Not yet settled. See here for instance, where I validate a Lexicoder-based tone measure using vector autoregression for analyses meant to detect partisan media bias in economic news dynamics. I find the opposite. 1/
journals.sagepub.com/doi/abs/10.117…
journals.sagepub.com/doi/abs/10.117…
Ditto here in a working paper with @alan_jacobs1 @jsmatthews99 and @HicksTM. Again the aim was to validate the tone measure, but this time for analyses on class bias in news media responsiveness. News media lead, rather than follow sentiment 2/ osf.io/preprints/soca…
We find a strong correlation between news tone and the fortunes of the affluent with much less responsiveness to those of the middle and lower class. This class bias is likely due to how traditional indicators of econ performance now better reflect the fortunes of the wealthy 3/
If the story was primarily that people learn about the state of the economy from their own experiences, and the news media follow along, we would not expect this pattern, nor the repeated failure of voters to punish incumbents for growing inequality at the ballot box 4/
And most convincingly of all, see this forthcoming @AJPS_Editor article by Marcel Garz and @gregmartinphd that identifies the effect of round number milestones in economic indicators on news coverage volume and incumbent electoral performance 5/
web.stanford.edu/~gjmartin/pape…
web.stanford.edu/~gjmartin/pape…
News media matter. That being said, both the media and the public are influenced by real world conditions, so they aren't all powerful. Causality may well run in both directions to some degree and the nature of this relationship may vary across contexts. More work is needed 6/
