Research

I adapt the Bayesian Local Projection framework of Ferreira, Miranda-Agrippino, and Ricco (2025) to a panel setting to jointly estimate the employment and wage responses to monetary policy shocks across sixteen supersectors, using the narrative shocks of Aruoba and Drechsel (2024). I document substantial heterogeneity in the sign, magnitude, and timing of responses, and find that price-rigidity measures explain little of it—instead, more upstream and more heavily unionized sectors respond most strongly. I construct counterfactual aggregate responses under historical versus contemporary employment weights to assess how the shift toward services has reshaped monetary transmission to labor markets.

We estimate zip-level “housing cost sensitivities” (HCS)—the percent change in local home prices or rents per 1% increase in national home prices caused by Ben-David et al. (2024) mortgage-rate shocks. Using Zillow home-value and rent indices for over 13,000 zip codes, we find wide heterogeneity. Across the interquartile range, local home prices rise between 0.3% to 2.0%, while rents range from −1.5% to 1.1%. We find the HCS are strongly regressive: lower-income zips exhibit substantially larger responses for both prices and rents. Zip-level home-price HCS are highly correlated with the city-level inverse supply elasticity of Guren et al. (2021), yet meaningful unexplained variation remains. Beyond supply constraints, we find that demand amplification (natural amenities), land-cost pass-through, and housing-market liquidity all shape HCS. Our results quantify where mortgage-rate movements most strongly transmit into local housing costs and clarify the mechanisms underlying that heterogeneity.

This paper presents a method for jointly estimating local projections when working with panel data. I use the S⁴ model selection routine of Koop and Korobilis (2016) to illustrate how cross-sectional heterogeneity, dynamic interdependencies, and static interdependencies can be jointly estimated with local projections, and compare the resulting impulse responses with those from an analogous panel VAR.