报告人：朱丹 教授（莫纳什大学）

时间：2025年09月10日 15:00-

地点：理科楼LA103

摘要：Recent evidence of a flattened Phillips curve calls for methods that capture more than average relationships. We propose a semiparametric time-varying parameter distributional regression (TVPDR) model that estimates full conditional distributions, allowing asymmetric and dynamic effects of macroeconomic drivers on inflation. Applied to U.S. inflation from 1982–2024, the model delivers superior point and density forecasts compared to mean- and quantile-based alternatives, especially during recessions. It identifies sharp spikes in deflation risk during downturns (exceeding 0.5 in 2009 and 0.6 in 2020) and surges in excessive inflation risk during the post-pandemic recovery (peaking near 0.9 in 2022). Shapley decompositions show that deflation risk is mainly driven by demand-side weakness and inflation persistence, while excessive inflation risk reflects supply-side shocks, particularly energy and food prices. Scenario analyses further reveal that unemployment fluctuations mainly affect the lower tail of the distribution, whereas energy prices dominate the upper tail. Taken together, the results show that the lower end of the inflation distribution remains sensitive to unemployment, while the upper end is less responsive and instead shaped by supply shocks. This asymmetric and time-varying sensitivity explains why the aggregate Phillips curve relationship appears flattened.

邀请人：张志民

欢迎广大师生积极参与！