Research

Abstract: Cross-sectional designs with time fixed effects are widely used in empirical macroeconomics, including Phillips curve estimation. We show that heterogeneous exposure to common shocks can bias slope estimates, conflate responses to idiosyncratic and synchronized slack, and overstate precision. We propose a two-step extension of interactive fixed effects that separates these responses under testable conditions. Using firms’ reported labor shortages across euro area sectors and countries, we find that a one standard deviation increase in a sector’s own tightness raises its inflation by about 0.25 percentage points – three times the time fixed effect estimate. A one standard deviation increase in synchronized tightness raises it by about 1.3 points, revealing a substantially stronger inflation response when shortages are pervasive.

Abstract: The literature established that a sovereign bond’s “convenience yield” premium diminishes when that country issues more debt. But how is this convenience yield affected when another country issues sovereign debt? Using high-frequency identification and debt management offices’ communication, we find that an increase in German or French debt reduces convenience yields across the eurozone. Spillovers to low-risk countries are one-for-one while those to riskier countries are smaller. To rationalize these findings, we develop a model with heterogeneous credit risk. Safe sovereign bonds are close substitutes to hedge against idiosyncratic risk, explaining large spillovers, while risky bonds are poor substitutes.

Abstract: Does faster productivity growth reduce inflation, and what are the implications for monetary policy in the AI era? Drawing on a panel of 17 advanced economies from 1890 and across six different monetary regimes, we show that it does -- but that the resulting disinflationary effect is almost fully offset by a commensurate increase in the real interest rate, leaving the appropriate nominal policy rate roughly unchanged. These two effects persist over time and across different monetary regimes and are robust to a wide range of alternative specifications. Applied to the current AI debate, our estimates suggest neither preemptive monetary easing nor preemptive tightening: over the long run, the two channels largely offset each other.