Author: Sunho Lee(Bank of Korea), Kyu Ho Kang(Korea University)

<요약>

Interactions between macroeconomic variables and the government bond market are of major interest to policymakers and researchers. This paper proposes a Bayesian large vector autoregression model of the yield curve and macroeconomic variables with the no-arbitrage restriction and develops a novel Bayesian method for its estimation. We demonstrate the efficiency of the proposed Markov chain Monte Carlo algorithm using an empirical application with the U.S. yield curve and 28 macro-financial variables, utilizing data from 1987 to 2022. To illustrate how the estimated model can be used to examine the interactions between macroeconomic variables and bond yields, we conduct scenario analyses as of December 2022, which reveal the following results. First, the FOMC members at the time held a more pessimistic view on inflation and a more optimistic view on the real economy compared to our model’s predictions, leading to a hawkish projected path of the federal funds rate. Second, a stress test conducted under a scenario of rapidly increasing financial uncertainty reveals a prolonged negative impact on factors of production. We provide the Julia package TermStructureModels.jl, which implements our methodology.