares) • Voter rational inattention, potentially limited bandwidth, potentially “sophisticated” about biased media • “worst case” biased media • An improvement on reach • reach: pr(voter exposed to this medium) • MP(K=1): pr(voter’s last exposure was to this medium) • Has all the strengths and weaknesses of reach measures Mergers in media distribution industries • Can’t use market power indexes • Can’t use media power indexes Media Power index is high • Media is pretty unconcentrated: might be no problem? • BUT, this paper points out, many voters use few media (most frequently TV networks) • And takes the view of media => voters of that Robustness • Rationally inattentive voters deciding on candidate quality • Various robustnesses within that • Emotive Appeals • Voter preferences over issues Some Positive Political Economy Things • Swing States • Too many messages from one source or too many uncontested messages from one source? • Getting out the base • Creating coalitions (those two definitions of conservative?)