Imperfect Information, Composition of Demand Shocks, and the Flattening of the Phillips Curve

We study why inflation responds differently to economic activity over time. Using survey data covering the universe of Japanese firms, we show that firms are unable to perfectly distinguish aggregate from sector-specific demand changes, leading to positively correlated expectations about these two components. We develop a model with imperfect information that reproduces this pattern and predicts that higher relative volatility of sector-specific demand reduces the sensitivity of inflation to changes in aggregate demand, thus flattening the Phillips curve. Testing this prediction with Japanese data from 1976 to 2022, we find that increases in the volatility of sectoral demand shocks explain significant changes in the Phillips curve slope over the sample period. Our results provide a novel explanation for the flattening of the Phillips curve: the composition of shocks -- not just their magnitude -- critically affects the sensitivity of inflation to aggregate demand.