Central banks lean on communication to anchor inflation expectations. But most households learn about the economy from the news rather than from policy statements, and in polarised societies, different outlets cover the same inflation very differently. Drawing on household panel data, television transcripts, and a randomised experiment in Turkey, this column shows that heavier inflation coverage on a viewer’s preferred channel raises their inflation expectations, an effect driven almost entirely by viewers of politically neutral channels. Partisan audiences barely respond, updating only to sources they already trust. Political segmentation of the media can therefore erode the common information base on which anchored expectations depend.
Over the past two decades, central banks have come to treat communication as a policy instrument in its own right, a way to steer the inflation expectations that drive spending and price setting. There is a catch, however: almost no household reads a monetary policy statement. People learn about inflation from the news, so the media sits between the central bank and the public.
A growing body of research shows that this intermediation matters a great deal. De Fiore et al. (2025) find that the tone with which US outlets cover Federal Open Market Committee (FOMC) decisions moves households’ inflation expectations, and that the effect is strongest precisely when inflation is high and volatile, while the underlying FOMC statements have little direct effect on households who never read them. How people make sense of inflation matters too. Wohlfart et al. (2021) show that the narratives households hold, many of them politically tinged, strongly predict whether they expect inflation to persist. And these beliefs are not a sideshow: they feed back into consumption and pricing decisions (Coibion et al. 2021), which is why communication that shifts them is valuable in the first place (Coibion et al. 2019).
These studies raise a question they were not designed to answer. What happens when the media itself is politically segmented, so that the very same inflation is reported very differently depending on which channel a household happens to switch on? In a polarised information environment, the media may not deliver a common public signal at all. It may instead deliver several, sorted by political allegiance. In recent work (Demiralp et al. 2026), we study exactly this, using Turkey as a natural laboratory.
A natural laboratory
Turkey is well suited to explore the question. Inflation has been high, volatile, and intensely political. Annual inflation in 2024 began at around 65%, peaked near 75% in May, and eased to about 44% by year-end. The television landscape is sharply polarised, and viewers largely watch channels aligned with their politics.
We combine three ingredients. The first is a new monthly household panel survey conducted by Koç University that records each respondent’s inflation expectations and their preferred prime-time news channel. The second is a corpus of prime-time news transcripts, scraped from public video archives, which we analyse with text methods to measure, channel by channel and month by month, how much each outlet talks about inflation and how optimistically or pessimistically it frames it. The third, described below, is a large randomised information experiment.
Two facts stand out before any regression. First, coverage of inflation differs dramatically across outlets and does not track actual inflation (Figure 1). Opposition-leaning channels mention inflation persistently and often; pro-government channels mention it far less; politically neutral channels sit in between.
Figure 1 Number of inflation mentions in prime-time TV news, 2024
Notes: Monthly average number of ‘inflation’ mentions per prime-time news programme, by channel (lagged one month). Pro-government channels are shown in orange, neutral channels in blue, and pro-opposition channels in black. Local elections took place on 31 March 2024.
Source: Demiralp et al. (2026).
Second, audiences sort along partisan lines (Figure 2). Voters of the incumbent party overwhelmingly watch pro-government channels, other voters watch opposition channels, and only about a fifth of each camp watches neutral outlets. The same macroeconomic reality is thus filtered through very different lenses.
Figure 2 Television viewership by 2023 vote
Notes: Share of respondents choosing each type of prime-time news channel, split by whether they voted for the incumbent party (AKP, orange) or not (black) in the 2023 general elections.
Source: Demiralp et al. (2026).
Does coverage actually move expectations?
The central challenge is that people choose their news, so a simple correlation between coverage and expectations could just reflect sorting. We address this by following the same individuals over time and asking whether changes in how much their initially preferred channel talks about inflation move their own expectations. This design absorbs fixed differences in political views and baseline beliefs, and we go further, comparing respondents with the same party affiliation, in the same province, in the same month, so that partisan messaging and local economic shocks are netted out.
The answer to the question – whether coverage actually move expectations – is yes. Greater inflation coverage on a household’s preferred channel leads to higher inflation expectations. The magnitude is economically meaningful but not extreme: a one-standard-deviation rise in inflation coverage raises expectations by roughly a tenth of their standard deviation, and the relationship survives even after we control for households’ own perceived past inflation. It therefore reflects genuinely new information that shifts forward-looking beliefs, not television merely echoing what shoppers already feel at the till. Consistent with this, coverage moves forward-looking expectations but not backward-looking perceptions of past inflation, which are anchored in personal experience. Media matters most precisely where personal experience offers the least guidance.
We find a similar, if anything larger, effect when we isolate plausibly exogenous variation: around local elections, campaign news crowds out economic reporting, temporarily lowering inflation coverage. Using this election-driven shift as an instrument confirms that coverage moves expectations rather than the reverse.
Strikingly, it is the sheer volume of coverage that matters, more than its tone. Once inflation is mentioned, optimistic reassurance does little to offset it, as if naming inflation cues memories of hard times. This helps explain why pro-government channels appear reluctant to raise the topic at all.
Who responds, and why
The average masks a sharp divide. The response is concentrated among viewers of politically neutral channels, for whom the effect is more than three times larger than average, while viewers of partisan channels barely react to changes in coverage. To understand why, we ran a pre-registered experiment that showed respondents short inflation narratives and randomly varied the outlet to which each narrative was attributed, mirroring how pro-government, pro-opposition, and neutral channels typically frame inflation.
The experiment clarifies the mechanism. Viewers of neutral channels update their expectations in response to the narratives regardless of tone or which outlet is named, consistent with weaker prior anchoring and greater openness to new information. Viewers of pro-government channels update mainly when the information is attributed to a pro-government source, a clear signature of politically aligned source credibility. Viewers of opposition channels do not systematically update to anything, consistent with strongly anchored priors or simple saturation, since inflation is already ever-present on the outlets they watch. Media effects, in other words, depend not only on the content of economic news but on the interaction between audience priors and perceived source credibility.
Why it matters beyond forecasts
Inflation expectations also carry political weight. Using the experimentally induced variation among neutral-channel viewers, we find that higher inflation expectations lower confidence in the government’s economic management and reduce stated support for the incumbent. The point estimates are suggestive rather than precise, but their direction is telling: a one-percentage-point rise in inflation expectations lowers the probability of supporting the incumbent by around 5% among neutral viewers, which even under conservative assumptions maps into a meaningful move in aggregate vote share. This is a plausible reason why partisan outlets have an incentive to minimise or maximise inflation coverage in the first place.
Implications for policy
The results point to an uncomfortable conclusion for expectation management. When inflation is high and expectations are weakly anchored, households lean heavily on intermediated information. But in a politically segmented media environment, the effectiveness of communication depends not only on the message but on who is perceived to be delivering it and on what audiences already believe. Strategies that rely on official announcements alone may fall flat where trust is politically uneven and where households inhabit distinct media ecosystems. Polarised media do not simply give people different facts; they weaken the common nominal anchor on which aggregate expectation formation depends. For central banks that have invested heavily in communication, that is a channel worth taking seriously.
Source : VOXeu
