ECB paper builds an inflation-risk gauge out of newspaper sentences
Serguey Asael Shinder2 min readThe ECB published Working Paper No 3288, Inflation narratives and risk premia, by Luís Fonseca, Giulia Martorana, Fabian Schupp and Giovanni Trebbi, in its working paper series on 21 September 2026.
The authors take a textual route to a quantity that is normally read off bond prices. Running a technique the paper calls Causality Extraction over press reporting on inflation, they classify each piece of coverage by what it names as the cause, and build an indicator — NetDemand — from the gap between the count of demand explanations and the count of supply explanations. The euro area series is drawn from the Financial Times, the US series from the Wall Street Journal. They then set that indicator against inflation risk premia across the maturity spectrum in both currency areas, and report an inverse relationship that survives controlling for the composite PMI and for the VIX.
What it means
The claim being tested is an old one in asset pricing and is easy to state. If inflation is expected to come from supply, it arrives alongside weak growth, so an asset whose payoff falls with inflation is bad insurance and investors want paying to hold it — a positive risk premium. If inflation is expected to come from demand, it arrives alongside strong growth, the same asset hedges a recession, and the premium can go negative. So the sign of the premium encodes a belief about which shock dominates.
What the paper adds is an independent reading of that belief. Until now it was inferred from the premium itself, which makes the theory difficult to test without arguing in a circle. Counting what newspapers say breaks the circle, because the text is generated by a process that has nothing to do with the bond market. That the relationship holds after the PMI and the VIX are controlled for is the part doing the work: it says the text is not simply standing in for activity or for volatility.
The obvious objection is that a newspaper is not the marginal investor, and the paper does not claim otherwise. What it claims is an association, measured, in two regions, across maturities, and stronger in the United States when investors are more averse to risk, and in the euro area when inflation itself swings more widely. A working paper carries the views of its authors and not of the ECB, which the series states on every cover — and this one is a measurement exercise, not a policy proposal.